Tonality: An Owner’s Manual

OXFORD STUDIES IN MUSIC THEORY

Series Editor Steven Rings

Studies in Music with Text, David Lewin

Metric Manipulations in Haydn and Mozart: Chamber Music for Strings, 1787-1791, Danuta Mirka

Songs in Motion: Rhythm and Meter in the German Lied, Yonatan Malin

A Geometry of Music: Harmony and Counterpoint in the Extended Common Practice, Dmitri
Tymoczko

In the Process of Becoming: Analytic and Philosophical Perspectives on Form in Early Nineteenth-
Century Music, Janet Schmalfeldt

Tonality and Transformation, Steven Rings

Audacious Euphony: Chromatic Harmony and the Triad's Second Nature, Richard Cohn

Music as Discourse: Semiotic Adventures in Romantic Music, Kofi Agawu

Beating Time and Measuring Music in the Early Modern Era, Roger Mathew Grant

Mahler's Symphonic Sonatas, Seth Monahan

Pieces of Tradition: An Analysis of Contemporary Tonal Music, Daniel Harrison

Music at Hand: Instruments, Bodies, and Cognition, Jonathan De Souza

Foundations of Musical Grammar, Lawrence M. Zbikowski

Organized Time: Rhythm, Tonality, and Form, Jason Yust

Flow: The Rhythmic Voice in Rap Music, Mitchell Ohriner

Performing Knowledge: Twentieth-Century Music in Analysis and Performance, Daphne Leong

Enacting Musical Time: The Bodily Experience of New Music, Mariusz Kozak

Hearing Homophony: Tonal Expectation at the Turn of the Seventeenth Century, Megan Kaes Long

Form as Harmony in Rock Music, Drew Nobile

Desire in Chromatic Harmony: A Psychodynamic Exploration of Fin de Siècle Tonality, Kenneth M.

Smith

A Blaze of Light in Every Word: Analyzing the Popular Singing Voice, Victoria Malewy
Sweet Thing: The History and Musical Structure of a Shared American Vernacular Form, Nicholas
Stoia

Hypermetric Manipulations in Haydn and Mozart: Chamber Music for Strings, 1787-1791, Danuta
Mirka

How Sonata Forms: A Bottom-Up Approach to Musical Form, Yoel Greenberg

Exploring Musical Spaces: A Synthesis of Mathematical Approaches, Julian Hook

The Musical Language of Italian Opera, 1813-1859, William Rothstein

Tonality: An Owner's Manual, Dmitri Tymoczko

Tonality: An Owner’s Manual
An Owner’s Manual

DMITRI TYMOCZKO
 Oxford University Press is a department of the University of Oxford. It furthers the University’s
objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a
registered trade mark of Oxford University Press in the UK and certain other countries.
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America.
© Oxford University Press 2023
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, without the prior permission in writing of Oxford
University Press, or as expressly permitted by law, by license, or under terms agreed with the
appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of
the above should be sent to the Rights Department, Oxford University Press, at the address above.
You must not circulate this work in any other form and you must impose this same condition on any
acquirer.
Library of Congress Control Number: 2022042508
ISBN 978–0–19–757710–3
eISBN 978–0–19–757712–7
DOI: 10.1093/oso/9780197577103.001.0001
 to anyone who cares enough
to make it all the way through
 Contents

Preface and Acknowledgments

1. Implicit Musical Knowledge

1. Gesualdo’s trick
2. The quadruple hierarchy
3. Philosophy
4. Statistics
5. Schema
6. Outline

Prelude: Transposition Along a Collection

2. Rock Logic
1. A melodic principle
2. A harmonic principle
3. A first loop family
4. Two more families
5. Shepard-tone passacaglias
6. Minor triads and other trichords
7. A fourth family
8. Other modalities
9. Function and retrofunction
10. Continuity or reinvention?

Prelude: The Tinctoris Transform

3. Line and Configuration
1. The imperfect system
2. Voice exchanges
3. Other intervals
4. The circle of diatonic triads
5. Voice exchanges and multiple chord types
6. Four-voice triadic counterpoint
7. Thinking within the chord
8. Seventh chords
9. Harmony and counterpoint

Prelude: Sequence and Function

4. Repetition
1. Repetition reimagined
2. Repeating contrapuntal patterns
3. The geometry of two-voice sequences
4. Three voices and the circle of triads
5. Three voices arranged 2 + 1
6. Four voices
7. Contrary-motion sequences
8. Melodic sequences and near sequences
9. Near sequences
10. Sequences as reductional targets

Prelude: Three Varieties of Analytical Reduction

5. Nonharmonic Tones
1. The first practice and the SNAP system
2. Schoenberg’s critique
3. Monteverdi’s “Ohimè”
4. The standardized second practice
 5. A loophole
6. After nonharmonicity

Prelude: Functional and Scale-Degree Analysis

6. The Origins of Functional Harmony

1. The logical structure of protofunctionality
2. Similarities and differences
3. Origin and meaning
4. Harmony and polyphony
5. The Pope Marcellus Kyrie
6. A broader perspective
7. “I cannot follow”

Prelude: Could the Martians Understand Our Music?

7. Functional Progressions
1. A theory of harmonic cycles
2. A more principled view
3. Rameau and Bach
4. Functional melody, functional harmony
5. Fauxbourdon and linear idioms
6. Sequences
7. Bach the dualist

Prelude: Chromatic or Diatonic?

8. Modulation
1. Two models of key distance
2. Enharmonicism and loops in scale space
3. Minor keys
4. Modulatory schemas
5. Up and down the ladder
 6. Modal homogenization and scalar voice leading
7. Generalized set theory

Prelude: Hearing and Hearing-As

9. Heterogeneous Hierarchy
1. Strategy and reduction
2. Two models of the phrase
3. Chopin and the Prime Directive
4. An expanded vocabulary of reductional targets
5. Simple harmonic hierarchy
6. The four-part phrase
7. Grouping, melody, harmony
8. Beyond the phrase: hierarchy at the level of the piece

Prelude: Why Beethoven?

10. Beethoven Theorist

1. Meet the Ludwig
2. From schema to flow
3. The Tempest
4. The Fifth Symphony
5. The “Pastorale” sonata, Op. 28
6. Schubert’s Quartettsatz
7. The prelude to Lohengrin

Conclusion

Appendix 1: Fundamentals
Appendix 2: Deriving the Spiral Diagrams
Appendix 3: Sequence and Transformation
Appendix 4: Corpus Analysis, Statistics, and Grammar
Terms, Symbols, and Abbreviations
Bibliography
Index
 Preface and Acknowledgments

This book grew like a city, each gleaming draft built over the ruins of a
previous version. The earliest stratum is pedagogical. After many years of
teaching tonal harmony, I had accumulated an unstructured miscellany of
ideas, techniques, and analyses. Some were the product of computational
corpus study—a practice I adopted around 2000, first using MIDI files and
Max/MSP, and then using Python, music21, and actual scores. My plan was
to use this material to put pressure on what I took to be theoretical
oversimplification. I wanted to demonstrate, against Schenkerian
orthodoxy, that harmonies in baroque and classical music followed a largely
nonhierarchical chord-to-chord grammar. I wanted to show that upper-voice
configurations in four-voice counterpoint obeyed a simple grammar of hand
positions. I wanted to argue that nonharmonic tones were not “merely
decorative,” and to trace the “irreducible seventh” back to the Renaissance.
I wanted to identify protofunctional tendencies in sixteenth-century music.
The goal was to provide more detailed information than could be found in
standard textbooks, augmented by conceptual heuristics I used in the
classroom.
This led to the aesthetic or critical layer. For the more I dove into the
grammatical minutiae, the more I began to question the value of
disinterested description: by avoiding issues of musical relevance, I was
tacitly accepting assumptions that should instead be put into question. This
felt increasingly problematic, pedagogically, aesthetically, and even
morally. Why should twenty-first-century musicians care about the detailed
grammar of sixteenth- or eighteenth-century practice? Who needs a truly
accurate account of largely extinct functional-harmonic conventions? What
is the aesthetic value of recapitulation? Though I love music, and consider it
intrinsically worthy of study, I felt increasingly strongly that aesthetic
significance should be a component of grammatical inquiry. Music theory
without philosophy is empty.
 The final, theoretical layer introduced the spiral diagrams that were to
play a central role in the completed book. Originally designed as
pedagogical simplifications, I gradually came to realize that they were
something more. For by freeing the mind from the burden of imagining
higher-dimensional voice-leading geometry, they left more energy for
understanding genuinely musical relationships, often in a way that
connected past and present. A major development here was my discovery
that Princeton graduate composers were interested in ideas I had taken to be
purely theoretical and historical. Their influence pushed me to connect
sequences to repetition and transformation more generally, to think about
how one might use complex motivic networks in contemporary contexts,
and to consider how pieces like the Waldstein might serve as models for
contemporary composition. More significantly, these connections led to the
idea of the quadruple hierarchy, one of the central themes in the manuscript
as it now stands. The result is that the oldest and newest sections of the
book are found side-by-side: for while I had long used the heuristics of
OUCH theory in the classroom (§3.6), I happened upon the general picture
of chords-as-scales only after I thought the book was essentially finished
(§3.7).
This last revision, so drastic and so late, raised the disturbing possibility
that I had ceased writing a book in any conventional sense: what had begun
as a journey had turned into a way of life, my manuscript evolving into a
repository for my latest thinking on all things music. Every time I learned
something new, or changed my mind, I updated the relevant file on my
computer. Having faced the prospect of spending the rest of my life
endlessly writing and rewriting the same words—along with the attendant
familial disapproval—I resolved to stop. The result is this snapshot of my
thinking as of the time of publication. Already I worry that readers will be
reluctant to climb so far out on my own little conceptual limb. And though I
have tried to be clear, I am aware that the material is extremely challenging.
Many people helped me along the way. Thirty-five years ago, John
Stewart sat next to me at the piano and showed me his version of OUCH
theory, planting seeds that were to bear fruit decades later. In a graduate
seminar at UC Berkeley, David Huron taught me about corpus study and
computational musical analysis. Richard Cohn introduced me to musical
geometry and provided valuable support when my thinking was in its early
stages. Clifton Callender, Noam Elkies, Rachel Hall, and Ian Quinn joined
me in working out the geometrical approach to voice leading.
More recently, Nick DiBerardino, Robert Andrew Scott, Jacob Shulman,
and Steve Taylor provided invaluable comments on drafts of the
manuscript. Robert’s incredibly detailed copyediting, of both music and
examples, went well beyond the call of duty. Jason Yust not only read the
manuscript, and shared portions of it with his class, but also served as a
useful sounding board for many half-baked thoughts. Mark Liberman was a
gracious consultant about linguistic and statistical matters. Michael Bruschi
and Christopher Peacocke read early drafts through chapter 5, the former
leading me to add preludes to each chapter. I have already mentioned the
graduate students in Music 527: Jenny Beck, Yihan Chen, Wei Dai, Nick
DiBerardino, Molly Herron, Natalie Miller, James Moore, Mauro Windholz,
and Zhoushu Ziporyn. The next semester, undergraduates in Music 306
user-tested portions of the book—leading to yet more rewriting. Others who
have helped me, by some combination of reading excerpts, answering
emails, and providing materials, include Giovanni Albini, Matthew Arndt,
Fernando Benadon, Ed Berlin, Eli Berman, William Caplin, David Cohen,
Michael Cuthbert, Harrison Davis, Donnacha Dennehy, Chris Douthitt,
Noam Elkies, David Feurzeig, Kenneth Forkert-Smith, Andrew Gelman,
Robert Gjerdingen, Aaron Grant, Dan Harrison, James Hepokoski, Julian
Hook, Nori Jacoby, Corey Kendrick, Megan Long, Rudresh Mahanthappa,
Nathan John Martin, Nathaniel Mitchell, Marco Buongiorno Nardelli, Drew
Nobile, Ali Rahmjoo, Peter Schubert, Ian Sewell, David Temperley, Yo
Tomita, and Dan Trueman. I am sure that this list is incomplete, and I
apologize to anyone I have missed.
My children, Lukas and Katya, allowed me to share their journey into
music, gently teaching me that I wasn’t as smart as I thought. Several of
their pieces turned into musical examples.
Last but not least comes my partner, emotional lodestone, and intellectual
hero Elisabeth Camp, who makes a cameo in a few footnotes. I am
embarrassed that neither of my books is good enough to be dedicated to her.
 1
Implicit Musical Knowledge

Twenty-first-century musicians inherit a canon of notated, largely triadic

music originating before Palestrina and stretching past Joplin. This triadic
tradition has long been central to classical concert life, the focus of
orchestral performances and chamber-music recitals. It is the music of the
movies, which in their most dramatic moments reach for the tropes of the
nineteenth-century orchestra. It occupies a central place in undergraduate
curricula, which often emphasize the imitation of earlier styles even though
analogous acts of mimicry are unknown in other areas. Its rich syntax
attracts significant scientific attention. And for all its distance from
contemporary music, it is the yardstick against which recent work is judged
—one that tells us our efforts are alternately too simple or too complex, too
intuitive or too cerebral. For many listeners, the baroque, classical, and
Romantic eras represent a pinnacle of musical achievement, the perfect
balancing of emotion and intellect. The intervening centuries may have seen
their share of intellectual progress, but the notated triadic tradition remains,
rightly or wrongly, the epitome of learned composition.
And yet this music remains oddly foreign. The central problem is that
musical knowledge is largely implicit—which is to say, embedded in
compositional, improvisational, and perceptual habits that may not be
available to conscious reflection. This is why composers can write without
being able to explain what they are doing and why listeners can understand
without being able to describe what they hear. And it is why music-making
has often been taught through ungrounded injunctions of the form “do this
when you find yourself in this situation.” Modernism brought with it a
distrust of inherited convention and a sense that substantial portions of
musical practice were outdated or dogmatic. One consequence was the
withering of a vast tradition of implicit knowledge, with composers
proposing artificial systems to replace the natural grammars of earlier times.
Another was the rejection of a whole host of very general concepts,
including scale, sequence, consonance, and nonharmonic tone, all on the
grounds that they were associated with the specific idioms of an outdated
past. A third was a shift in the economic value of implicit and explicit, as
musical culture came to value the ability to describe functional-harmonic
grammar more than the ability to speak the language. Finally, lacking an
embodied compositional counterweight, pedagogy and scholarship
gravitated toward strict laws and speculative generalizations. Where
“composer theory” leans toward the piecemeal and empirical, “theorist
theory” inclines to the systematic a priori.
Compounding these difficulties is the fact that our musical ancestors
inhabited a vastly different intellectual world, bounded by conventions that
we now consider unjustified: near-universal avoidance of parallel perfect
intervals, drawing a sharp boundary between consonance and dissonance,
using only some of the available triadic progressions, limiting scalar
vocabulary to a small number of modes, satisfying the formal demands of
fugue and sonata, returning to the opening key at the end of the piece, and so
on. These conventions were born of a Platonic conception of music as
objective and extra-human—once basic science, part of the quadrivium
alongside mathematics and astronomy, music has been demoted to a kind of
fancy fun, its “rules” reconceived as conventions rather than laws of nature.
The result is a past that is simultaneously paradigmatic and problematic. Our
pedagogy tells a story of self-alienation: introductory harmony textbooks
devoted to functional-harmonic conventions that have largely lost their
force, counterpoint manuals shockingly similar to those of the sixteenth
century, and advanced analytical methodologies explicitly designed to
establish the aesthetic superiority of the White Germanic Genius.
All of which leads to a dilemma. On one side is the imperative of
diversification: recognizing the many varieties of musical pleasure,
acknowledging the systematic prejudice that led to a White male canon, and
grappling with pedagogical traditions that no longer seem appropriate to
contemporary musicians. On the other is the imperative of conservation,
expressed by the belief that for all its problems, the notated tradition is vital
and important in a way that contemporary markets may not always recognize
—and that it can still be worthwhile to engage in the relatively unprofitable
act of performing and studying this music, or composing new pieces that
show its influence. We need a new way of connecting our past to our future,
reconceptualizing music not from the standpoint of the native speaker, but
from the standpoint of the sympathetic foreigner who is interested in
something new. We need to scrub away the irrelevant dogma and reveal the
core practices of continued interest.
I have written this book because I believe recent music theory allows us to
understand music in a way that is more responsive to contemporary
intellectual and aesthetic values, explicitly describing a vast field of formerly
implicit knowledge. Indeed, I think we can now generalize the core
techniques of earlier music, including voice leading, counterpoint, sequence,
motivic development, nonharmonic tones, and modulation, so that they
apply to any chords and scales whatsoever—recognizing and using those
techniques in a vast array of contexts, from Renaissance polyphony to jazz,
rock, and contemporary notated composition. To call this a “generalized
tonal theory” would be a misnomer, for its techniques are broad enough to
embrace extremes of consonance and dissonance, encompassing both
tonality and atonality. My goal in this book is to explain this generalized
theory in a way that is both comprehensible and comprehensive, starting
with the analysis of existing pieces and considering how these ideas might
also be used to make new music.
The properties I am interested in have clear and audible effects, producing
distinctive sonic signatures that are easily identified: the characteristic sound
of rock harmony or J. S. Bach’s counterpoint, Chopin’s distinctive approach
to chromaticism, Beethoven’s fondness for contrary-motion sequences. But I
am primarily interested in musical knowledge as it is revealed by
compositional and improvisational behavior, and my starting point will
typically be the composer’s perspective rather than the listener’s. The
following chapters will explore four distinct characterizations of musical
knowledge. First, it is mathematical, a matter of knowing how to move
through abstract spaces of musical possibility. This aspect is highly
susceptible to generalization and modeling, revealing surprising structure
across a very wide range of practices. Second, and somewhat opposed to
this, it is particular, a matter of internalizing idioms and tendencies
conforming to no general laws. Like knowledge of fashion or slang,
knowing a musical style is often a matter of knowing to do things this way
and not another, for no particularly good reason. Third, musical knowledge
is often implicit—which is to say, embedded in compositional and
improvisational practice in a way that is not available to conscious
reflection. In this respect, it is comparable to a natural language whose
grammar displays complex structure going beyond the explicit
understanding of its speakers. Fourth and finally, musical knowledge has
often been dogmatic, shaped by avoidances we now consider unmotivated or
superstitious.
Each of these aspects of musical knowledge calls for different theoretical
tools. The theory of voice leading, pioneered by Richard Cohn, uses
geometry and mathematics to elucidate the connection between harmony and
melody, revealing fascinating structure in specific pieces, in the work of
individual composers, and across genres. Statistical corpus studies,
cultivated by David Huron and brought to the masses by Michael Cuthbert,
reveal tendencies invisible to conventional analytical methods. The
particularity of musical knowledge is the purview of schema theory,
deriving from Leonard Meyer by way of Robert Gjerdingen. Finally, there is
analytic philosophy, introduced into music theory by Milton Babbitt and
helping us separate dogmatic assertion from valid logical argument.1
Together these different strands of inquiry allow us to pursue music theory
as continuous with cognitive science, modeling the internal mental
representations that underlie compositional competence—or what we know
when we know how to compose.2
This knowledge is technical, detailed, and difficult to acquire. It is the
knowledge of exactly which shades of paint combine to produce a vivid
sunset, of which signs mark an animal’s track through a forest, of how to
write a compelling legal brief. It is being able to navigate your house in the
dark. This book exists because I love peering over other composers’
shoulders to see intelligible patterns emerging from the tens of thousands of
decisions that jointly produce a musical composition. (And, of course, trying
to hear those patterns as I listen or play at the piano.) Those who are bored
by detailed analysis should close this book immediately. Those who stay,
hopefully, will find it exhilarating to go beyond pedagogical platitudes,
understanding musical practice in real and specific detail—not least because
it reveals a music that is less systematic, and sometimes differently
systematic, than textbooks make it out to be.
Done well, analysis gives a sense of ownership, allowing us to understand
how flesh-and-blood humans might find their way to composing beloved
works. I am aware that this metaphor can suggest rapaciousness, capitalism,
or untrammeled power.3 Yet we speak of taking ownership of our own
words, and in our society people sometimes come to own objects of
significant public concern—a historic home, a well-known work of art, an
ecologically significant wetland, or a beloved sports team. In these cases, to
be an owner is to be a custodian, temporarily inhabiting an office with
responsibilities and duties that we might not have chosen. My suggestion is
that contemporary musicians are in an analogous position. My hope is that a
vigorous reexamination of the past may ease the sense of conflict between
conservation and diversification, allowing us to see the commonalities
between such figures as Carlo Gesualdo and Rudresh Mahanthappa,
Heinrich Schütz and Carole King, Fanny Mendelssohn Hensel and Meredith
Monk, Dmitri Shostakovich and McCoy Tyner. This is in keeping with my
belief that jazz is a natural continuation of an extended common practice
beginning before Palestrina and including musicians like Debussy and
Stravinsky—a repository for the sort of implicit knowledge that was once
pervasive.

1. Gesualdo’s trick

Gesualdo’s “Moro Lasso” opens by sliding from C♯ major to A minor to B

major to G major (Figure 1.1.1). Where earlier theorists might have seen
arcane contrapuntal wizardry, or perhaps a specific chromatic idiom, twenty-
first-century musicians will recognize a general technique: outer-voice tenths
moving downward by semitone, forming triads by alternately adding a minor
third above and below this dyadic core.4 One could perhaps label the first of
these a “progression between hexatonic poles” and the second a “standard
iv6–V,” but this would bifurcate a unified sequence into two unrelated
harmonic moves.5 The pattern appears in both three and four voices, and can
be found throughout the Western tradition (Figure 1.1.2).
Figure 1.1.1. The opening of Gesualdo’s “Moro Lasso” (1611). Underneath, I analyze the underlying
three-voice schema: two voices moving down by semitone, harmonized by a minor third either above
or below this interval. This pattern is then reregistered so that the parallel thirds are always in the
lowest voices.
Figure 1.1.2. Gesualdo’s trick in Beethoven’s Seventh Symphony (1812), IV, mm. 136–146;
Schubert’s “Morgengruß” (1823), mm. 12–15; and Debussy’s Prelude to the Afternoon of a Faun
(1895), m. 107. Beethoven and Debussy use three voices; Schubert uses Gesualdo’s four-voice
version.

When confronted with a passage like this, I try to follow what I call the
Prime Directive:
whenever you find an interesting musical technique, try to generalize it to every possible
chordal and scalar environment.

In this case we have two voices moving in parallel with the third alternating
between prime and inverted forms of a sonority, arrayed in close position
and without voice crossings. Thus generalized, Gesualdo’s trick can be
found in many other places: Figure 1.1.3 shows Mozart using it in the first-
movement development of his C major piano sonata K.309, now with
descending minor rather than major thirds; Figure 1.1.4 shows one of its
several appearances in The Rite of Spring, with the parallel voices forming
major sevenths; Figure 1.1.5 shows Rudresh Mahanthappa using it in the
twenty-first century. Figure 1.1.6, meanwhile, applies the technique in
diatonic space whose unit of distance is the scale step. Here two voices are
separated by diatonic third, moving downward in parallel, with the last voice
alternately a fifth and sixth above the bass. What results is the descending-
third “Pachelbel” sequence (rechristened the “Romanesca” by Robert
Gjerdingen), a central idiom of both modal and functionally tonal
composition.6 From the standpoint of the Prime Directive, Gesualdo,
Stravinsky, Mahanthappa, and Pachelbel are all doing the same thing.
Figure 1.1.3. In mm. 73–76 of K.309, I, Mozart uses Gesualdo’s trick but with parallel minor thirds
rather than major.
Figure 1.1.4. The second measure of R114 in The Rite of Spring. Here the parallel voices form major
sevenths and articulate a melody; the third voice is alternately a tritone above the bottom voice and a
tritone below the top.
Figure 1.1.5. Rudresh Mahanthappa’s 2006 composition “The Decider” linearizes Gesualdo’s trick,
moving the perfect fourth in parallel, the middle voice alternately a major second below the top voice
and a major second above the bottom.

Figure 1.1.6. The Pachelbel sequence moves two voices in parallel diatonically, alternately adding the
third above and below.

We can go further. Any chord can be inverted so as to preserve any two of

its notes; by moving these in parallel we can construct analogues to the
three-voice passages we have just considered. Figure 1.1.7 shows a
prominent theme from The Rite of Spring, in which superimposed major and
minor thirds create a delicious polytonal clash. Inversion around the middle
interval exchanges upper and lower thirds.7 Rather than mechanically
alternating between these two chord-forms, Stravinsky moves the upper
voices diatonically along the white-note collection, superimposing three
distinct scales. (Here he exploits the fact that the diatonic scale’s two-step
intervals, its “thirds,” are either three or four semitones large, the top and
bottom intervals of his chord.) The end product is an unusual blend of
horizontal and vertical coherence: the harmonies all form 014 trichords
while the melodies move along three diatonic scales. For all its near-
parallelism, it is a remarkable feat of counterpoint.
Figure 1.1.7. At R57 of The Rite of Spring, Stravinsky uses Gesualdo’s trick, inverting a four-voice
sonority so as to preserve its eight-semitone middle interval.

Collectively, these analyses exhibit several features of my approach to

music theory. First, a composer-centered perspective that uses modern tools
to generalize and extend implicit cognition: theory as “rational
reconstruction” of the compositional process. Typically, my goal will be to
produce sketches that are simple enough that readers could imagine
inventing them, while also being rich enough that readers could imagine
expanding them into the finished piece. Though this activity is not science, I
imagine it to be within shouting distance of that enterprise, broadly
conceived. Second, a fluid understanding of stylistic boundaries such that
modal, functionally tonal, and nontonal music participate in a shared
dialogue. Third, the identification of specific techniques used throughout the
history of Western music, a fundamental language of counterpoint grounded
in the abstract structure of musical space, and encompassing both schematic
routine and idiosyncratic innovation. Fourth, analyzing triadic music using
dyadic logic, embodied by the parallel voices in the preceding analyses.
Fifth and finally, an implied aesthetic attitude: I find beauty in such
connections, and write for those who take joy in the thought that similar
techniques can be found from Gesualdo to the present.
The Prime Directive functions as an analytical maxim encouraging us to
ask whether superficially different idioms arise from a unified musical logic.
It is not necessary that composers themselves understood that logic; indeed,
it is more likely that their practice developed intuitively and without any
abstract theoretical framework. Yet the Prime Directive can still serve an
explanatory function, allowing us to understand why certain idioms might
have survived. (For example, the preceding examples all combine harmonic
consistency with parallel stepwise motion in two voices, features valued by
composers in many different styles.) It can also serve a generalizing
function, allowing us to adapt earlier techniques to new harmonic domains.
But we can sometimes find moments where musicians seem to exhibit a
more general understanding: in Figure 1.1.8, for example, McCoy Tyner
builds a stepwise descending line using close-position triads in two different
scales, F mixolydian and E♭ minor pentatonic; because of the scale change,
the pentatonic “triad” looks superficially like a fourth chord.8 In §9.3 we will
find Chopin doing something similar, using the same voice-leading
techniques in a variety of different contexts. Here we see a more abstract
understanding of contrapuntal possibility, one that detaches musical
techniques from specific chordal and scalar environments.
Figure 1.1.8. In mm. 18–23 (1’27” ) of his solo on “Passion Dance” (from The Real McCoy), McCoy
Tyner plays a descending line that moves from close-position diatonic triads to close-position
pentatonic triads and back, deploying the same technique in different scales. Black noteheads on the
bottom staff are not played.

2. The quadruple hierarchy

The Prime Directive leads to my guiding idea, that Western music displays a
hierarchical structure in which the same basic procedures occur on multiple
levels simultaneously. This is the quadruple hierarchy of surface voices
moving inside chords that move inside more familiar scales that are
themselves moving through chromatic space—or as I will sometimes call it,
the collectional hierarchy, as the number of levels can vary. Figure 1.2.1
presents a clear example, the pianist’s right hand moving sequentially along
three voices that articulate the familiar “omnibus” schema.9 It is not
unreasonable to think that these triadic voices form a three-note scale, a
series of melodic positions available throughout the piano’s range, with
right-hand melody moving systematically along this scale very much like
“doe a deer” moves along the diatonic collection. Yet these three voices form
triads that would normally be considered chordal.
Figure 1.2.1. In mm. 38–41 of Beethoven’s Op. 54, II, voices move systematically along the top three
notes of the omnibus pattern, as if they formed a three-note scale.

The quadruple hierarchy opens a new analytical project of determining the

precise combination of background and foreground motion that reproduces a
given passage. In some cases, this can lead to unexpected insights. Figure
1.2.2 analyzes the verse of the Beatles’ “Eight Days a Week”; if we are
willing to treat the harmonies as small scales, then we can say that the
melody alternates between two different scale degrees, 64–63, 64–63, and
63–64, moving up to degree 65 to finish the phrase.10 What is interesting is
that the melody returns to its starting pitch but not its starting scale degree:
the F♯ that starts the verse is, on this analysis, scale degree 64, while the F♯
that finishes it is scale degree 65. Melody and harmony thus cancel out, the
chords continually sinking while the melody ascends in order to stay fixed in
register—countervailing motion on two hierarchical levels. Here, modeling
chords as scales allows us to assert two seemingly contradictory
observations: that the music repeats more or less exactly, and that it
articulates a continuous descent at a deeper, scalar level. The result is a
psychologically complex repetition that also continues an ongoing process of
scalar descent.11
Figure 1.2.2. “Eight Days a Week” returns to the same pitch but a different scale degree.
 One obstacle to recognizing the quadruple hierarchy is our difficulty
imagining how transposition can be applied to notes outside the relevant
collection. For example, it may seem impossible to transpose a figure like
C–D–E along the C major triad, since the note D is foreign to that harmony.
Here it helps to realize that musical objects can contain intervals at multiple
hierarchical levels. Figure 1.2.3 shows two interpretations of the C–D–E
figure, one on which the D is an incomplete neighbor a scale step above the
initial C, and another on which it is a scale step below the final E; either
interpretation allows us to move the figure along the C major triad.12 This
same strategy works when we need to transpose chromatic figures such as
C–C♯–D along the diatonic scale, or virtually any collection of notes along
any other. Indeed, we will see that this idea is implicit in a wide range of
practices from motivic development to the “tonal answer” of baroque
counterpoint.

Figure 1.2.3. Transposing the motive C–D–E along the C major triad. The top line interprets the
second note as being a scale step above the first; the bottom line interprets it as a scale step below the
second. These lead to different forms on G.

Another obstacle is what might seem like an ontological gulf between

chord and scale. The paradigmatic chord is a concrete object localized in
both pitch and time, its notes situated in specific octaves and sounding
together. The paradigmatic scale is instead a collection of melodic slots that
are localized neither in time nor in register. Imagine hearing C4, E4, G4 in a
C major context. While we can typically answer questions like “what voice
is sounding the chord’s fifth, and what octave is it in?” we cannot answer the
analogous questions about notes not currently sounding. In what octave is
the leading tone? The question makes no sense: the leading tone is a
potentiality, a melodic location existing a step below every tonic and two
steps above every dominant. It is in every octave and none, a pitch class
rather than a pitch.
However, this distinction is determined by function rather than inherent
musical structure. In twentieth-century languages such as impressionism or
jazz, seven-note collections can act as concrete chords. Conversely, smaller
collections can act like scalar fields of play. Figure 1.2.4 shows the first half
of the remarkable mid-sixteenth-century motet Heu me Domine, likely
written by Vicente Lusitano. This triadic background is as abstract and
rigorous as anything in Ligeti or Reich, ascending semitonally from E3 to
A6 before falling back to the cadence. Virtually all the melodic lines can be
generated from this structure, which does not explicitly appear in the music;
instead, the concrete musical voices switch between “background” scale
degrees so as to remain within their registral bounds, generating musical
variety while avoiding contrapuntal malfeasance (e.g., second-inversion
triads and forbidden parallels). This is the inverse of the pattern we found in
“Eight Days a Week,” the concrete lines moving downward along the ever-
ascending background—like walking on a treadmill, expending energy just
to stay in the same place. The result is a paradoxical combination of registral
stability and endless ascent, reminiscent of the barber-pole illusion, known
to musicians as “Shepard tones.”13
Figure 1.2.4. A reduction of the first half of Lusitano’s 1553 motet Heu me domine. Rhythmic
durations have been halved. The background accounts for virtually all the notes in the piece.

This leaves analysts with a choice between two different kinds of

reduction. In a piece like The Well-Tempered Clavier’s first C major prelude,
we can understand figuration as a purely rhythmic phenomenon, outlining a
series of concrete voices whose notes belong to particular registers. In Heu
me domine, the reduction produces something more abstract, a scale-like
structure whose voices can appear in any octave. Schenkerian theory does
not recognize this distinction, its reductions typically delivering concrete
pitches, with octave shifts understood as a matter of “register transfer”—
displacing a chorale’s voices from their “true” or “obligatory” home. When
we reduce to a scalar background, there is no “obligatory register” since the
background object is a scale existing in every octave and none. As we will
see, this perspective opens a host of new analytical possibilities, allowing us
to reconceive some portion of Schenkerian practice in a more scientifically
reputable way.14
Implicit in these examples is the idea that scales can be linked by
particular voice leadings. Typically, it is the regularity of the surface that
leads us to postulate voice leadings at the scalar level. It is natural to read
Figure 1.2.5 as presenting a surface voice moving regularly along a scalar
field whose notes are themselves shifting. That is: the abstract melodic slot F
moves up by semitone to F♯ while the surface voice continues to articulate
the same series of melodic intervals (“up by two, down by one”). In A
Geometry of Music, I used the metaphor of a frog hopping along a circular
arrangement of lily pads, making regular patterns—for example, “clockwise
by two pads, counterclockwise by one”—while the pads themselves drift.
The result is a deep analogy between scale degrees and musical voices: at
some level, scale degrees are voices, articulated not by concrete musical
instruments but by abstract structures of melodic possibility.
Terminologically this means we need to be careful to distinguish concrete or
surface voices, which articulate perceptible melodies, from abstract voices,
which are scale degrees or melodic slots. The right-hand melody in Figure
1.2.1 presents a single surface voice moving along three abstract voices.

Figure 1.2.5. Here the voice moves according to a regular pattern (“up by third, down by step”) along
a series of melodic slots that are themselves moving. The example is drawn from the first movement
of Mozart’s G major piano sonata, K.283.

Central to the idea of the quadruple hierarchy is the claim that

transposition and inversion are available not just at the chromatic level, as in
atonal set theory, or at the scalar level, as in diatonic set theory, but also at
the chordal level—where they form a new, “chordal set theory” governing
the motion of voices. Figure 1.2.6 attempts to visualize this situation,
numbering chromatic, scalar, and chordal degrees so that middle C is 60.
These numbers allow us to represent intervals using addition and subtraction
at all levels. Motion along the chord occurs on this top layer: the chordal
interval +1 in Figure 1.2.1 moves one step to the right, –2 moves two steps
to the left, and so on. The chord in turn provides a translation or mapping
from its own degrees to those of the scale. Intervallic patterns like the (+2, –
1) of Figure 1.2.5 occur on this level. The scale in turn maps its degrees to
the chromatic degrees at the bottom of the graph, offering yet a third
opportunity for collectional motion. (The chromatic neighbors of Figure
1.2.1 move along this level.) Together the collections form a recursive
architecture that will be more familiar to computer scientists than to music
theorists, yet it is one that every musician intuitively grasps.

Figure 1.2.6. The quadruple hierarchy.

Motion at different levels of the hierarchy gives rise to a diverse range of

phenomena: motion of scale within aggregate produces modulation, motion
of chord within scale produces voice leading, motion of voice within chord
produces motivic development, chordal inversion, and what I call
“structured arpeggiation” (§4.8). Particularly important is the technique of
counteracting, or nearly counteracting, an operation at one level with its
analogue at another: combining transposition along chord and scale to
produce efficient voice leading, or combining inversion along both chord
and scale to produce Gesualdo’s trick (appendix 3).15 Musicians have
learned to manipulate the hierarchy in special cases, typically mediated by
idioms and schemas. The general picture, however, is likely to exceed our
intuitive comfort. Here then is an opportunity for theory to add value—
starting with the analysis of earlier music, abstracting its techniques in
accordance with the Prime Directive, and showing how hierarchical set
theory can be applied in a range of musical environments.

3. Philosophy

Another motivation for this book is my sense that music theory presents
philosophical challenges whose complexity is not always appreciated. As a
discipline that crosses the boundary between science and art, it requires both
musical intuition and methodological sophistication, yet it is extremely
difficult for a single human being to acquire expertise in both domains. The
result is a tendency to neglect some of the most perplexing issues in what is
an inherently challenging intellectual enterprise.
The most obvious issue is the divergence between eye and ear, between
structure apparent in musical scores and what can be heard by even well-
trained listeners. This is a familiar problem in twentieth-century contexts,
exemplified by worries about the audibility of twelve-tone and other
modernist modes of musical organization.16 But analogous problems
surround earlier music as well: for instance, both introspection and
experimental research suggest that long-term tonal closure (the return to the
tonic at the end of a long piece) is not particularly salient for most listeners,
yet the architecture of sonata form seems to presume its importance—the
purported function of the recapitulation being to resolve the “large-scale
dissonance” created by the appearance of the second theme in a “foreign”
key.17 Were earlier composers fixated on structures that even well-trained
listeners cannot hear? Did they wrongly expect their audience to experience
recapitulation as resolving a dissonance? Was classical music written for a
small elite with absolute pitch? These are uncomfortable questions, and
contemporary theorists are understandably reluctant to confront them head-
on, yet they are essential for understanding the aesthetic significance of
classical music.
 The divergence between eye and ear undermines the common music-
theoretical inference from patterns-in-scores to listeners’ awareness of those
patterns. This what-you-see-is-what-you-hear principle is central to projects
as different as Fred Lerdahl’s a prioristic psychology, which assumes that
“experienced listeners” recreate in their minds a completely accurate copy of
the musical score, and Gjerdingen’s historicist “schema theory,” which
proposes score analysis as a tool for reconstructing the “situated
psychology” of earlier listeners.18 Thus Leonard Meyer writes, “the structure
of the affective response to a piece of music can be studied by examining the
music itself,” a statement echoed by Gjerdingen and Byros.19 We know this
statement is false in the case of long-term tonal closure, and we have reason
to worry that it fails in other cases as well: for instance, earlier composers
avoided parallel perfect intervals even in cases where they are very difficult
to hear (e.g., when tucked away in inner voices). Consequently, any
identification of what is seen with what is heard will underestimate the
complex mixture of aural and conceptual, of heard and imagined, at play in
the Western notated tradition.20
To be clear, I accept that musical styles are characterized, in part, by
recurring regularities and patterns, and that our ability to perceive such
patterns underwrites our ability to appreciate and understand music.
Furthermore, I agree that theory can enrich our perceptual experience by
sensitizing our ears to structure: having written this book, I regularly find
myself hearing the idioms and patterns I discuss, even when listening to
unfamiliar pieces without a score. But I reject the claim that structure-in-
scores is ipso facto available to listeners’ conscious reflection, or even that
an ideal listener should consciously try to track compositional technique.
Gjerdingen compares the listener to a figure-skating judge observing a series
of standard moves like the Lutz or triple axel. I will argue that music
perception is more complex than that: composers can navigate spaces of
musical possibility without being able to say exactly what they are doing just
as listeners can understand what they are hearing without being able to
describe their understanding in words. Music is written for audiences, not
judges.
One way of putting this point is that there is no fixed relationship between
compositional technique and communicative meaning, or how something is
made and what it says. In some circumstances, artists efface their methods,
deliberately aiming to create illusions. (Outside of music we have magic
tricks and special effects; inside it, we have Chopin’s arpeggiation patterns
or Ligeti’s “touches bloquées,” simple physical gestures creating outsized
impressions of complexity.) In other cases, technique is meant to be
apprehended only holistically and tacitly: for example, sound engineers and
record producers make countless decisions about reverb, panning, filtering,
microphone placement, and amplification, all in service of creating a
coherent and aesthetically pleasing sonic field, but with no expectation that
audiences will consciously track these decisions (“wow, nice use of a high-
pass filter”). In still other circumstances there is a closer connection between
technique and content, as when composers intend listeners to recognize the
return of a theme. But many practices are difficult to categorize: when it
comes to sonata form there is a genuine question about whether we should
expect an idealized listener to consciously track specific technical events
such as the “medial caesura” or whether we should think of those events as
tools for creating compelling energetic trajectories. Much the same might be
said for galant schemata, which some theorists might compare to words in a
language and others might consider technical tools for generating a certain
sort of musical flow.
If theory’s fundamental challenge is the gap between eye and ear, then the
runner-up is inferring aesthetic norms from musical evidence. This is an
issue that can arise even with perfect perception of every musical detail. For
suppose we notice that rock and classical music use different harmonic
progressions: should we conclude that rock musicians are deliberately
rejecting classical norms, or simply speaking a novel musical language?21
Or we observe Beethoven effacing standard sonata-form signposts such as
the medial caesura: should we read this as a deliberate rejection of the past
or as a new set of formal conventions that can be understood on its own?22
Do popular music’s pervasive accents on beats 2 and 4 represent a deviation
from a genuine psychological expectation that beat 1 should carry the
strongest accent, or an alternate metrical norm? Here it is all too easy to
project inappropriate expectations onto earlier styles, powering up the
hermeneutic engines when our oversimplifications fail.
Consider the nineteenth-century theorist Ebeneezer Prout’s approach to
one of the more vexed problems in historical theory, the prohibition on
parallel perfect intervals in the presence of nonharmonic tones. Prout
endorses the common view that suspensions represent (or “substitute” for)
their tones of resolution, and thus cannot be used to mask parallels. He then
immediately concedes that this norm was not obeyed by the “Old Masters”
in the case of perfect fifths, proposing a sonic explanation—that this
violation is “less unpleasing” and “quite unobjectionable” (Figure 1.3.1).
This notion of a musical rule not followed in practice should make us
suspicious, for how are we to know that a norm exists, if not through
practice? Textbooks are unreliable, as authors might just be wrong about the
music they intend to describe. (Consider the prohibition of split infinitives, a
more-or-less illusory rule with no grounding in actual linguistic behavior.23)
Perhaps instead of describing composers as habitually violating their own
rules we should look for rules that accurately describe compositional
behavior. Or perhaps our so-called rules are generalizations theorists have
made about complex musical traditions—with apparent departures better
described as limitations of the generalizations themselves. Here it is
important that earlier composers were trained from an early age and in a
hands-on manner, a pedagogical approach that can preserve idioms resisting
systematic generalization.24 In replacing this hands-on tradition with
textbooks and college classes, we tend to reify the regularities, endowing
them with a metaphysical weight they do not deserve.
Figure 1.3.1. Prout’s 1889 discussion of suspension-masked parallel perfect intervals.

Prout’s struggles point to an ambiguity in the notion of a grammatical

“rule.” In both music and language, it is easy to come up with
generalizations that are true yet incomplete: music theorists rightly say that
nonharmonic tones do not generally license parallel fifths, just as linguists
correctly say that English does not allow one word to be inserted into
another. Yet there are rare but systematic exceptions to these generalizations
(e.g., “unbebloodylievable”). Grammatical norms, in other words, are
multilayered, with one and the same utterance simultaneously violating a
true generalization while also conforming to a more nuanced set of rules.
Misunderstanding this point sometimes leads theorists to insist on
universality against the evidence of our senses—as when Prout postulates
persistent violations of an imagined rule forbidding suspension-masked
parallels. Conversely, it can lead us to dismiss rules just because of the
occasional exception—as when contemporary theorists reject Roman
numerals because of their various limitations.25
A third philosophical issue involves norms of linguistic meaning, and in
particular a set of antiquated assumptions inherited from figures like
Schenker and Schoenberg. These can lead to dogmatic stipulations about
how words must be used: that the label “I6” necessarily presupposes
inversional equivalence, that “V–I” implies the primacy of the bass voice, or
that “modality” must mean something more than a largely diatonic language
with a range of possible tonal centers.26 By contrast, for the better part of a
century, philosophers, psychologists, and cognitive scientists have been
suspicious of strict definitions, emphasizing that words often acquire their
meanings by virtue of complex and ever-changing networks of practice. A
single term can thus support a range of uses varying with context, speaker,
and time.27
Wittgenstein famously wrote about the difficulties of defining “game.”28
Many of his arguments could also be applied to terms like “seventh chord.”
It is a truism of music history that seventh chords came to be accepted as
genuine sonorities somewhere around the time of Monteverdi; before that, it
is said, they are mere byproducts of nonharmonic tones. But the history of
seventh chords is like a river with many tributaries, and chapter 5 will
describe the “Renaissance seventh,” a common idiom in which sounding
seventh chords cannot be analyzed as nonharmonic decorations of a
consonant background; this idiom evolves seamlessly into the ii65–V
cadence of baroque music, where the seventh was explicitly recognized. In
other words, the history of music presents us with a continuous evolution in
which one and the same vertical structure, the irreducible seventh, gradually
acquires more and more independence, occurring initially only in one
context and slowly acquiring new uses. Are the earlier instances “truly”
seventh chords? The question, like the quest for a simple definition of
“game,” is not particularly productive. We can if we like try to demarcate the
“true” seventh chords from the false, but this is to impose discreteness where
history gives us continuity. It seems more fruitful to say that the vagueness
of the term “seventh chord” reflects the continuity of compositional practice;
instead of a bright line separating seventh from non-seventh, we have a
central core of paradigmatic cases surrounded by a penumbra of increasingly
problematic examples.29 Much the same could be said for terms like
cadence, dominant, second theme, sonata, tonic, meter, and indeed the entire
practice of Roman numeral analysis.
If we respond to this fuzziness by insisting on strict definitions, we will
invariably fall into dogmatism. When William Caplin says that a classical-
style authentic cadence must involve a root-position dominant chord that
lasts right up until the arrival of the tonic, he is committing himself to
withholding the term “cadence” from a host of moments that may sound or
feel cadential.30 When James Hepokoski and Warren Darcy assert that a
sonata’s second theme must as a matter of definition be preceded by a
“medial caesura,” they commit themselves to denying what may for all the
world sound like second themes—or alternatively, to defining “medial
caesura” as whatever comes before a second theme.31 The issue here is not
so much moral as intellectual, for the risk is that we will transform important
generalizations (“the passages we feel to be second themes are very often
preceded by a pause”) into trivial assertions of linguistic convention (“I will
refuse to use the term second theme unless I can identify a preceding
pause”).32 This is very close to what philosophers call the “no true Scotsman
fallacy.”33
In this book I will try to honor the fuzziness of conceptual boundaries in
two ways. First, by avoiding, inasmuch as possible, fruitless disputation
about whether a musical object “really” deserves a particular name or not—
whether a certain passage “really” is a second theme, or whether a certain
progression “really” is a cadence, whether two figures “really” are versions
of the same motive, and so on. Many of these questions simply have no
objective or interesting answers, much like the question of whether a hot dog
really is or is not a kind of sandwich.34 We encounter no intellectual or
methodological difficulties if we acknowledge that there are grey areas
where concepts, from “bald” and “sandwich” to “cadence” and “closing
theme,” simply fail to grip.
Second, I will try to favor “thin” uses of music-theoretical terms involving
only minimal presuppositions, as opposed to “thicker” or methodologically
freighted definitions. For me, the term “modality” will refer to largely
diatonic music with a wide range of possible centers, making no assumptions
about the presence of reciting tones or the utility of the plagal/authentic
distinction. Thus “D phrygian” means “music that generally uses the two-flat
diatonic collection while articulating D as its tonal center.”35 Similarly,
Roman-numeral labels like “I6” will mean “scale degrees 1, 3, and 5 with 3
in the bass” carrying no presumptions about roots or inversional
equivalence. This is, in the first instance, because I think simpler terms are
better suited to a slowly evolving musical practice in which stylistic
boundaries are inherently fuzzy. But it is also because the insistence on
“thick” terminology has the effect of balkanizing us into distinct linguistic
communities, baking our beliefs into words themselves, so that disputes
about music are complicated by bickering about language. The risk is that
the already-small field of music theory will be even further divided, with
historicists, Formenlehre theorists, Schenkerians, and empirical
musicologists each cultivating their own little gardens of terminological
presupposition.
A final philosophical issue involves humility and openness to criticism.
Some intellectual communities are relatively hospitable to robust debate,
placing a premium on empirical confirmation and an open-minded
consideration of alternatives. Others, such as politics and the law, are more
likely to view argument as a tool for establishing predetermined goals—the
point being for your side to win rather than to discover unexpected truths. I
believe there is too much of this second attitude in music scholarship. The
field labors under the shadow of figures who made claims far outstripping
the evidence available to them, holding to their views even in the face of
reasonable counterargument.36 This has led to a culture in which pointed
criticism is sometimes taken to be impolite or even insolent. Both inclination
and upbringing lead me to believe that vigorous give-and-take is our best
mechanism for improving understanding, and for the most part I have tried
to engage with theorists who I feel have genuinely contributed to our
understanding; in disagreeing, at various points, with Fétis, Schenker,
Schoenberg, Gjerdingen, Cohn, Caplin, or Hepokoski and Darcy, I do not
mean to belittle their ideas, but rather to honor their contributions by trying
to refine them. Nor, for the most part, do my criticisms undercut their
fundamental achievements, instead identifying points where they may have
drawn overly broad conclusions or gone too far. Understanding is a process,
and criticism is the engine that makes it go: future theorists will no doubt
find lacunae and errors in this book, refuting some claims and improving
others.
 4. Statistics

Where earlier scholars worked intuitively, twenty-first-century theorists can

be more systematic: with computers it is possible, for example, to list every
instance of parallel perfect intervals found on the surface of Palestrina’s
mass movements. With a little work we can go further, identifying all the
parallels that appear in the “harmonic skeleton” that has been shorn of
nonharmonic tones. This gives us hope that, for the first time in history, we
can produce an empirically adequate account of the prohibition on parallel
perfect intervals. Corpus analysis thus provides a “ground truth” against
which we can evaluate traditional theoretical claims, much as historians of
science can evaluate previous theories against what is now known.37
Computational corpus studies thus stand alongside geometry as an important
twenty-first-century addition to the music theorist’s toolbox.
The new data allow us to extend our conception of “music theory” beyond
the explicit statements contained in written treatises, to the theories
implicitly encoded in musical works themselves. Language users can be said
to have implicit theories of grammar, revealed for example by whether they
allow themselves to split infinitives—theories absorbed through imitation
and implicit learning. The history of English grammar is constructed by the
interaction between these implicit theories and the explicit theories of
grammarians, which both affect and are affected by the implicit theories.
(Sometimes we may even find cases where a single person’s implicit theory
diverges from their explicit beliefs.38) In much the same way, composers
have implicit views about theoretical issues both small and large, which we
can recover (if imperfectly) through a careful study of their music—thus
revealing an intuitive “composer theory” that stands alongside the explicit
statements of “theorist theory.”39 Corpus data allow us to reconstruct the
dialogue between explicit and implicit, a history of musical ideas written
jointly by theorists and composers.
Unfortunately, musical scores are for many computational purposes
opaque: D–F–A can be a tonic in D minor, a supertonic in C major, a mere
agglomeration of nonharmonic tones deserving no chordal label, and many
other things as well. Linguists dealt with analogous issues by constructing
substantial corpora of hand-analyzed data that served both as primary objects
of study and as training sets for automatic language-parsing algorithms.40
For this book, I have followed their lead, creating machine-readable Roman-
numeral analyses of more than one thousand pieces stretching from Dufay to
Brahms; as I explain in appendix 4, all this material is freely available on the
internet—music, data, code, and instructions for reproducing all the graphs,
tables, statistics, and analyses in the book. Over the last decade I have
carefully checked this corpus, improving it as I use it.41 Handmade
annotations greatly increase the power of computational analysis, but at the
cost of introducing subjectivity—an issue that will be explored in later
chapters.
The recent explosion of musical data has the potential to transform our
understanding of a wide range of questions. Consider Paul von Hippel and
David Huron’s investigation of the traditional claim that melodic leaps tend
to lead to a change in direction—a principle found in Western counterpoint
texts both ancient and modern.42 Given the resources available when they
wrote, von Hippel and Huron had to explore this question using a limited
sample of just over 200 heterogeneous works: 35 Schubert lieder and 176
folk songs from diverse cultures. Finding no compelling evidence for
reversal in this small dataset, they proposed that “post-skip reversals” might
instead be explained by tessitura: large leaps, they observe, are more likely
to take vocalists to their registral extremes, and registrally extreme notes are
likely to move toward the center of their range. This led them to a memory-
free mechanism whereby all notes seek the center of a vocalist’s register, in
contrast with the traditional view that melodic direction depends on the size
of the preceding interval.
Figure 1.4.1 combines the vocal parts in virtually all of Palestrina’s mass
movements, a sample large enough that we can inspect melodic behavior at
every point in the tessitura. We see a substantially greater tendency for leaps
to change direction throughout the range, a tendency that is also present in
the fugues of The Well-Tempered Clavier, while being severely reduced in
artificial “memory-free” music that uses the same distribution of melodic
intervals at every point in the range.43 It is now easy to perform this sort of
analysis on a wide range of corpora from the thirteenth century to the
present, a vastly greater sample than von Hippel and Huron were able to
access. Such analyses suggest that melodic leaps often do preferentially
change melodic direction over and above what can be explained by register
alone—in Josquin, Palestrina, Monteverdi, J. S. Bach, Beethoven, Sacred
Harp vocal polyphony, rock music, and other genres as well (Figure 1.4.2).
This tendency is greater in some styles than in others, and particularly
pronounced in Western notated music, both vocal and instrumental; but it
does not seem to be entirely an artifact of tessitura.

Figure 1.4.1. The tendency for leaps (solid lines) and steps (dotted lines) to reverse direction in
Palestrina’s masses, the fugues from The Well-Tempered Clavier, and artificial music based on these.
In the real music, there is a substantial tendency for leaps to reverse direction, even in the middle of
the tessitura; this tendency is much smaller in the artificial music.
Figure 1.4.2. The tendency of melodic leaps to reverse direction in a variety of repertoires. Due to the
frequent use of the pentatonic scale, a three-semitone interval has been classified as a step in Sacred
Harp and rock.
 Such evidence needs to be viewed with caution, not just out of general
principle but also because musical concepts, including “voice” and “step,”
can be difficult to pin down. In sophisticated tonal music we find melodies
alternating between two logical voices, arpeggiating chords (i.e. moving by
step along a nonscalar collection), and inserting embellishing tones between
leap and reversal (Figure 1.4.3). These examples remind us that the principle
of post-skip reversal is native to the context of counterpoint instruction, a
simplified environment in which many musical techniques are off the
table.44 Simple counting will not necessarily answer the musical question.

Figure 1.4.3. Skips that do not immediately reverse direction due to multiple abstract voices, stepwise
motion along a triadic alphabet, and melodic embellishment.

What we can say is that computational analysis often provides limited and
defeasible evidence in favor of music-theoretical common sense. Leaps lead
disproportionately to changes of melodic direction in many repertoires. Vast
tracts of Renaissance polyphony decorate triads with the nonharmonic tones
of traditional contrapuntal theory. And, of course, countless functionally
tonal passages exhibit the tonic-subdominant-dominant-tonic organization
described in harmony texts. Here and elsewhere empirical investigation
undercuts the more extreme forms of music-theoretical skepticism—a
complete rejection of “post-skip reversal,” or of Roman-numeral analysis, or
of the very notion of nonharmonicity.45 In the past, such skepticism has
often served theoretical ambition, the rejection of received wisdom paving
the road for more radical alternatives. It is difficult to see how this approach
can survive the era of abundant musical data.
At the same time, however, the following chapters will uncover numerous
challenges to textbook verities: nonharmonic tones that cannot be “reduced”
to consonances, or classical idioms that do not conform to any simple model
of harmonic progression. Conversely, we will find untheorized regularities in
a variety of genres. Perhaps the most consequential are the continuities
between sixteenth-century modality and eighteenth-century functionality—a
very gradual harmonic simplification beginning in the Renaissance and
continuing through the classical era, often invisible to the analytic eye. As a
result, there is hardly any music that is purely “modal” in the sense of being
untainted by functional principles, nor any that is purely “functional” by
virtue of having shorn all connections to earlier practice. Consistent with this
picture are the differences between dialects of functional tonality, with
baroque music retaining more vestiges of Renaissance practice. Corpus
analysis thus suggests we replace the binary opposition of modality and
functional tonality (and perforce the assumption that these represent separate
and coherent “systems” of musical thinking) with a more continuously
changing set of contrapuntal and harmonic conventions.46
We are, in other words, led in equal and opposite directions: on the one
hand to the conclusion that traditional theory is generally correct, and on the
other to the realization that its basic assumptions sometimes fail. One might
draw an analogy to Newtonian physics, which provides accurate descriptions
of the macroscopic world despite being fundamentally wrong about its
microscopic structure. To the engineer, these failures are minor and can often
be ignored. To the philosopher, they are cataclysmic, for the failure of
Newtonian physics points to a fundamentally counterintuitive picture of the
universe. For all the success of classical mechanics, the world simply is not
made of tiny billiard balls proceeding along deterministic paths. Something
altogether different is going on.

5. Schema

A good candidate for this something is the idiom or schema—contemporary

music theory’s answer to its earlier Newtonianism. Here the pioneering
figure is Robert Gjerdingen, who proposes to scrap much of the apparatus of
traditional theory in favor of a “construction grammar” in which composers
recombine atomic patterns.47 Musical behavior, on this view, is not lawlike
at its deepest level—instead it is an unruly and messy enterprise whose basic
objects are specific gestures, often tied to particular formal contexts. In
many ways, Gjerdingen favors a return to a pre-Rameauian conception of
musical structure, rooted in the inherently particular injunctions of
seventeenth- and eighteenth-century pedagogy.48
 Figure 1.5.1 shows an unusual cadential formula that can be found
throughout J. S. Bach’s work. The main point of interest is the bass, which
unsettles two fundamental presuppositions about eighteenth- and nineteenth-
century music:

(1) That dissonant chordal sevenths invariably resolve downward by step to

a chord tone; here 4̂ moves through passing 3̂ and 2̂, with the next
chordal note a fourth below.49
(2) That nonharmonic tones are essentially decorative and irrelevant to
syntacticality; for this figure almost never involves the bass 4̂ leaping to
the tonic. Apparently, Bach felt that the stepwise motion was important
to the sense of resolution.

Gjerdingen’s work offers an antidote to two powerful music-theoretical

temptations. The first is to declare unusual moments nonsyntactical or
ungrammatical simply because they violate oversimplistic generalizations;
instead, he suggests that idioms are as important to Bach’s idiolect as the
procedures of textbook harmony. The second temptation is the construction
of ad hoc analyses seizing on particular features of the local context to
justify the “violations of the rules”—hermeneutic narratives that purport to
explain why Bach might have turned outlaw at just this particular moment.
For what initially seems like a violation may simply be a second-order
default, a rare but legitimate syntactic option. Schema theory encourages us
to consider that “exceptions to the rules” might simply be features of a
practice, explained by no more general principle or law. Here one could
draw an analogy to English phrases such as “a whole nother kettle of fish” or
“unbebloodylievable,” exceptions to the general truth that English words
cannot be inserted inside one another.
Figure 1.5.1. An idiom from the first B♭ minor fugue in The Well-Tempered Clavier, mm. 21–24. The
bass of the V2 steps down to the tonic without resolving.

The task is then to characterize Bach’s idiom accurately: is it just a fact

that his sevenths sometimes move down by fourth? Or does this happen only
with dominant seventh chords or only in cadential contexts? How can we
improve upon the largely correct generalization that sevenths usually resolve
down by step to chord tones? We cannot dismiss these questions out of a
distaste for abstract theorizing, for the question is precisely how to
characterize “this” idiom—as a specific cadential gesture, a technique for
resolving the seventh, or in some other way. It is just here that statistical
generality can help schema-theoretic particularity: given a set of scores
along with a labeling of their chords, it is possible to identify all the
moments in which sevenths resolve anomalously; these can then be
examined individually to determine the scope of a composer’s actual
practice. In Bach’s chorales we find that this idiom usually occurs in the
context of this particular schema (V2–I–V–I, often with the cadenza doppia
in one of the voices), largely at cadences, but occasionally at other points in
the phrase, and very occasionally in inner voices (Figure 1.5.2).50 The
cadential schema thus stands at the center of a network surrounded by
progressively more uncommon variants.51 Without detailed empirical
investigation it is simply impossible to describe the scope of the idiom.
Figure 1.5.2. Unusual resolutions of the seventh in Bach’s chorales. In “Christus, der uns selig macht”
(Riemenschneider 198, BWV 283), mm. 7–8, and “Ich dank dir, Gott, für all Wohltat”
(Riemenschneider 223, BWV 346), m. 11, the tenor steps down to the tonic. In “O Welt, sieh hier dein
Leben” (Riemenschneider 117, BWV 244.10), mm. 9–10, the bass steps down by fourth but on
unusual scale degrees; the seventh is resolved in register by another voice.

Gjerdingen has provided an important corrective to speculative overreach,

and I fully endorse his claim that musical style is idiomatic at its core,
employing schemas and patterns that sometimes violate broader grammatical
generalizations. But I reject the thought that generalization is inherently
suspect, for it seems entirely likely that composers learn both specific idioms
and also more general and systematic principles: the “rules” of functional
harmony—traditionally expressed by precepts such as “V does not go to
ii”—are generalizations of this sort, learned implicitly in something like the
way that children learn the grammar of their native language.52 Something
similar can be said for the precepts of traditional counterpoint theory, which
accurately describe techniques for augmenting consonant harmonies with
passing tones, suspensions, and the like. These “rules,” or rather
approximately true generalizations, are broader than Gjerdingen’s idioms,
more conducive to genuine creativity, and roughly comparable to the
implicitly learned principle that English has a subject-verb-object norm.
Schema theory is best understood as augmenting rather than replacing the
generalizations of familiar theory.
Consider, as a point of contrast, galant composers’ preference for I6 over
iii in the nonsequential “Romanesca” (Figure 1.5.3). Gjerdingen explicitly
disavows any attempt to explain this preference by “chord grammar,” the
“rise of tonality,” or “other grand abstractions,” instead proposing that
eighteenth-century musicians play by “eighteenth-century rules” grounded in
the solfege and unfigured-bass traditions.53 By contrast, I think that we can
often explain why schemas are the way they are. Indeed, the ionian-mode
preference for I6 long predates the eighteenth century (Figure 1.5.4); this
general favoring of I6, I will claim, is both a rudimentary manifestation of
“chord grammar” and an early stage in the development of functional
harmony. From this point of view, the puzzle is not the use of I6 in the
nonsequential “Romanesca” but rather the persistence of iii in sequential
contexts: indeed, the Pachelbel sequence was originally just one of many
related idioms in which descending thirds alternately support 53 and 63
sonorities (Figure 1.5.5); the gradual pruning of these options, to the point
where only the Pachelbel progression remains, was part of the lexical
narrowing that produced functional tonality. These might be grand
abstractions, but they are grounded in concrete data.

Figure 1.5.3. Sequential and nonsequential variants of Robert Gjerdingen’s “Romanesca” schema.
Figure 1.5.4. The percentage of triads above 3̂ that are in first inversion.

Figure 1.5.5. Three ways to harmonize stepwise descending thirds with alternating 53 and 63
sonorities. The first is common in virtually all triadic music; the second can be found in the baroque
but is rare thereafter; the third is common only in modal music. Here “D4A2” means “descending
fourth, ascending second.”

One of my goals, therefore, will be to combine the schematic approach

with an openness to regularity and generalization. To that end I will argue
that many schemas originate in constraints inherent to our materials: rather
than the figure skater’s codified moves, they are responses to the contours of
musical space—a rock climber taking advantage of a particularly prominent
outcropping. The musical analogue of the rock face is a contrapuntal
geometry that funnels intuitive exploration toward certain solutions. These
can take on different rhetorical roles in different styles: Gjerdingen’s
“Prinner,” for example, may function as a precadential “riposte” in the galant
period, and a cadence in the Renaissance. This difference is built atop a
deeper level of constraint arising from the fundamental logic of musical
possibility.54 Thus we should distinguish the “deep Prinner,” a very general
voice-leading schema used from the Renaissance to the present, from the
“shallow Prinner” that is a specific galant mannerism.
Another general principle, tacitly understood by composers in many
genres, is that contrary-motion sequences are most balanced when they
move by about half an octave. The idiom in Figure 1.5.6 appears in virtually
all modal and functionally tonal styles, the two voices moving stepwise to
enumerate the diatonic consonances—fifth, third, octave, sixth. The pattern
can be continued in two obvious ways, either lowering the top voice to move
the pattern down by fifth (or, in the other voice, up by fourth), or raising the
bass to move up by fifth (down by fourth in the other voice).55 Repeating
one of these options produces a sequence, while using them in alternation
produces a quasi-sequence that returns to its starting point after two
iterations (Figure 1.5.7). Thus what might initially seem like an arbitrary
figure of speech, drawn from an infinite reservoir of equally good
alternatives, can be plausibly described as an optimal path through a
preexisting geometrical space of musical possibilities. The advantage of the
general approach is that it allows us to connect traditional practices with
those of our time. In Figure 1.5.8, for example, we find Meredith Monk
using diatonic contrary motion to generate the same sort of fifth relationship
found in Western classical music, only now featuring seconds and fourths. I
think it is useful to recognize the continuity between this excerpt and the
more schematic contrary motion patterns of the classical style. This is
possible only if we ground schematic particulars in broader potentialities.

Figure 1.5.6. Two contrary-motion sequences featuring diatonic consonances. They can be reversed to
create registral expansion.
Figure 1.5.7. Two examples of a contrary-motion schema alternating thirds and sixths: Fanny
Mendelssohn Hensel’s “Januar,” mm. 15–18 (from Das Jahr, H. 385); and a passage from Klaus
Badelt’s film score Pirates of the Caribbean: Curse of the Black Pearl (2002, heard at 2:07:32). The
first forms a brief fifth sequence, while the second combines an ascending and descending fifth to
return to its starting point.
Figure 1.5.8. A motive from Meredith Monk’s Anthem (2020), analyzed in relation to the traditional
contrary-motion schema. Both feature “antiparallel” motion where the upper voice ascends by fifth
(T4 or four diatonic steps), while the bottom voice descends by fourth (T–3).

Without general principles, the schematic approach can devolve into a

picture of composition as the mindless stitching-together of precomposed
chunks, reminiscent of jazz textbooks that teach only licks or “artificial
intelligence” programs that recombine preexisting music to produce the
illusion of something new. Genuine musical thinking requires novel
solutions to high-level compositional problems: Fanny Mendelssohn Hensel
harmonizing a traditional diatonic schema in an unusual way, Meredith
Monk adapting a familiar technique to dissonant dyads. The goal of this
book is to describe some of the general techniques that can be found
throughout Western practice: “thinking within the chord,” hierarchically
nested transposition, repeating contrapuntal patterns, and many others as
well. These general principles are the source of the particular schemas that
help to define specific styles.
Schema theory shares with Schenkerian theory a focus on outer-voice
contrapuntal patterns, privileging line and dyad over triad and seventh chord;
it differs from Schenkerian or Roman-numeral theory in favoring particular
idioms over general principles. I will try to triangulate between these
approaches, describing general principles governing the interdependence of
harmony and counterpoint. To that end I will introduce a family of
geometrical models that allow us to view any given passage under a dyadic,
triadic, seventh-chord, or scalar lens. What results is a generalized analogue
of schema theory, combining linear ideas from Schenker with harmonic
ideas from Rameau, and revealing connections between a wide range of
practices from the Renaissance to the present day.

6. Outline

This book is framed by two largely analytical chapters. Chapter 2 argues that
a variety of rock progressions, all unusual from the standpoint of classical
theory, reflect straightforward features of musical geometry—and hence that
the intuitive competence of the rock musician is in part a matter of knowing
one’s way around the space of chromatic triads. This chapter is meant to
provide an accessible introduction to my general approach. Chapter 10
makes a broadly similar point about Beethoven, focusing on what I call the
“Ludwig” schema, and considering some of the philosophical challenges his
music poses. This is more of a culmination, linking technical issues in voice-
leading geometry with philosophical questions about analysis. Together, the
two chapters suggest that a range of different musical styles can be linked by
a subterranean geometrical logic.
Sandwiched in between are seven more theoretical chapters examining
various features of modal and functionally tonal syntax. Chapter 3 considers
the voice-leading system, arguing that counterpoint is not fundamentally a
matter of avoiding parallels, or even balancing of independent melodic lines,
but of understanding the interdependence between harmonic and melodic
forces. As we will see, general linear impulses such as the preference for
descending stepwise melodies will tend to generate different kinds of
harmonic progressions in different contexts—so that one and the same
melodic tendency produces modal results in one situation and functional
results in another. My approach here recalls the Schenkerian idea that
melodic phenomena generate harmonic patterns “at the musical surface.”
But unlike Schenker, I argue that the voice-leading system often generates a
distinctively modal logic which reverses classical harmonic norms.
Chapter 4 then zeroes in on the particular subject of repetition, proposing
a new theoretical device, the repeating contrapuntal pattern, that can be
found from the Renaissance to the present day. This is perhaps the newest
chapter of the book, using abstract notions of symmetry to identify an aspect
of musical competence crossing stylistic boundaries; its argument continues
in appendix 3, which connects classical sequences to twentieth-century
transformations. Chapter 5 considers the nonharmonic system coordinating
the behavior of dissonant and nonharmonic tones. I argue that nonharmonic
tones are not simply “decorative,” particularly in the case of suspensions.
This leads to a more general consideration of the ways in which
nonharmonic tones can serve harmonic ends—and hence the impossibility of
drawing a clean distinction between harmonic and nonharmonic realms. The
discussion again serves as a case study for my broader approach: traditional
theories of nonharmonicity work well in general, while also breaking down
in alarming ways.
The latter part of the book turns to the harmonic system, a set of initially
implicit norms that were eventually codified by figures such as Rameau and
Riemann. Chapter 6 interprets the rise of harmonic functionality as a gradual
process spanning more than two centuries, using analysis and corpus study
to tease out a simpler form of protofunctionality appearing early in the
sixteenth century. Chapter 7 analyzes the local procedures of mature
functionality as the product of three independent subsystems—harmonic
cycles, fauxbourdon, and sequences. Here the main innovations are (1)
tracing functional practice to basic affordances of geometrical space; (2)
using the circle of diatonic thirds to highlight the quasi-sequential nature of
functional harmony; (3) including fauxbourdon as a core functional
subsystem; and (4) expanding our conception of sequential practice. Chapter
8 discusses scales and keys, which I present as chordlike objects operating at
a higher hierarchical level; I also use voice leading to propose a new theory
of enharmonicism. Chapter 9 turns to higher-level organization, considering
melodic strategies, harmonic recursion, and eight-bar phrase structure.
Each chapter is prefaced by a short prelude setting the stage for the more
detailed investigation to follow; for some readers, these may provide the
most concentrated source of interesting material. Where appropriate, I end
each chapter by considering twentieth- and twenty-first-century applications
of its techniques. Some readers will want to chart a path reflecting their
particular interests: the examination of earlier music is mostly confined to
chapters 2–6, while functional tonality is the focus of chapters 7–10. Those
who are interested in analysis can find detailed discussions of Monteverdi’s
“Ohimè, se tanto amate” in §5.3, the opening of the Rondo-Burlesque from
Mahler’s Ninth Symphony in §5.5, the Kyrie of Palestrina’s Pope Marcellus
Mass in §6.5, Marenzio’s “Ahi Dispetata Morte” in §6.7, four “dualist”
pieces by J. S. Bach (invention, prelude, chorale, and fugue) in §7.7, several
Chopin pieces in §9.3, and, in chapter 10, five nineteenth-century
movements by Beethoven, Schubert, and Wagner. Once the basic framework
of chapters 2–4 is taken on board, chapters can be read out of order, as
loosely connected essays on different aspects of musical structure.
Any theory that is concerned with voice leading needs to grapple with
Heinrich Schenker, whose musical ideas are in equal parts insightful and
infuriating—and whose moral and aesthetic views are often downright
repulsive. One of the major surprises for me, over the course of my writing,
is how often I have found myself close to Schenkerian insights.56 Like
Schenker, I believe music is often governed by a deep contrapuntal logic that
is obscured when we focus too closely on roots and Roman numerals. I
endorse the Schenkerian idea that chords can exist “below the musical
surface,” though for me this means treating them as small scales along which
voices move. I likewise share Schenker’s perception that sequences are
fundamentally connected to certain sorts of linear intervallic structures, or
“repeating contrapuntal patterns” as I call them. I frequently adopt a broadly
Schenkerian strategy of analyzing music by “reducing” it to familiar
templates or schemas, eliminating surface detail to reveal a simplified
backbone. On a more detailed level, I often follow Schenker by assigning
voice exchanges to the musical surface.57 Finally, I use voice leading as a
lens for understanding higher-level phenomena such as modulation. In all
these ways, I will try to infuse the relatively flat models of voice-leading
geometry with a richer hierarchical structure.
But beyond these broad similarities are a host of important differences.
The biggest is that I am fundamentally skeptical about the long-range
connections that characterize so much Schenkerian practice—the idea that
there are aesthetically and perceptually meaningful connections between
notes separated by dozens or hundreds of measures. In my view this
represents a fantastical disregard for human perceptual limitations.58 Almost
as important are two different conceptions of musical reduction: for me it is
a process of modeling, speculative hypotheses about the cognitive structures
that might produce a musical passage, whereas for Schenker it reveals
something like a musical essence.59 Furthermore, I view scales (both
familiar and generalized) as important bearers of harmonic structure, while
Schenkerian theory affords them no privileged status. I also think sequences
can provide alternatives to the chorale-like backgrounds of Schenkerian
analysis, allowing us to understand an irregular musical surface as distorting
a more-regular background. Finally, I am largely unconcerned with unity as
an aesthetic value and with reductional stages beyond the first. All of which
is to say that I try to incorporate Schenkerian insights in ways that are both
empirically informed and epistemologically modest, tempered by geometry,
corpus study, and schema theory.
The notated tradition has often been a source of undue anxiety. We are bad
listeners if we do not have a sense of resolution when a sonata recapitulates
its second theme in the tonic key. We are bad historians if we hear tonic-
dominant resolution in sixteenth-century music. We are bad composers if we
need the piano, or fail to imagine an entire piece in a single blazing instant.
We are aesthetically unsophisticated if we happen to prefer rock to Wagner.
Over and over again, we are told that we are Ordinary while the dead
composers are Geniuses—and that we should subordinate our preferences to
those of our aesthetic betters. One possible response is to reject the learned
tradition in favor of simpler and more accessible styles. My hope is that this
book could point toward another: that by generalizing and demystifying
basic compositional techniques, it will allow us to appreciate the notated
tradition’s virtues while also defusing some of its elitism—giving us a
middle path between uncritical acceptance and outright rejection. This is the
final and most unlikely meaning of my title.

1
Babbitt was a partisan of logical positivism, whose narrowminded and aggressive skepticism
drove many theorists away from analytic philosophy as such; however, I believe that music theory has
much to learn from its emphasis on clear reasoning.
2
This is not the only project that music theorists can undertake: we can study perception, either
using experiments or introspection; we can study the historical concepts that have actually guided
music creation; we can engage in various forms of criticism and evaluation. I will be concerned with
these issues only secondarily, focusing instead on the intuitive knowledge of basic pitch structure.
3
McCreless 2011 considers the metaphor of musical ownership from many different points of
view.
4
Note that Gesualdo substitutes G major for the G minor implied by the sequence (§4.9).
5
Cohn 1996 and 2012, pp. 106–9.
6
Hook 2002 reports that John Clough (2000) offered a similar analysis.
7
Stravinsky’s four-voice passage involves the same abstract registral configurations as Gesualdo’s,
with alto doubling bass in one chord form and tenor doubling soprano in the other. In both passages,
the four-voice sonorities are related by pitch-space inversion.
8
Later in the same solo, Tyner uses the technique with four-note pentatonic subsets (Figure A3.11).
9
Yellin 1998, Telesco 1998, and §4.6.
10
Throughout this book I notate scale degrees so that middle C is as close to 60 as possible.
11
Lewin (1986, p. 355, note 19) makes a similar point about Schubert’s “Morgengruss.” Whether
the Beatles were explicitly thinking in these terms is a complex and likely unanswerable question.
12
If we interpret D as a passing tone connected by step to both C and E, then it is unclear how to
start the figure on G.
13
Shepard tones can be used to produce a sequence that appears both to fall continuously and also
to remain fixed in register; the illusion is accomplished by gradually fading in higher-register
harmonics while fading out the lower-register harmonics (Shepard 1964).
14
Peter Westergaard (1975) asserted that leaps of a third or more inherently marked a change of
voice below the surface, a claim that initially sounds counterintuitive. But when background voices
are triadic and connected by stepwise voice leading, the statement is almost tautologous: since the
background of Heu me Domine contains almost no leaps, leaps on the surface are necessarily due to
motion from one background strand to another, precisely as Westergaard suggests.
15
The investigation of hierarchically embedded musical “alphabets” begins with Simon and
Kotovsky 1963, continuing with Simon and Sumner 1968, Deutsch and Feroe 1981, and Lerdahl 2001.
The idea of extending set-theoretical techniques to arbitrary collections is implicit in Clough 1979a
and 1979b. This book fuses these ideas with the theory of voice leading, considering arbitrary
hierarchies of arbitrarily moving collections, with analogous transformations occurring at each
hierarchical level, and motion at one level often counteracting motion at another.
16
Lerdahl 1988.
17
Charles Rosen writes, “material presented outside the tonic must have created, in the eighteenth
century, a feeling of instability which demanded to be resolved” (1971, p. 74, my italics). The
experiments of Cook 1987 and Marvin and Brinkman 1999 suggest that this may not be true.
18
The diagram in Lerdahl 2020, p. 23, formalizes the inference from the musical surface to the
“heard structure.” See Tymoczko 2020a for more.
19
See Meyer (1956, p. 32), Gjerdingen (1996, pp. 380–81, and 2007, pp. 16–19), and Byros (2012,
pp. 282–84, which is admirably clear about these issues, and from which several of the preceding
references are drawn).
20
Aural comprehension is sometimes used as a criterion of musical content—which we might call
the what’s-there-is-what-you-hear principle. For example, Hadjeres, Pachet, and Nielsen 2017 use
listener data to judge whether a computer can be said to “compose like Bach.” I doubt those listeners
were sensitive to the structural details that interest theorists, composers, and analysts.
21
Stephenson 2002, chapter 5. The issue here is comparable to questions about whether a speaker
uses “bad grammar” or speaks a slightly different dialect; chapter 2 interprets rock music with
reference to specifically modal norms.
22
Hepokoski’s notion of “dialogic form” (2010) places a very heavy weight on inferred cultural
norms; chapter 10 interprets Beethoven as less beholden to existing convention.
23
Pullum and Huddleston 2002. Weber (1817–1821) 1846 is impressively nondogmatic about
musical practice.
24
Gjerdingen 2020.
25
This point is further obscured by the music-theoretical conflation of the grammatical and the
pedagogical: introductory textbooks simultaneously function as our most complete descriptions of
musical syntax, leaving us with an extensive literature articulating true but simplistic generalizations,
and comparatively little scholarship which tries to describe musical norms at a more granular level.
26
See the prelude to chapter 6 for specific references.
27
Margolis and Laurence 2003. Hilary Putnam notes that “momentum equals mass times velocity”
was once true by definition; yet in special relativity this statement is false, true only in the limit of
very small velocity (Putnam 1988, p. 8).
28
Wittgenstein 1953.
29
See Rosch 1978, which develops Wittgenstein’s observations in a scientific context. Caplin
(2013, p. 127) associates this attitude with “postmodernism,” but I think it is close to the consensus
among analytic philosophers and cognitive scientists, two groups not known for their postmodern
attitude. Lawrence 2020 observes that Caplin operates with a “classical” theory of concepts.
30
For instance, Caplin 2004, example 6, Mozart K.310, III, mm. 173–174, and the opening themes
of Beethoven’s Op. 127, I and IV. Richards’s term “closural function” (2010) is virtually synonymous
with the pre-Caplin “cadence,” underscoring the potentially terminological nature of the debate.
Burstein 2015 also critiques Caplin’s approach.
31
Effacing the medial caesura is, as Mark Richards (2013a, 2013b) has noted, a common tactic in
Beethoven’s music (Figure P10.4, §10.3).
32
I think Caplin is right to point out that 5̂–1̂ bass motion typically plays an important role in
signaling strong cadences in the classical style, just as I think Hepokoski and Darcy are right to note
that second themes are very often preceded by pauses. But in making these true observations into
definitions, they reject the very possibility of counterexamples. I see Brown 2005 as declaring that he
will only use the word “tonal” to describe pieces that are susceptible to Schenkerian analysis.
 33
The fallacy begins with a claim that “no Scotsman would do x” (Flew 1975). When presented
with a counterexample, a Scotsman who does in fact do x, the claimant responds by arguing that the
doer is “no true Scotsman.” In this way it becomes definitionally true that no Scotsman would do x,
since anyone who does x is not a (true) Scotsman. This renders a potentially true generalization into a
trivial fact about linguistic usage.
34
For sandwiches see Lund 2014 and Scherer 2015; Burstein (2015, p. 105) notes that traditional
theorists often emphasized the importance of judgment in concept application.
35
Powers 1992b, p. 212: “Nowadays professional folksingers and composer of folksongs talk
glibly about their Dorian and Mixolydian tunes, and so too sophisticated jazz practitioners and
textbooks alike, just as composers and theorists of art music used to compose using ‘dominants’ with
‘flatted leading-tones’ and call it ‘modal harmony.’ ”
36
Both Schoenberg and Schenker, for example, placed a heavy emphasis on musical relationships
of dubious perceptibility, justifying their methods by appeals to a nebulous and problematic notion of
“genius” (Arndt 2013).
37
There are interesting analogies between the history of science and the history of music theory,
which both struggle to chart a path between the twin dangers of triumphalism (or “whiggishness”) and
relativism. I suspect that the history of music theory, having survived whiggishness in the form of a
Hegelian-Schenkerian progressivism, is currently in the late stages of its relativist or historicist phase,
with a more realist, computational era on the horizon. See Wootton 2015.
38
Labov 1975 describes a case in which a speaker used constructions he declared to be
ungrammatical.
39
Compare Weber (1817–1821) 1846, pp. 453–54: “Theorists really possess incomparably less of
theory than practitioners themselves. For the former teach false rules while the latter act according to
those which are true. The rules which the former lay down, infinitely more often prove erroneous than
correct; while the latter produce elevated works of art, from which we might and should long since
have deduced better rules.”
40
McEnery and Wilson 1996.
41
Mistakes no doubt remain, however, and readers are invited to send corrections by email.
42
Von Hippel and Huron 2000.
43
The artificial datasets are amnesiac or “Markov twin” versions of the musical data that have the
same (first-order) melodic statistics: note x in the Markov twin uses the same intervals, with the same
probabilities, as note x in the original music, but has no memory of the previous interval. For more,
see von Hippel and Huron 2000. Note that the artificial datasets do exhibit a tendency toward “post-
skip reversal” at their registral extremes; Huron and von Hippel are not wrong about the existence of
this effect, just about its magnitude.
44
Von Hippel and Huron were concerned both with the universal claims of Leonard Meyer and
Eugene Narmour, and also with the situated pronouncements of Western pedagogues. Their data more
directly bears on the former.
45
Schenker rejected (local) Roman-numeral analysis while Schoenberg rejected the very notion of
nonharmonicity. These views are echoed by contemporary figures such as Robert Gjerdingen (1996)
and Ian Quinn (2018).
46
For the “two systems” see Dahlhaus 1990, pp. 59–65.
47
Gjerdingen and Bourne 2015. Gjerdingen’s project is influenced by Meyer 1996.
48
The connection is explicit in Holtmeier 2007.
49
If we assert that the bass F (3̂) is a chord tone, then the upper-voice E♭4 on the second quarter is
the seventh of iii7 (F–A♭–C–E♭), which moves up by step.
 50
The “cadenza doppia” is the formula in the top voice of Figure 1.5.1, where the leading tone
moves up to the tonic, often forming a suspension, before ultimately moving back to the leading tone
and resolving once again to the tonic.
51
In discussion, theorists have sometimes suggested that Bach’s cadential idiom “works” because
the bass and upper voices become temporarily out of sync, with the seventh resolving to 3̂
prematurely, before the upper voices reach the tonic. But consider this explanation from the point of
view of a student attempting to imitate Bach: here it is crucial that the “falling out of sync” (if that is
what it is) almost always occurs in a particular context and on a particular set of scale degrees. This
limitation is apparent only when we switch perspectives from that of the passive listener, confronting a
fixed body of earlier music, to that of the composer, who needs to know when the idiom is
permissible.
52
Temperley 2006 also argues for abstract schemas, while Byros 2017 distinguishes “microtheory”
from “macrotheory.” Rabinovitch (2018, 2019, and 2020) also tries to systematize schematic thinking,
though from a different point of view. Holtmeier 2011 notes that eighteenth-century musicians had
easy access to an elaborate theoretical discourse containing generalizations about topics such as
sequences, the rule of the octave, and the fundamental bass.
53
Gjerdingen 2007, p. 34.
54
Holtmeier 2011 makes a similar point.
55
These two intervals, the fourth and fifth, are the diatonic scale’s closest approximation to a half-
octave.
56
Burstein (forthcoming) describes the difficulties of assigning a univocal view to Schenker
himself or Schenkerians more generally. In this book, I will typically be concerned with views shared
by many different scholars—what Burstein describes as a “Schenkerian practice” originating with an
“idealized” Schenker.
57
Cf. Brown 2005, p. 78.
58
Forte 1959 aptly compares Schenker to Freud, whose work was considered much more scientific
in 1959 than it is now.
59
Cf. Schenker (1935) 1979, p. xxiii: “The musical examples which accompany this volume are
not merely practical aids; they have the same power and conviction as the visual aspect of the printed
composition itself (the foreground). That is, the graphic representation is part of the actual
composition, not merely an educational means.”
 Prelude
Transposition Along a Collection

If I had to identify the single most important concept in music theory, I

would pick transposition along a collection, the operation that shifts musical
material along a chord, a scale, or any other set of notes. Here the collection
acts like a musical ruler, providing a unit of musical distance that allows us
to move objects by one or more steps. Alternatively, we can think of a
collection as a slightly uneven ladder: the conceptual trick is to measure
distance with ladder rungs (steps) rather than some fixed unit like inches
(semitones). Figure P2.1 turns “doe a deer” into “ray a drop” by shifting
each note upward one rung along the diatonic ladder; this operation
preserves scalar distance, or distance as measured in scale steps. In “West
End Blues,” Louis Armstrong uses the same operation to move his motive
successively downward along an E♭ major triad; this preserves distances as

measured in triadic steps, or steps along the three note “scale” E ♭ –G–B ♭
(Figure P2.2). Actually, that oversimplifies: it is only the last two notes of
the motive that move along the triad, with the first a neighbor note one scale
step above the second note; to analyze the passage we therefore need to
juggle scalar and chordal distances simultaneously.
Figure P2.1. Moving a three-note motive up by step along the diatonic scale (T1).

Figure P2.2. In m. 7 of “West End Blues,” Louis Armstrong transposes the second two notes of his
motive downward by chordal step along the E♭ major triad (t–1), with the first note a neighbor one
scale step above the second.

In Figure P2.3, Domenico Scarlatti transposes along both chord and scale,
the right hand moving up by one triadic step while the left hand ascends by
two scalar steps. (For the sake of illustration, I am ignoring the slight
difference between the hands’ patterns and considering some notes to be
nonharmonic.) We see that for A–C–E, transposition by triadic step (C–E–A)
is almost equivalent to transposition by two diatonic steps (C–E–G). Readers
can explore this idea by shifting arbitrary motives along the intrinsic scale
defined by their own notes; this produces a range of interesting results,
particularly when combined with transposition along the scale, or when one
declares some notes to be nonharmonic. The passages in Figure P2.4 lie
somewhere between traditional motivic development and modern set theory;
as a group, they suggest that musical identity is up to us, determined by an
object’s transformational properties rather than fixed for all time by its
intrinsic constitution.
Figure P2.3. Domenico Scarlatti’s Sonata in A minor, K.3, mm. 3–6. The right-hand motive moves up
by chordal step (t1), while the left-hand variant moves up by two scale steps (T2), with the two
transpositions producing very similar results. (Here and elsewhere, I use p and n for passing and
neighboring notes.) Since t1 and T2 are almost the same, t–1 almost cancels T2 so that the combination
t–1T2 does very little. The bottom system analyzes the efficient voice leading between the top system’s
(A, C, E) and the bottom system’s (G, C, E) as a combination of transpositions along both chord (t–1)
and scale (T2).
Figure P2.4. Transposing two motives along the intrinsic scale defined by their pitches. Two different
stepwise transpositions are shown, one treating all notes as harmonic, the other treating some as
nonharmonic. Boldface n labels potential chromatic neighbors, regular n labels potential diatonic
neighbors. The last measure of the third line transposes a triad along its fifth, interpreting the second
note as a third above the first.

While motion-along-a-collection is most familiar in melodic contexts, it

plays an important role in the harmonic domain as well. Consider the notion
of “registral inversion.” Students are taught to construct inversions by
moving a chord’s bass to its soprano, a procedure that works well in the
conventionalized language of classical music, where harmonies are always
familiar, bass and melody are paramount, and the precise configuration of
inner voices is secondary. In other contexts, the procedure becomes
problematic. The chord in Figure P2.5 has one high note and two low notes
close together. As we move bottom note to top, we destroy its characteristic
spacing: the first inversion has one low and two high notes, while the second
inversion has all three notes high and close. Worse yet, another inversion
does not bring us back to our starting point but to a configuration in which
all notes are close together.
Figure P2.5. Forming registral inversions by moving a chord’s bottom note to the top.

An alternate strategy moves each voice upward by step along the chord
itself. First we compress the chord’s notes into a single octave, forming a
scale from its notes; then we move each voice by the same number of steps
along that scale. This yields a series of voicings sharing the same abstract
registral arrangement and returning to their initial configuration in a higher
octave (Figure P2.6).1 Once again, transposition along a collection preserves
intervals as measured by that collection: here, the middle note is always one
chordal step above the lowest, while the top is four chordal steps above the
middle. This approach is familiar in guitar pedagogy, where it is used to
construct inversions that keep the left hand in roughly the same position; in
chapter 3 we will find that it plays a role in traditional figured-bass theory as
well.

Figure P2.6. Forming registral inversions by moving each voice along the intrinsic scale formed by
the chord’s notes.
 Transposition along the chord is significant here because it preserves a
tangible aspect of musical structure: voicing, or the pattern of registral
spacing measured in steps along the chord itself. It is also important because
the most efficient voice leading between two chords—the mapping that
collectively moves the voices by the smallest overall amount—will always
connect chords that are voiced in the same way. Thus if we want to find
efficient voice leading between two chords, we should arrange them using
the same spacing-in-chordal steps. Figure P2.7 maps the preceding
example’s C3–D3–A4 to a minor triad spaced in exactly the same way.
There are three fundamental possibilities depending on whether the C in the
initial chord is mapped to the root, third, or fifth of the minor triad.2 We
conclude that chordal-step distance will be relevant whenever composers are
concerned with efficient voice leading, which is much of the time. Indeed,
composers might find themselves using transposition-along-the-chord
unknowingly, as a byproduct of trying to minimize contrapuntal motion. In
motivic contexts, composers intentionally move musical objects along the
chord; in contrapuntal contexts, these same transformations can arise
implicitly and without conscious knowledge. The phenomenon of
transposition-along-a-collection spans the border between implicit and
explicit.

Figure P2.7. Three spacing-preserving voice leadings from C3–D3–A4 to a minor triad; the first maps
the C to the triad’s root, the second to the triad’s third, and the last to the triad’s fifth. All chords are
voiced with the middle note one chordal step above the bass and four chordal steps below the soprano.

The deep point is that any collection of notes, from chord to motive to set,
can be associated with two different scales: an external or enclosing scale—
chromatic, diatonic, or some other contextually relevant collection—and the
intrinsic scale formed from its own notes. To transpose along the first scale
is to transpose in the ordinary sense, preserving intervals as measured along
the scale; to transpose along the second is to generalize the notion of chordal
inversion as just described, preserving intervals as measured along the chord.
Recent music theory has largely concerned itself with the first of these
transformations, whereas traditional theory often concerned itself with the
second, either implicitly (e.g., as a means of obtaining efficient voice
leading) or explicitly (e.g., close and open position). The intrinsic scale is a
missing link between traditional tonal theory and modern set theory,
obscured by the habit of conceiving sets in exclusively chromatic terms.
Once we train ourselves to think within a variety of different scales, starting
with “diatonic set theory” and generalizing from there, we will eventually
find our way to the intrinsic scale and the concepts implicit in traditional
theory.
There are many other uses for the concept of the intrinsic scale. One, to be
discussed in the next two chapters, harmonizes melodies using transposition
along both chord and scale. Another connects chords related by pitch-class
inversion and voiced in the same way; such spacing-preserving progressions
will invariably preserve the distance between at least one pair of voices.3
What results are exactly the passages we explored in §1.1, where two voices
move in parallel while the remaining notes alternate to create similarly
spaced, inversionally related chords. In other words, the seeming
contrapuntal wizardry of Gesualdo et al. is a straightforward consequence of
a concern for efficient voice leading, or attention to chord voicing as
measured in chordal steps (Figure P2.8). The recent discipline of “neo-
Riemannian theory” is largely concerned with these progressions, and
particularly those where the parallel voices are stationary; the concept of the
intrinsic scale gives us an intuitive handle on this complex literature, and on
the musical practices motivating it.4
Figure P2.8. Spacing-preserving progressions from a chord to its pitch-class inversion. Intrinsic
spacing is listed to the right of each staff. Open noteheads show the parallel voices. The progressions
on the left move to different registral inversions of the same target chord in a voicing-preserving way.
The progressions on the right also transpose the destination chord chromatically.

Transposition along the chord and transposition along the scale combine
to form a kind of doubly parallel motion along a pair of hierarchically nested
collections. I will notate x-step transposition along the chord as tx, with a
lowercase “t,” and y-step transposition along the scale as Ty, using a
boldface T for chromatic transposition and regular T for transposition along
scales such as the diatonic.5 Together, the two operations generate a wide
range of harmonies from a single starting point, all broadly similar in
arranging the same intervals in the same abstract registral configuration;
indeed these two transpositions are the only operations that preserve both
chord type and distance in chordal steps.6 For an n-note chord in an o-note
scale, transposition by n chordal steps is equivalent to transposition by o
scale steps, so we can always replace tn with To and vice versa. The two
operations commute, so txTy (tx and then Ty) is the same as Tytx (Ty and then
tx). Subscripts add as we would expect, with txTy and taTb combining to form
tx+aTy+b. We will see that there are even circumstances in which we might
want to consider inversion along both chord and scale, notated ixIy.
Remarkably, there is a simple and intuitive geometry of hierarchically
nested transposition, equally applicable to voice in chord, chord in scale, and
scale in chromatic aggregate. We can use the following recipe to represent
transposition along both an n-note collection (“chord”) and an o-note
collection (“scale”) containing it.
A. Draw a spiral with n loops, attaching its end to its beginning. In this
book I will always a draw clockwise spiral beginning at 12 o’clock and
moving inward until it reaches 9 o’clock for the nth time, then
connecting the end of the spiral clockwise to its starting point (Figure
P2.9).
B. Mark off o equally spaced points along this spiral, labeling them with
consecutive scale tones. To do this, divide the circle into o equal pie
slices; move along the spiral placing a point at the border of every nth
slice, with n the size of the chord: every border for one-note chords,
every other border for two-note chords, every third for three-note
chords, and so on. In this book I will usually place the C chord at the
innermost point at 12 o’clock and move clockwise along the spiral in a
descending fashion, labeling B, B♭, A . . . for the chromatic scale, B, A,
G . . . for the diatonic, and so on.

Figure P2.10 shows the resulting diagrams for major chords and diatonic
scales in chromatic space. Each point represents an entire collection of notes,
a complete major chord or complete diatonic scale. (In my experience,
readers can be confused by this point, so I will repeat it: in this book, points
on the spiral diagrams will always represent entire chords rather than single
notes.) The major-chord spiral winds around the circle three times, with
chords positioned at 12, 3, 6, and 9 o’clock (every third clock position); the
scalar spiral winds around the circle seven times, with chords placed at every
seventh slice. Readers can verify that a chord’s angular position corresponds
to the sum of its pitch classes, with each one-slice clockwise motion
decreasing the chord’s sum by one.7 Given a diagram, we can determine the
size of the chord from the number of loops and the size of the scale from the
number of points. Do not be concerned by the line’s self-intersection, as this
is an artifact of depicting an intrinsically higher-dimensional geometry on a
flat piece of paper; as appendix 2 explains, the actual line does not intersect
itself.
Figure P2.9. Spirals representing chords with one, two, three, and seven notes.
Figure P2.10. Spiral diagrams for major triads in the chromatic scale and seven-note diatonic
collections in the chromatic scale. Each point represents a complete chord: “Cmaj” in the first example
represents the C major triad, while “Cdia” in the second represents the C diatonic scale.

There are three rules for moving around the space.

1. Sliding along the spiral corresponds to transposition along the larger
scale-like collection, shifting the chord’s notes upward or downward by
the same number of scale steps. In this book clockwise motion descends
and counterclockwise motion ascends. (This is because descending
musical motion and clockwise circular motion are both defaults.) One
needs to move n slices clockwise to transpose an n-note chord down by
scale step.
2. A full loop around the center of the space corresponds to a transposition
along the smaller chord-like collection, moving each note upward or
downward by one chordal step as if the chord were a scale. A clockwise
loop moves notes downward by one chordal step while a
counterclockwise loop moves notes upward by one chordal step. (This
requires “jumping rings,” or leaving the spiral at some point, as
discussed in appendix 2.) A double loop moves each voice by two steps,
a triple loop by three steps, and so on. An n-fold loop transposes each
voice by octave and is equivalent to sliding along the complete spiral.
Unless otherwise noted, I will always use “loop” to mean loops
enclosing the center of the spiral; paths that return to their starting point
without enclosing the center leave each voice exactly where it began.
3. Two paths on the spiral represent the same voice leading when they
begin and end at the same points and involve the same total quantity of
angular motion. This total quantity of angular motion will always be
equal to the sum of the intervals moved by the voices.8

Figure P2.11 illustrates. The third rule is in many ways the most important,
as it allows us to understand every path as a combination of loops and slides,
or chordal and scalar transpositions. Before using a particular model, we will
typically calculate the voice leading linking nearby chords; this is the
combination of loops and slides that counteract each other as much as
possible, and it changes from diagram to diagram. Chapter 2 will explain
how to do this.
Figure P2.11. In the spiral diagrams, clockwise motion descends and counterclockwise motion
ascends (– and + respectively). Path a slides along the spiral and transposes along the scale. Path b
loops around the space and transposes along the chord; this requires leaving the spiral at some point.
Paths c and d represent the same voice leading as they start and end at the same point and involve the
same total quantity of angular motion (90° clockwise). Path c shows that this voice leading is T–5t1,
combining a clockwise loop (t1) with a five-step counterclockwise slide T–5.

The rest of this book will use the spiral diagrams to model intuitive
musical knowledge—and particularly the efficient voice leadings that result
when an operation-along-a-chord nearly counteracts the corresponding
operation-along-a-scale. We will focus on a small collection of spaces,
analyzing rock music with the 3-in-12 graph of chromatic triads, early music
with the 2-in-7 and 3-in-7 graphs of diatonic dyads and triads, and
functionally tonal modulation using the graph of 7-in-12 diatonic scales. In
modeling melody, we will sometimes consider very small scales such as the
2-in-3 graph of two-note triadic subsets. We will also use multiple spiral
diagrams to describe nested musical motion, the same basic structure
representing the motion of voice in chord, chord in scale, and scale in
aggregate. With a little practice, the reader can learn to translate geometrical
path to musical notation and vice versa; with a little more practice, the reader
can construct and manipulate the spiral diagrams for any chord in any scale.
The spiral diagrams can also be extended to include multiple chord types and
a wider range of voice leadings, including those with voice exchanges, and
those in which chord tones are doubled.9 No mathematics is required, and no
particular music-theoretical skill beyond an ability to manipulate
hierarchically nested transposition. My hope is that the spiral diagrams can
function as the music-theoretical equivalent of a consumer product, allowing
readers to enjoy the benefits of sophisticated musical geometry without
mastering its details—much as we can use GPS without fully understanding
the general-relativistic calculations it performs.
Readers who want to dive straight into the technical details can turn to
appendices 1 and 2; those who prefer a more analytical introduction should
instead proceed to chapter 2.

1
If we start with a chord whose notes are compressed within a single octave, the new algorithm
returns the same results as the traditional process of moving bottom note to top.
2
The next chapter’s prelude will explain how to derive the remaining spacing-preserving voice
leadings from these three options.
3
NB: “inversion” here means “pitch-class inversion,” the turning-upside-down of a chord’s
intervals, rather than “registral inversion,” the rearranging of its notes in register. In this case, the
identity of the parallel voices is determined by the registral inversions of the chords in the progression.
4
Spacing-preserving voice leadings between inversionally related chords can be obtained by
inverting a chord twice, along both the intrinsic and extrinsic scale; see appendix 1, appendix 3, and
Tymoczko 2020b for more.
5
This notation was first used by Julian Hook (2003, 2008).
6
Here I am using “chord type” to mean “transpositional set class.” If we consider pitch-class
inversions (like the major and minor triad) to be the same, then we need to include the “generalized
neo-Riemannian voice leadings” that send a chord to its similarly spaced inversional partner as shown
in Figure P2.8 (e.g., the “parallel” voice leading that lowers a major triad’s third by semitone or raises
the minor triad’s third by semitone). These act as half a transposition-along-the-chord (Tymoczko
2020b).
7
Here we need to use “clock arithmetic,” where we take the remainder when divided by the size of
the scale. So, for instance, the C major triad contains C, E, and G or 0, 4, 7 in pitch class notation,
which together sum to 11 (0 + 4 + 7). A♭ major contains A♭, C, and E♭, or 8, 0, 3, which also sum to

11. E major is E, G♯, and B, or 4, 8, and 11, which sum to 23, which is 11 when we divide by 12 and

take the remainder. B major, G major, and E♭ major are three pie-slices clockwise, since their notes
sum to 8 modulo 12.
8
Rule 3 is equivalent to the statement that two paths correspond to the same voice leading if one
can be smoothly transformed into the other without breaking the path or moving its endpoints; such
deformations are studied by topologists (Tymoczko 2020b). In summing the intervals in each voice,
we do not use clock arithmetic but rather real numbers representing paths in pitch-class space
(appendix 1, Tymoczko 2011a).
9
Readers may find it useful to visit the website https://www.madmusicalscience.com, where they
can find movies and software.
 2
Rock Logic

In many musical styles, melodic steps tend to descend while leaps tend to
ascend.1 In Figure 2.0.1, for example, the singer steps downward from 3̂ to
1̂, descending by step within the chord before leaping back to 3̂ to start the
next verse. It is reasonable to think that the melody motivates the song’s
inversion of traditional harmonic norms: if we decide to associate each
melodic note with a major triad in the D major scale, the I–V–IV–I
progression is one of a small number of possibilities. The subversion of the
tonic-subdominant-dominant paradigm, in other words, may not be a
rebellion against classical harmonic practice, but rather a straightforward
solution to a basic musical problem.

Figure 2.0.1. “Helpless” by Neil Young.

This chapter will use this idea to motivate a model of rock harmony, one
that emphasizes the melodic origin of its harmonic idioms. Though the
underlying techniques apply equally well to ascending or leaping melodies,
my main focus will be on stepwise descent. The goal is to understand the
implicit knowledge of talented but possibly untrained musicians—principles
that they might plausibly absorb from thousands of hours of listening and
playing. This project is broadly comparable to that of constructing a
grammar for natural language, another formal system that humans learn
without explicit instruction. This embodied and intuitive grammar contrasts
with the explicit doctrines of the notated tradition, from species counterpoint
to twelve-tone composition.

1. A melodic principle

Suppose a melody embellishes a stepwise descent from one tonic-triad note

to another. For simplicity, imagine that we decide to harmonize each note
with a root-position major triad—ignoring minor chords for the moment.
What sorts of progressions should we expect to find? Figure 2.1.1 identifies
all the two-chord progressions beginning with C major and supporting a
stepwise melodic descent, with points representing major triads and lines the
different steps: thus, C and G are connected by a thick solid black line
indicating that the root of C major can descend by semitone to the third of G
major, and a thin dotted black line indicating that the third of C major can
descend by major second to the fifth of G major. The figure can be
understood as a musical “gameboard” describing the local moves available
to musicians interested in descending steps. Simple tabulation shows that
descending steps preferentially give rise to what are sometimes called
“weak” root-progressions: ascending fifths rather than descending fifths,
ascending minor thirds rather than descending minor thirds, and descending
rather than ascending steps (Figure 2.1.2).2
Figure 2.1.1. A musical gameboard depicting the possibilities for stepwise melodic descent between
triads. Each letter represents a complete major triad. Solid lines represent the possibility for a note to
descend by semitone while dotted lines represent a major-second descent. A thick dark line descends
from the root, a thin dark line descends from the third, and a lighter line descends from the fifth.

Figure 2.1.2. “Weak” root progressions (ascending fifths and thirds, descending steps) generally
permit more opportunities for stepwise descent, with the major third being the only interval equally
balanced between strong and weak.
 The model has Schenkerian overtones in suggesting that harmonic
patterns can arise out of fundamentally melodic aims—in this case, phrase-
level descending motion from one tonic-triad note to another. But rather than
trying to show that stepwise descending melody necessarily reproduces the
conventions of functional harmony, I will suggest that it leads to different
kinds of progressions in different harmonic contexts. In A Geometry of
Music, I used complicated geometrical models to try to capture this
interdependence of harmony and melody, describing twisted, singular, and
higher-dimensional spaces in which each musical voice is represented by its
own dimension. Here I will use the simpler spiral diagrams for the same
purpose. Figure 2.1.3 reproduces the voice-leading space for major triads in
the chromatic scale. For reasons to be explained in §2.6, I label points using
the triad’s root; readers should keep in mind, however, that each point
represents an entire chord.
Figure 2.1.3. The spiral diagram for chromatic triads. Each point represents a complete major triad.

We know that slides along the spiral correspond to transposition along the
scale while loops represent transposition along the chord. Unfortunately, it is
not obvious from these principles how to understand radial motion between
nearby chords, say from C major vertically outward to E major. This is a
general problem: for each new spiral diagram, we will need to calculate the
voice leadings between nearby points. Here, transposition-along-the-chord
largely counteracts transposition-along-the-scale, leaving efficient voice
leading as the result.
 There is a general recipe that works for any spiral diagram: to find a direct
path between two chords, slide your finger in the desired direction
(clockwise or counterclockwise) from one to the other along the spiral; if
you are looking for a purely radial path, then both directions are equivalent.3
The number of chords you touch, not counting the first, is the transposition
along the scale (Tx), with positive and negative values of x corresponding to
counterclockwise and clockwise motion respectively. Meanwhile, the
number of times you revisit your initial angular position, y, becomes the
transposition along the chord t±y, with its sign opposite that of the scalar
transposition. The two combine to form Txt–y or T–xty, a composite
representing the most direct voice leading between the two chords.
Example 1. Radial motion from C to E. Imagine sliding counterclockwise
along the spiral from C major to E major. Not counting our starting point, we
touch four chords for T4. (Remember that I use boldface T for chromatic
transposition, and regular T for transposition along a scale such as the
diatonic.) We return once to our initial angular location of 12 o’clock, for t–1,
with a negative sign because our scalar transposition is positive. The
combination of these two motions, shown in Figure 2.1.4, moves C down to
B, keeps E fixed, and moves G up to G♯; this is the LP voice leading of neo-
Riemannian theory.4 The same reasoning shows that radial motion from E
major out to A♭ major, and from A♭ major inward to C major, also produces
T4t–1 or LP.
Figure 2.1.4. Calculating the radial path from C to E. We move counterclockwise along the spiral by
four steps to reach E, for T4; we pass our original angular position once, for t–1.

Example 2. Radial motion from E to C. Now we slide clockwise from E

down to C, returning once to our starting point for T–4t1. This is the opposite
of the voice leading in the previous example: G♯ moves down to G, E stays
fixed, and B moves up to C. This is the PL voice leading of neo-Riemannian
theory. Similar reasoning shows that radial motion from A♭ to E, and from C

to A♭, also produces T–4t1, or the PL voice leading. This allows us to express
all motions on the diagram in musical notation (Figure 2.1.5).
Figure 2.1.5. The voice leadings on circular chord space represented in notation: vertical motion
corresponds to t–1T4 or t–1T–4 (LP or PL) while horizontal motion corresponds to chromatic
transposition (Tx).

Example 3. 90° counterclockwise motion from C to F. We could simply

reapply the algorithm, sliding counterclockwise five steps for T5 and passing
our original radial position once for t–1. Alternatively, we can add a
chromatic transposition to one of the radial voice leadings just calculated.
For instance, to calculate the counterclockwise path from C to F, we start
with the outward radial motion from C to E, representing the voice leading
(C, E, G) → (B, E, G♯).5 To this we add a quarter turn along the spiral from
E to F, transposing each note up by semitone and giving us the composite (C,
E, G) → (C, F, A). Symbolically, T4t–1 + T1 = T5t–1. This voice leading
holds root fixed, moves third up by semitone, and fifth up by two semitones.
A quarter-turn clockwise from F to C has the opposite effect, moving F
down to E, A down to G, and holding C fixed (t1T–5). Similarly, clockwise
motion from C to G moves C down to B, E down to D, and holds G fixed.
Clockwise motion from C to B♭ (T–2) lowers each voice by two semitones,
while counterclockwise motion between the same two points generates the
voice leading (C, E, G) → (D, F, B♭), or (t1T–2).
Proceeding in this way, we see that the descending stepwise melodies
recorded on our earlier “graph of melodic possibilities” (Figure 2.1.1) are
exactly those that are produced by radial or short clockwise motion on the
new spiral graph. It follows that a musical preference for descending
stepwise melodies will be modeled by a geometrical tendency to move
radially or clockwise. The spiral diagram also encodes a musically useful
notion of distance, since larger motions represent larger voice leading.
Finally, the graph gives us a notion of betweenness, showing how larger
voice leadings can be factored into smaller moves: for example, Figure 2.1.6
shows that the two-step clockwise progression from C major to B ♭ lowers
each voice by major second; there are exactly three chords lying between C
and B♭ on the spiral, each preserving the stepwise descending voice leading.
The model’s geometrical structure thus encodes the musical intuition that the
G major chord lies “on the way” from C major to B♭ major, as do B and E♭
major. Here, a picture is worth a thousand words in the sense that these
intuitions of “betweenness” are packed into a single percept, rather than
many disconnected verbal descriptions.6
Figure 2.1.6. Geometrical representations encode relations of “betweenness.” Here we see that B
major, E♭ major, and G major are all between C major and B♭ major; this means we can insert them

between C and B♭ without disrupting the stepwise descending voice leading.

In exploring voice leading, it is often useful to consider what I call the

“Principle of Musical Approximation,” which encourages us to think of
nearly even chords as perturbations of completely symmetrical chords.
Figure 2.1.7 shows a sequence of semitonally descending augmented triads,
which can be labeled in a number of ways: C+–B+–B♭ +, C+–E♭ +–G♭ +, C+–
G+–D+, and so on. Since the augmented triad is completely symmetrical this
is a conceptual rather than sonic difference. The major triad is nearly
symmetrical—just a single semitone away from the augmented triad—and
divides the octave into three almost-even parts. We can obtain a variety of
descending major-chord progressions by lowering any note of each
augmented triad: C–B–B♭, C–E♭–G♭, C–G–D, and so on. These are the same
root progressions we obtain in the augmented case, only now they are
sonically distinctive; any members of this collection can be related by
applying major-third substitution (t1T–4 or t–1T4) to their constituent chords;
because of the triad’s near symmetry, this major-third substitution has only a
small effect on the voices. Musically, we can think of the resulting
progressions as slight distortions of a continuous parallel descent, an
asymmetrical major-chord analogue to the perfectly parallel augmented-
chord model.7 We can repeat this derivation to obtain descending sequences
of minor triads (raising an augmented-triad note), diminished triads, and so
on.

Figure 2.1.7. (top) A descending sequence of augmented triads that can be labeled in a variety of
ways. (bottom line) By lowering a note of each augmented triad, we can obtain a variety of root
progressions exhibiting descending stepwise voice leadings. These progressions all move one step
clockwise on the spiral diagram.

A sophisticated treatment of melody is necessarily hierarchical, allowing

voices to move inside a chordal field that is itself moving. This can be
modeled by combining two separate spiral diagrams: a 1-in-3 diagram
representing the motion of voice inside chord, and a 3-in-12 diagram
representing the motion of chord inside chromatic collection. For the sake of
simplicity, this chapter will generally adopt an alternate approach in which
the melodic voice is stationary relative to the triadic field, being “carried
along” by the triadic motion. (We can think of the inner voices as moving in
doubly parallel motion with the melody, using the same combination txTy.)
This allows us to avoid the more complex hierarchical picture, treating 3-in-
12 motions on the circle as metonymic for the melodic motions they
produce. From this point of view, it is the bass that provides contrary motion,
moving relative to the shifting triadic field. Figure 2.1.8 shows that we can
think of the bass as following its own double transposition, sharing the same
chromatic transposition Tx but a different scalar transposition ty. We will
return to this thought in §3.7.

Figure 2.1.8. In this chapter, we will generally imagine inner voices moving in doubly parallel motion
with the melody while the bass moves independently.

2. A harmonic principle

Progressions like C→B and C→G ♭ are contrapuntally plausible but

musically rare, at least in major-mode songs. This is presumably because
they are felt to move between distant chords, implying very different
diatonic scales.
Figure 2.2.1 formalizes the notion of “diatonic distance” using the circle
of fifths: any diatonic scale is distance 0 from itself, and distance 1 from
those collections a fifth away: thus C and G diatonic are one unit apart, since
just one semitonal shift (F→F ♯ ) relates them, while C and D diatonic are

two units apart, since it takes two semitonal shifts (e.g., C→C♯ and F→F♯)
to transform one into the other. This conception of “diatonic distance” is a
voice-leading distance that reflects how far the scale’s notes have to move to
get to their destination; this is similar to, but not quite the same as, a model
based on common tones. Scalar voice-leading distance can be represented by
the circular graph of diatonic scales in chromatic space, where motion
between adjacent points represents the single-semitone voice leading
between fifth-related diatonic collections, and angular distance is equivalent
to scalar distance (Figure P2.10). We will return to this idea in chapter 8.

Figure 2.2.1. Voice-leading distance between diatonic collections.

I will define the harmonic distance between two major triads as the
smallest distance from a scale containing one chord to a scale containing the
other (Figure 2.2.2). Thus the C major triad is distance 0 not just from F
major and G major (since the C, F, and G triads all belong to the C diatonic
scale), but also from D major (since the G diatonic scale contains both C
major and D major triads) and B♭ major (since F diatonic contains B♭ major

and C major). Meanwhile, the C major triad is one unit away from the E ♭

major triad (since C major belongs to F diatonic, and E♭ major belongs to B♭

diatonic, and the B ♭ diatonic scale is one step away from F diatonic). By
similar reasoning the C and A major triads are one unit away from each
other. Major-third-related triads, like C and E, or C and A ♭ , are two units
away from each other, whereas semitonally related triads are three steps
away. Maximally separated are tritone-related triads, which are four steps
apart. This notion of triadic distance is very closely related to the circle-of-
fifths distance between triads, here justified by the fact that fifth-related
triads create coherent diatonic backgrounds or macroharmonies.8 It is, in
other words, a voice-leading distance at another hierarchical level.

Figure 2.2.2. Diatonic distance between major triads, defined as the minimum distance between a
diatonic scale containing the first chord and a diatonic scale containing the second.
 The distance traversed by a repeating chord progression can then be
defined as the total number of modulatory steps involved in the entire
progression, with the proviso that we return to the initial chord and scale
(Figure 2.2.3). My proposal is that rock musicians favor chord progressions
that are diatonically close—that is, lying in a single diatonic scale or a small
collection of closely related scales. If this is so, then we should expect more
fifth-related triads like C and F, and major-second related triads, like C and
B♭, than semitonally or tritone-related triads (like C and B or C and F♯).9
The suggestion is that rock musicians have implicitly absorbed a weak
diatonic norm through extensive exposure to Western music.10

Figure 2.2.3. Measuring the total diatonic distance of a progression; the progression D–A–E–G has
diatonic distance 2, since it requires modulating from D diatonic to A diatonic and back.

Because this norm is often learned implicitly, it is not absolute; we can

therefore expect to find progressions between triads not belonging to the
same diatonic scale—with progressions by minor third more prevalent than
major third, major third more prevalent than semitone, and semitone more
prevalent than tritone. This is what we see in Figure 2.2.4, which records the
root progressions between major chords in the top 200 of Rolling Stone’s
“500 greatest songs of all time.”11 (Note that this figure does not distinguish
ascending intervals from descending intervals; we will ignore that issue for
the moment.) The ordering of intervals is precisely what the model suggests,
with fifths being the most frequent root motion, followed by major seconds,
minor thirds, major thirds, and minor seconds.12
Figure 2.2.4. Progressions connecting two major triads in Rolling Stone’s “500 Greatest Songs of All
Time.” Ascending and descending motions are grouped together, P5 is “perfect fifth,” M2 is “major
second,” and so on. The columns show the percentages in all the pieces, the major-mode pieces, the
minor-mode pieces, and an average of the two modes.

Figure 2.2.5 rewrites the spiral diagram using Roman numerals, with
darkness representing harmonic proximity to the tonic chord. Once again,
the figure leads us to expect an abundance of progressions by ascending
fifths, ascending and descending major seconds, and, to a lesser extent,
ascending minor thirds. It does not give us reason to expect many
descending fifths or descending thirds, progressions traditionally associated
with functional tonality. For this reason, I will say that the gameboard
embodies a retrofunctional norm.13 Starting with two very basic musical
preferences—descending melodic steps and a preference for diatonically
close progressions—we have derived substantive expectations about rock
harmony, expectations that are both different from those appropriate to
classical music, and (as we will shortly see) consistent with what we find in
the repertoire.
Figure 2.2.5. The spiral diagram for major triads, rewritten with Roman numerals and with dark
chords being harmonically closer to I.

Conversely, the rules of functional harmony are precisely not the rules you
would come up with if you were to construct a musical syntax on your own
(if, say, you were a talented but not formally trained musician sitting in your
bedroom with a guitar). Instead, you would be more likely to come up with
the inverse—a retrofunctional norm, favoring ascending rather than
descending fifths and ascending rather than descending minor thirds. This
retrofunctional norm, I will argue, is an intuitive counterknowledge arising
from the deep structure of musical space, a subterranean musical practice
distinct from, and to some extent opposed to, the culturally sanctioned
routines of the notated tradition.

3. A first loop family

Neil Young’s “Helpless” features simple but powerful musical forces

operating in concert, its melody descending from third to root, while its
harmonies articulate one- and two-step clockwise motions on the 3-in-12
spiral (Figure 2.3.1). The result is an endlessly repeating, endlessly falling
root-position I–V–IV–I, a diatonic progression that is exquisitely logical yet
alien to more than two centuries of functionally tonal music-making. It is,
one might say, both natural and forbidden—natural in being an obvious
musical solution to a deep musical problem, and forbidden because it
violates strong cultural norms.

Figure 2.3.1. “Helpless” graphed on the spiral diagram.

 The Beatles’ “Eight Days a Week” traces an equally forbidden path
through the space, replacing the D–A–G with D–E–G (Figure 2.3.2). Once
again, the melody outlines a 3̂–2̂–1̂ descent, though now embellished with
additional notes that imply multiple voices. Classically oriented music
theorists have puzzled over this progression, wondering whether the E major
chord should be understood as “predominant” or “thwarted secondary
dominant.” I reject this dichotomy: rather than feinting toward a dominant
that never appears, the II chord is a thing-in-itself, largely unknown in
classical harmony and not reducible to its categories. Like “Helpless,”
“Eight Days a Week” embodies a nonclassical norm, harmonizing
descending stepwise melodies with closely related major triads.14

Figure 2.3.2. “Eight Days a Week.”

Elliott Smith’s “Rose Parade” features a I–V–II–IV progression that takes

one chord from each radial position (Figure 2.3.3). This progression fuses
the chords of “Helpless” and “Eight Days a Week” precisely as they appear
on the circle—almost as if they had been combined in the popular-music
equivalent of a particle accelerator, smashed together to produce an
undiscovered byproduct. Once again we have an unadorned 3̂–2̂–1̂ melody
in the voice, with Smith’s self-destructive lyrics annihilating the cheerful
harmony, bubblegum pop concealing an almost unprecedented expression of
rock-star self-hatred.15 Like Kurt Cobain, Smith dares us to take pleasure in
his expressions of self-torment—bringing to mind Dorothy Rowe’s
description of depression as “a prison where you are both the suffering
prisoner and the cruel jailer.”16 The power of this song lies precisely in the
use of iconic musical material to express extreme emotion.

Figure 2.3.3. “Rose Parade.”

It is interesting to think of “Rose Parade” as a maximal progression, an

archetype which generates the others by elision: eliminating II gives us
“Helpless” while eliminating V gives us “Eight Days a Week.” From this
point of view, the four positions on the chromatic chord circle can be thought
of as “slots” that can be either filled or left open—a purely contrapuntal
analogue to the familiar harmonic categories we call “tonal functions.”17 A
third possibility is to eliminate IV, producing I–V–II–I (Figure 2.3.4). This is
rare, though a variant does appear in Bob Dylan’s “Knockin’ on Heaven’s
Door.” (The use of the minor ii, shown in Figure 2.3.5, does not unsettle the
descending stepwise voice leading, as we will discuss.) That song again
harmonizes 3̂–2̂–1̂, alternating the I–V–II of “Rose Parade” with the I–V–IV
of “Helpless”—as if hinting at the full “Rose Parade” progression without
ever stating it. Here the ii–I has the character of an interruption, a musical
“comma” demanding the more definitive IV–I close.
Figure 2.3.4. A family containing all the progressions that can be generated by eliminating a nontonic
chord from the “Rose Parade” progression.

Figure 2.3.5. “Knockin’ on Heaven’s Door.”

Dylan’s song uses an important technique that can be found throughout

the triadic tradition, harmonizing its melody by transposing along the
prevailing chord. My example shows that the background voices shadow the
main melody at a distance of two steps along a changing sequence of
harmonies—producing a scalar interval of a sixth or fifth depending on the
chord. This is essentially just parallel motion within a (modulating) scale,
but using three-note scales rather than seven-note scales. Once again we see
the importance of a flexible approach to musical distance: measured in
chordal steps, the two melodies move exactly in parallel, maintaining their
two-step distance; measured in diatonic or chromatic steps, they diverge.
Similar parallelism can be found whenever voices maintain the same spacing
along the sounding harmonies. Chapter 3 will use this idea to develop a
general theory of four-part counterpoint.
 Collectively, these examples show how the spiral diagrams can help us
perceive relationships that would not otherwise be apparent: on first hearing,
one would not think to associate the progressions of “Helpless” and “Eight
Days a Week”; but in the context of “Rose Parade” and “Knockin’ on
Heaven’s Door” the similarities become much clearer. This is characteristic
of what philosophers call a “family resemblance”—like a genetic family, our
“chord-loop family” (or “loop family”) is united by features that are more
obvious when we consider the group as a whole than when we attend to pairs
of individuals. By revealing the different paths through chord space, the
spiral model gives us a collection of loop families that can be used to
categorize and conceptualize rock norms. It also hints at a picture of rock
music as nearly sequential, repeatedly taking short clockwise motions
through an abstract harmonic space. These are not exact sequences as there
are a variety of options at each angular position; instead they represent a
disguised or distorted regularity that we will encounter elsewhere.

4. Two more families

“Sympathy for the Devil” uses the same progression as “Helpless,”

reinterpreted as a mixolydian I–♭VII–IV rather than ionian V–IV–I.18 This is
one of the most characteristic progressions in hard rock, featured in the
Who’s “I Can’t Explain,” AC/DC’s “Back in Black,” and many other songs.
Mick Jagger’s melody vacillates between 5̂ and 4̂ until the chorus, where it
descends to 3̂ over a V7–I progression (Figure 2.4.1). The music thus
combines rock retrofunction in its verses with functional tonality in the
chorus, using harmonic syntax to reinforce its form. (“I Can’t Explain” does
something similar, moving to V just before the old-fashioned I–vi–IV–V
chorus.) The mixolydian I–♭VII–IV can be embedded inside a I–♭III–♭VII–
IV which, like the progression of “Rose Parade,” is geometrically maximal
—selecting one harmony from each of the four angular positions in the spiral
diagram. This progression appears in Status Quo’s “Pictures of Matchstick
Men,” Seals and Crofts’s “Summer Breeze,” the White Stripes’ “The Air
Near My Fingers” (shown in Figure 2.4.2), and countless other songs. As
before, we can generate three-chord progressions by eliminating one of the
nontonic chords (Figure 2.4.3). All three possibilities are reasonably
common: I– ♭ VII–IV–I is ubiquitous (e.g., “Tangled Up in Blue”), I– ♭ III–

♭ VII–I is pervasive in hard rock, often with a minor tonic (“Stairway to

Heaven,” “Locomotive Breath,” “Paranoid,” see §2.6 for discussion of
minor triads), and I–♭III–IV–I, though not quite as popular as the others, is
also found (“I Can See for Miles”). Again, we might not think to associate
these songs without the spiral model.

Figure 2.4.1. “Sympathy for the Devil.”

Figure 2.4.2. “The Air Near My Fingers.”

Figure 2.4.3. Progressions that can be generated by eliminating a nontonic chord from the “Air Near
My Fingers” progression.

At the level of the phrase, rock songs tend to adopt one of two strategies.
“Helpless” and “Back in Black” return to the tonic at the start of each two-
bar unit, overlapping their phrases to generate a feeling of constant forward
progress. Here, harmonic arrival coincides with melodic departure, and the
music is largely devoid of cadences in the sense of phrase-terminating
pauses. “Eight Days a Week,” “Sympathy for the Devil,” and “I Can’t
Explain” instead return to the tonic at the end of each two-bar unit; this
creates a much stronger sense of arrival, allowing for a feeling of rest
between each repeating unit. Here it is reasonable to speak of cadences in a
relatively traditional sense. 19
The third loop family is built on the I–V–♭VII–IV progression featured in
songs like “Natural Woman” (written for Aretha Franklin by Goffin and
King) and Duran Duran’s “Rio.” The former is shown in Figure 2.4.4, the
descending voice leading clear on the musical surface. This 6–5 progression
is the retrograde of the classical tradition’s ascending 5–6 sequence (Figure
2.4.5), giving rise to a descending chromatic line. Unlike the other
progressions we have considered, it plays a significant role in the classical
tradition—appearing in the music of Beethoven (Op. 31, no. 1, the Waldstein
sonata), Schubert (the late G major quartet, etc.), and many other composers.
In classical contexts it usually articulates a descent from tonic down to
dominant, whereas in popular music it often returns plagally to the tonic
after just four chords.20 Thus classical composers embed the progression
within a broadly functional movement from tonic to dominant whereas
popular music is more likely to use it as an alternative to tonic-dominant
functionality. The four-chord pattern generates two three-chord progressions
we have already encountered, I– ♭ VII–IV–I and I–V–IV–I, and one new

progression, I–V– ♭ VII–I, used in the Beatles’ “I’m a Loser” and Guns-N-
Roses’ “Sweet Child O’ Mine.”
Figure 2.4.4. “Natural Woman.”
Figure 2.4.5. Ascending and descending 6–5 sequences.

5. Shepard-tone passacaglias

Figure 2.5.1 offers two interpretations of the repeat in Neil Young’s

“Helpless”: according to the first, we take two steps clockwise around the
circle before rewinding, moving back to the start via ascending voice leading
that resets all voices to their initial position; according to the second, we
return from subdominant to tonic by way of another clockwise motion,
completing a full circle in chord space. This circle, we know, has the effect
of transposing down one step along the chord, sending F ♯ down to D, D

down to A, and A down to F♯. Musically, it represents a hearing in which D

of the IV continues as an inner voice of the following tonic—a hearing that
is supported by a subtle feature of Neil Young’s Massey Hall performance,
the guitar’s high G4 (4̂) entering just as the vocal melody arrives at its D4.
This suggests that the G major IV chord continues downward to the D major
tonic, the entire piece an endless descent. As in “Eight Days a Week,” the
melodic voice has to ascend relative to this descending background in order
to remain in place, returning to the same pitch but a different scale degree.
Figure 2.5.1. Two ways of conceiving the repeat in “Helpless.”

Figure 2.5.2 traces this endless Urlinie throughout the progression: the
note D4, which ends the vocal line, continues downward by step in the guitar
to C♯ (over V) and then B (over IV) before arriving at A3 at the next repeat
—whereupon it continues downward to G3 before shifting upward by octave
to the guitar’s high G4. The music thus embeds a hidden canon or round
endlessly descending by step (Figure 2.5.3). Unlike a traditional round,
however, where the melodic lines are articulated by actual instruments, this
canon is constructed by listeners—supported by rich, strummed guitar
chords that suggest but do not completely determine specific voices. It is we
who connect the D4 in the vocal melody to the subsequent D4 in the guitar,
we who hear the common tones and stepwise motions linking each chord to
the next. Nevertheless these are reasonable connections to make: the
stepwise motions in Figure 2.5.2 are really there in the music, and can be
identified even by the casual listener. (Sing the canonic line along with the
recording and you will realize how obvious it is.) To me this is one of the
most fascinating differences between voice leading in classical music and
rock—in the former, it is embodied in notation while in the latter it is
partially constructed by the audience.21

Figure 2.5.2. Each voice in “Helpless” connects to the next-lowest voice across the repeat.

Figure 2.5.3. “Helpless” embeds a three-voice continuously descending canon. Voices enter every two
measures.

“Helpless” thus presents an endless stepwise descent, sinking ever

downward while somehow remaining stationary. This description once again
evokes the Shepard tone, an endless glissando that somehow remains in
place. David Feurzeig coined the term Shepard-tone passacaglia to refer to
progressions in which voices connect nontrivially across the repeat, as when
soprano becomes alto, alto tenor, and so on: here, what initially seems like
repetition can also function as a continuation, in this case because the lines
continue their trajectories.22 Expressively, this endless, tumbling descent—
moving always forward yet also staying the same—functions as an apt
representation of helplessness, capturing the persistence of memory
alongside inexorable progression of time, the chords staying fixed as the
years tick by. It is, perhaps, one of the reasons we are inclined to listen to
these three harmonies over and over, always progressing yet never reaching
a conclusion.
These Shepard-tone passacaglias are easy to find once you start to listen
for them: indeed our three loop families all give rise to analogous canons
(Figure 2.5.4). The canon will sometimes be obscure and hard to follow, but
sometimes very close to the musical surface—as in the Beatles’ “You Won’t
See Me,” where the main melody encircles the background vocals to create a
continuous scale (Figure 2.5.5). That descent was apparently a focus of
compositional attention: the verse ends with a turn to V/IV, allowing the
composite melody to descend by almost a complete octave, reaching the low
tonic just as the chorus arrives.

Figure 2.5.4. The canon in “Eight Days a Week.” Voices enter every two measures.
Figure 2.5.5. “You Won’t See Me” explicitly articulates a continuous stepwise descent.

It is natural to wonder how many of these repeating, pseudocanonic rock

progressions there are—that is, how many progressions (a) begin and end at
the tonic; (b) move clockwise through a single turn in chord space, selecting
one of the three chords at every angular position; and (c) contain no
harmonically “distant” chords whose roots are related by semitone or
tritone.23 It turns out there are just four: the three families we already
considered and I–V–II–VI, a sequence of ascending fifths. This last
progression is almost entirely unknown in popular music, at least as a
repeating four-chord loop; this is likely because VI–I makes for an
unsatisfying return to the tonic, particularly in comparison with the IV–I that
ends the other three progressions.24
Remarkably, these four progressions are all identical under rotation. That
is, they all can be expressed as a repeating sequence of major triads D–A–E–
G, but with a different choice of tonic: if D is tonic, we have the “Rose
Parade” progression (I–V–II–IV); if A is tonic it is “Natural Woman” (I–V–
♭VII–IV); if E is tonic it is “The Air Near My Fingers” (I–♭III–♭VII–IV); and
if G is tonic, it is the unused I–V–II–VI (Figure 2.5.6). What this means is
that our three song families are all in some deep sense the same, and in fact
they can be superimposed to good effect. Figure 2.5.7 tries, lamely, to
illustrate. (It is much better to sing it with friends.) Here we have a strange
sort of polytonality without dissonance, the different keys largely a function
of where one feels the four-bar hypermetric accent. If you know the original
songs, it is possible to flip back and forth between interpretations, creating
an unstable situation where scale degree is a function of attention—a
musical analogue of the famous duck/rabbit illusion.25 Figure 2.5.8
represents the canon implied by this ubiquitous four-chord sequence, a line
that can be described, only half-jokingly, as the “fundamental canon of
rock.” It is what we might call a deep schema, rarely grasped explicitly but
existing in the musical unconscious—and manifesting in many seemingly
unrelated ways.

Figure 2.5.6. A single progression heard in four different ways.

Figure 2.5.7. Three songs in three keys played at the same time without dissonance.

Figure 2.5.8. The fundamental canon of rock. Voices enter every two measures.

One moral is that rock centricity is often a matter of phrasing and musical
emphasis rather than abstract pitch relationships. O-Zone’s “Dragostea Din
Tei” provides a marvelous example, modulating from C major in the chorus
to A minor in the verse without ever departing from its repeating C–G–a–F
loop (to be discussed shortly). This is initially accomplished hypermetrically,
prolonging the G chord for two measures so that the verse begins on A minor
rather than F major (16 bars into the music, at about 0'30"). The remaining
modulations operate entirely by orchestration, the bass synthesizers dropping
out a bar before each section change; their return creates a musical accent
which, along with the melody, helps shift the sense of tonal center back and
forth between A minor and C major, even while preserving the four-bar
harmonic rhythm. These modulations suggest that centricity can be a matter
of delicate hints rather than unmistakable signposts.26
Our progression C–G–D–A–(C), with its three ascending fifths and one
ascending minor third, is cousin to the most famous minor third in all of
music theory, Rameau’s double emploi. Rameau interpreted functional
harmony using a sequence of four descending diatonic fifths d–G–C–F
whose phrase structure was offset relative to its harmonic structure: C–F–d–
G–(C) rather than d–G–C–F–(d). His “double emploi” is the minor-third
juncture between the ends of the fifth sequence, sometimes elided into a ii65
composite. These two chord loops are retrogrades: three fifths and a minor
third, ascending in rock and descending in functional harmony. This
resemblance underscores the sense in which rock music is genuinely
retrofunctional, its chords inverting classical expectations—not out of
cussedness but for reasons of contrapuntal logic.

6. Minor triads and other trichords

By now, readers will be wondering how to fit minor triads into the spiral
model. A first answer is that Figure 2.1.3 can be used to represent any type
of three-note chord in chromatic space: major chords, minor chords,
diminished triads, 026 trichords like C–D–F♯, three-note chromatic clusters,
and so on. (This is why I labeled the figure by root, without specifying the
complete chord type.) That is, its basic principles are completely
independent of a chord’s specific intervallic content, depending only on the
size of the chord and the size of the scale: sliding along the spiral transposes
along the scale, while looping around the space transposes along the chord,
so that moving radially outward, in any twelve-tone scale, combines
transposition down one chordal step with transposition up four chromatic
semitones (t–1T4). For the C minor triad, radial motion corresponds to the
voice leading (C, E♭, G) → (B, E, G); for an 026 trichord it is (C, D, F♯)→

(A♯, E, F♯); and for a chromatic cluster it is (C4, D♭4, D4) → (F♯3, E4,
F4).27 (Inward radial motion is the inverse, transposing one step upward
along the chord and four semitones chromatically downward.) In each case,
radial motion involves the same total amount of upward and downward
semitonal motion, so that the sum of the paths in all voices is zero semitones.
This universality is one of the most remarkable features of the geometry, a
single n-in-o graph describing any n-note chord in any o-note scale.
There is, however, one important difference between these various spiral
diagrams: the closer the chord is to the augmented triad, the tighter the
connection between clockwise motion and descending stepwise voice
leading. Figure 2.6.1 distorts augmented triads to obtain the 025 trichords on
the spiral diagram. With a nearly even chord, like the major or minor triad,
we do not need to distort the augmented triad very much, and the result is
smooth descending voice leading (e.g., Figure 2.1.7). The 025 triad requires
a larger amount of distortion, and this makes for a larger amount of contrary
motion as we move clockwise along the spiral diagram. In the aggregate,
adding the motion in all voices, descent still dominates: each clockwise turn
produces a voice leading whose melodic intervals sum to –3. But as chords
become more and more uneven, the connection to stepwise melodic motion
is less and less clear (appendix 1).
Figure 2.6.1. The position of 025 chords on the spiral diagram.

Of course, we would really like to represent both major and minor chords
at the same time. One way to do this is to superimpose two versions of our
graph on top of each other, as in Figure 2.6.2.28 To use this new figure we
need to know that the direct clockwise path between parallel major and
minor triads is the “parallel” voice leading that alters the chord’s third by
semitone: thus moving clockwise from C major to C minor lowers E to E♭,

moving from A major to A minor lowers C ♯ to C, and so on.29 (Naturally

enough, the reverse paths raise the minor triad’s third.) This allows us to
calculate the voice leading corresponding to any other path on the graph: for
example, to figure out the voice leading from C major to E minor, we
combine the radial voice leading from C major to E major, (C, E, G) → (B,
E, G♯), with the “parallel” voice leading (B, E, G♯) → (B, E, G); the result,
(C, E, G) → (B, E, G), is the “leading-tone exchange” that lowers the root by
semitone.30 We can use this graph to represent chromatic triadic sequences
such as those in Gesualdo and Mozart (§1.1).
Figure 2.6.2. Superimposing major and minor spiral diagrams.

Another option is to use Figure 2.1.3 to represent major or minor triads,

with the single point “C” standing for both C major and C minor. This
allows us to use a simple graph to represent a family of closely related
progressions, each equivalent up to the exchange of parallel major and
minor. Figure 2.6.3 uses this approach to model the upper voices in the
passacaglia from Shostakovich’s E minor trio. The phrase begins with
ascending fifths and stepwise descending melodies: B ♭ , F, C, G, with an
upper voice C diminished chord between C and G; the next chord should be
D major, but we instead get the augmented chord F ♯ –B ♭ –D, a single
semitone away from each of the triads at the bottom of the circle (here
forming the top of a G minor-major seventh chord); we then proceed to the
expected A, with its voices exactly where they would be had the sequence
continued. The top voice of the final chord, B diminished, leads naturally to
the tenor B ♭ across the repeat; its doubled F serves the same musical
function as the G4 in Neil Young’s “Helpless,” melodic overlap creating the
illusion of endless descent. This resonance is neither coincidence nor a
matter of influence, but rather the independent rediscovery of similar
solutions to similar problems.

Figure 2.6.3. The passacaglia from Shostakovich’s E minor piano trio. The numbers above the figure
show how to interpret these motions on Figure 2.1.3, using a single point to represent major, minor, or
diminished triads with that root. For augmented triads, the three radial points collapse into a single
chord.

In the rest of the chapter, I will mostly follow the strategy of grouping
major and minor together, sacrificing detail in the name of graphical
simplicity; readers who prefer more precision can use Figure 2.6.2 instead.
Chapter 3 will introduce the diatonic analogues of these structures, in which
major and minor triads appear as transpositionally related species of the
same genus, the diatonic triad. This provides yet a third strategy for
simultaneously modeling major and minor triads.
 7. A fourth family

My last loop family involves five chords again arranged in an ascending-

fifth pattern so as to be maximally close diatonically. Its best-known
representative is “Hey Joe,” made famous by Jimi Hendrix.31 When I first
learned the song, I was puzzled by its combination of dark lyrics and
“bright” ascending-fifth major chords. (Why I thought of them as ascending
fifths rather than descending fourths is a bit mysterious, but I did.) It was not
until decades later, when I focused on its voice leading, that I understood the
musical logic: closely related triads and stepwise descending melody, an
endlessly sinking lament (Figure 2.7.1). Here, however, C major and E
major occupy the same registral position, producing an ascending semitone
B→C across the repeat. The loop can be found in any number of classic-rock
songs including the Beatles’ “A Day in the Life” and “Here Comes the Sun,”
Al Green’s “Take Me to the River,” and “Time Warp” from The Rocky
Horror Picture Show. It is notable that the progression can be played with
the five major triads that do not require barring in standard guitar tuning (C,
G, D, A, E).

Figure 2.7.1. “Hey Joe.” Despite the ascending semitone B–C, we can hear a canonic descent E–D–
C♯–B–(C), with the melodic E descending to the inner-voice C across the repeat.
 Figure 2.7.2 shows some of the many songs that can be derived from this
pattern by eliminating chords and replacing major with minor. These are
some of the most important progressions in all of rock, including i–VI–VII–i
and a–F–C–G, which can appear either as i–VI–III–VII in A minor or I–V–
vi–IV in C major.32 Since the outer chords in the five-chord sequence
occupy the same angular position, they are linked by symmetrical voice
leading in which one note ascends and one note descends; from the
standpoint of the spiral diagram, neither belongs “before” or “after” the
other. We can therefore exchange them to obtain E–G–D–A–C. Neil Young’s
“Keep on Rockin’ in the Free World” uses the progression’s three-chord
subset e–D–C in its verse and the four-chord e–G–D–C in the chorus, the
minor tonic creating a completely diatonic progression while allowing for
stepwise descent from VI to i (Figure 2.7.3). Another four-chord subset, E–
G–A–C, appears in Dave van Ronk’s version of “House of the Rising Sun,”
to be discussed in §2.10. Boyce and Hart’s “Steppin’ Stone,” recorded by
Paul Revere and the Raiders, uses van Ronk’s progression in the context of
an explicit and continuous melodic descent: the backing melody 8̂– ♭ 7̂–6̂–
♭6̂ connects smoothly to the main vocal 5̂–4̂–♭3̂–1̂, forming a descending
octave (Figure 2.7.4).33 Interestingly, the melodic composite articulates the
canon implicit in the version of the progression that has a minor tonic; it is
not inconceivable that the writers heard this line in the progression, writing
the melody in the way a listener might track the Shepard-tone passacaglia in
the final product. From this perspective, the melody’s G–e descent, rather
than a departure from a stepwise norm, is faithful to the progression’s inner
logic.
Figure 2.7.2. Progressions in the “Hey Joe” family. “Heart-Shaped Box” is starred because its
progression usually features ascending voice leading.

Figure 2.7.3. Neil Young’s “Keep on Rockin’ in the Free World” uses subsequences of the progression
e–G–D–A–C.
Figure 2.7.4. The vocal parts in “Steppin’ Stone” articulate a continuous stepwise descent, while the
harmonies present a four-chord subsequence of e–G–D–A–C. The vocal pitches are implied by the
triads’ descending voice leading.

It is extraordinary how many iconic rock progressions can be derived

from ascending fifths—perhaps with some chords eliminated, and perhaps
with the phrase structure misaligned with the fifths-sequence. That this is so
attests not just to the inner consistency of rock harmony, but also to the way
ascending fifths mediate between the demands of harmony and counterpoint,
generating descending melodies while keeping triads diatonically close. My
own musical education deemphasized ascending fifths in favor of the
descending fifths of functional music, with the latter treated as “normal” and
the former exceptional. One of my goals in this book is to correct this
misapprehension, restoring the ascending fifth to its central place in both
modal and functionally tonal theory.
At this point it is useful to pause for a moment of metatheoretical
reflection. Grant for the sake of argument that rock music exhibits a
retrofunctional regularity, with descending melodic steps generating
harmonic patterns in the way I have described. According to schema
theorists like Meyer, Gjerdingen, and Byros we should be able to use these
objective musical patterns to draw conclusions about the listening public’s
subjective experience. But I think this is misguided; after all, we are the
historical audience for this music, and we can simply ask whether we have
always been explicitly aware of these structures. In my case, the answer is a
resounding “no!” While I may have had some vague and intuitive sense of
“what rock harmony sounds like,” I have been genuinely surprised by the
degree of structure to be found in music I have spent my life playing,
writing, and listening to. Here we have clear regularities in a genre,
explained in part by the structure of musical space, but not explicitly present
in the mind of at least one historically situated listener. If objective musical
structure were a reliable guide to listener psychology, then this sort of
surprise should be impossible. This suggests we need a more nuanced
account of the relation between observable musical patterns and the
psychology of composers and listeners.

8. Other modalities

Figure 2.8.1 shows the spooky prayer near the end of Heinrich Schütz’s
setting of Psalm 84 (“Wie lieblich sind deine Wohnungen”), the eighth of his
Psalmen Davids (1619, SWV 29). It is an intensely powerful rendering of
the Psalm’s most direct address to God, eleven consecutive root-position
triads of uncertain tonal provenance—the unmetered homophonic chant a
sharp contrast with the preceding polyphony. When I first encountered this
music, I heard resonances with rock that led me to gently question my sanity.
Why should Schütz’s early seventeenth-century composition remind me of
Jimi Hendrix or Aretha Franklin? It took some time before I was ready to
accept these resonances as something more than accidental; and it took
longer still before I was confident enough to imagine that they might reflect
deep musical values—a shared interest in closely related harmonies and
stepwise descending melody. I now think it is not so surprising that we
should find resemblances here: rock musicians and early seventeenth-
century composers both operated in a context where functional-harmonic
constraints were comparatively weak, available options rather than
inviolable laws. In such environments, musicians are free to explore the
fundamental geometry of triadic voice leading, and in particular the
connection between descending voice leading and retrofunctional harmony.
Figure 2.8.1. “Wie lieblich sind deine Wohnungen,” mm. 136–146.

In other words, this music represents a largely unrecognized species of

quasi-chromaticism. We are accustomed to using “chromaticism” in
connection with the nineteenth century, where efficient voice leading
provides shortcuts between remote harmonic locations; Schubert moving
directly from C major to A ♭ major, exploiting efficient voice leading to
modulate between distant keys.34 I am arguing, in effect, that similar
strategies can be found in repertoire that is more resolutely diatonic. My
claim has two components. First, this diatonic music, like nineteenth-century
chromaticism, allows contrapuntal forces to operate relatively free of
harmonic constraints: in the nineteenth century this led to surprising
chromatic shifts; while in modal music it leads to “nonfunctional”
progressions that may or may not lie within a single diatonic scale. Second,
insofar as this music involves a diatonic norm, it is weak, controlling the
departures from the prevailing key and limiting chromatic excursions. The
progressions we have been exploring constitute a species of heavily diatonic
chromaticism in which the diatonic scale exerts a greater gravitational force
than it does in the nineteenth century. Yet the underlying principles are
similar.
What is true of this early music is equally true of later modality. Figure
2.8.2 shows the Intermezzo of Grieg’s “Watchman’s Song” (Op. 12, no. 3,
written in about 1866).35 The spooky i–VII–♭II–IV–V progression represents
the Spirit of the Night, nearly parallel four-bar phrases outlining an 8̂–7̂–6̂–5̂
melodic descent. In classical music this line would likely be accompanied by
a familiar functional progression; in Grieg it breaks free, the unusual D–F–
a–B harmonies more obviously related to Schütz and rock than to Corelli or
Haydn.

Figure 2.8.2. The Intermezzo of Grieg’s Op. 12, no. 3.

The counterintuitive association between early and late modality is one of

the main motivations for this book. My idea is that there is a triadic “state of
nature” that is most directly accessible when harmonic function is weak, and
which is correspondingly harder to see in more harmonically regimented
styles. This proto-grammar arises not from any human convention but from
the underlying geometry of three-note chords—a geometry which is
intrinsically neutral with respect to harmonic progression, acquiring
directionality from external factors such as the preference for descending
melody. Because this geometry is deep, rooted in constraints both
mathematical and psychological, it produces resonances across very
different musical styles—Shostakovich and Neil Young, the Beatles and
Grieg. My goal is to take these resonances seriously, allowing them to guide
us toward a better understanding of music outside the classical tradition.
As a general rule, these sorts of retrofunctional progressions are
uncommon in baroque and classical styles. Nevertheless, we can find traces
of this nonfunctional default woven throughout that music—in the
ascending-fifth sequences that so often feature in developments (§8.5), in the
retrofunctional progressions that make up the “fauxbourdon rule of the
octave” (§7.5), in familiar yet harmonically anomalous idioms such as the
Waldstein sequence (§3.7), and in many other undertheorized corners of the
tradition. Retrofunctional techniques thread like a hidden cantus firmus
throughout functionality, an unruly strip of wilderness surrounded by the
well-charted territory of rule-governed harmony.
Occasionally, this musical Other rears its head. Figure 2.8.3 shows Mozart
using stepwise descending voice leading to answer IV–I in the tonic with
IV–I in the subdominant, a striking departure from harmonic decorum.36
Figure 2.8.4 shows the retrofunctional ending of one of J. S. Bach’s
phrygian-melody chorales, the music modulating from D to b along an
ascending-fifth sequence; once again, we have the impression of a temporary
suspension of functional harmony, the music briefly reverting to earlier
norms.37 These moments underscore the importance of viewing functional
harmony against the background of modal practice, a more general default
that can help explain and contextualize these exceptional moments. Such
“retrofunctional” progressions are utterly clear in their contrapuntal logic,
and not intrinsically strange. What is unusual is just to find them here—tiny
moments where harmonic convention fails to grip, allowing us to glimpse
the world beyond the borders of textbook harmony.
Figure 2.8.3. Retrofunctional motion in Mozart’s piano sonata K.333, III (mm. 40ff).

Figure 2.8.4. A retrofunctional sequence of ascending fifths at the end of Bach’s “Es woll uns Gott
genädig sein” (BWV 311, Riemenschneider 16).
 9. Function and retrofunction

One of the most interesting features of rock is the coexistence of modal and
functional harmony. Songs like “Sympathy for the Devil” and “I Can’t
Explain” juxtapose mixolydian I– ♭ VII–IV–I verses with functional
dominants in their choruses. The Beatles’ “With a Little Help from My
Friends” does something similar, beginning with ascending fifths supporting
a two-voice compound melody (Figure 2.9.1).38 The retrofunctional fifths
then reverse direction to produce a functional ii–V7–I progression, now
supporting 2̂–3̂ in the melody. In the figure, I interpret this 2̂–3̂ as moving
between voices, with the arrival on 1̂ implied but not explicitly stated. (The
tonic scale degree arrives in measure 3 but without harmonic support, while
the accompaniment clearly states the 4̂–3̂ descent.) The chorus harmonizes a
strong 2̂–1̂ with an archetypal rock-mixolydian progression, ♭ VII–IV–I.
This ascending-fifth pattern then shifts up by fifth, from ♭VII–IV–I to IV–I–
V, arriving on a functional half-cadence with the guitar part recapitulating
the phrase’s 5̂–4̂–3̂–2̂ descent. The verse’s three-chord units thus alternate
ABAB between function and retrofunction. In the second verse, this ABAB
alternation is highlighted by the vocals, with Ringo Starr taking the modal
unit (I–V–ii) and the background vocals singing the functional unit (ii–V7–
I). This association continues through the bridge, where the background
singers ask a functional question (vi–V/V) which the lead vocal answers
retrofunctionally (I–♭VII–IV). The contrast between harmonic styles is thus
reinforced by the orchestration.
Figure 2.9.1. The Beatles’ “With a Little Help from My Friends” alternates functional and
retrofunctional motion.

“In rock music the harmonies tend to go backwards, except when they
don’t.” This sounds like a parody of untestable theorizing . . . yet it is
entirely possible that rock music does in fact juxtapose distinct harmonic
subsystems, one functional and one modal, each with its own characteristic
repertoire of harmonic moves.39 Insofar as the systems are kept separate,
occurring for instance in different sections of a song, then it is not
unreasonable to declare the music to be harmonically polyvalent. Such
juxtapositions have been dubbed “code switching” by Richard Cohn,
borrowing a term from linguistics to describe nineteenth-century
chromaticism.40 Just as bilingual speakers can move smoothly between two
different languages, so too do popular musicians shift between the default
voice leading of the modal system and more conventionalized functional
routines. To be sure, the line between languages can vary from individual to
individual: to my ears, tunes like “With a Little Help from My Friends”
really do move back and forth between distinct harmonic styles, but for
others the song may be unified creole, a coherent fusion of elements that are
no longer distinct.
I hear a similar sort of code switching in a piece such as Thomas Morley’s
“April Is in My Mistress’ Face” (1596), whose 38 measures are divided into
four short sections, each setting a single line of text. The opening is strongly
functional, with paired voices articulating T–(S)–D–T progressions that
would be unobjectionable in the high baroque (Figure 2.9.2).41 The second
phrase shifts to B ♭ , its ascending-fourth progressions including a series of
ascending leading tones. The third phrase is homophonic and declamatory,
starting with primary triads in B ♭ major and shifting to G minor as the
thought of fall is introduced. Here it is possible to perceive the major/minor
opposition as carrying some of the “happy/sad” valence it acquires later.42
Figure 2.9.2. Thomas Morley’s “April Is in My Mistress’ Face” begins functionally, but ends
retrofunctionally.

After so much functionality, the last 16 measures make a striking shift

away from functionality. The music uses a common Renaissance idiom, the
diatonic version of the descending 6–5 sequence, here decorated with a voice
exchange that creates an outer-voice canon—one that veils but does not
obliterate the descending voice leading (Figure 2.9.3). I hear a sharp musical
contrast, the ending dark and modal where the opening was bright and
functional. The root-position chords throw the nonfunctional progressions
into relief: ascending thirds and descending fourths that counter functional
expectations.

Figure 2.9.3. The end of Morley’s piece is based on a diatonic version of the descending 6–5
decoration, embellished with a voice exchange. The voice exchange highlights the canon in the outer
voices.

Compare this music to Grieg’s “In My Homeland,” the nineteenth of the

Lyric Pieces (Op. 43, no. 3, written in about 1886 and shown in Figures
2.9.4–5). Once again, the functional outlines are generally clear even when
the local progressions avoid strong dominants. The twelve-measure phrase
begins with a version of Morley’s descending 6–5, outer voices hinting at the
very same canonic structure (though substituting I6 for the expected
mediant); this proceeds to a strongly functional I–vi–V/V–V cadence. The
dominant then moves nonfunctionally across the phrase boundary to IV,
beginning an “up by step, down by third” sequence in sevenths.43 That
phrase ends with another stereotypical cadence, leading to a proto-
Debussyian alternation of ii and V. Grieg thus uses two of the three
Renaissance sequences that typically harmonize stepwise descending thirds
(cf. Figure 1.5.5), reviving progressions that had been neglected for
decades.44 His melodic logic is so clear that one may not notice the absence
of dominant-tonic progressions.45

Figure 2.9.4. The twelve-bar core of Grieg’s Op. 43, no. 3.

Figure 2.9.5. A reduction of Grieg’s Op. 43, no. 3.

These cross-stylistic resonances suggest a new starting point for twenty-

first-century pedagogy. Harmony textbooks often begin and end with
functionality, as if the TSDT archetype were the norm and anything else a
departure. But the eighteenth century’s rigid harmonic laws are more
exception than rule: in seventeenth-century music, as in contemporary rock,
we find alternative models in which functionality is just one option among
many—music in which the deep logic of voice leading sometimes takes the
lead. It would make sense to begin our curricula with these harmonically
flexible styles, which often feel more relevant than the more rule-governed
language of chorale and sonata. One suspects that our reluctance to do so is
not a matter of considered deliberation, but of the lack of appropriate
conceptual tools: we know how to teach the principles of functional
harmony but not the deep contrapuntal logic that animates other music.46
The theory of voice leading can help fill this gap, giving us a more rigorous
understanding of contrapuntal possibility—and in the process dampening the
pedagogical echoes of classical-music chauvinism.

10. Continuity or reinvention?

The connection between early music and rock is in one sense obvious: bands
had names like “Renaissance,” wrote lyrics about knights and royalty, and
played folksongs like “Scarborough Fair.” Amidst political turmoil, an
idealized past provided an alternative to an unsatisfactory present, with
modal harmony its sonic emblem. This was clear to conservative musicians,
who largely rejected modality in favor of functional harmony—to the point
where one can have a pretty good sense, even before listening, of the
harmonic vocabulary of a song like “The Ballad of the Green Berets” or
“Stand by Your Man.” What results is a strange historical inversion whereby
an older modal practice was associated with progressive politics while newer
functionality symbolized traditional values.
When I was young, I assumed the triadic modality of rock was linked by a
continuous tradition to the modality of the Renaissance. Surely, I imagined,
people never stopped strumming dorian and mixolydian tunes in informal
venues such as pubs and homes, uncorrupted by the learned tradition. What
happened in the 1960s, I imagined, was simply that this marginalized and
ubiquitous musical tradition was raised to public consciousness. But having
looked for evidence of this continuity, I find myself much less convinced
than I used to be. American and English folk recordings made before 1950
tend to be harmonically spare, featuring either unaccompanied melodies,
drones, or perhaps oscillating between two chords. It is surprisingly difficult
to find recordings that are (a) clearly modal, (b) made before 1950, and (c)
feature three or more distinct triads.47 Insofar as these recordings are
representative, folk modality tends toward monody.
Furthermore, the picture of folksongs as rigid vessels for modal practice is
at odds with the extraordinary flexibility of the vernacular tradition:
surveying early twentieth-century recordings, one finds the same tunes
appearing in different forms—changing modes and time signatures, and
acquiring a variety of harmonizations. Consider the earliest recording of
“The House of the Rising Sun,” made for Vocalion in 1933 by the
Appalachian musicians Clarence (“Tom”) Ashley and Gwen Foster, under
the name “Rising Sun Blues.” The piece, transcribed in Figure 2.10.1, is a
sixteen-bar major-mode blues with an ABBC structure, a standard twelve-
bar blues that repeats its middle four bars. As such it is largely functional,
oscillating between I and IV before reaching the V–I cadence. Its most
unusual chord—a iv2 that occurs under a melodic ♭ 7̂ not present in the
harmony—can perhaps be seen as a minor subdominant (an instance of
“modal mixture”) though it also has an affinity with the VI that appears in
later versions of the tune. (In some verses of the Ashley/Foster recording,
this iv2 moves to V43 on beat three.) A 1935 variant by Homer Callahan
(under the name “Rounder’s Luck”) has simple i–V–i harmony, while other
versions are monophonic (Georgia Turner, 1937), or blues-based (Roy Acuff
and His Smoky Mountain Boys, 1938). These performances are considerably
more functional than most of the songs in this chapter, and indeed a later
recording by Ashley and Doc Watson sands down the rough edges of the first
version, softening the iv2 and hewing more closely to blues convention.

Figure 2.10.1. Ashley and Foster’s 1933 recording of “The House of the Rising Sun.”

The route to the rock harmonization of “Rising Sun” is instructively

circuitous. Bob Dylan’s 1962 acoustic recording borrows its chords from
Dave van Ronk; it features an a–C–D–F progression over an 8̂–7̂– ♯ 6̂–6̂
descending bass that sounds for all the world like a tonally functional
lament, albeit interrupted by a root-position tonic where a classical composer
would use i64 (Figure 2.10.2).48 Here there is no dramatic rejection of
functional norms: the subdominant, so central to the blues, is subsumed
within the descending bass, the reharmonization producing an ABAC
harmonic structure at odds with the vocal line’s ABBC. If anything, Dylan’s
version feels more conventional than Ashley’s, the blue notes of the original
melody now supported triadically.49 By contrast, the Animals’ 1964
recording speaks the language of rock: Dylan’s stepwise descending lament
is moved to an upper voice and its harmonies placed in root position—an
aggressive progression that feels much less functional than its predecessor.
Far from being the relic of an ancient tradition, these retrofunctional triads
are a recreation, a reimagined past as artificial as Debussy’s.50

Figure 2.10.2. Bob Dylan’s 1962 harmonization of “The House of the Rising Sun.” The melody is an
idealization that does not try to capture the nuances of Dylan’s performance.

A similar story can be told about “Scarborough Fair,” initially recorded in

1955 by Gordon Heath and Lee Payant (Figure 2.10.3). Their major-mode
version features simple I–V–I harmonies, the jaunty vocal declamation
evoking a comedic Renaissance figure. Like much vernacular music, it is in
a “loose duple” that permits the occasional extra beat; my transcription
locates this beat in the third measure of each phrase, resetting the downbeat
to the final syllable and allowing for a comfortable pause between lines.51
This version is followed in 1957 by Ewan MacColl and Peggy Seeger’s
harmonized dorian version, its melody collected ten years earlier.52 Audrey
Coppard made a monophonic dorian-mode recording in 1958, and Shirley
Collins released a very similar take in 1959. Meanwhile Peggy Seeger
released a functional major-mode recording in 1960, while Marianne
Faithfull provided a functional minor-mode harmonization in 1965. The tune
appears with its now-canonical dorian chords on Martin Carthy’s 1965 debut
album; Carthy taught it to Simon and Garfunkel, who used it in their 1966
medley—a top-ten hit that eventually became standard.

Figure 2.10.3. Gordon Heath and Lee Payant’s 1955 version of “Scarborough Fair.”

Such examples suggest that, in the absence of recordings, folk songs are
an unlikely vehicle for the faithful transmission of modal harmony. Instead
of fixed and unchanging songs we find endless interpretations, united chiefly
by lyrics and secondarily by melody. As far as I can tell, early twentieth-
century American and English folk music is if anything more functional than
its descendants: triadic modality, with its blatantly nonfunctional
progressions, is largely a phenomenon of the late 1950s and 1960s, not
obviously continuing any widespread folk practice.53 In this sense, the
modality of the 1960s is best understood as the delayed echo of modernism,
popular musicians rejecting functionality a half-century after classical
composers did.
Thus if there are resonances between Renaissance and rock, this is likely
due to intrinsic triadic logic rather than actual historical connection. What
sustains the modal “state of nature,” what leads to its reappearance in
disconnected eras, is that it offers straightforward solutions to basic musical
problems—stepwise descending melodies and closely related triads. In other
words, it is a “syntax” supported not by convention but by constraints
inherent in our musical materials. Musicians need not be aware of these
constraints to be bound by them; they simply need to be interested in writing
music that feels right (with their intuitions prioritizing stepwise melody,
diatonically close harmonies, and so forth). By analyzing this feeling of
rightness we can transform the implicit into the explicit, revealing hidden
roads connecting different musical languages. If we are lucky, this new
understanding will suggest new compositional possibilities as well.

1
This asymmetry can also be found in birdsong and human speech, suggesting a potential origin in
respiration, the sharp intake and gradual release of breath required for both speaking and singing. See
Tierney, Russo, and Patel 2008 (music and speech), Tierney, Russo, and Patel 2011 (the “motor
constraint hypothesis”), and Savage, Tierney, and Patel 2017 (birdsong and music).
2
For the terms “strong” and “weak,” see Meeùs 2000. Stephenson 2002 makes a similar
observation about the prevalence of weak progressions in rock, though he explains the phenomenon
differently.
3
In general, the initial direction is important. If you want a short clockwise path from C to F, you
need to start by sliding clockwise; the algorithm in this paragraph will produce the path T5t–1. If you
were to instead slide counterclockwise, the algorithm will calculate the shortest counterclockwise path
from C to F, T–7t1; this is a different but equally important voice leading. For a purely radial path the
two directions produce the same result.
4
The “LP” voice leading moves a major triad’s root down by semitone and fifth up by semitone;
“PL” moves a major triad’s third down by semitone and fifth up by semitone. I call this “the major-
third system” (Tymoczko 2011a). For a comparison between voice-leading geometry and neo-
Riemannian theory see appendix C of Tymoczko 2011a and Tymoczko 2020b.
5
As appendix 1 explains, this notation means “C moves to B,” “E moves to E,” and “G moves to
G♯.”
6
Camp 2018.
7
Yust 2015a explores distorted and quantized sequences.
8
See Tymoczko 2011a.
9
To be sure, there are some styles, like heavy metal, that emphasize these diatonically distant
progressions precisely for their ominous and unusual quality; we will ignore this “contratonal” music
for the time being. For “contratonality” see Huron 2007a, p. 339.
10
Of course, many rock musicians have absorbed this norm through musical training as well.
11
De Clerq et al. 2011. See also Figure 3.4 in Temperley 2018. Pinter 2019 surveys the early
history of ♭VII in popular music.
12
Note that there is not a single tritone progression on the list—and that there is a huge gap
between the diatonic intervals of a fifth and major second, and the rest. The relatively greater
popularity of fifths over major seconds, which both have diatonic distance 0, is likely explained by the
longstanding norm of fifth motion. A secondary factor may be the fact that fifth-related triads are
diatonically “superclose” since each pair belongs to two different diatonic collections, whereas major-
second-related pairs belong to just a single collection.
13
Stephenson 2002 notes that the chord progressions in rock music are the opposite of those in
classical music, and suggests that this is the explanation for their popularity: rock harmony, in his
view, is fundamentally oppositional, a matter of doing what was not done in earlier styles. My
explanation instead proposes that rock musicians are making affirmative use of the possibilities
afforded by deep musical relationships; if anything, rock harmony is more intuitive than classical
harmony. Their oppositional quality is more byproduct than motivation—though unfamiliarity is likely
part of their attraction.
 14
See §3.9 for more on the origins of this progression. It represents a popular-music instance of
what Samarotto 2004 describes as the “sublimating” of ♯ 4̂.
15
“When they clean the street, I’ll be the only shit that’s left behind.”
16
Rowe 1983.
17
My argument is not that ♭III, V, and VII are harmonically similar, the way dominant chords like
V and vii° are; instead, they are contrapuntally similar in the sense that any of them can be inserted
between I and ♭ VII without disrupting the stepwise descending voice leading. For the claim that
major-third relations can play similar harmonic functions, see Cohn 1999; this view, as Cohn notes, is
cousin to Lendvai’s minor-third-based “axis system” (§3.8).
18
Temperley (2018, p. 35) surveys the many writers who have commented on the dual meanings of
this progression.
19
See Moore 2012, p. 85, on “open” and “closed” phrases.
20
In the Eagles’ “Hotel California,” the descending 6–5 moves from tonic to dominant in tonally
functional fashion. Temperley 2018, p. 57, notes the descending chromatic line. One can interpret this
as a variant of Gesualdo’s trick where a single voice moves chromatically while the other two alternate
between root position and first inversion.
21
The composer Paul Lansky has spoken about composing so as to allow the listener to chart their
own way through a piece (Perry and Lansky 1996).
22
Feurzeig 2010 cites examples from Miles Davis (“Blue in Green”) to Philip Glass (the
“Spaceship” harmonies from Einstein on the Beach) and Ligeti (the fourth movements of the Horn
Trio and Violin Concerto), including the Shostakovich passage analyzed in §2.6.
23
There are various other ways to characterize this class of progressions, for instance as those
involving a diatonic distance less than 6, or those whose roots lie in a pentatonic scale (Biamonte
2010).
24
See Temperley 2011b. In rock, IV chords are almost as likely to go to I as V chords are. The Alt-
J song “(Interlude 3)” from An Awesome Wave presents a four-chord ascending-fifth sequence F–C–
G–D, but I hear G and not F as the tonic despite the hypermetric accent on F.
25
This is a picture that can be perceived either as a duck or a rabbit, but not both at the same time.
26
One can make a similar point about Boston’s “Peace of Mind,” which may seem to shift into C♯
minor at the start of the chorus despite its generally E major tonality. See Harrison 2016, pp. 70–72
(on “Cherry Cherry”), Doll 2017, and Nobile 2020.
27
In this last example, I use pitches to show how voices move, as discussed in appendix 1.
28
This graph is similar to familiar music-theoretical constructions such as the Tonnetz and
Douthett and Steinbach’s “Cube Dance” (1998). Here we reconstruct these graphs by superimposing
spiral diagrams (appendix 1).
29
To determine this, we can ask what spacing-preserving voice leading from C major to C minor
moves its voices by intervals collectively summing to –1; see Tymoczko 2020b.
30
As before, different paths between the same points represent different voice leadings: the
counterclockwise path from C major to E minor, for example, corresponds to the voice leading (C, E,
G) → (E, G, B).
31
The song is credited to Billy Roberts but likely derives from Niela Miller’s 1955 “Baby, Please
Don’t Go to Town.”
32
These and many other progressions are discussed in Doll 2017, an extensive survey of rock
harmony. Richards 2017 focuses on the “axis progression” a–F–C–G.
 33
Minor Threat’s version expresses the continuity between these two vocal lines, the single
vocalist alternating between them.
34
See Cohn 1996, 1999, 2012, Tymoczko 2011a. Agmon’s “within-key chromaticism” (2020) is
very close to my own.
35
For an insightful treatment of Grieg’s late style, see Taylor 2019.
36
The chord sequence C–G–F–C, with descending voice leading, evokes the “fauxbourdon rule-of-
the-octave” idiom, I–V6–IV6–I64, though here it occurs over a different bass and in the wrong key
(§7.5).
37
Bach frequently modulates from major to relative minor by a sequence of ascending fifths
(§7.7). The bass here instantiates something very close to Byros’s “le–sol–fi–sol” schema (2012) but
on the wrong scale degrees.
38
Steve Forbert’s “Goin’ Down to Laurel,” from Alive on Arrival, uses the same progression.
39
Many other theorists have commented on rock harmony’s polystylistic qualities, including
Everett 2004, Osborn 2017, and Temperley 2018 (pp. 54ff).
40
Cohn 2012.
41
The F♮ in m. 9 provides a hint of modality.
42
See Meier 1988, Part II, chapter 7.
43
The voice leading recalls that of mm. 10–11 in Wagner’s Tristan, which is (C, F, A♭, D) → (B,

F♯, A, D♯) when we remove crossings. Here, the A♭→A voice becomes G♯→B.
44
The parallel thirds appear in the first sequence’s alto and tenor (disrupted by the substitution of
6
I for the iii) and as sixths in the second sequence’s soprano and tenor.
45
Note the ubiquitous parallel tenths between outer voices.
46
Ewell (2020) has emphasized that this focus on functional harmony is also a reflection of elitist,
ethnocentric, and racist forces.
47
There are a number of continuous modal traditions that are unlikely to have served as a historical
bridge between Renaissance and rock. Gregorian chant is one. Irish folk music is another, with modal
elements present from the transcriptions of Edward Bunting (1773–1843) to the recordings of the early
twentieth century.
48
Rings (2013) notes a similar progression in the work of Len Chandler, an oboist in the Akron
Orchestra, who would likely have been familiar with the learned modality of the early twentieth
century. See Grieg’s Elegie (§8.7) for the sort of descending bass line that might be the ancestor of
Dylan’s lament.
49
Wagner 2003.
50
Later uses of the progression, such as “Stairway to Heaven,” further detach it from its origins in
the lament: here, the i–III64–IV6–VImaj7 returns to i, decorating the idiomatic modal-rock VI–VII–i
rather than the functional-harmonic i–V–i.
51
Compare measure 3 of Figure 7.2.6.
52
Kloss 2012 presents a thorough archaeology of the song, showing that both major and minor
variants have an extensive history. The earliest published transcriptions resembling the standard
modern version appear in the very late 1870s, though in aeolian rather than dorian.
53
We see a similar progression when we consider other folk traditions, such as shape-note
polyphony: the clear modality of its eighteenth-century tunes is gradually regularized and made more
functional in the nineteenth century, returning only as twentieth-century practitioners become more
interested in historical authenticity. Thanks here to Rachel Hall.
 Prelude
The Tinctoris Transform

Students are taught to understand counterpoint as the braiding of

independent lines, with melodies the primary objects of attention and
harmonies the byproducts of their interaction. This chapter will introduce a
complementary perspective that is central to a good deal of musical thinking
both practical and speculative, one that emphasizes vertical shapes or
configurations having different locations or centers of gravity—so that
harmonies are salient and melodies more like byproducts. Where the first
perspective deals in descriptions like “one voice is at A4 and the other at
D4,” the second instead uses phrases like “one voice is a fifth above the
other, forming a vertical interval centered on F4.” This has the result of
highlighting the contrast between parallel motion, which preserves the
vertical configuration, and contrary motion, which changes it. The two
perspectives are mathematically equivalent, and one can switch between
them without loss of information; indeed, I suspect many musicians
regularly do so without realizing it.
We can think of these as different musical coordinate systems. In both
cases, we begin by numbering pitches in some enclosing scale, using
chromatic or diatonic or other pitch-labels as appropriate. We then choose
some arbitrary order for our musical voices. In the familiar system, pitch
numbers are used to label each voice’s position, yielding a list containing the
notes in the first voice, second voice, and so on. In the second system, we
instead label voices by their distance from one “reference” voice (usually the
first), whose coordinate is necessarily 0; this gives us a configuration, or
collection of relative distances that can be instantiated anywhere in pitch
space. To specify the configuration’s location, we need another number
recording the average of the pitches; I will call this its center of gravity.1 The
difference between the two systems is illustrated by Figure P3.1. The second
system is closely related to what physicists call the center-of-mass
coordinate system.2
Figure P3.1. Two musical coordinate systems.

Though it may seem newfangled and mathematical, the configuration-

based approach is in fact very old. Johannes Tinctoris used it to analyze two-
voice counterpoint in his 1477 Liber de arte contrapuncti, arranging part of
the book so that chapters correspond to starting intervals and subsections
destination intervals, with each section describing how the combination can
appear.3 The result is a tedious verbal description of what is essentially a
matrix or spreadsheet (Figure P3.2). Contemporary musicians will be
tempted to dismiss Tinctoris’s approach as unbearably pedantic, a residuum
of the scholastic obsession with enumeration and lists. (C. S. Lewis once
wrote that, of all the inventions of modernity, the medieval mind would be
most impressed by the card catalogue.4) But there is purpose to Tinctoris’s
pedantry, and something to be learned from his approach.
Figure P3.2. Some of Tinctoris’s recommendations in matrix form. The row labels record the starting
configuration, measured in diatonic steps from tenor to counterpoint voice; the column labels record
the destination configuration measured in the same way. Numbers in the cells refer to the motion of
the tenor in diatonic scale steps. Thus, if the harmonizing voice is a third above the tenor in the first
chord (row label 2), moving to a third below the tenor in the second chord (column label –2), then the
tenor can ascend by 0–4 scale steps (i.e., up to a fifth). Similarly, if the harmonizing voice starts a third
below (row label –2) and is moving to the unison (column label 0), the tenor can either stay fixed ( 0)
or descend by step (–1).

One advantage of the configurational perspective is that it clarifies the

phenomenon of efficient voice leading. This is because the most efficient
voice leading between two vertical configurations will necessarily minimize
the change in center of gravity. This simple observation helps elucidate some
important patterns found throughout Western music: among the center-of-
gravity-preserving progressions in Figure P3.3 we see the major-third triadic
progressions beloved by Schubert, the minor-third tetrachordal progressions
of Wagner, Debussy, and Stravinsky, and a host of familiar fifth-
progressions; each combines a changing vertical shape with a relatively
stable center of gravity. The example shows that transpositionally related
dyads can be connected by efficient voice leading when their roots move by
tritone or near tritone (i.e., fourth or fifth), three-note chords when their roots
move by major third or near major third (i.e., minor third or perfect fourth),
four-note chords when their roots move by minor third, tritone, or intervals
nearly equal to these, and so on.5 These relationships underwrite many
important idioms from the Renaissance to the present day.

Figure P3.3. Efficient voice leading minimizes the change in a chord’s center of gravity. The example
presents dyadic, triadic, and seventh-chord progressions in chromatic space; above each I identify the
change in center of gravity, equivalent to the average of the melodic motions in all voices.

Figure P3.4. The diatonic consonances spanning an octave or less, arranged from small to large.
Composers often move along the arrows.
 They are also manifest in the structure of our geometrical models, where
the amount of angular motion corresponds to the change in center of gravity.
For example, on the 3-in-12 diagram (Figure 2.1.3), radial motion from C
major to E major lowers the root by semitone while raising the fifth by
semitone; since the ascending and descending semitones cancel, center of
gravity is preserved. In general, two chords will have the same angular
position when they can be linked by voice leading in which the change in
center of gravity due to looping (transposition along the chord) exactly
counteracts the change in center of gravity due to sliding (transposition
along the scale): in this case T4 raises the center of gravity by one third of an
octave, while t–1 lowers it by that same amount. Readers who look ahead to
Figures 4.2.7 and 3.8.4 will see that tritone-related perfect fifths occupy the
same radial position in 2-in-12 space, while minor-third-related seventh
chords occupy the same radial position in 4-in-12 space; this is the
geometrical origin of the idioms in the previous paragraph.
Efficient voice leading also requires that vertical configurations
themselves change by only a small amount. Figure P3.4 arranges the
diatonic consonances by size. Two-voice consonant counterpoint tends to
move by short distances on this graph: in the two-voice passages of
Palestrina’s masses, for example, about 90% of the fifths move to sixths or
thirds, while about 80% of the unisons move to thirds.6 (Note that this claim
implicitly uses the new coordinate system: to say “fifths often move to
sixths” is to say that voices configured as a fifth often move so that they end
up configured as a sixth.) By contrast, imperfect consonances often move to
other imperfect consonances by parallel or contrary motion. Putting these
points together, we can formulate a useful rule of thumb: if you are on a
perfect consonance, move along Figure P3.4 to a nearby imperfect
consonance (or, less likely, jump over an imperfect consonance to reach the
nearest perfect consonance); if you are on an imperfect consonance, move
about half the time along the dotted lines to another imperfect consonance,
and otherwise to a nearby perfect consonance. This positive counterpoint
gives musicians an affirmative goal, a set of plausible options going beyond
a negative approach built on prohibitions. It aims to capture the intuitive
knowledge of a composer who knows what to do next: “if I am located at
this point in musical space, then my most obvious destinations are . . .” Such
knowledge becomes increasingly important as additional voices bring more
contrapuntal options, or when composers are constrained by harmonic rules.
 A second advantage of the configurational perspective is that it helps us
focus on structural relationships that are independent of transposition; in this
respect it is the earliest forerunner of what we have come to call musical set
theory.7 Two progressions can be said to belong to the same category if they
exemplify the same series of vertical configurations, differing only in
transpositional level: “an octave followed by a major sixth,” “a major triad
in 53 position followed by a minor triad in 64 position,” “a root-position
half-diminished seventh in close position followed by a root-position
dominant seventh in close position,” and so on. In 1614, Thomas Campion
used this idea to categorize four-voice voice leadings between root-position
triads, extending the Tinctorian perspective from two voices to four (Figure
P3.5).8 In A Geometry of Music I said that such progressions are individually
transpositionally related; here I will say they are related by the Tinctoris
Transform.
Figure P3.5. Two of Thomas Campion’s contrapuntal categories, along with his notation of the shared
vertical configurations. The arrows and T labels are my own, showing that progressions within each
category are related by the independent transposition of each chord.

Paths on our spiral diagrams are related by the Tinctoris Transform if they
differ only in the amount of sliding involved; this is true whether we are
considering a simple diagram representing only the transpositions of a single
chord type or a more complex diagram superimposing multiple chord types.
By ignoring slides, we can group together a large number of superficially
different possibilities, revealing similarities that may not be intuitively
obvious. The 3-in-12 diagram of chromatic major triads, for example,
contains only three fundamental kinds of motion, the clockwise loops of 0,
1, or 2 turns (t0, t–1, t–2); all other motions can be obtained by adding slides
to these (Figure P3.6). This sort of simplification can be analytically useful:
for example, the three canonic sequences in Figure P3.7 repeatedly apply a
transformation of the form Txt1, differing only in the amount of sliding
involved. While they might initially seem unrelated, they are similar from
the configurational point of view.

Figure P3.6. Voice leadings from the central C major triad to triads on each line differ only in their
transpositional component.
Figure P3.7. Three canonic sequences that repeatedly apply the same voice-leading pattern, differing
only in their diatonic transposition.

Finally, configurations can help us understand the logic behind some

seemingly arbitrary prohibitions of traditional pedagogy. For example, many
textbooks prohibit similar motion into perfect consonances (“direct fifths
and octaves”), but Figure P3.8 shows that similar motion is rare regardless
of destination interval: the general scarcity of similar motion is at least as
striking as the difference between the perfect and imperfect consonances.
Our new perspective helps explain why this is so: the change in center of
gravity will be zero when two voices move in exact contrary motion; as the
change in center of gravity increases, the voice leading becomes more and
more unbalanced and the largest leap gets larger and larger (Figure P3.9).
This situation persists until we reach the boundary of oblique motion, where
one voice remains fixed and all the motion is in the other voice. As we move
into similar motion, the total motion in both voices starts to increase. It
follows that there is a basic incompatibility between similar motion and the
principle that voices should move by small distances: given any instance of
similar motion, we can reduce the melodic intervals in each voice by
reducing the change in center of gravity. Much the same point could be made
about “voice overlap,” a form of similar motion where one voice crosses
another’s starting point.9 Our new coordinate system allows us to understand
the don’ts of negative counterpoint in terms of affirmative dos—in this case,
minimizing melodic motion by keeping the center of gravity approximately
fixed.

Figure P3.8. Approaching perfect and imperfect consonances in Ockeghem, Josquin, and Palestrina.
Values are percentages, with each column adding up to 100 (within rounding).
Figure P3.9. (left) An abstract diagram relating a dyad’s change in center of gravity to its overall size.
When the change is zero, the voice leading involves perfectly contrary motion and the overall voice-
leading size is minimized; voices move in a balanced way by approximately the same distance. As we
increase the change, the melodic intervals become more unbalanced, until we reach oblique motion,
where all the movement is in one voice; beyond this, we have similar motion where the overall size of
the voice leading increases. (The line of obliqueness is half the change in interval size.) (right) A
concrete example using imperfect consonances. Underneath each progression, I count the total number
of melodic steps in all voices.

Hierarchical set theory supplements the Tinctorian perspective by

providing a new way to measure the intervals in a vertical configuration. We
can formalize the intuitive notion of a voicing as a pattern of spacing in
chordal steps: a “close position” voicing is a voicing in which each note is
one chordal step above its lower neighbor (e.g., C4–E4–G4) while an “open
position” voicing places each note two steps above its lower neighbor, as in
C4–G4–E5. We can think of voicings as set classes, collections of pitch
intervals measured along the intrinsic scale: for triads, close position is the
(pitch-set) set class 012, measuring along the intrinsic scale from bass to
soprano, open position is the (pitch-set) set class 024, and so on. Thus any
chord presents multiple set classes simultaneously: a chromatic set class
determined by its pitch classes, a diatonic set class determined by the
extrinsic scale, and an intrinsic pitch-set set class determined by its voicing
(Figure P3.10). In different musical situations different sets can be more or
less salient, with pitch-class content often more perceptible for small chords
in close position, and intrinsic spacing more important as chords grow larger
or are voiced in unusual ways.
Figure P3.10. Both Bill Evans’s “So What” chord and Schoenberg’s “Farben” chord are voiced as
open-position pentachords, spaced as two-step intervals along the intrinsic scale. The tables show how
the two five-note chords each present three set classes; “CHR” measures the size of their chromatic
intervals, “DIA” is the scalar size along the white-note or A melodic-minor scale, and “CHORD” is
the spacing along the chord’s intrinsic scale. Since the chordal spacing is identical, the chords are
voiced in the same way.

The Tinctorian tradition of configurational thinking continues into the

eighteenth century, reflected in figured-bass pedagogy’s terms for describing
the shapes played by the keyboardist’s right hand. It lives on in
contemporary guitar manuals that categorize voicings in terms of their
intrinsic spacing (“drop 2,” etc.) and jazz-theory books that enumerate the
open-position voicings of the pentatonic scale.10 Its most comprehensive
manifestation is set theory, whose even-more-general vocabulary of musical
shapes is sufficient to describe any conceivable collection of notes. Set
theory shares with its antecedents the factoring out of transposition, leaving
behind a description of relative relationships that can be instantiated at any
transpositional level: an interval, a figured-bass hand position, a voicing, a
“set class.” The connection between these traditions has been obscured by
three factors. First, traditional set theory is unusually indifferent to registral
position, disregarding vertical spacing entirely to focus on pitch-class
content alone.11 Second and related to this, the early set theorists were not
concerned with voice leading, a core topic of traditional theory. Third and
finally, set theory originally assumed chromatic space whereas traditional
theorists delt with the diatonic world. In recent years, however, theorists
have started to extend set-theoretical concepts to include both diatonicism
and voice leading, uncovering surprising connections between different
intellectual traditions. Chapter 3 will continue this line of inquiry.

1
Since the reference voice has position 0, the two systems both use n changing numbers to label
the notes of an n-note chord. We could also measure the configuration’s location using the sum of the
chord’s pitches instead of their average.
2
Transposition is a structure-preserving perceptual symmetry, much as “translation” (physical
movement) is a symmetry of 3D space. By factoring out transposition and translation, we can simplify
musical or physical problems.
3
Tinctoris measures an interval’s location relative to the tenor (his “reference” voice), rather than
the interval’s center of gravity; this difference is not fundamental, as the location of the tenor can be
computed from the center of gravity and vice versa.
4
Lewis 1964, p. 10.
 5
These root motions correspond to 12/n, where 12 is the size of the octave and n is the size of the
chord. As we move away from these optimal intervals, we increase the size of the resulting voice
leading.
6
Motion from unison to unison, fifth to fifth, or octave to octave will produce parallel perfect
intervals so long as the center of gravity changes.
7
See Nolan 2003 and Schuijer 2008 for histories of set theory that begin more recently.
8
This example comes from Schubert 2018, who hypothesizes that Thomas Tallis may have used it
in composing his 1570 Spem in Alium. I think we can find evidence for the strategy throughout the
sixteenth century, most obviously in those repertoires that make extensive use of close-position root-
position triads (§6.1).
9
Here the reasoning is more complicated. The prohibition on voice overlap dictates that large
melodic intervals be harmonized not with near-parallel motion but significant changes of spacing.
These changes of spacing increase voice independence and are often, but not always, more efficient
than the near-parallel alternatives.
10
For example, Laukens 2019 or Levine 1989.
11
This amounts to discarding some of the information contained in the relative configuration, for
instance by treating its numbers as unordered and conceiving them modulo the octave. Morris 1995
incorporates both pitch and pitch-class information.
 3
Line and Configuration

Musicians learn their way around abstract spaces of musical possibility,

developing an intuitive understanding of the contrapuntal paths between
sonorities of various kinds. This knowledge is typically schematic and style-
specific, the relevant pathways changing from one circumstance to another.
This chapter will use spiral diagrams to provide a more general perspective.
The central claim is that a wide range of musical practices subsist on an
underlying geometrical structure: across many styles, the desire for certain
kinds of melodies, such as those that move by step, combines with a desire
for consistent harmony to constrain compositional possibilities. These
constraints, which are as much mathematical as aesthetic, shape the syntax,
idioms, and schemas that characterize many different genres. Something
similar can be said for the idiosyncratic progressions found in composers
like Gesualdo or Mahler. Both rule and exception can be understood
geometrically.
On a technical level, I will analyze polyphony not just as the
superimposition of independent melodies but also as successions of vertical
configurations obeying their own higher-level laws. This perspective has
long been used to provide practical heuristics for creating idiomatic
counterpoint—as in §3.6, where it underwrites a simple system for teaching
four-part chorale harmonization. But it is also intimately connected to set
theory and geometry, leading to a convergence between modern theories of
voice leading, which arose in response to nineteenth-century chromaticism,
and traditional diatonic pedagogy, which in turn inspired contemporary
schema theory. Attending to vertical configurations can contribute to our
embodied musicianship while also deepening our appreciation for subtle
musical mathematics.
The plan is to gradually increase complexity, starting with two and three
voices before considering four-voice composition with both triads and
seventh chords. At a certain point, this purely quantitative increase of voices
leads to a qualitative change in perspective: for while it is possible to
understand two- and three-voice counterpoint using nonhierarchical models,
four-voice triadic counterpoint is more consistently hierarchical, with voices
often moving inside an abstract and scale-like harmony (§3.7). As we
progress, we will encounter several intertwining themes, including the
persistence of two-voice logic in three- and four-voice contexts, the
geometrical foundations of common idioms (e.g., the “Quiescenza” or the
common-tone diminished seventh), a collection of “wrong-way” resolutions
in which familiar objects move contrary to our expectations (the ascending-
fifth resolutions of the diatonic third, the subdominant triad, and the
dominant seventh), and a set of practical tools for understanding traditional
counterpoint (“OUCH theory”). Inspired by the Prime Directive, we identify
similar phenomena reappearing in different environments, the same abstract
relationships manifested in increasingly sophisticated ways.
A word of warning: this is one of the most dense and difficult chapters in
the book, discussing almost all the geometrical spaces needed later. Readers
who want a break can read the opening of chapter 4 after they finish part 1,
returning to part two before reading §4.4. Similarly, the opening of chapter 5
can be read without absorbing this material, though part 3 is needed in §5.5.

Part 1. Two Voices

1. The imperfect system

Some of the deepest mysteries of counterpoint arise from the geometry of
imperfect consonances. Figure 3.1.1 presents the spiral diagram for thirds
and sixths in diatonic space.1 Here, the unit is one scale step and “diatonic
seconds” such as C–D and E–F have size 1 (appendix 1).2 Since our chords
have two notes, the line of transposition winds twice around the circle before
returning to its starting point; since the diatonic scale has seven notes, there
are seven equally spaced points along the spiral. Sliding along the spiral
transposes along the scale, looping around the spiral transposes along the
chord, turning thirds into sixths and vice versa.
Figure 3.1.1. The spiral diagram for diatonic dyads.

We can calculate the effect of single-step clockwise motion from C to F

using the recipe in §2.1. Sliding clockwise from C to F touches four points
for T–4, passing our initial radial position once for t1. Figure 3.1.2 shows that
these two motions combine to form the voice leading (C, E) → (A, F),
moving the third’s root down by two scale steps and its third up by one step.
I call this the descending basic voice leading for diatonic thirds, since it
connects every point on the graph to its nearest clockwise neighbor.3 Single-
step counterclockwise motion generates the inverse voice leading, sending
the third up by two steps and the root down by one step. I call this the
ascending basic voice leading. (When context is clear, I will not specify the
direction.) Repeatedly applying the basic voice leading moves through the
two inversions or “modes” of the interval (third and sixth), touching on each
inversion of every imperfect consonance before returning to the initial chord,
now with the notes transposed by octave (Figure 3.1.3).

Figure 3.1.2. Moving one clockwise step from C to F combines T–4 and t1. The third moves up by
step and the root moves down by third.
Figure 3.1.3. A series of clockwise moves in the space produces a descending-fifth progression in
which thirds and sixths alternate.

There are two nonobvious ways to analyze motion along the diagram. The
first is to assign every path a number corresponding to the angular distance
traveled, or, equivalently, the number of basic voice leadings required to
generate that path. Since there is only one chord at every angular coordinate,
radial motion is completely determined by angular motion; as we will see,
this allows us to collapse the spiral into a circle. Alternatively, we can adopt
the Tinctorian strategy of decomposing voice leadings into combinations of
transposition along both chord and scale, or loops and slides. From this
perspective, there are just two basic possibilities, corresponding to those that
transpose along the chord and those that do not; these use an odd or even
number of basic voice leadings, respectively. Within each category, voice
leadings differ only by a change in center of gravity, controlled by the scalar
transposition Tx (Figure 3.1.4).
Figure 3.1.4. The voice leadings along the dyadic circle expressed in musical notation; the top staff
applies an odd number of basic voice leadings, combining transposition along the scale with one-step
transposition along the dyad (t1); the bottom staff applies an even number of basic voice leadings and
involves no transposition along the dyad.

It follows that purely contrapuntal relationships will tend to privilege fifth

motion: if we are in two-note diatonic space, then fifth-related dyads are
literally adjacent to one another. This dyadic “circle of fifths” is independent
of both the acoustic circle of fifths and the circle of fifths connecting
diatonic scales (chapter 8). We should therefore expect that linearly oriented
music will contain a plethora of fifth-related imperfect consonances, even
when roots are not the explicit objects of compositional attention. Figure
3.1.5 shows that the opening of Palestrina’s mass Ave regina coelorum
contains four instances of the diatonic third’s basic voice leading. The fifth-
motion here arises from two distinct sources, the contrapuntal logic of
diatonic thirds and the converging voices of the cadence. The rise of
functional tonality involves the integration of these two, assigning harmonic
significance to the fifth-progressions that are inevitably produced by the
system of diatonic dyads. This is more or less explicit in a passage like
Figure 3.1.6, where the dyads of the main theme articulate the basic voice
leading while the root progressions conform to tonally functional
expectations, the descending fifths of the cadence of a piece with those
within the phrase.
Figure 3.1.5. The opening of the Palestrina’s mass Ave regina coelorum. The numbers below the
example indicate angular motion on Figure 3.1.1, with negative numbers corresponding to clockwise
(descending) motion.

Figure 3.1.6. The start of the last movement of Beethoven’s E♭ major piano sonata, Op. 27, no. 1.

The dyadic logic of the 2-in-7 spiral interacts with the triadic system in
complex ways. In the descending voice leading the initial dyad is almost
always the root and third of a triad, while the second dyad can either be
root/third or third/fifth, producing authentic and deceptive resolutions
respectively (Figure 3.1.7). A passing tone is often inserted into the melodic
third to create a resolving tritone. The ascending-fifth voice leading embeds
into a wider range of progressions; here again a passing tone yields 6̂–7̂–1̂
against 4̂–3̂, combining plagal root motion with an authentic resolution of
the tritone (7̂–1̂/4̂–3̂). In baroque music this figure is often harmonized as
shown in the middle column of Figure 3.1.7, producing a pair of eighth-note
harmonies over a stable bass (ii–vii°6 or IV–vii°64).4 The rightmost column
instead supports the voice leading with a changing bass, an option that is
increasingly preferred as functional harmony develops. The contrast between
these two approaches reflects a shift from a dyadic-contrapuntal conception
of functionality, where linear patterns are predominant, to a later and more
triadic orientation.

Figure 3.1.7. Triadic harmonizations of the diatonic third’s basic voice leading. The descending form
typically appears as root and third of the initial triad (top line, “1–3”), moving either to root-third or
third-fifth of a second triad. The ascending form can also appear as third-fifth of the initial triad
(bottom line, “3–5”). In both cases, a passing tone can create a resolving tritone. The rightmost
column harmonizes the dyadic progression with a change of bass.

Attending to these dyadic patterns can reveal subterranean levels of

musical structure obscured by Roman numerals. In the second theme of the
first movement of Beethoven’s Op. 23 violin sonata, for example, the root
progressions are mysterious while the sequence of basic voice leadings is
completely clear, moving up and down the ladder of fifths (Figure 3.1.8).5
This dyadic organization is concealed by the third’s changing position,
appearing initially as the triad’s third-fifth rather than root-third. In the
recapitulation, Beethoven normalizes the music so that roots follow upper
voices, replacing a–G–d with C–G–d and clarifying the theme’s sequential
structure.6 Similarly, Figure 3.1.9 shows Clara Wieck Schumann presenting
a period whose two endings exploit the two different basic voice leadings of
the dyad (C♯, E); the surrounding music embeds the ascending basic voice
leading into three of its four possible triadic positions, appearing as the
bottom thirds of successive triads, the top thirds, and the top and bottom.
Here the logic is one of chordal subsets rather than complete triads. Roman
numerals are not so much wrong as incomplete, taking a back seat to other
relationships.

Figure 3.1.8. The second theme of the first movement of Beethoven’s Op. 23 violin sonata. While the
root progressions are mysterious, the sequence of basic voice leadings is systematic, ascending by
fifth, descending, and then ascending again.
Figure 3.1.9. The opening of the middle section of the second mazurka in Clara Wieck Schumann’s
Soirées Musicales, Op. 6. The top staff shows various appearances of the ascending and descending
basic voice leadings, designated by open and closed noteheads respectively. The annotations indicate
whether the basic voice leading appears as the bottom (b) or top (t) third of the triad.

The dyadic circle is also the source of a number of schemas central to

functional tonality. Figure 3.1.10 analyzes four patterns discussed in
Gjerdingen’s Music in the Galant Style: the converging cadence takes a
single clockwise step from 2̂–4̂ to 5̂–7̂, augmented by passing tones in each
voice. The “Fonte,” or descending-fifth sequence, makes a series of
clockwise moves. The “Monte,” or ascending-step sequence, alternates
clockwise steps with three-step counterclockwise moves, transposing up by
step at every other chord. The “Quiescenza” alternates rightward and
leftward steps, tonicizing a pair of fifth-related harmonies with ascending
and descending basic voice leadings. To these I have added a common I–IV–
V–I pattern not identified as schematic by Gjerdingen; like the Quiescenza,
it juxtaposes rightward and leftward motions. Such examples suggest that
the basic idioms of functional composition are not so much arbitrary
conventions as responses to the basic affordances of musical space: beneath
the conventional and schematic surface lies a deeper geometrical logic
featuring short-distance motions in the space of imperfect consonances.7

Figure 3.1.10. Common schemas and patterns analyzed as paths on the imperfect consonance’s spiral
diagram.

2. Voice exchanges
Any voice leading can be decomposed into two parts: a crossing-free
component that preserves spacing as measured along the intrinsic scale and a
voice exchange rearranging the notes of a single chord, with paths
collectively summing to 0.8 (As we will see, this factoring of voice leadings
into voice exchanges and a crossing-free residue is a point of contact
between Schenkerian theory and voice-leading geometry.) The diatonic third
has two elementary voice exchanges, a small one moving each voice by two
steps (third down to root and root up to third), and a large one moving each
voice by five steps (third up to root and root down to third). More
complicated voice leadings—for instance, keeping the third fixed while
moving the root up by octave—can be generated from these two, perhaps in
combination with one or more transpositions along the chord.9 Figure 3.2.1
uses curved arrows to represent the elementary voice exchanges. This new
geometry can represent every two-voice voice leading connecting any
imperfect consonances in the same diatonic collection.

Figure 3.2.1. Curved arrows represent voice exchanges: dotted and solid lines are the two- and five-
step exchanges, c0 and c1.

The voice leadings on this augmented graph are again fundamental to

functionally tonal composition, schematic but more general and flexible than
idioms as we tend to think of them. A single voice exchange combines with a
single basic voice leading to produce “antiparallel” motion in which the
voices move in contrary motion by almost the same amount. For instance,
the basic voice leading (C4, E5) → (A3, F5) combines with the two-step
voice exchange (A3, F5) → (F3, A5) to produce (C4, E5) → (F3, A5), with
the root falling by fifth and the third ascending by fourth; that same basic
voice leading combines with the five-step voice exchange (A3, F5) → (F4,
A4) to produce (C4, E5) → (F4, A4), now with the root ascending by fourth
and the third descending by fifth. These voice leadings are contrapuntally
balanced in the sense that they move their voices by approximately the same
distance—a fourth and a fifth. If we move further along the circle, say by
combining a three-step clockwise motion with one of the voice exchanges,
we get more and more unbalanced motion (Figure 3.2.2). Here again
diatonic fourths are contrapuntally privileged: fourth-related thirds, as
geometrical neighbors, allow for antiparallel motion in which the voices are
almost exactly balanced, and hence minimize the largest distance moved by
either voice; other root motions produce less-balanced counterpoint, with
larger leaps in one of the voices. This is why we find fifth-relations in
diatonic contrary-motion passages from Monteverdi to Meredith Monk (e.g.,
Figure 1.5.8).10
Figure 3.2.2. (top) Antiparallel motion decomposed into a basic voice leading and a voice exchange.
(bottom) Antiparallel motion between fifth-related thirds involves the most balanced intervals in the
two voices; as we move away from the fifth, the motion moves to one voice.

We can also apply the two-step voice exchange separately from the basic
voice leading to produce the sequence in Figure 3.2.3. Here the voices move
apart in near contrary motion, the bass descending by thirds while the top
voice alternates steps and thirds. Figure 3.2.4 shows the opening of the last
movement of Beethoven’s A♭ major piano sonata, Op. 26. The music begins
with the basic voice leading in a descending-fifth sequence that continues
across the textural change. (Here the schema appears in the context of
shifting scales tonicizing ii and vi, adding chromatic variety without
disrupting the contrapuntal logic.) Beethoven follows this with the two-step
voice exchange, moving the hands apart so that the tenor plays A♭ while the
top voice plays C; the basic voice leading then produces F–D, or vii°6/V, at
which point the pattern restarts between alto and bass under an upper-voice
pedal. The opening music now returns but with the two hands in contrary
motion—vertical dyads cycling through the two-step voice exchange, the
“antiparallel” voice leading, the five-step voice exchange, and a second
antiparallel voice leading. Here we have all the imperfect system’s atomic
moves: the basic voice leading, the two voice exchanges, and their
combination. This fusion of contrapuntal and harmonic is typical of
Beethoven, its joyous simplicity laying bare the fundamental mechanisms of
functional coherence.

Figure 3.2.3. Alternating the basic voice leading and the two-step voice exchange leads to a wedge
sequence in which harmonies move by fifth.
Figure 3.2.4. The start of the last movement of Beethoven’s A♭ major piano sonata, Op. 26.

One way to internalize these relationships is to focus on the voices’ third-

or-sixth arrangement, which can either oscillate back-and-forth to preserve
the dyad’s spacing-in-chordal-steps or form a contrary-motion wedge. Figure
3.2.5 shows a few common idioms that can be obtained by moving one hand
along some familiar pattern—such as a scale or arpeggio—while the other
moves from third to sixth and back, augmented with occasional parallel
motion. (Parallel motion and shifts between third and sixth generate all the
moves on the expanded diagram of Figure 3.2.1.) My practical,
improvisational understanding of functional tonality improved significantly
when I developed an intuitive feel for these patterns. Both Schenkerian
theory and schema theory have emphasized outer-voice idioms of this sort;
here we take the additional step of tracing them to the spiral geometry of
musical possibility, an abstract logic constraining musicians’ linear impulses
and leading them toward a small collection of musical possibilities.
Figure 3.2.5 Third-and-sixth patterns that can be generated with the expanded circle of Figure 3.2.1.
On the left, parallel patterns that preserve spacing in chordal steps; on the right, contrary-motion
patterns.

3. Other intervals
Chapter 2 showed that a single graph can be used to represent voice-leading
relationships among the chromatic transpositions of any three-note chord.
The same is true of the other spiral diagrams: to generate Figure 3.1.1 we
need only the size of the chord, which tells us how many times our “line of
transposition” wraps around the circle, and the size of the scale, which tells
us how many equally spaced points to place on the line. This means that the
graph represents any two-note interval in any seven-note scale: no matter
what scale we choose, we can label its notes with the letters A–G, redefining
the familiar names to refer to the notes of our new scale, and no matter what
interval we choose, we can form a descending “basic voice leading”
combining four-step descending transposition along the scale with one-step
ascending transposition along the chord, or T–4t1 (Figure 3.3.1). Though
superficially different, these “basic voice leadings” move their voices in
structurally similar ways.

Figure 3.3.1. The basic voice leadings for diatonic fifths, seconds, and unisons. The figure shows how
they all combine four-step transposition downward along the scale with one-step transposition upward
along the chord, or T–4t1.

The development of the first movement of the first Beethoven piano

sonata superimposes two basic voice leadings, the outer voices in thirds and
the inner voices in fifths; because of their structural similarity, the two dyads
stay in sync with each other, moving down by step every other chord and
combining to produce triads with doubled root (Figure 3.3.2). The resulting
four-voice pattern is ubiquitous throughout functional harmony, one of the
main voice-leading schemas used for descending-fifth sequences. Figure
3.3.3 shows some other formulae containing the diatonic fifth’s basic voice
leading: a standard 7–6 fauxbourdon pattern, a descending-fifth sequence of
incomplete seventh chords, common in virtually all functional music from
the baroque to jazz, and a descending-fifth sequence of seventh chords
fusing these two. Baroque sequences often trade on the similarity of these
patterns, blurring the boundary between “merely linear” fauxbourdon and
robustly harmonic descending fifths (§7.5). Here again, functional schemas
exploit the structure of a preexisting space of musical possibilities.
Figure 3.3.2. A sequence from the development of Beethoven’s Op. 2, no. 1 (mm. 73–80), combining
basic voice leadings for thirds and fifths.
Figure 3.3.3. The diatonic fifth’s basic voice leading can be embedded into a fauxbourdon 7–6
sequence, a sequence of incomplete descending-fifth seventh chords, or a four-voice descending-fifth
sequence of root-position and second-inversion seventh chords. The last sequence contains two
separate copies of the pattern.

Of course, thirds and sixths predominate in the Western tradition, partly

because composers favor imperfect consonances and partly because all the
derived voice leadings are permissible—parallel motion, antiparallel motion,
and the basic voice leading itself. Perfect consonances require more cautious
treatment due to both forbidden parallels and the ambiguous status of fourths
and tritones. However, there are genres that are less squeamish about these
matters: Figure 3.3.4 uses the spiral to analyze a phrase from the Sacred
Harp tune “Heavenly Spark,” exulting in its perfect-consonance harmony.
More complex are the opening eight bars of Thelonious Monk’s
“Pannonica,” where the upper voices outline a diatonic fifth that sinks
slowly downward, much like the previous chapter’s rock songs (Figure
3.3.5); the left hand subposes various bass notes under this right-hand core
so that the upper voices move from third-seventh to root-fifth and back, and
then up to sixth/ninth.11 These basic diatonic patterns are further elaborated
by a variety of scale shifts, the music implying a variety of black-note keys
before returning to C major. Jazz often involves similarly sophisticated
manipulation of diatonic geometry, exploiting the link between descending-
fifth harmonic progressions and efficient voice leading.

Figure 3.3.4. A phrase from “Heavenly Spark” represented on the circular space for diatonic fourths.

Figure 3.3.5. The opening of Thelonious Monk’s “Pannonica” decorates a pair of upper voices
forming fourths and fifths (bottom staff).
 Beethoven superimposed two basic voice leadings in a musical context,
and we can do the same in the theoretical domain. Figure 3.3.6 combines the
circular spaces for all three diatonic consonances, with unisons and octaves
on the outside, fourths and fifths in the center, and thirds and sixths between
those two. Angular position on this graph represents the sum of the chords’
pitch classes, while radial position mirrors the prelude’s abstract graph of
diatonic consonances; small radial motions thus change the vertical
configurations by small amounts (Figure 3.3.7). This is essentially the
Möbius strip representing all two-note diatonic consonances, reconstructed
as the sum of three separate spiral diagrams.12 Expressed in musical
notation, its basic motions appear as shown in Figure 3.3.8, horizontal lines
embedding the contrary-motion pattern of Figure 1.5.6 (a version of the
“Ludwig” schema [§10.1]) and the vertical direction representing parallel
motion. This is a Tinctorian approach to two-voice counterpoint,
emphasizing parallel and contrary motion rather than the locations of the
individual voices.
Figure 3.3.6. Superimposed spiral diagrams for the three diatonic consonances. Changes in interval
size correspond to radial motion in this space; changes in center of gravity correspond to angular
motion.
Figure 3.3.7. The cross section of dyadic chord space embeds the graph of dyadic consonances from
Figure P3.4.
Figure 3.3.8. Consonant, two-voice diatonic counterpoint decomposed into parallel and contrary
motion. Open noteheads show imperfect consonances.

Figure 3.3.9 presents an even more abstract version of the space, taking
advantage of the fact that diatonic spirals contain exactly one chord at every
angular position: the new space compresses the spiral’s two strands into a
single circle, each angular segment representing a basic voice leading and
switching between the interval’s two inversions (fourth and fifth, third and
sixth, unison and octave). Despite its abstraction, it represents the very same
voice leadings as the earlier spiral diagrams, providing a simple model of
many different intervals (Figure 3.3.10).
Figure 3.3.9. An abstract graph of two-note diatonic consonances. Set class corresponds to radial
position while center of gravity corresponds to angular position.
Figure 3.3.10. Three sequences with descending-fifth root progressions. In the first, the upper voices
move clockwise along the middle ring from FD to GB to EC; in the second they move along the outer
ring from FC to FB to EB. The third mixes these two as shown by the dotted lines.

Part 2. Three Voices

4. The circle of diatonic triads

We start our study of three-voice counterpoint with the spiral diagram
representing diatonic triads (Figure 3.4.1). Since a triad has three notes, the
line of transposition winds three times around the circle; since the diatonic
scale has seven notes, this line contains seven equally spaced points. To find
the basic voice leading we slide clockwise along the spiral from C down to
E, touching five chords for T–5 and passing by our initial radial position
twice for t2; the two operations combine to produce T–5t2 (or equivalently
T2t–1), a voice leading in which the triad’s root descends by one diatonic
step, lowering the chord’s center of gravity by ⅓.13 This links diatonic triads
in a chain of thirds connected by single-step voice leading, reaching the
descending-step transposition at every third iteration:
(C, E, G) → (B, E, G) → (B, D, G) → (B, D, F) → and so on.
As before, iterating the basic voice leading produces all possible
combinations of transposition along the chord and transposition along the
scale—or in other words every three-voice triadic voice leading that
preserves spacing in chordal steps.14 Once again, the same graph represents
any three-note chord in any seven-note scale.

Figure 3.4.1. The circle of diatonic triads, along with its basic voice leading. Repeated applications of
the basic voice leading pass through each inversion of each triad before returning to the initial
configuration transposed by octave.

The previous chapter used a chromatic version of this space to explore

how stepwise descending melodies can be combined with triadic harmonies.
There, contrapuntally close chords such as C and E major could be
harmonically quite distant. The circle of diatonic triads is special insofar as it
encodes a kind of harmonic similarity as well: adjacent chords on the circle
are close both contrapuntally (linked by single-step voice leading) and
harmonically (sharing two of three notes). It follows that the diatonic
analogue of chromatic third substitution will preserve a good degree of a
chord’s harmonic quality (Figure 3.4.2). This linking of harmonic and
contrapuntal similarity is crucial for the development of functional harmony
(§7.2).
Figure 3.4.2. Chromatic “major third” substitution compared to diatonic “third substitution”; the
former changes chord-quality dramatically while the latter preserves its harmonic character. As a
result, the progressions on the right sound reasonably similar while those on the left sound very
dissimilar.

The triadic circle allows us to deploy the theoretical strategy of chapter 2

in a fully diatonic context, revealing how melody and harmony constrain one
another. Consider the “Et incarnatus est” passage from the Credo of
Palestrina’s mass Spem in alium, shown here as Figure 3.4.3. The music is
largely homophonic and almost entirely made up of root-position chords
whose upper voices are in close position. As the top voice descends C–B–A,
we hear a series of ascending fifths F–C–G, with the last chord leading to A
minor as a modal subtonic, producing a VI–III–VII–i progression common
in popular music.15 We then hear a symmetrical i–iv–i before the melody
ascends and the harmonies reverse, leading to a i–VII–III–VI progression
that sounds rather like vi–(IV6)–V–I–IV in C. Figure 3.4.4 models the
palindromic upper voices as a series of mostly clockwise moves (F–C–G–a–
d) that unwind themselves exactly (d–a–G–F–C). The resulting music is
extremely clear in its association of melody and harmony, the descending
melodic steps leading to ascending fifths while ascending melodic steps
generate a much greater sense of functionality.
Figure 3.4.3. The “Et incarnatus est” from the Credo of Palestrina’s mass Spem in alium. Its
harmonies form an embellished inverted palindromic arch.
Figure 3.4.4. The path traced by the first five chords in the “Et incarnatus est”; the music then retraces
its steps.

Renaissance music is suffused with “generalized fauxbourdon,” or three-

voice passages in which triads descend through a variety of triadic
inversions. In Figure 3.4.5 Ockeghem alternates 63 and 53 chords to produce
a descending 6–5 sequence; on the triadic circle, the passage moves by one
and two clockwise steps. Figure 3.4.6 presents a series of ascending-fifth
progressions from C to G to D to A, moving two steps clockwise on the
triadic circle. (Here I consider the d64 chord to be harmonic, a reading that
will be discussed in chapter 5.) These are diatonic analogues to the passages
in chapter 2, and indeed there is plenty of rock music that can be modeled on
the diatonic spiral—for example, Israel Kamakawiwoʻole’s version of “Over
the Rainbow,” where the melody descends almost by octave while the
harmony loops around the spiral three times (Figure 3.4.7).16 Something
similar could be said about the McCoy Tyner passage in Figure 1.1.8, though
that music passes through the pentatonic scale as it descends.

Figure 3.4.5. Ockeghem, Gloria from the mass De plus en plus, mm. 127ff. Numbers represent steps
on the triadic spiral diagram (positive counterclockwise, negative clockwise).

Figure 3.4.6. Palestrina, Kyrie from the mass Descendit angelus Domini, mm. 39–40.
Figure 3.4.7. Israel Kamakawiwoʻole’s “Over the Rainbow.”

Of the many possibilities for generalized fauxbourdon, a few survive to

become functional schemas. Figure 3.4.8 shows three idioms featuring a pair
of stepwise voices: the Pachelbel and Prinner progressions, joined here to
descend through a melodic octave; the standard “rule of the octave” ascent;
and what I call the “fauxbourdon rule of the octave,” a mostly parallel-sixth
pattern that often harmonizes descending bass lines (§7.5).17 Things are
complicated by the fact that descending melodic motion tends to correlate
with retrofunctional progressions (Figure 3.4.9). Thus where modal
composers could freely use short clockwise steps on the triadic circle,
functional composers need to restrict themselves to a small number of
harmonically acceptable options.18
Figure 3.4.8. Three common functional patterns involving consistent clockwise or counterclockwise
motion on the triadic circle.
Figure 3.4.9. The possibilities for stepwise descent, starting from a C diatonic triad. The first line
shows clockwise steps on the circle of diatonic thirds; the next line shows the number of voices that
descend by step; the third shows the interval connecting the roots. For each interval, “weak”
progressions permit more descents than “strong” progressions: ascending-third progressions allow for
one descending voice whereas descending thirds permit none; ascending fifths allow for two descents,
while descending fifths allow for one; and descending steps allow three descents, whereas ascending
steps allow only two.

Two further points about functional harmony: first, the primary triads I,
IV, and V divide the triadic circle nearly evenly, into arcs of length 2, 3, and
2; this means that as a group they collectively minimize common tones,
articulating three maximally distinct harmonic poles. There are two other
nearly even divisions of the circle that contain the tonic triad, I–ii–V–I and
I–IV–vii°–I, both paradigmatic T–S–D–T progressions; this suggests a
geometrical interpretation of harmonic functions as maximally disjoint
regions of triadic space, involving triads with the fewest possible common
tones.19 Second, there is an interesting symmetry between the two fifth-
progressions IV–I and V–I: each has a direct voice leading in which two
voices move stepwise, ascending for the dominant and descending for the
subdominant, and an indirect voice leading in which two voices move by
third and one moves by step (Figure 3.4.10). Both of these indirect voice
leadings can be embellished with passing tones to create a diminished triad,
producing a standard descending-fifth progression and an “ascending
subdominant” respectively. The dualistic symmetry between these two kinds
of fifth motion, and the tension between functional tonality and descending
stepwise voice leading, will be topics for chapter 7.
Figure 3.4.10. Direct and indirect voice leadings for V–I and IV–I progressions.

To see how these relationships can play out, consider “Liebst du um

Schönheit,” the fourth song in Liebesfrühling, Robert Schumann’s Op. 37
and Clara Weick Schumann’s Op. 12. Clara’s music, shown in Figure 3.4.11,
presents a sixteen-bar ABAC period whose A sections are arranged as four-
bar sentences. Both consequents harmonize outer-voice tenths, the first using
the diatonic fifth’s descending basic voice leading to suggest the seventh-
chord idiom at the bottom of Figure 3.3.3; the second embellishing the
Pachelbel/Prinner hybrid with the diatonic third’s ascending basic voice
leading, here tonicizing the second chord in each ascending-fifth pair.20
Meanwhile, the join between the period’s two halves exploits the chromatic
proximity of minor-third-related seventh chords, moving directly from F7 to
A ♭ 7 (§3.8 below). The resulting music is both effortless and dense, a
spontaneous yet conceptually intricate combination of linear, harmonic,
dyadic, triadic, and seventh-chord patterns. How did earlier composers
manage to become fluent in such an extraordinarily difficult musical
language?

Figure 3.4.11. An outline of Clara Wieck Schumann’s “Liebst du um Schönheit,” Op. 12.

The answer, I think, is that its basic conceptual structures are neither the
homogenized harmonic patterns of Roman-numeral analysis, nor the equally
homogenized linear patterns of Schenkerian analysis, nor the highly
particular idioms of existing schema theory. Instead, they are somewhere in
between, linear/harmonic hybrids arising from the fundamental geometry of
diatonic objects: a small set of moves on various spiral diagrams. These
hybrid structures are more general than schemas and more specific than
“dominant-tonic progressions” or “parallel motion in tenths”; instead, they
are that subset of contrapuntal possibilities that is compatible with Roman-
numeral harmonic grammar, pruned from the larger vocabulary of earlier
music. A major lesson of voice-leading geometry is that this subset is both
small and surveyable: that the linear drive to write parallel tenths, if it is to
be compatible with functionally harmonic grammar, will manifest itself in a
limited number of ways. These are not idioms so much as possibilities. It is
not that harmony or counterpoint is primary; instead, only a few moves have
both harmonic and contrapuntal virtues.

5. Voice exchanges and multiple chord types

The Tinctoris Transform gives us an alternate perspective on two-voice
counterpoint: instead of independent musical lines entwined like affectionate
snakes, it depicts a series of verticalities waxing and waning as their shared
center of gravity rises and falls. To model three-voice counterpoint we could
replace vertical intervals with registral configurations such as 53, 63, and 64,
using two numbers to record the relative positions of three voices. However,
it is traditional to adopt broader categories representing more general classes
of vertical configurations or hand shapes. Eighteenth-century figured-bass
pedagogues recognized three positions for a keyboard’s right hand: the
default “close position,” with three voices articulating a complete triad and
spanning less than an octave (encompassing 53, 63, and 64), the “doubled
interval” position in which two adjacent voices sound the same pitch with
the third note less than an octave away, and a “half open” position in which
two voices are an octave apart with the third voice between them.21
If we require that voices sound at least two distinct pitch classes, and that
adjacent voices be separated by an octave or less, then there are two other
possibilities, the “open” configuration where the voices sound a complete
triad, with each voice two triadic steps above the next-lowest voice, and the
“open octave” position where two voices are an octave apart with the third
note outside this interval, less than an octave from the nearest voice. This
yields the five possibilities of Figure 3.5.1. These can be divided into the
complete triadic voicings (close and open) and incomplete voicings in which
one note is doubled (doubled interval, half open, open octave). Within
categories, configurations are related by octave displacements: we can turn
DI into H, C into O, and H into OO, by transposing the middle voice by
octave. As mentioned in the prelude to this chapter, these generalized hand
positions are vertical configurations measured along the intrinsic scale
defined by the voicing’s notes.22

Figure 3.5.1. Five upper-voice configurations and a diagram of their most common connections.

Though these categories were originally designed to describe the upper

voices of four-voice figured-bass accompaniments, we can repurpose them
for counterpoint in three voices. Once again we describe three-voice music
using a center of gravity and an upper-voice configuration. In Figure 3.5.2
the center of gravity undulates gently, rising from 60.33 to 62.33 in m. 5,
descending to 58.33 in m. 7, and ending near its starting point on 60.
Meanwhile upper voices are largely in O and OO positions, with just three
shifts to C in mm. 3–4 and 8.23 The four O→O voice leadings move their
roots by fifth, taking two steps on the triadic circle. The open-octave chords
always move to complete triads, usually open but occasionally close. Most
of the open-octave chords are doubled thirds, with the exception of the two
doubled fifths in the final phrase.
Figure 3.5.2. The Kyrie of William Byrd’s Mass in Three Voices (~1591).

This pedagogical theory of “hand positions” maps more-or-less directly

onto the concepts of set theory and musical geometry. Figure 3.5.3 shows all
possible diatonic three-voice consonances: as in Figure 3.3.9, one-step
angular motion represents the basic voice leading, increasing or decreasing
the center of gravity by one third of a diatonic step; radial motion,
meanwhile, corresponds to changes of vertical configuration.24 Though the
figure looks similar to the earlier dyadic graph, it is in fact somewhat more
abstract, its radial dimension modeling a space that is intrinsically two-
dimensional. Figure 3.5.4 presents a more faithful representation of the two-
dimensional cross section of the three-dimensional space of ordered diatonic
trichords. Here our five consonant configurations are disjoint regions
arranged left-to-right, while changes in center of gravity move in the third
dimension, out of the page.25 Once again, there is a direct line from figured-
bass terminology to set theory and geometry: thinking in vertical
configurations is geometrical thinking, with terms like “close position”
corresponding to compact regions in the space of musical possibility.
Figure 3.5.3. An abstract model of three-voice diatonic counterpoint.
Figure 3.5.4. The five three-voice hand positions correspond to regions in trichordal ordered pitch
space. These contain all configurations of three diatonic voices ordered in pitch with no two adjacent
voices more than an octave apart.

We can sharpen our understanding of these configurations by considering

the issue of parallel perfect intervals. One close-position triad can move to
another without creating parallels except when both are in root position.
Similarly, open-position triads can be linked without parallels except when
both are in root position or both are in first inversion. Half-open chords can
progress to half-open chords only when the outer-voice octave stays fixed.
Doubled-interval and open-octave positions are slightly more flexible, since
the doubling can shift voices; however, such motions are rare.26 As a general
rule, then, we can expect that three-voice progressions will conform to the
map on Figure 3.5.1—either moving from close to close, open to open, or
horizontally by one step. Thus DI typically moves to C, OO moves to O, and
H moves to C or O. About 75% of the three-voice transitions in Palestrina
and Victoria follow these rules; for Josquin the number is closer to 70% as
crossings are more common.27
Theorists since Zarlino have enjoined composers to write complete triads
whenever possible.28 However, three-voice composition involves an
unusually strong tension between harmony and melody: the desire for
harmonic richness pushes toward complete triads, while the desire for
melodic elegance pulls toward incomplete sonorities. As a result, doubled
thirds are more common than first-inversion triads (Figure 3.5.5). One might
say that the fundamental three-voice triadic harmony is the third, which is
accompanied either by a fifth, a doubling, or a sixth (in that order).
Furthermore, it is a mathematical fact that any three-voice voice leading
between consonant diatonic sonorities will have two voices connecting
imperfect consonances unless one of the sonorities contains no imperfect
consonances at all.29 This means that three-voice voice leadings will often
have two voices behaving according to the “imperfect system” of §3.1,
moving between imperfect consonances by some combination of parallel
motion, voice crossings, and a basic voice leading.
Figure 3.5.5. Configurations of three-voice consonances in a range of Renaissance music; numbers
are percentages with each column adding to 100 (within rounding).

It can sometimes be useful to attend to these voices. In Byrd’s “Kyrie,” for

example, we find that the first two phrases share the dyadic sequence (F, A)
→ (E, G) → (F, A) → (D, B♭)→ (F, A), with the lower voice moving from
bass to tenor in the repeat: we begin with I–V6–I–IV6–(vii°)–I, with the
parenthesized chord the basic voice leading’s resolving tritone; in the repeat,
the new bass line instead gives us I–V–I–ii–vii°6–I. It is easy to imagine that
Byrd heard the V chord in the second phrase as being similar to the V6 in the
first, and the IV–(vii°) pair as analogous to ii–vii°6, a modal precursor to
inversional equivalence arising from dyadic voice leading. The third phrase,
meanwhile, is less dyadic and functional, featuring a descending 6–5
progression overlapping with a series of ascending fifths.30 The sequence
should lead back to (D, F) → (C ♯ , E) in the outer voices, but the middle
voice instead leaps to take the bass’s role, allowing for a root-position
cadence. As a whole, the piece is a nice example of triadic music’s ability to
juggle different logics, moving from dyadic organization in the first two
phrases to triadic organization in the third.
Our five upper-voice configurations are sufficient to represent any
consonant three-voice diatonic voice leading in which adjacent voices do not
cross and are no more than an octave apart. We can model voice crossings by
considering the three fundamental voice exchanges that swap a chord’s
adjacent pitch classes along equal and opposite paths: for the triad, these are
the two-step voice exchanges that swap root and third or swap third and
fifth, and the three-step voice exchange that swaps root and fifth (Figure
3.5.6). Voice crossings are fairly rare in the Renaissance and even rarer in
later music.31 Furthermore, there is a tendency for the ear to filter out
crossings, associating voices registrally rather than by instrument.32
Nevertheless, these are important possibilities that we sometimes need to
consider. Readers who care about contemporary music might find it
interesting that the five positions discussed here, when conceived as patterns
of spacing-along-a-chord, can be applied to any trichord in any scale.
Figure 3.5.6. (top) The three basic pairwise voice crossings, exchanging root and third, third and fifth,
and root and fifth. (bottom) We can factor any triadic voice leading into a combination of a crossing-
free voice leading (here connecting close-position E minor and B diminished triads) plus a series of
pairwise voice exchanges.

Part 3. Four Voices

6. Four-voice triadic counterpoint

While two- and three-voice counterpoint can be usefully analyzed using
spiral diagrams, four voices require a different approach. This is because the
spiral diagrams model doubly parallel motion along both chord and scale.
Any appearance of contrary motion is thus a byproduct of the unevenness in
the chord’s internal structure, and in some sense not counterpoint at all.33
Genuine contrary motion instead arises from motion between chord types
(represented by the different rings on Figures 3.3.6 or 3.5.3) or voice
exchanges (represented by the curved arrows on Figure 3.2.1). Thus triadic
counterpoint has to change as we add voices: with two or three voices,
composers have access to multiple types of consonance, but with four or
more voices, they are largely restricted to a single sonority and hence more
reliant on voice exchanges.34 Historically, two and three voices often feature
in counterpoint classes, with four voices relegated to the domain of
harmony; at a more fundamental level, however, this is a difference within
counterpoint itself, a boundary we inevitably cross when the number of
voices increases beyond the size of our harmonies.
The philosophical implications of this change will be discussed in the next
section. Here I want to introduce a simple and broadly useful strategy for
writing in four voices: modeling the upper three voices as in the previous
section, and treating the bass as an independent actor. Thus the top voices
can be conceived as moving along Figure 3.5.1, typically from C to C, O to
O, or between adjacent positions such as DI and C. The most common
situation has the upper voices moving from close position to close position,
or open position to open position; this produces the situation in Figure 3.6.1,
where two groups of voices articulate different dual transpositions—three
upper voices moving in doubly parallel motion along both chord and scale,
and a bass moving independently.35 This reduces the four voices to two
fundamental lines, alto and tenor shadowing soprano in a nonobvious
fashion.
Figure 3.6.1. A schematic diagram of C→C and O→O counterpoint.

Figure 3.6.2 shows the distribution of upper-voice configurations in the

four-voice passages from Palestrina’s masses and J. S. Bach’s chorales.
Palestrina’s upper voices are generally close, concentrated toward the left
side of the graph, while Bach focuses on the C↔H↔O subsystem. (This
reflects the greater importance of voice crossings in Palestrina.) Figure 3.6.3
shows the most popular triadic upper-voice transitions in the chorales: more
than 30% connect close-position triads, with another 12% connecting open
to open; the next seven transitions connect adjacent positions on the model.
The only surprise is the frequency of motion between C and O, which is
slightly more popular than motion to and from the open-octave position.
Figure 3.6.2. Upper-voice configurations in the four-voice passages of Bach’s chorales and
Palestrina’s mass movements; size is proportional to frequency.
Figure 3.6.3. The most popular triadic upper-voice transitions in Bach’s chorales.

Figure 3.6.4 updates our upper-voice model by incorporating C↔O

transitions and relabeling DI and OO as “U” for Unusual. When I teach
Bach-style harmony, I use the mnemonic “OUCH theory” to refer to these
possibilities, providing students with a simple and symmetrical set of
guidelines: UDI goes to C, C goes anywhere except UOO, H goes to C or O,
O goes anywhere except UDI, and UDI goes to O. (In these contexts, the “U”
label is useful in producing a snappy mnemonic while reinforcing the
configurations’ rarity.) I introduce the configurations gradually, asking that
students first use only close-position triads, then only open position, then
open and close, and finally open, close, and half open. Only after they are
comfortable with these possibilities do I introduce the rest.

Figure 3.6.4. A grammar of triadic upper-voice configurations, extended to allow for C↔O
transitions. DI and OO have been relabeled as Unusual (U) to reflect their rarity in Bach.

These rules account for about 85% of the four-voice triadic voice leadings
in the Bach chorales and about 75% of those in the four-voice passages of
Palestrina’s masses. The approach simplifies the long lists of voice leadings
sometimes found in textbooks, allowing students to produce idiomatic part-
writing with a minimum of pain: the focus on upper-voice configurations
helps them avoid parallels, encourages them to think about spacing, and
gives them a positive set of options to pursue.36 They know, for example,
that if they are in H position they will be moving to either C or O on the next
chord; furthermore, they know that it is impossible for upper voices to form
parallel octaves when following these rules, and unlikely for them to form
parallel fifths.37 The rules generalize traditional figured-bass pedagogy,
extending its upper-voice configurations to all possible arrangements
(consistent with the principles of good spacing), and not just those that lie
comfortably under the hand. More importantly, they clarify that changes in
one line have implications for the others, arising from norms of vertical
spacing.
 It can be instructive to label vertical configurations in preexisting pieces.
Let us imagine that Bach began composing Figure 3.6.5 with the bass and
harmony—perhaps improvising at the keyboard while paying particular
attention to the outer voices. Now observe that the upper voices use a
default, close-position harmonization right up to the point at which this
would produce parallels (C♯–D in the tenor and bass). Here he shifts to the
half-open position by moving the tenor down one triadic step, to A rather
than D. At this point one might expect him to return to close position, but
given his harmonic choices, this would double the leading tone and produce
an augmented fourth in the tenor. A second half-open voicing would create
parallel octaves while O and UDI are equally problematic; Bach’s best option
is to move the tenor down another triadic step to F ♯ and an open-octave
voicing. At this point, he is free to return to close position, which he does for
the rest of the phrase. So far there is nothing particularly marvelous about
the individual melodies; it is only when Bach seizes the opportunity to add
tenor-voice passing tones that the counterpoint comes alive, transforming an
awkward harmonic skeleton into something attractive. “OUCH theory”
provides a plausible if speculative genealogy of the passage—one that
suggests how an attention to vertical configurations can help create a
genuine counterpoint of melodic lines. My suspicion is that composers rarely
if ever conceive of counterpoint in purely linear terms, always keeping one
eye on the subtle play of vertical configurations.
Figure 3.6.5. A hypothetical derivation of the opening of J. S. Bach’s “Ich dank’ dich, lieber Herr”
(BWV 347, Riemenschneider 2).

The “Hosanna” from Palestrina’s Spem in alium mass evinces similar

thinking in a different style. On Figure 3.6.6 I have labeled the upper-voice
configuration of every four-voice chord, marking crossings with a ×. The
passage is divided into five phrases, each supporting a repetition of the text
(“Hosanna in excelsis”) and featuring a motive that moves from an
ascending interval (usually a step, occasionally a third) to a stepwise
descent. Overall, about 65% of the chords are in close position, 14% open,
and 11% half open, with the doubled interval and open-octave positions
together accounting for 5% of the sonorities; the only move not on Figure
3.6.4 is the m. 18 leap from C to OO. The first, third, and fourth phrases
exhibit a similar structure, starting with widely separated voices and moving
to a close-position cadence: we have O→H→C in the first and fourth phrase,
and OO→O→H→C→DI in the third, systematically enumerating the five
positions. In each case, the compression of the voices is driven by an inner-
voice entry of the “Hosanna” motive. Harmonically, the music is suffused
with retrofunctional progressions: ascending fifths (especially the six
connecting B♭ major to E major in mm. 20–23), ascending thirds (F–a–C in
mm. 14–15), and ascending and descending steps. This is a beautiful
example of counterpoint leading harmony, an idiomatic Renaissance
analogue to the patterns we found in rock.
Figure 3.6.6. The “Hosanna” from Palestrina’s mass Spem in alium.

OUCH theory works because upper voices often form complete triads
while the bass moves independently. This naturally raises the vexed question
of doublings. Bret Aarden and Paul T. von Hippel have produced an
astonishing survey of pedagogical advice on the subject, considering forty
different texts, all of which seem to disagree.38 A particular focus of
controversy is the permissibility of doubling the bass in first-inversion major
triads such as I6; the question is whether these chords can support upper
voices in C or O position, or whether they should instead use one of the
other configurations. For all the theoretical disagreement, compositional
practice is surprisingly clear: Figure 3.6.7 shows that in Bach’s chorales, the
bass is the most frequently doubled note except when it is a leading tone.39
And in Palestrina’s masses we find something even more remarkable. If we
conceive vertical configurations as combining bass, middle element (a third
or fourth above the bass), and top (fifth or sixth), then Palestrina’s practice is
to double the bass most often, the middle element next, and the top last.40
The major exception is V6, where the leading tone is less likely to be
doubled. Palestrina’s reluctance to double the top voice may reflect its status
as an active or variable note, free to move between fifth and sixth while still
forming a consonance; doubling this note would eliminate that freedom by
turning one of these options into a dissonance.
Figure 3.6.7. Doublings in four-voice passages in J. S. Bach’s chorales and selected Palestrina masses.
Each number is a percentage representing the proportion of chords of that type with the given
doubling.

I know of no substantial body of music that avoids doubling the bass of

first-inversion chords. Certainly, the practice of Palestrina and Bach quashes
any attempt to ground this supposed prohibition in acoustics or aesthetics:
Palestrina and Bach knew a good deal about sonority, and it is reasonable to
follow their practice rather than the opinions of pedagogues.41 My own
suspicion, shared by Gottfried Weber, is that the avoidance of the doubled
third arose not out of aesthetics but rather out of the exigencies of figured-
bass realization: here doubling the bass of a first-inversion chord can lead
the unwary improviser into parallels, and the half-open position generally
allows for more melodic options (Figure 3.6.8).42 It is possible that this
purely pragmatic consideration gradually acquired an illusory aesthetic
weight, mutating into an analytically unsupportable prohibition on doubled
thirds.

Figure 3.6.8. Using half-open chords to avoid parallels. In the first measure, upper-voice close-
position triads create parallels between bass and alto; replacing the second close-position triad with a
half-open voicing removes the problem.

7. Thinking within the chord

The heuristics in the previous section divide bass from upper voices in a way
that may be untrue to a piece’s intrinsic organization. This is not necessarily
a problem for the practitioner, as they generate a very wide range of musical
possibilities—again, all those voice leadings without voice crossings in pitch
space and in which upper voices are no more than an octave from their
neighbors. But they are theoretically unsatisfactory, allowing us to produce
four-voice counterpoint without necessarily understanding it very deeply.
We can obtain a richer perspective by allowing for more flexible
groupings of voices, so that any pair or triple can move in doubly parallel
motion. The first passage in Figure 3.7.1 uses the standard OUCH strategy,
the three upper voices moving against an independent bass; the second
instead groups voices into pairs; while the third groups the bottom three
voices together. The soprano and bass are the same in each passage; the only
difference is which voice the inner voices follow. Underneath I identify the
motion of voices inside the triad (tx) and the motion of the triad inside the
scale (Tx). We can assign these to different musical levels, imagining the
chord to move along the scale in the background while the voices move
along the chord in the foreground.

Figure 3.7.1. Three progressions grouping voices in different ways. In each case, the top group moves
by T–1 and T1t–1 while the bottom group moves by T–1t1 and T1. Underneath, I analyze these motions
as combining motion of chordal background inside scale, and motion of voices inside chordal
background.

In other cases, we might want to postulate background voice leadings that

themselves combine transposition along chord and scale: abstract voices that
function as scale degrees along which the surface voices move. Back in
Figure 1.2.1, we saw upper voices ascending along a three-note scale that
participated in the “omnibus” schema—a pattern so familiar and
commonplace that it is recognizable despite the right-hand figuration. In that
example, an upper-voice melody moved along a triadic background that
itself moved by t1T–3 from the start of one sequential unit to the next. Figure
3.7.2 uses this strategy to analyze the opening of Beethoven’s Waldstein. On
the bottom staff we have a triadic background that generates three scale
degrees which I number from bottom to top: abstract voice 1 descends
semitonally, sounding C–B–B♭–A, while abstract voices 2 and 3 descend in
whole steps, sounding E–D–D–C and G–G–F–F respectively. The pianist’s
left hand is stationary relative to this structure, its motion entirely produced
by the background. The surface voices of the right hand, meanwhile, ascend
along the background to form a chain of ascending thirds E–G–B–D–F–A–
C, with the alto one or two chordal steps below the soprano.43 (For
simplicity, my analysis omits the tonicizations of G and F, which exploit the
diatonic third’s ascending basic voice leading; I also rectify octave
displacements to show the continuing pattern of ascending thirds.) Familiar
chromaticism is here embellished hierarchically, the stepwise voice leading
sublimated into the background.
Figure 3.7.2. The opening of the Waldstein. The background voice leading is shown in the lowest
staff, creating three abstract scale degrees. The left hand always sounds degrees 1 and 3 while the right
hand ascends along this three-note scale.

We can model this geometrically with the smallest nontrivial spiral

diagram, that of two voices moving within a triad. Its basic voice leading
moves the voices by alternating triadic step, producing a series of one- and
two-step verticalities (Figure 3.7.3). Beethoven’s right hand outlines a series
of short counterclockwise motions in this space, moving between thirds,
fourths, and fifths while the left hand remains motionless (Figure 3.7.4).
This, however, only represents the voices’ motion within the triadic
background. To represent the background’s motion, we need a second spiral
diagram representing triads in chromatic space. This background is very
similar to the models in chapter 2, now augmented by an additional
hierarchical layer that creates an expanding wedge. For the first time we
need two diagrams rather than a single spiral on its own. This hierarchical
depth is characteristic of four-voice music.
Figure 3.7.3. The spiral diagram for two voices inside a three-note triad. Here t represents
transposition along the triad while τ represents transposition along the dyad.
Figure 3.7.4. The Waldstein opening plotted on a pair of spiral diagrams. Letters above the music
correspond to labeled paths on the diagrams. Here there are three levels of transposition: τ refers to
transposition along the right-hand dyad, t to transposition along the triad, and T to chromatic
transposition.
 I call this thinking within the chord: an abstract chorale provides scale-like
collections within which surface voices move. One can imagine a
metaphorical harp (or autoharp) sounding a chord’s notes in every octave,
providing a field of contrapuntal play.44 Those slots shift as the chord
changes, typically by efficient voice leading; the surface voices’ motion-
within-the-chord is unperturbed by the shifts, like a melodic sequence inside
a modulating scale (e.g., Figure 1.2.5). This description resonates with the
Schenkerian insight that voice leading often takes place in a musical
“background” simultaneously real and yet not always obvious in the score.
More specifically, it allows us to implement the Schenkerian strategy of
relegating doublings and voice exchanges to the surface, as artifacts of
voices moving within a simpler background.45 My approach gives
nonmetaphysical meaning to the claim that the right-hand B3 and D4 in mm.
3–4 of the Waldstein are “the same” as the left-hand B2 and D3—that is, the
same abstract scale degrees articulated by different concrete voices.
The Waldstein’s ascending thirds are produced by voices ascending within
a tertian background that slowly descends. The Tristan prelude opens with a
similar chain of ascending thirds arising in a similar way (Figure 3.7.5); the
technique reappears at the prelude’s climax when the outer voices expand in
an uncontrolled frenzy (Figure 3.7.6).46 In later chapters we will encounter
analogous passages in Chopin’s “Revolutionary” Etude (§9.3), Schubert’s
Quartettsatz (§10.6), the prelude to Wagner’s Lohengrin (§10.7), and, in the
conclusion, a composition exercise of my own devising. These are all
structurally analogous to the Waldstein, ascending melodic thirds refracting
the tertian structure of standard harmony through the distorting lens of a
nontrivial background.

Figure 3.7.5. The opening of Tristan. The lower staff shows a background voice leading with voices
or scale degrees numbered from bottom to top. The melodic “Desire” motif ascends along these scale
degrees, sounding voice 1, voice 2, voice 3, voice 4, and then voice 1 in a higher octave.
Figure 3.7.6. The climax of the Tristan prelude, mm. 77–83. The lowest staff shows the background
against which the outer voices move.

OUCH theory thus treats as universal what is really just one particular
way of organizing musical lines, setting bass against upper voices; this is
why it struggles when voices are divided into pairs or when the soprano is
opposed to the lower voices. With more parts come even more possibilities:
the first forty-voice voice leading in Thomas Tallis’s Spem in alium traces a
very large number of paths from G major to C major (Figure 3.7.7). It would
boggle the mind to try to analyze this progression in forty-note chord space
where every triad appears more than fifty thousand times, each
corresponding to a different doubling. Much more intuitive is the idea that
the forty voices move within a shifting three-note scalar field, pursuing their
own independent combinations of scalar and chordal transposition. As in
Figure 3.7.1, we can imagine harmonies moving in parallel along the scale
(Ty) while voices move inside them (tx).47
Figure 3.7.7. The first forty-voice voice leading in Spem in alium. Boldface numbers show how many
voices sound each interval. The voices together articulate four voice leadings on the triadic circle, two
of which are shown: the clockwise five-step path and the counterclockwise two-step path (labeled –5
and 2 on the figure). The remaining paths, labeled –12 and 9, can be obtained by sliding
counterclockwise and clockwise along the spiral respectively. Together, the four voice leadings
contain all the intervals in which voices move by fifth or less.

There are three basic points here. The first is that as a very general matter,
triadic counterpoint in four or more voices often sets groups of voices
against one another: bass in contrary motion to the upper voices, soprano
and alto against tenor and bass, and so forth. The second point is that over
the course of Western music history composers seem to have become
increasingly aware of the hierarchical possibilities inherent in this
arrangement, writing music in which surface voices move along abstract
backgrounds whose scale degrees themselves articulate nontrivial voice
leadings, a significant musical development occurring within the functional
era. The third point, which will be the focus of chapter 8, is that functional
tonality adds yet another hierarchical level, with voices moving inside scale-
like chords that move within seven-note scales that are themselves
modulating through chromatic space: wheels within wheels, spiral diagrams
nested like the epicycles of Ptolemaic astronomy.

8. Seventh chords
A diatonic seventh chord contains a fifth between root and fifth, and another
fifth between third and seventh. Figure 3.8.1 shows that as we move
clockwise along the spiral, we apply a basic voice leading to each of the
fifth-pairs in turn. It is as if the seventh-chord diagram divides the dyadic
fifth-diagram in half: two steps on the seventh-chord diagram takes one step
on each fifth-pair’s dyadic diagram; where the fifth’s basic voice leading
links fifth-related harmonies, the seventh’s basic voice leading links third-
related harmonies, with the diatonic third being half the diatonic fifth.
Figure 3.8.1. Clockwise one-step motion on the circle of diatonic sevenths produces a sequence of
descending-third progressions in which the seventh descends (T–2t1). Two-step motion produces
descending-fifth progressions (T–4t2, e.g., from Cmaj7 to Fmaj43), while three-step motion produces
ascending steps (T–6t3, Cmaj7 to Dmaj2). The stems show that the four-voice voice leadings can be
factored into two pairs of voices each articulating the diatonic fifth’s basic voice leading.

Because of this, seventh chords again allow for a seamless integration of

melodic and functional-harmonic logic: where clockwise steps in triadic
space feature ascending-third roots, clockwise steps in seventh-chord space
produce descending-third root progressions; two clockwise steps make a
descending fifth. This geometry is the basis of many familiar songs: “All the
Things You Are” oscillates between the notes of one fifth-pair in the
descending-fifth sequence, descending by almost a complete octave (Figure
3.8.2); songs like “Autumn Leaves” have a very similar structure. Rock and
pop occasionally use the same strategy, as in “I Will Survive” (Figure 3.8.3),
where the melody descends stepwise from tonic to subdominant before
ascending. This is exactly the sort of pattern we found in chapter 2, only now
with seventh chords rather than triads. The addition of a fourth voice
magically converts retrofunctionality into functionality.

Figure 3.8.2. “All the Things You Are” outlines the descending-fifth sequence of seventh chords.
Open noteheads show melodic tones.
Figure 3.8.3. “I Will Survive,” written by Freddie Perren and Dino Fekaris and sung by Gloria
Gaynor, also follows the descending-fifth logic. During the C major chord, the melody descends to a
lower voice.

Meanwhile, the space of chromatic sevenths is vital to understanding a

range of nineteenth- and twentieth-century idioms. Figure 3.8.4
superimposes the circular voice-leading spaces for dominant and diminished
seventh chords. The trapezoidal boxes enclose dominant sevenths sharing
three notes with the same diminished seventh, a relation that lends them a
degree of harmonic similarity. The situation here is intermediate between the
space of chromatic triads (where nearby chords, like C major and E major,
can be harmonically quite different) and diatonic triads (where nearby
chords are harmonically similar): minor-third- and tritone-related dominant
sevenths, while not quite equivalent, have a definite harmonic
resemblance.48
Figure 3.8.4. Dominant and diminished seventh chords.

Progressions from a diminished seventh to its in-box dominant-seventh

neighbors can function as vii°7–V7, with a single voice descending by
semitone. Clockwise motion from diminished seventh to the nearest out-of-
box dominants produce authentic resolutions, usually acting as vii°7/V–V7.
Counterclockwise steps, meanwhile, are the “common-tone resolutions” of
second-semester theory, with three voices ascending by semitone; the
counterclockwise direction reflects the progressions’ predominantly
ascending voice leading. Figure 3.8.5 shows that these familiar voice
leadings are all equivalent under the Tinctoris Transform, differing only in
the transposition of the destination chord.

Figure 3.8.5. Motion from diminished to dominant seventh: one counterclockwise step corresponds to
the “common-tone” resolution (+3); in-box clockwise motion preserves three common tones and can
function as vii°7–V7 (–1); one out-of-box clockwise step corresponds to an authentic resolution (–5).

One-step clockwise motion between dominant sevenths produces the four

quasi-authentic voice leadings in Figure 3.8.6. Three of them are reasonably
common: the first is a standard descending-fifth progression between
dominant sevenths, while the second is its tritone substitution, initially
associated with the resolution of the German augmented sixth. The third, (C,
E, G, B♭)→ (C, D, F♯, A), occasionally functions as a deceptive resolution,
particularly in J. S. Bach (V7→V2/ii in the old key, IV7→V2 in the new).49
The final progression (C, E, G, B♭)→ (C, E♭, G♭, A♭) is quite unusual, and I
do not know of many examples from the literature. All four feature
predominantly descending voice leading and a seventh resolving down by
step.

Figure 3.8.6. One-step clockwise motion from one dominant seventh to another. This produces either
an authentic resolution, its tritone substitution, a deceptive motion from V7 to V2/ii, or a descending
major-third progression.

One-step counterclockwise motion produces a package of unusual

resolutions that evoke the common-tone diminished seventh: each of the
progressions in Figure 3.8.7 simply replaces the standard common-tone
diminished chord with a dominant seventh sharing three of its notes. For that
reason, I call these chords “common-tone dominant sevenths,” even though
some of the progressions feature no common tones!50 One is the reverse of
the dominant-seventh resolution, with seventh rising upward to the leading
tone—a dramatic deceptive motion that becomes increasingly important over
the course of the nineteenth century (Figures 3.8.8–9). A second has the
dominant seventh acting as a neighbor to the dominant seventh a major
second below; it appears in Haydn and Mozart and has a clear resemblance
to the common-tone diminished seventh (Figure 3.8.10; Figure 3.8.11 shows
a variant resolving to a triad).51 The third option, a semitonal lower
neighbor, (B, D♯, F♯, A) → (C, E, G, B♭), is relatively rare, though with a
triadic arrival it becomes a deceptive resolution of the dominant seventh.
One of my favorite nineteenth-century idioms—perhaps the most colorful of
its many altered and substitute dominants—combines this pseudo-common-
tone resolution with a dominant-tonic bass, fusing dominant and common-
tone functions (Figure 3.8.12).52 The fourth possibility is often found with a
triadic chord of resolution, usually as Ger65 →I64 and but occasionally as a
more generic neighboring progression. Late nineteenth- and early twentieth-
century American music often used these ascending resolutions to harmonize
blue notes, explicitly linking them to the common-tone diminished sevenths
of the classical tradition (Figure 3.8.13).
Figure 3.8.7. The “common-tone dominant sevenths.” Each replaces a common-tone diminished
seventh with a dominant seventh sharing three of its notes.
Figure 3.8.8. Ascending-fifth progressions in Mahler’s Ninth Symphony, I, mm. 408–409, and
Beethoven Op. 59, no. 3, I, mm. 42–44.
Figure 3.8.9. A dominant seventh on IV in Grieg (Op. 68, no. 3, mm. 26–34).

Figure 3.8.10. V/ii acting as a neighboring chord to V7 in Mozart’s C major piano sonata K.309, I,
mm. 3–4.
Figure 3.8.11. Supertonic dominant sevenths moving directly to the tonic in Grieg (Op. 68, no. 4, mm.
52–53) and John Williams (“Hedwig’s Theme” from the Harry Potter movies). In the last example,
the common-tone dominant seventh replaces the cadential dominant.
Figure 3.8.12. Common-tone voice leading over an authentic V–I bass: Beethoven uses this pattern in
Op. 59, no. 1, I, mm. 361–362; in Chopin’s Op. 50, no. 2, mm. 77–79, the common-tone diminished
seventh arrives while the bass is still on 5̂, and is sustained through the tonic arrival.

Figure 3.8.13. Three passages that link common-tone dominant and diminished sevenths, in contexts
that evoke the blues IV7: Scott Joplin’s Pine Apple Rag (1908, mm. 2–4 of the trio); Charles
Trevathan’s racist “Bully Song” (1896, mm. 1–10, harmonizing ♭3 with VI7 rather than IV7); and
Antonio Maggio’s “I Got the Blues” (1908, mm. 8–11).

Finally, one diminished seventh can ascend semitonally to another—a

move particularly characteristic of Beethoven (Figure 3.8.14).53 In the
context of the previous examples we can understand this as yet another
variant of the common-tone diminished seventh—indeed, we would hardly
question the association in cases where a dominant seventh briefly
intervenes between the two diminished sevenths.54 The four-note spiral
diagram thus allows us to generalize the “common-tone diminished seventh”
to contexts where common tones are no longer present, moving beyond a
specific idiom to a more general form of tetrachordal retrofunctionality in
which seventh chords resolve contrary to our expectations.

Figure 3.8.14. A semitonally ascending diminished seventh in Beethoven’s Op. 2, no. 1, IV, mm. 28–
29.

Our model sorts seventh-chord progressions into categories: the pseudo-

authentic resolutions moving one step clockwise, the neutral voice leadings
moving radially, and the retrofunctional or quasi-common-tone resolutions
moving one step counterclockwise. As the nineteenth century progresses,
seventh chords are gradually liberated from their contrapuntal obligations,
the “dissonant” seventh increasingly permitted to ascend. This in turn allows
for a more thorough exploration of tetrachordal geometry. There is, I think, a
strand of nineteenth- and twentieth-century music in which the diagram’s
categories are analytically relevant, with minor-third-related dominants
having a kind of equivalence, and clockwise and counterclockwise motion
corresponding to more and less expected progressions.55 The evolution
culminates in the twentieth century, when musicians begin speaking of the
intersubstitutability of minor-third and tritone-related dominants.56

9. Harmony and counterpoint

Analysts are accustomed to finding elaborate contrapuntal machinations in
nineteenth-century music, where chromatic voice leading provides an
alternative to functional-harmonic organization. Chapter 2 identified similar
processes in the unlikely setting of popular music, describing a quasi-
chromaticism in which the diatonic scale exerts greater gravitational force.
This chapter has extended the argument to purely diatonic counterpoint from
the sixteenth century onward, using spiral diagrams to model the
interdependence of harmony and counterpoint in a range of modal and
functionally tonal styles. The ubiquity of these relationships reflects the
power and generality of voice-leading geometry: we can construct spiral
diagrams for any chord in any scale, allowing us to observe how similar
linear forces lead to different harmonic outcomes in different musical
environments.
With diatonic thirds, a preference for descending melodic motion
produces descending-fifth progressions and a distinct sense of functionality;
indeed, a random series of short clockwise motions in the spiral diagram will
tend to produce the impression of harmonic function (Figure 3.9.1). By
contrast, triadic geometry associates descending melodies with
retrofunctional progressions, leading to sharper conflicts between melodic
and functional norms: random clockwise steps in the space of chromatic
triads sound more rock than classical, while clockwise steps in the space of
diatonic triads evoke the weightless progressions of the Renaissance. Four-
note diatonic music once again lends itself to functionality—and indeed, this
is one reason for the seventeenth-century embrace of seventh chords (§7.4).
Thus there is no simple relationship between linear procedure and harmonic
outcome, but rather a shifting and context-dependent set of implications.
Figure 3.9.1. Simple dyadic counterpoint randomly generated by a series of mostly clockwise steps on
the dyadic circle.

This is beautifully illustrated by the reappearance of nonfunctional triadic

patterns in four-voice functional contexts: for example Randy Newman’s
“When She Loved Me” uses a descending-fifth sequence of seventh chords
that embeds the eerie descents of Gesualdo’s “Moro Lasso” (Figure 3.9.2,
compare Figure 1.1.1).57 What might initially seem like a mere curiosity
instead reflects the fact that descending voice leading tends to be
retrofunctional with triads, but functional with seventh chords. A similar
relationship connects the II–IV–I progression of “Eight Days a Week” to its
functional ancestor, the V7/V–iiø7–V7–I of Bobby Rydell’s “Forget Him.”58
Rydell’s progression is a Romantic trope, combining a secondary dominant
with so-called modal mixture and exploiting descending semitonal voice
leading.59 In translating this idiom into retrofunctional triads, the Beatles
turned it into something genuinely new (Figure 3.9.3).60 Our two examples
thus move in opposite directions, Randy Newman incorporating a
nonfunctional idiom within a functional progression, the Beatles drawing
nonfunctional triads out of functional sevenths.

Figure 3.9.2. A descending-fifth seventh-chord progression embedding the nonfunctional triads from
“Moro Lasso.” Randy Newman uses the descending-fifth progression in “When She Loved Me” from
Toy Story 2 (at “still I waited for the day”).
Figure 3.9.3. The Beatles’ I–II–IV–I derived from a functional seventh-chord idiom.

This approach contrasts with more fixed conceptions of the relation

between linear and horizontal. Schenkerians tend to favor a “compatibilist”
view according to which contrapuntal forces, acting on their own, inevitably
generate the progressions of functional tonality.61 Richard Cohn, meanwhile,
leans toward a more incompatibilist outlook, proposing chromaticism and
functional harmony as separate and opposed musical forces.62 I think these
perspectives each capture only part of a larger story: the relationship
between horizontal and vertical is fundamentally context-dependent, with
different geometrical environments providing different harmonic
affordances. This in turn reflects a more general continuum of musical
practice: on one side we have idioms in which counterpoint and functional
tonality work together smoothly; toward the middle, linear forces generate
occasional harmonic hiccups that do not dislodge our sense of tonal
functionality; and at the other extreme, we find passages that totally efface
the sense of functionality and sometimes even centricity.63 The spiral
diagrams allow us to visualize this spectrum of musical practice, showing
exactly how harmony and counterpoint relate.
Composers can use these ideas to explore unfamiliar combinations of
chord and scale. Figure 3.9.4 uses a spiral diagram to generate a chorale
from my piece The Thousand Faces of Form. The basic chord is a gapped
stack of fifths, F–C–G–D–A–E–[B]–F♯, initially appearing with G–D–A–E

in the bass and F ♯ –C–F at the top: a dominant-seventh sharp nine over a
tonic pedal and with an additional major ninth. The chorale takes a series of
clockwise steps that pass through seven inversions, repeatedly applying the
basic voice leading before returning to the one-semitone transposition of the
opening chord. This is abstractly parallel motion, with each chord belonging
to the same set class and spaced in the same pattern of chordal steps—and
though the progression is complex and clangorous, I hear its chords as
audibly similar, reflecting their similar spacing and intervallic constitution.
One could use this chorale to generate even-more-complex passages in
which different groups of voices combine different transpositions along the
chord tx with the same transposition along the scale Ty, as in §3.7. Here we
start to glimpse how we might generalize traditional compositional
procedures to arbitrary chord-and-scale environments, precisely as suggested
by the Prime Directive.
Figure 3.9.4. A chorale built from the basic voice leading for a seven-note chord in twelve-note space;
each chord is T7t–4 of its predecessor and is voiced 3, 5, 3, 2, 3, 3 in intrinsic steps.
 1
Visit https://www.madmusicalscience.com/cs.html for interactive versions of this space, and
https://www.madmusicalscience.com/graphs/2in7.pdf for a printable version of the graph. Other
graphs can be found by replacing “2” and “7” with the appropriate numbers.
2
For more on this perspective, see chapter 4 of Tymoczko 2011a.
3
This voice leading reduces the dyad’s center of gravity by one: if we number notes so that C = 0,
D = 1, E = 2, and so on, then the chord (E, G) is (2, 4), with a sum of 6 and center of gravity (or
average) of 3; the basic voice leading sends (E, G) to (C, A), or (0, 5), with a sum of 5 and an average
of 2.5.
4
As Nicolas 2019 points out, Étienne-Denis Delair’s 1690 Traité d’accompagnement pour le
théorbe, et le clavessin recommends the ascending basic voice leading for ascending-fifth bass lines,
often producing a quasi-functional result.
5
The end of the phrase ends with an e-b ascent that continues the G–d–a pattern of the opening
phrase, though I would not put much emphasis on this connection.
6
See m. 182; this is one of countless instances of Beethoven transforming a theme so as to reveal a
disguised sequence.
7
Rabinovitch (2018, 2019, 2020) likewise grounds schemata in deeper contrapuntal principles;
though he starts from very different theoretical principles, I believe his approach is largely consistent
with mine. The link between galant schemas and dyadic geometry is also supported by Holtmeier’s
observation that galant composers favored outer-voice imperfect consonances (2011).
8
The voice leading (E4, G4, C5) → (D4, A4, E5) can be decomposed into the voice exchange (E4,
G4, C5) → (C4, G4, E5) and the spacing-preserving (C4, G4, E5) → (D4, A4, E5); the latter has both
chords in “open position,” with each voice two steps above its predecessor. When chords are related
by transposition, the spacing-preserving component is a transposition along the chord.
9
To move the root up by octave while keeping the third fixed, we can transpose both notes up
along the chord, sending C up to E and E up to C. We then apply the five-step voice exchange to move
E up to C and C down to E. The result leaves E fixed and in the same voice while raising C by octave
in the other. In other words, we have (C4, E4) → (E4, C5) → (C5, E4), for a composite of (C4, E4) →
(C5, E4).
10
The fifth is the closest diatonic approximation to the half-octave. In a scale with an even number
of notes, such as the chromatic or octatonic, then tritone transposition allows for perfectly balanced
contrapuntal motion, connecting maximally close thirds.
11
Thanks here to Gunnlaugur Björnsson.
12
In this space, the line of unisons acts like a mirror, “reflecting” paths backward into the interior
of the graph. See Tymoczko 2011a, chapter 3.
13
T–5t2 is equivalent to T2t–1 since T7t–3 does nothing and T–5t2 + T7t–3 = T2t–1. Three applications
of T–5t2 produces T–15t6 or T–1 since T–14t6 also does nothing. The descending basic voice leading
turns (E, G, B), or (2, 4, 6) in diatonic note-numbers, into (D, G, B), or (1, 4, 6), decreasing the sum
from 12 to 11 and center of gravity from 4 to 3.66.
14
These voice-leading patterns were the subject of explicit discussion in early figured-bass
treatises; see, for example, the examples on p. 77 of Buelow 1992 or the discussion on pp. 202–3 of
Bach (1762) 1949.
15
Chapter 6 will discuss the application of Roman numerals in an early-music context. These near-
palindromic phrases are fairly common in the Renaissance (e.g., Figure 8.6.11 or the start of
Palestrina’s “Stabat Mater”).
16
Note that the top line of the ukulele jumps above the vocal at the dominant, producing the
familiar effect of overlap.
 17
The Pachelbel/Prinner combination is an exact A2D4 sequence that has been truncated and
repeated: b°–C–G–a–e–F–C–[b°]. Caplin (2015, p. 51) mentions this progression.
18
Free use of small clockwise steps can sometimes be found in functional music, for example the
opening of Mozart’s G minor quintet K.516 (g–f♯°6–f6–C64–E♭ 6–D6), the opening of the Waldstein
(§3.7), or the Chopin excerpts in §9.3.
19
Agmon 1995 and Quinn 2005.
20
Compare Figure 3.4.8 and Figure 3.1.7; see also the first measure of J. S. Bach’s chorale “O
Haupt voll Blut und Wunden” (BWV 244.54, Riemenschneider 74), which uses the same
Pachelbel/Prinner hybrid, starting on 4̂ rather than 6̂.
21
See Buelow 1992 and Bach (1762) 1949. Robert Gauldin (1997) calls my “half open”
configuration the “open/octave” position.
22
For example, imagine we form a scale containing two notes in all possible octaves (i.e., . . . C3,
G3, C4, G4 . . . ). Measured in steps along that scale, DI is 001 or 011 and corresponds to collections
like C3, C3, G3 or G3, G3, C4. In this same scale H is 012 (i.e., C3, G3, C4 or G3, C4, G4) while OO
is 013 (C3, G3, G4) and 023 (C3, C4 G4). For C and O, we form a scale containing three distinct notes
in every octave (i.e., . . . C3, E3, G3, C4, E4, G4 . . . ). Three consecutive notes in that scale form a
close-position triad (C, or 012 in steps along that scale), while three notes such as C3, G3, E4 form an
open-position triad (O or 024).
23
The chord at the end of m. 4 is either half-open or open, depending on how we read the melodic
A4–G4–A4 figure (§5.1).
24
For instance, the descending basic voice leading moves (D, D, A) or (1, 1, 5) to (C, F, F) or (0, 3,
3), changing the chord’s sum from 7 to 6 and the center of gravity from 2.33 to 2.
25
Again, efficient voice leading requires small changes in configuration: when every voice moves
efficiently, the total configuration moves by short distances on the graph.
26
For example, (D4, D4, F4)→ (C4, E4, E4).
27
Readers might notice that this figure is remarkably similar to the graph of two-note consonances
(Figure P3.4), with close position playing the role of thirds and open position playing the role of sixths
—a curious and largely coincidental resonance between two rather different music-theoretical
constructions.
28
Zarlino (1558) 1968, Schubert 2008.
29
A three-voice diatonic consonance contains either zero or two imperfect consonances. If both
chords in the voice leading contain two consonances, then at most one of the three pairs of voices in
each chord contains an interval that is not an imperfect consonance. So the third pair must contain an
imperfect consonance in both chords.
30
As discussed in §5.2, I consider the 64 chord to be harmonic. The phrase also features a
conspicuous point of imitation, with the middle voice picking up the top voice’s “Kyrie” a bar later.
31
A simple computer survey suggests they occur in less than 3% of Renaissance voice leadings.
See also Huron 1991.
32
Huron 2016.
33
When a chord divides the scalar octave exactly evenly, “spacing in chordal steps” is equivalent
to “spacing in scale steps.” As the chord divides the octave less and less evenly, the two notions
diverge.
34
It is interesting to ask whether (C, G) and (C, C, E) should be considered sonorities in their own
right or subsets of a triad; Renaissance and functional composers would likely give different answers.
35
This is comparable to Figure 2.1.8, only now with a wider range of inversions in the bass.
36
For examples of long lists of options, see Kostka and Payne 2003, who borrow from McHose
1947.
 37
For example a DI↔C transition can produce parallel fifths only if moving between a close, root
position triad and a fifth with one note doubled; H↔O transitions can produce parallel fifths only
between adjacent voices, as in (D4, A4, D5) → (C4, G4, E5).
38
Aarden and von Hippel 2004.
39
Note that the bass is doubled if and only if the upper voices are in C or O position.
40
Huron 2016 tries to derive doubling practices from principles of upper-voice spacing, but he
exaggerates the rarity of the DI position; in Palestrina, DI is very common, and in Bach, it is
reasonably common. I describe it as “unusual” only in a pedagogical context.
41
Aarden and von Hippel 2004 contains a good summary of the justifications given for various
doubling rules. Schoenberg ([1911] 1983, p. 59) argues that the doubled third is acoustically defective.
42
Weber (1817–1821) 1846, pp. 190–92.
43
A fact that first came to my attention a few years back, when I awoke from a deep sleep, sat bolt
upright in the middle of the night, and said to myself, “Holy cow, how did I never notice the ascending
thirds in the Waldstein?” Whereupon my waking self had to reconstruct this analysis to figure out what
I meant.
44
Autoharps play triads, which are the collections relevant to this discussion, but provide no voice
leading from one chord to the next; harps supply voice leadings via their pedals, but sound seven-note
collections.
45
Previous theorists have tried to translate Schenker’s metaphors of “surface” and “depth” into
perceptual terms, postulating that humans naturally experience harmony in hierarchical fashion
(Lerdahl and Jackendoff 1983, Lerdahl 2001, and Rohrmeier 2020).
46
Milton Babbitt (1987, p. 148) emphasized that the opening of Tristan spells out a “Tristan” chord
in its upper voice (G ♯ –B–D–F ♯ on Figure 3.7.5). I see this as the byproduct of the same general
relationships that produce the ascending thirds in the Waldstein.
47
Schubert 2018 proposes a similar analysis.
48
Erno Lendvai (1971) argues that major or minor chords have the same function when their roots
belong to the same diminished seventh, at least in Bartók’s music. I find this somewhat plausible in
the case of dominant-functioning chords, less so in the case of tonic and subdominant.
49
See, for example, m. 12 of the second E major fugue from The Well-Tempered Clavier or mm.
19, 31, and 71 of the second F major fugue.
50
The name is meant to emphasize their relation to the common-tone diminished seventh, not the
presence of common tones in the actual progression. Clendinning and Marvin 2016 use the term
“common-tone German sixth” to describe one of these progressions.
51
Other examples can be found in Haydn’s Symphony no. 100 (“Military”), I, mm. 195–197 (VI65
→V ), Beethoven’s Symphony no. 2, I, mm. 40–43 (V65/V→vii°7/V→V43/IV), and Kabalevsky’s
7

“Sonatina” (Op. 27, no. 18, mm. 33–36, II2→I).

52
This chord originates in the baroque minor-mode use of III+maj7 as a dominant, which sometimes
progresses deceptively to VI rather than i (e.g., the second C minor prelude in The Well-Tempered
Clavier, mm. 3–4 and 24). This deceptive III–VI then becomes an authentic major-mode V–I in the
nineteenth century (e.g., “Estrella” from Schumann’s Carnaval, mm. 5–6 and again in m. 7). The
major triad on VII appears in early rock-and-roll as a semitonal lower neighbor (“Jailhouse Rock”); it
is interesting to try to hear a connection to these earlier idioms.
53
Other examples include the endings of the first D major and second G major preludes in The
Well-Tempered Clavier, m. 80 of the WTC’s first A minor fugue, chorales such as BWV 40.3
(Riemenschneider 43, m. 14), 309 (R166, m. 5), 60.5 (R216, m. 5), 272 (R340, m. 3), and 248.42
(R360, m. 10), Mozart’s Clarinet Concerto K.622, I, mm. 141ff, Beethoven’s Op. 10, no. 3, II, mm. 5–
8 (which interposes a dominant seventh between diminished sevenths), and Figures 10.3.4 and
10.4.11.
54
I hear these diminished sevenths on 4̂ as having a dominant function which is then thwarted by
the ascending-semitone resolution.
55
See, for example, Grieg’s “Salon” (Op. 65, no. 4).
56
See Okazaki 2015, p. 98.
57
Thanks to Zhoushu Ziporyn for this example. Cohn 2012, p. 148, discusses Wagner’s use of this
progression in Parsifal.
58
The first Beatles song to use II–IV is “She Loves You,” presenting it as both a dominant
approach and as a plagal progression in its own right; it was almost certainly influenced by Rydell’s
song. See Everett and Riley 2019, pp. 64–65.
59
On modal mixture, see Tymoczko 2011a, §6.6; Rydell’s idiom appears in Figure 9.3.1.
60
The IV–♭VI–I progression in Otis Redding’s 1965 “I’ve Been Loving You Too Long” is perhaps
comparable, a triadic analogue of IV–iv7–I. The I–II–IV progression also appears at the end of
Mendelssohn’s Op. 120, no. 1, first as part of a “fauxbourdon ROTO” idiom (§7.5) and then in root
position. The bass voice converts more-or-less idiomatic music to something more unusual.
61
Headlam 2012 and Yust 2018, chapter 10. For the term “compatibilism” see Tymoczko 2011a,
chapter 7.
62
See Cohn 2012, pp. 199–203. Cohn’s book is largely concerned with chromatic triads, focusing
on an unusually retrofunctional environment. This leads him to describe the progression in Figure
3.9.2 as “feign[ing] a red-herring tonality” (2012, p. 148) rather than exemplifying the compatibility of
tetrachords, descending semitonal voice leading, and functional progressions.
63
See §9.3. As Cohn notes, a range of figures from Dahlhaus (1967) to Lerdahl (2001) to Charles
Smith (1986) advocate for a more unified perspective.
 Prelude
Sequence and Function

The last movement of Beethoven’s Pathétique contains a brief tutorial in the

many varieties of repetition. Its A ♭ -major episode opens with a canonic
sequence whose melodic intervals move from voice to voice, the
counterpoint almost entirely generated by “antiparallel” voice leading in
which one voice ascends by fourth and the other descends by fifth, with the
motions switching hands after every dyad. Despite this alternation of
intervals, the left hand always sounds the root and the right hand the third.1
Figure P4.1 graphs the passage on the 2-in-7 diagram, where it moves
clockwise at each turn, combining the descending basic voice leading with
alternating two- and five-step voice exchanges. Zooming out to the level of
the bar we find a transpositional sequence that moves each voice down by
step.
Figure P4.1. Beethoven’s Pathétique sonata, III, mm. 79–86, in notation and on the dyadic circle.
Dotted and solid curved arrows represent the five- and two-step voice exchanges respectively, each
occurring before the radial motion.

Since the diatonic scale has an odd number of notes, the voices exchange
intervals when they return to the tonic: where the right hand initially
ascended from C by fourth, it now descends from C by fifth, and where the
left descended from A♭ by fifth, it now ascends from A♭ by fourth (Figure
P4.2). The humor lies in the impression of invertible counterpoint: in the
consequent, each hand plays the intervals of the antecedent’s other hand, but
on the same notes as before. To my ear, contour and rhythm trump pitch-
class content, and the theme seems to switch hands. (The quarter-note delay
contributes to this impression by directing our ears to the left hand.) The
passage manages to sound trivial and brilliant at once, a paradoxical effect of
“counterpoint” that is hardly worth the name.
Figure P4.2. The opening phrase as a single sequence.

On the brilliant side of the ledger, the final left-hand fourth is decorated
by stepwise contrary motion, producing Gjerdingen’s “converging cadence.”
In the opening, this forms a half cadence, tonicizing E♭ while reinforcing its
status as dominant; it also serves the rhythmic function of filling four
measures with seven chords. In the consequent the voices start converging
one chord earlier, on F rather than B ♭ —a result of the hands having
exchanged melodic intervals. (Rhythmic displacement conceals this, the
consequent’s left hand only one quarter-note ahead of the antecedent.) This
creates a stronger sense of B♭-as-dominant, and hence a stronger tonicization

of the final E♭, reinforced by the melodic E♭. This is a nice example of the
power of schematic vocabulary: the two phrases are nearly identical,
descending by fifth from A♭ to E♭, yet because of the different positions of
the converging cadence they manage to form a period, moving from a half
cadence to a full cadence on the dominant. At which point the sequence has
returned to its original state so that the repeat follows organically—meaning
that the restatement is also a continuation of the underlying process. The
result is almost deconstructive in its simplicity, the nearly pure unfolding of
an algorithm that generates invertible counterpoint, a canon, a tonally
functional period, a half cadence, a modulation to the dominant, and the
conventionally mandated repeat. It is simultaneously funny and profound.
 The four-bar middle section, meanwhile, presents a near sequence that
states a brief melody on E♭, D♭, and A♭.2 Figure P4.3 shows that the motive
outlines an octave-displaced version of the diatonic third’s ascending basic
voice leading. The circular pattern of roots E ♭ –B ♭ –D ♭ –A ♭ –E ♭ outlines a

reordered fifth sequence D♭–A♭–E♭–B♭–(D♭) perturbed by the insertion of a

minor third (§2.5). Here the thirds on B♭ and D♭ are superimposed to form a

B♭ minor triad, producing the three-chord sequence E♭–b♭–A♭–E♭. (We can

also imagine the basic voice leading displaced by triadic step, appearing as
third-fifth of B ♭ minor rather than root-third.) This is a nice example of
triadic music’s ability to juggle two- and three-note logics, embedding a
dyadic sequence in nonsequential triads: E ♭ –b ♭ –A ♭ –E ♭ , the unusual root
progressions normalized by tonicization. This music reverses the descending
fifths of the opening measures, moving upward along the ladder of fifths,
though the perturbation ensures that we end just where we began.
Figure P4.3. The contrasting middle section, with the diatonic third’s ascending basic voice leading
stated successively on E♭, D♭ (the upper third of B♭ minor), and A♭.

The final measures present a one-bar contrary-motion sequence that

moves the hands an octave farther apart with each iteration; this is
accomplished by altering the opening music so that one hand consistently
falls by fifth rather than alternating fifths and fourths (Figure P4.4). This
one-bar sequence is embedded inside a four-bar canonic sequence, with the
descending scalar eighths switching hands from left to right. The larger
sequence is not a contrary-motion sequence since one hand stays fixed (C6
to C6) while the other descends by three octaves (A♭5 to A♭2); a complete
second statement would therefore move the right hand down by three
octaves while keeping the left hand fixed, returning the hands to their
original configuration. But the four-measure pattern breaks off in favor of a
standing-on-the-dominant in C minor, preparing the main theme’s return.
Figure P4.4. The episode’s final section features a one-measure contrary-motion sequential unit
embedded inside a four-measure canonic sequential unit.

Beethoven’s rounded-binary episode is thus composed of three sequences

in close succession, a canonic sequence, a near sequence, and a contrary-
motion sequence, all presented with insistent simplicity, the music wearing
its structure on its sleeve. Its omnipresent repetition illustrates Beethoven’s
love of patterns while hinting at a conception of functional harmony as
inherently repetitive, its patterns disguised, distorted, and blurred to the point
where they are no longer obvious. Chapter 4 will provide a comprehensive
framework for analyzing this sort of transformed repetition, considering its
manifestations in Renaissance, classical, and modern music.
If we sometimes underestimate this sort of repetition, that is because we
tend to conceive counterpoint negatively, in terms of rules and prohibitions,
rather than positively, in terms of possibilities and affordances. It is just this
habit that led Charles Rosen to claim that writing counterpoint is “not after
all a very difficult craft.”3 Rosen was right that pretty much anyone can
follow rules, avoiding forbidden parallels and awkward leaps and so forth.
But at the same time genuine contrapuntal understanding can be quite
difficult to obtain.4 Beethoven’s deliberate simplicity is compelling for
reasons that have nothing to do with schoolbook rules; instead, it presents a
kind of embodied mathematics, drawing musical consequences from a small
number of postulates. For me, this embodied aspect is essential to its beauty:
this is a mathematics that grows out of practice rather than being imposed
upon it by external fiat. Its theorems are contrapuntal schemas, embedded in
works of art and untheorized even to this day. Understanding this music is
difficult and creating it doubly so, just as living a meaningful life requires
more than avoiding arrest.

1
See Caplin 1998, chapter 16, for more on these rounded-binary interior episodes. Steve Taylor
observed that this passage echoes the consequent of the slow-movement opening theme; it is also very
similar to the third variation of Op. 1, no. 3, III (discussed in §4.10), and the second theme of Op. 18,
no. 5, IV (mm. 36ff).
2
It is also a microscopic version of what we will call, in chapter 8, the up-and-down-the-ladder
schema.
3
Rosen 1998, p. 545.
4
C. P. E. Bach advertised his father’s “original thought” in the contrapuntal domain (Wolff 2001, p.
8). One component of this originality is Bach’s distinctive approach to the diatonic third’s ascending
basic voice leading (§3.1, 4.6, 7.7).
 4
Repetition

Voice leadings are atomic musical moves that can be combined arbitrarily.
One of the central discoveries of the Renaissance was that they could be
repeated, chained together algorithmically to convert an isolated event into
an ongoing process. Repeating contrapuntal patterns are an important feature
of sixteenth-century music, where they are often subtle and disguised (e.g.,
the ascending fifths in Figure 3.6.6, whose sequential structure is not
immediately obvious).1 This disguised repetition evolves into the overtly
repetitive sequences of the baroque, which embed repeating contrapuntal
patterns into larger musical blocks. Nineteenth-century composers make
increasing use of contrary-motion “wedge” sequences, sometimes nontonal,
paving the way for the very flexible transformations of modern music—a
style that permits almost any conceivable transformation at any level of the
quadruple hierarchy. Appendix 3 considers this practice as an extension and
generalization of the traditional sequences that are the focus of this chapter.

1. Repetition reimagined

If we want to repeat a passage, the most obvious thing to do is play it

unchanged. Less obviously we can transform it so as to preserve some
essential features of its structure. The most familiar structure-preserving
transformations are transpositions moving the music through pitch and
permutations moving it through the instrumental ensemble; these are the
basis of sequences and rounds respectively (Figure 4.1.1). Combining
transposition and permutation produces a transposing round where the lines
move through both pitch and ensemble, as in the third passage of Figure
4.1.1. While transposition and permutation are not the only possible
transformations, they are central to a great deal of music and will occupy us
for the rest of this chapter.2
Figure 4.1.1. In repeating a passage, we can transpose it to form a sequence, permute its voices to
form a round, or do both to form a transposing round.

A general model of repetition should therefore combine three elements, a

“cell” or “unit” consisting in a set of melodic lines to be repeated, a
collection of transpositions applied at each repeat, and a permutation or
“twist” moving the lines among the available voices. Figure 4.1.2 adapts a
notation originating with the Russian theorist Sergei Taneyev, using arrows
to generate new sequential units from their predecessors.3 Each arrow points
from source to destination and is labeled with a transposition: thus, whatever
music is in the top voice is transposed by x and placed in the second voice,
and whatever music is in the second voice is transposed by y and is sent to
the top; the music in the third voice is transposed by z and remains in that
voice. In most sequences, the arrows form a stable structure that stays fixed
as the music repeats. In a transpositional sequence, all the transpositions
have the same subscript, and arrows do not cross: each voice retains its own
melodic material and is transposed in exactly the same way. Crossed arrows
create canonic sequences, partitioning the voices into groups sharing the
same melodic content. A three-voice sequence can either have no crossings
and hence no canon, a pair of crossed arrows forming a two-voice canon, or
all three arrows crossing to form a three-voice canon (Figure 4.1.3).

Figure 4.1.2. An abstract model of repetition in which a cell’s melodic lines are transposed and
permuted as they repeat.
Figure 4.1.3. Three examples of the schema. (top) An ascending-step sequence from Marianna
Martines’s “Vo solcando un mar crudele” (from Scelta d’Arie Composte per suo diletto, 1767), mm.
122ff. (middle) “Three Blinde Mice” (1609) uses permutation but not transposition. (bottom) A
sequence with both transposition and permutation from the first G♯ minor fugue in The Well-Tempered
Clavier, mm. 86ff.
 This model resolves some longstanding difficulties surrounding the
identification of repeating sequential units. We can distinguish the simple
period, which is the length of the minimal block of music sufficient to
generate the sequence, or the cells as I have defined them, from the grand
period, which is the time it takes for all voices to cycle back to their initial
melodic positions, all transposed by the same amount (not counting octave
displacements that may differ from voice to voice). Previous theorists have
often focused on the grand period to the exclusion of the simple period; in
large part this is because they have neglected permutation as a sequence-
generating operation.4 In the presence of permutation, however, the grand
period will contain multiple simple periods: the sequences in Figure 4.1.3 all
have a simple period of one bar, but it takes three bars for “Three Blinde
Mice” to cycle back to its original melodic configuration and two bars for
the G♯ minor fugue to do so. In some cases, different groups of voices will
have different simple periods (e.g., Figure 4.5.7).
Sequences can be stable or unstable in two different ways. A collection of
transposing arrows is harmonically stable when its transpositions differ only
by octave. These sequences are structure-preserving in a very general way,
producing transpositionally related blocks regardless of their cells’ content.
Harmonically unstable sequences create contrary-motion “wedge” patterns
that move through unrelated vertical states (Figure 4.1.4). Here again the
grand period will typically contain multiple simple periods, the voices
cycling through different transpositions until they return to their initial
alignment. In earlier music, unstable sequences often exploit unique features
of the sequential unit to create a consonant result. This typically involves the
sort of invertible counterpoint found in Figure 4.1.5.5
Figure 4.1.4. Schoenberg’s “Angst und Hoffen” (The Book of the Hanging Gardens, Op. 15, no. 7)
contains a wedge sequence that transposes the voices by different amounts.

Figure 4.1.5. An invertible-counterpoint sequence in mm. 57–60 of the second E minor prelude in The
Well-Tempered Clavier. The sequence uses invertible counterpoint at the twelfth (mm. 1–2).

Registrally stable sequences preserve distance between voices from one

grand period to the next, while contrary-motion sequences alter those
distances by one or more octaves. A collection of canonic voices, whose
parts are interlinked by arrows, is always registrally stable, as each line
passes through the same sequence of transpositions with each grand period.
In the bottom passage of Figure 4.1.3, for example, the music in the top
voice moves down four steps and then up three steps while the music in the
middle voice moves up three steps and down four; in both cases the total is –
1 and the lines transpose down by diatonic step. The bass voice, meanwhile,
moves down by eight diatonic steps with each grand period, forming a
contrary-motion sequence. Note that my term “contrary motion”
encompasses both oblique and similar motion from one grand period to the
next—as in this last example, where the top voices descend by step while the
bass descends by a ninth.
Readers are encouraged to be patient, as it can be genuinely difficult to
wrap one’s mind around the full range of effects that can be obtained from
this simple model: while working on this book I have been repeatedly
surprised to discover that yet another familiar passage could be described
with transposing arrows. I take some comfort in the thought that the
difficulty is not so much theoretical as it is musical—it is not that I am
overcomplicating something simple, but rather that composers have made
extremely sophisticated use of transformed repetition.6 To be sure, I am
trying to understand this practice abstractly, replacing particular schemas
with a general framework. (And to be fair I sometimes exercise the analyst’s
privilege of asserting permutations not found on the musical surface—as in
§2.5, where I identify canons created by the listener’s ear rather than the
composer’s pencil.) But the real issue here is the wide range of effects that
can be obtained by simply doing the same thing over and over.

2. Repeating contrapuntal patterns

Look at the left side of Figure 4.2.1 while covering the right side with your
hand. If you wanted to repeat this voice leading, how would you do it?
Figure 4.2.1. Two ways to repeat the voice leading on the left.

Rather than a single correct answer there are two possibilities: continuing
the melodic motions within each voice to form a contrary-motion wedge, or
permuting the melodic intervals so that they move from voice to voice.
Either yields a sequence with a repeating cell that is just one note long.7 I
will call these one-note sequences repeating contrapuntal patterns.
Repeating contrapuntal patterns can be associated with phrases like “voice X
moves up by 3 semitones and then this melodic interval moves to voice Y;
voice Y moves down by one semitone and then this motion shifts to voice
X.” We can think of them, loosely, as combining two separate voice
leadings, one that determines the melodic intervals (“move voice X up by
three semitones”) and another that permutes these melodic intervals among
the voices (“its motion moves to voice Y”).8 Figure 4.2.2 contains repeating
contrapuntal patterns from Josquin, Beethoven, and Stravinsky.
Figure 4.2.2. Three sequences whose units have just a single chord. (left) Josquin, from the Sanctus of
the mass L’Ami Baudichon, mm. 14ff; (middle) the Beethoven passage in Figure P4.1; (right) one of
the main motives of Stravinsky’s Firebird.

There is something potentially confusing about the notion of a single-note

sequential unit. With larger units, the lines’ internal structure allows us to
draw transpositional arrows on the basis of just two iterations: looking back
at Figure 4.1.3, it would be easy to reconstruct the arrows were they to
disappear. With a repeating contrapuntal pattern, however, we need at least
three chords to identify the sequence. After all, the left side of Figure 4.2.1
contains two chords, yet permits two equally good continuations. This is
because repeating contrapuntal patterns are characterized by the melodic
intervals that occur between units—in Figure 4.2.1, the ascending minor
third and the descending semitone. It takes two notes to identify those
intervals, and three for us to observe them moving around the texture. When
analyzing repeating contrapuntal patterns, I will therefore include each
voice’s initial melodic interval in parentheses, measured in diatonic or
chromatic steps as appropriate; these intervals move around the texture
following the arrows. This interval-and-arrow notation is slightly more
general than the transposition-and-arrow notation, as it is insensitive to the
notes’ initial registral positions (Figure 4.2.3). For any initial registral
position, however, melodic intervals and transpositional labels carry the
same information, and we can compute one from the other.
Figure 4.2.3. When voices shift octaves, the melodic-interval notation stays the same while the
transpositional notation changes.

I have said that a sequence is harmonically stable if its transposition

arrows differ only by some number of octaves. For a repeating contrapuntal
pattern, this means, first, that it connects transpositionally related chords;
and second that the melodic intervals are linked to specific chordal elements
(e.g., root, third, and fifth). Thus we can associate harmonically stable
repeating contrapuntal patterns with phrases like “shift the root and third
down by step while keeping the fifth constant.” Any such sequence will
necessarily correspond to a type of motion in one of our spiral diagrams.
For we have seen that motion on the spiral diagrams can be described by
abstract instructions for moving chordal elements: for instance, in the 2-in-7
graph of imperfect consonance (Figure 3.1.1), every clockwise step moves
the root down by third and the third up by step (T–4t1, the descending basic
voice leading). It follows that a harmonically stable repeating contrapuntal
pattern is just a repeated pattern of motion through a spiral diagram. We can
highlight this connection with a change in notation: rather than using scalar
transpositional arrows to link notes across voices, we can instead use
hierarchically nested transpositions to link notes within voices, as in Figure
4.2.4. This clarifies that little-t transposition is inherently canonic, cycling
voices through different chordal elements. Indeed, even a trivial passage like
Figure P2.6 has an imitative structure, its three voices ascending C–D–A in a
one-beat canon.
Figure 4.2.4. Two interpretations of the same repeating contrapuntal pattern. The first uses diatonic
transpositions to connect notes in different voices; the second uses double transposition along chord
and scale to relate notes within voices.

The upshot is that there are deep connections between the seemingly
distinct domains of repetition, canon, and voice leading. These are reflected
by two different ways we can conceive of repeated moves on a spiral
diagram: as harmonic sequences and as generators of canons. This
connection is in the first instance theoretically surprising, as one might not
expect that repetition had anything to do with canon. It is also interesting
historically, as it is likely that the canonic aspect was initially paramount: in
a passage like Figure 4.2.5, Palestrina was presumably more concerned with
imitation than the ascending-fifth pattern of harmonies.9 As time passed,
however, these moments of close imitation perhaps came to be valued for
their harmonic patterning as well; thus when J. S. Bach or Mozart make
sequential use of the diatonic third’s basic voice leading, its harmonic aspect
is ascendant (Figure 4.3.2). In this way, a concern with counterpoint and
imitation could gradually yield to a focus on harmonic repetition without any
clear line dividing the two. Chapter 6 will argue that repeating contrapuntal
patterns could also spark awareness of abstract structural concepts such as
“root” and “fifth,” thus setting the stage for inversional equivalence.

Figure 4.2.5. The ascending basic voice leading in the top two voices of the Sanctus of Palestrina’s
mass Ave regina coelorum, mm. 18–20.

In large-unit sequences, repeating contrapuntal patterns connect analogous

chords in adjacent units. For example, the expositional sequence of The
Well-Tempered Clavier’s first E minor fugue articulates the diatonic third’s
ascending basic voice leading across units (Figure 4.2.6). The repeating core
of Ligeti’s Passacaglia Ungherese is structurally if not expressively similar,
a series of major thirds and minor sixths that compose to form a one-step
transposition along the chord (Figure 4.2.7). (Note that one will not
necessarily find the same voice-leading relationship between all analogous
pairs of chords in a large-unit sequence: changes of spacing within the unit
can change the Txty relationship between analogous chords, as in Figure
4.2.8.) This explains why we can describe sequences either using the crossed
Tx arrows of §4.1, or by identifying the voice leading connecting analogous
chords in successive units, as with the uncrossed arrows of Figure 4.2.4;
when considering large sequential units we typically want to use transposing
arrows, but when concerned with repeating contrapuntal patterns, we may
prefer the latter.10 In either case, there is a two-way mapping between
sequences and voice leadings: every sequence determines a repeating
contrapuntal pattern between analogous chords of successive units; and
every voice leading can be expanded into a larger-unit sequence. The theory
of voice leading and the theory of sequence are one and the same.

Figure 4.2.6. The exposition of the first E minor fugue in The Well-Tempered Clavier (mm. 11–12) is
a sequence with two-measure units, whose initial notes articulate the basic voice leading from one unit
to the next.

Figure 4.2.7. The round in Ligeti’s Passacaglia ungherese makes a complete circle in the space of
chromatic major thirds.
Figure 4.2.8. In a large-unit sequence, not all chords need be related by the same voice leading. Here,
the first chords in each unit relate by T–4t2, while the second add the complicated voice exchange
crtcrfcrtctf, with crt the pairwise voice exchange of root and third, crf root and fifth, and ctf third and
fifth.

Indeed, there is a general recipe for constructing a sequence from a voice

leading between transpositionally related chords: starting with any such
voice leading, draw transpositional arrows between the notes, and then use
those arrows to form either a repeating contrapuntal pattern or a larger-unit
sequence containing it. Figure 4.2.9 goes through the process for a voice
leading using nontertian tetrachords. It is somewhat surprising that
contemporary musicians do not learn this recipe, for it is a useful technique
that could easily be incorporated into the university theory curriculum—one
that allows us to explore the full spectrum of transposed and permuted
repetition. Its absence reflects our tendency to think about sequences
schematically, as a set of idioms characteristic of functional tonality, rather
than as manifestations of a general transformational technique.
Figure 4.2.9. A recipe for converting a voice leading into a sequence. We start with a voice leading
connecting chromatic 0457 chords, or major triads with added fourth. The grand period is three chords
long.

3. The geometry of two-voice sequences

The isomorphism between sequence and voice leading allows us to use the
spiral diagrams to explore the geometry of sequential possibility. For
suppose we have a two-voice, harmonically consistent repeating
contrapuntal pattern: if it has no voice exchanges then it is either canonic or
transpositional, depending on whether it takes an odd or even number of
steps on the dyadic spiral; if it has voice exchanges, then it is either canonic
or contrary-motion, again depending on how many steps it takes (Figure
4.3.1).11 Readers will recall that these sequence-types—transpositional,
contrary-motion, and canonic—were all discussed in this chapter’s prelude.
The spiral diagrams show that they are the only harmonically consistent two-
voice possibilities. Geometry also justifies our decision to include
permutation alongside transposition as a sequence-generating operation:
after all, any harmonically stable voice leading generates a sequence, and
some of these permute their voices.

Figure 4.3.1. Sequences generated by repeated clockwise motion on the dyadic circle. The top line
does not involve any voice exchanges while the bottom line incorporates the two-step voice exchange
(symbolized by the asterisk; see the dotted curves on Figure 3.2.1). This turns parallel sequences into
canonic sequences and canonic sequences into contrary sequences.

With no voice exchanges, canonic sequences take an odd number of steps

on the dyadic circle and transpositional sequences take an even number of
steps. (Refer back to Figure 3.1.4, where the top-line voice leadings generate
canonic sequences, with each voice alternating root and third, while the
bottom-line voice leadings generate parallel sequences.) The most common
pattern is just single-step motion on the spiral, producing the diatonic third’s
basic voice leading; this schema appears in a substantial proportion of
baroque and classical sequences, embellished lightly if at all. Classical
sequences usually feature descending fifths, whereas earlier styles more
frequently use ascending fifths, often embellished with passing tritones as
described in §3.1 (Figures 4.3.2–3).12 We can find three-step motion on the
circle as well (e.g., Figure 4.5.4).13 Larger motions, such as five or seven
steps, make for awkward melodic leaps and are correspondingly rare.
Meanwhile, sequences that take an even number of steps on the circle are
transpositional; these typically occur in explicitly parallel passages (e.g.,
fauxbourdon) or else at the beginning of longer sequential units.
Figure 4.3.2. The development sequence of Mozart’s K.545, I, mm. 37–38, and a sequence from the
first C minor fugue in The Well-Tempered Clavier, mm. 9–10.
Figure 4.3.3. The diatonic third’s ascending basic voice leading in the second E minor prelude of The
Well-Tempered Clavier, mm. 23–28.

For sequences with an odd number of voice exchanges, canons result from
an even number of steps on the dyadic circle, while antiparallel motion
results from an odd number of steps (bottom staff of Figure 4.3.1). The most
common sequences use the two-step voice exchange: on its own it produces
a simple round; combined with one-step clockwise motion it produces
antiparallel motion in which one voice moves by fourth and the other moves
by fifth in the opposite direction. The combination of stepwise motion with
the two-step voice exchange is also reasonably common, particularly in its
ascending form (e.g., Figure 7.6.10). For a sequence with an even number of
voice exchanges, the situation is reversed: an odd number of steps produces
a canonic sequence, and an even number produces antiparallel motion.
Sequences with larger motions on the dyadic circle tend to be rare as they
involve more unbalanced motion between the two voices.14
We now turn to harmonically unstable sequences. Any two-voice canonic
repeating contrapuntal pattern, whether harmonically stable or not, will
preserve its canonic structure when either of its dyads is moved by exact
contrary motion.15 This means that any such pattern can be generated by
applying contrary-motion distortion to either parallel motion or Pathétique-
style alternating antiparallel intervals, depending on the interval of
transposition as measured from one grand period to the next (Figure 4.3.4).16
For a particular interval, the various possibilities will be represented by a
matrix such as that in Figure 4.3.5. The most common harmonically unstable
canon is the dyadic 5–6; it can be embedded into a triadic sequence as root-
fifth, fifth-third, or both simultaneously—in which case we have three
canonic voices, one sounding third and fifth, another fifth and root, and
another root and third (Figures 4.3.6–7).17 This structure can be transformed
via dyadic expansion into a 5–10 sequence (e.g., the outer voices on the right
side of Figure 2.9.3). The music in Figure 4.3.8 makes sophisticated use of
these possibilities: my reconstruction begins in the top line with an
antiparallel ascending-fifth sequence; every other chord is then subjected to
contrary-motion contraction, the top three voices lowered by step while the
bass rises by the same amount; finally we apply “third substitution” to the
last three chords, lowering the bass by third and raising the fifth of the
upper-voice triad. This produces a familiar harmonic loop, four descending
fifths and one third, b–e–A–D (§2.5).
Figure 4.3.4. Two-voice canonic sequences derived from either parallel or antiparallel motion by
expanding or contracting every other chord. This produces sequences with even or odd intervals of
transposition respectively. On the bottom left, a pattern from m. 56 of the second B♭ minor fugue in
The Well-Tempered Clavier.
Figure 4.3.5. The matrix of two-voice, two-chord canonic patterns repeating down by step. The
examples can be read right-to-left to produce ascending-step patterns.

Figure 4.3.6. The 5–6 voice-leading pattern is an example of invertible counterpoint at the tenth, with
the G–A voice moving down by fifth to C–D while the C–C voice moves up by sixth to A–A. This is
equivalent to transposing G–A down by tenth and transposing the whole pattern up by sixth. This
figure can be embedded in a triadic 5–6 sequence in two separate ways.
Figure 4.3.7. Two ascending-step sequences from The Well-Tempered Clavier, each involving a
canonic relation among two voices. The top system is a reduction of mm. 13–19 of the second C major
fugue, and has the canon between the third-fifth and fifth-root voices; the bottom is from the second
F♯ minor fugue, mm. 47–48, and has its canon between the root-third and fifth-root voices.
Figure 4.3.8. Barbara Strozzi’s “Il Contrasto dei Cinque Sensi,” mm. 3–4. The top staff presents an
antiparallel fifth sequence; on the second staff, the voices of every other chord are moved together by
step; the third staff applies third substitution to the final three chords, lowering the bass by third and
raising the alto by step.

The interesting conceptual point is that canon and contrary motion arise as
byproducts of the basic mechanics of sequential voice leading, rather than
requiring deliberate compositional artifice: they are outcroppings that the
rock-climber may choose to grasp or not. The presence or absence of
canonic patterning is thus a matter of whether composers decide to draw out
the canonic potential latent in their material, with noncanonic sequences
often obscuring preexisting canonic relationships (Figure 4.3.9).18 In the
case of contrary motion, the handhold is less commonly grasped, though
there are circumstances where it is analytically useful to be aware of it: a
nice example comes from the opening of Beethoven’s Tempest sonata, a
sequence that could form a contrary-motion wedge, though Beethoven
chooses not to (Figure 10.3.2).19 These facts allow us to unpack metaphors
like “the will of the tones,” connecting low-level musical affordances to
high-level compositional devices like canon and contrary-motion.20 An
intuitive sense for this connection lies behind a good deal of compositional
expertise.

Figure 4.3.9. Measures 9–12 of the trio of Maria Szymanowska’s Polonaise no. 1, from 18 Dances of
Different Genres. The figuration disguises rather than highlighting the canon inherent in the voice
leading.

All of this has an obvious resonance with Schenkerian theory. Schenker

famously denied that sequences exist, preferring instead to speak of linear
motion.21 But what are these linear motions if not repeating contrapuntal
patterns? Schenker was certainly right that it can be useful to focus on the
voice leading connecting chords at the start of successive sequential units,
for if we do, we will almost always find common patterns that, when iterated
enough times, transpose each voice by the same amount—which is to say,
linear motion. (Indeed, it is definitionally true that the chords at the start of
each grand period are always related by transposition.) This reveals an
important mathematical link between the local voice leadings connecting
adjacent chords, and nonlocal relationships occurring across larger
timespans. It is all too easy to pass from this mathematical truth to the
musical conclusion that sequences are essentially linear devices with no
harmonic significance, or that a sequential unit represents its initial chord, or
perhaps even that “sequences” (as distinct from linear motion more
generally) do not exist. While I do not subscribe to these views, I suspect
they are based on genuine insight. Perhaps Schenker looked at countless
sequences and noticed that their initial chords frequently spelled out familiar
contrapuntal moves that often occur on the musical surface. If he drew
questionable conclusions from this observation, then that may be because he
lacked the mathematical tools needed to unravel this extremely intricate
knotting of harmonic and contrapuntal forces.

4. Three voices and the circle of triads

With larger chords, the subject of voice exchanges becomes much more
complex. Setting them aside for the moment, we can repeat the preceding
analysis: for an n-note chord, an n-step motion on the spiral diagram, or
motion by any integer multiple of n, will produce a transpositional sequence
in which all voices move in parallel; any other motion will produce canonic
sequences. Figure 4.4.1 generates three-voice sequences by taking one, two,
four, five, and seven clockwise steps on the triadic circle. The resulting
schemas are all very common. Because the triad divides the octave very
evenly, short motions on the triadic circle produce little melodic activity and
the canonic imitation is very subtle.
Figure 4.4.1. Repeating contrapuntal patterns generated by short clockwise motion along the triadic
circle. Each column is related by the Tinctoris Transform. The –2 and –4 patterns contain descending
three-note scale fragments, while the –5 and –7 patterns contain descending triadic arpeggios. The
examples can be read backwards to produce ascending sequences (+1, +2, etc.).

The five- and seven-step sequences, by contrast, create the clear

impression of canon. Five-step motion produces a fifth-sequence whose lines
sound an ascending triadic arpeggio, typically filled out with passing tones.
Figure 4.4.2 shows the pattern over an independent bass, its third entry
divided between alto and tenor. Figure 4.4.3 shows a sequence from The
Well-Tempered Clavier’s second C♯ minor fugue; this is the same idiom in
its descending form, disguised by a suspension and incorporated into a
longer sequential unit. Seven-step motion embeds the triadic arpeggio within
a round that moves by chordal step at each repeat. In “L’amante modesto,”
Barbara Strozzi uses this sequence to create a bravura expansion whose aural
complexity belies its straightforward structure (Figure 4.4.4). The technique
is reminiscent of the “Shepard-tone passacaglias” in §2.5, only ascending
rather than descending, and directly embodied in the score rather than
constructed by the listener.

Figure 4.4.2. An ascending-fifth canonic sequence from Barbara Strozzi’s “Gli amanti falliti,” mm.
32–33. This uses the +5 sequence, the retrograde of the –5 pattern in the previous example.
Figure 4.4.3. A sequence from the second C♯ minor fugue in The Well-Tempered Clavier, mm. 10ff.
The sequence uses the descending-fifth canonic voice leading labeled –5 on Figure 4.4.1.

Figure 4.4.4. The opening of Barbara Strozzi’s “L’amante modesto” uses the +7 sequence, the
retrograde of –7 on Figure 4.4.1.

Remarkably, these two patterns can be combined—each new triadic

arpeggio either a fifth above or at the same pitch level as the previous entry
(Figure 4.4.5). This combined pattern appears from Josquin to Beethoven,
often at moments of expressive urgency; for reasons that will become clear
at the end of chapter 6, I call it the Chase schema. Figure 4.4.6 shows
Orlando Gibbons using it to generate a quasi-sequential pattern of ascending
fifths F–C–g–d. The unison entries function to slow the harmonic rhythm, to
avoid second-inversion triads, and to extend the melodic lines into longer
scales. Figures 4.4.7–8 show two of the countless examples found
throughout the literature.22 The Chase schema is both sequence and canon,
appearing in virtually every Renaissance composer’s music and surviving
well into the nineteenth century. From the standpoint of the Prime Directive,
it is essentially similar to the two-voice canon generated by the diatonic
third’s basic voice leading.

Figure 4.4.5. (left) Each new entry can appear either a fifth above the previous entry, or at the same
pitch level. (right) Using repeated entries to avoid second-inversion triads.

Figure 4.4.6. The Chase schema in Orlando Gibbons’s In Nomine a 4, mm. 17–19.
Figure 4.4.7. In mm. 190ff of Josquin’s “Benedicite, omnia opera,” we have a descending-fifth
sequence featuring two superimposed instances of the diatonic third’s descending basic voice leading,
with all four voices in canon. After a brief ascending-step sequence, where the upper voices continue
to inhabit the circle of imperfect consonances, we have seven entries of a canonic sequence based on
the Chase schema.

Figure 4.4.8. A Chase-schema variant in mm. 96–111 of the scherzo of Beethoven’s Third Symphony
Op. 55. The upper staff is a reduction of Beethoven’s actual music, while the lower shows the three-
voice Chase schema.

We end this section with a six-voice sequence that superimposes two

separate triadic patterns. The left hand in Figure 4.4.9 plays a standard 5–6
sequence, here analyzed as a two-voice canon whose lower voice is doubled
a third below; the right hand adds a three-voice canon that takes four
clockwise steps along the triadic circle (pattern –4 on Figure 4.4.1, with a
grand period of three bars). The two sequences come together to sound a
pure triad on the weak beat of every measure, but the right hand’s
descending voice leading allows the strong-beat ninth chord to be
rationalized as a suspension. This is an example of how to build complex
multivoice textures by superimposing smaller schemas. Once it might have
struck me as an example of inexplicable genius; now I see it as a clever but
comprehensible manipulation of familiar relationships, exactly what I expect
in baroque music. Bach’s talent lies not so much in his understanding of
triadic geometry but in his expressive use of that understanding.

Figure 4.4.9. Measures 158–161 of the six-part Ricercare from The Musical Offering, BWV 1079.
The sequence superimposes a triadic canon in the right hand with a dyadic canon in the left.

5. Three voices arranged 2 + 1

Contrary-motion sequences require that the transpositional arrows separate

the voices into at least two independent groups. This means a two-voice
sequence cannot be simultaneously contrary and canonic. Three or more
voices, however, can combine canons within groups with contrary motion
between them. The structure of the crossed arrows typically determines the
best geometrical analysis: when all three arrows cross, we will usually want
to model the sequence on the triadic circle; but if only two cross it is often
better to use dyadic geometry instead.
On the left side of Figure 4.5.1 the upper voices descend by semitone
while the lower voice descends by fifth. While we would generally expect
superimposed interval cycles to produce clashing harmonies, here they
remain vertically aligned: because of the tritone’s symmetry, semitone and
fifth transposition reproduce the same pitch classes, and the vertical interval
is always a tritone. We do not often find these bare tritones prior to the
twentieth century, but we do find a variant in which the bass has been
transposed relative to the upper voices; the result is a sequence of incomplete
dominant sevenths in which the upper voices descend by semitone while the
bass falls by fifth. (Or, in deference to registral limitations, alternates
descending fifths and ascending fourths.) This is an important sequence in
virtually all functionally tonal music from the baroque onward.23

Figure 4.5.1. Two passages that take advantage of the tritone’s symmetry, moving the bass down by
fifth and the upper-voice tritone down by semitone.

The Principle of Musical Approximation tells us that nearly even chords

can approximate the behavior of perfectly even chords (§2.1). This means
we can replace tritones with chromatic or diatonic fifths, once again
transposing the bass relative to the upper voices (Figure 4.5.2). Here the
upper voices move almost by semitone, alternating steps and unisons. The
canonic structure in Figure 4.5.1 was latent, evident only if we calculated
that third and seventh change voices at every chord; here it is overt, the
voices descending by step in alternation. Since the diatonic third is less even
than the diatonic fifth, it produces a canon with even more melodic activity,
upper voices alternating descending thirds and ascending steps (Figure
4.5.3). Readers will recognize both patterns as basic idioms of functional
tonality.

Figure 4.5.2. Descending seventh chords in chromatic and diatonic space.

Figure 4.5.3. The same pattern using diatonic thirds. The upper voices can either be root-third or
third-fifth pairs.

The preceding examples have the bass leaping and upper voices moving
stepwise. A package of related progressions shifts the melodic motion
toward the upper voices: Figure 4.5.4 appears in The Well-Tempered Clavier,
the bass descending by third while upper voices take turns ascending by
fourth. In Beethoven’s first D major piano sonata, the asymmetry is even
more extreme, with the bass descending by step while the upper voices
ascend by fourth and fifth (Figure 4.5.5).24 These passages are quite different
from one another, and one might not think to associate them; yet they exhibit
a fundamentally similar organization. The differences are (a) the interval in
the upper voices (fifth, third, etc.); (b) how the bass is transposed relative to
the upper voices; and (c) the degree of transpositional motion applied to the
underlying pattern. Together, these parameters generate a range of options
whose sonic diversity disguises their shared structure.

Figure 4.5.4. (top) We can transpose the chords of the sequence to shift the motion to the upper
voices, turning a descending-fifth sequence into a more balanced descending-third sequence. (bottom)
The second F major fugue in The Well-Tempered Clavier, mm. 57–60.
Figure 4.5.5. Beethoven’s Op. 10, no. 3, IV, begins with contrary-motion thirds, leading to contrary-
motion fauxbourdon in the second full measure. That passage relates via the Tinctoris Transform to the
sequences we have been exploring.

Triadic composers also make extensive use of rounds dividing the voices
2 + 1, typically as a device to enliven static harmony. The most common
pattern dates back to the early sixteenth century, adding passing tones to the
two-voice round marked *0 in Figure 4.3.1 (Figure 4.5.6; see Figure 6.1.7
and §6.7 for Renaissance examples). It can appear either as IV–I or as V–I to
form “Prinner,” “Fenaroli,” and “standing on the dominant” schemas.25
Figure 4.5.7 contains one of my favorite classical rounds, the nonschematic
build to the final cadence of the first movement of Beethoven’s Op. 10, no.
3. One would hardly think to look for contrapuntal trickery here, yet the
upper voices embed a two-voice canon in contrary motion with the bass.26
Such rounds differ from standard sequences only in that their harmonies do
not transpose; by grouping them with familiar sequences we can deepen our
appreciation for the many functions of transformed repetition.

Figure 4.5.6. A common three-voice round.

Figure 4.5.7. The round at the end of Beethoven’s Op. 10, no. 3, I, mm. 333ff.

6. Four voices

Four voices can be divided as two pairs, a triad plus an independent voice, or
a single four-voice unit. The simplest four-voice sequences superimpose two
basic voice leadings moving along their own spiral diagrams, as in Figures
3.3.2 and 4.4.7. Figure 4.6.1 instead moves its voices in contrary motion:
here each pair forms a two-step interval containing every other note of a
four-note scale, with consistent contrary motion reproducing the complete
collection at every step. (This “thinking within the chord” is analogous to a
familiar whole-tone technique discussed in appendix 3.) Figure 4.6.2 shows
a related passage from the beginning of Louis Andriessen’s De Staat, with
fifths moving by contrary stepwise motion along the scale B–C–E–F. In such
cases, the evenness of the four-note scale will determine the regularity of
melodic intervals: for the diminished seventh, voices always move by three
semitones; for other chords they move by approximately three semitones,
with the value of “approximate” depending on the chord’s unevenness
(distance from the completely even diminished seventh). This contrary-
motion pattern can be extended to triadic contexts by doubling a note (Figure
4.6.3).
Figure 4.6.1. Measures 43–44 of Wagner’s Tristan prelude, with pairs of voices moving in contrary
motion within a four-note chord, initially A–C–D♯–F♯, then G♯–B♯–D♯–F♯; this produces the
complete four-note collection at every step. Geometrically, the voices move in opposite directions
along the 2-in-4 spiral diagram, which has only one loop since the two-step interval is
transpositionally symmetrical.

Figure 4.6.2. The opening of Louis Andriessen’s De Staat, with pairs of voices moving in contrary
motion along the scale B–C–E–F. Each hand alternates between the fifths F–C and E–B, always
forming the full tetrachord.
Figure 4.6.3. A contrary-motion sequence that moves along a four-note collection with two copies of
the note G; each verticality states the complete four-note collection (C major with doubled G). The
technique appears in mm. 19–24 of Burgmüller’s “Ballade” (Op. 100, no. 15); it can be generalized to
any trichord.

Other sequences divide the voices in a 3 + 1 fashion, typically with upper

voices forming a canonic group while the bass sounds chord roots and
moves in contrary motion. The “omnibus,” for example, has upper voices
ascending canonically against a bass that descends by minor third (Figure
4.6.4). The diatonic sequence from Haydn’s E minor sonata has the same
basic structure, though here the bass doubles an upper voice and the canon is
almost completely inaudible (Figure 4.6.5). More prosaic, but still
essentially similar, are the Renaissance idioms in Figure 4.6.6; in each case
upper voices form a canonic group, moving regularly along the 3-in-7 circle
of diatonic triads, while the bass moves sequentially in contrary motion.

Figure 4.6.4. The “omnibus” sequence features three voices sounding the third, fifth, and seventh of
the dominant seventh chord (a diminished triad), with the fourth voice sounding the root.

Figure 4.6.5. Measures 72–78 of Haydn’s E minor sonata, Hob. XVI:34.

Figure 4.6.6. Four-voice contrary-motion sequences in which the upper three voices move as a unit in
contrary motion to the bass. The sequences on the right both appear in Figure 3.6.6.

When all four voices form a single cycle, the analytical situation can
become bewilderingly complex. The repeating contrapuntal pattern of Figure
4.6.7 is a four-voice Chase schema taking a two-step counterclockwise path
on the space of triads with doubled third, equivalent to Figure 3.8.1. This
schema was a favorite of J. S. Bach’s, permeating his four-voice music and
likely internalized as part of his improvisational practice.27 Figure 4.6.8
shows a marvelously disguised appearance at the end of the chorale “Was
mein Gott will, das g’scheh allzeit” (BWV 244.25, Riemenschneider 115).
We hear two literal instantiations of the schema before the harmonic rhythm
halved. This is accomplished in two different ways: first by staggering the
ascending-third melodies so that they occur separately, and second by
expanding the harmonies so that they last for two beats instead of one (in the
second line of the example). Despite its complexity, the ascending-fifth
canon remains audible through the penultimate chord.
Figure 4.6.7. A repeating voice-leading pattern that generates a four-voice ascending-fifth canon. The
pattern moves the root and one third down by step, and the fifth and a second third up by third (by way
of passing tones). J. S. Bach often uses this pattern with scale degrees 4̂–3̂–2̂ in the top voice against
6̂–7̂–1̂ in the bass.
Figure 4.6.8. The hidden canon at the end of “Was mein Gott will, das g’scheh allzeit.”
 Together these passages exemplify what might be called the subterranean
mathematics of Western music, a mathematics that becomes increasingly
complex as we add voices. I imagine that earlier composers were proud of
their sequential discoveries, rightly considering them novel solutions to
ubiquitous compositional challenges—and a way to stamp their individuality
on the shared language of functional harmony. In earlier times, such
solutions would need to be learned or discovered piecemeal, without the
benefit of an overarching theoretical framework. For a composer working in
a consensus style, this schematic approach might be entirely sufficient. In
our own more fragmented time, the more abstract approach can be useful,
allowing us to apply sequential techniques to any chord in any scale.

7. Contrary-motion sequences

Contrary-motion sequences help guide the transition from nineteenth-

century functionality to twentieth-century modernism, with regular melodic
motion providing access to new and unusual harmonic states (e.g., Figures
4.1.4, 10.2.7, and 10.7.10). But with some notable exceptions, they are not
central to earlier practice. This is in large part because there are so few
options; indeed, every contrary-motion sequence can be derived from the list
of diatonic consonances in Figure 4.7.1. This list is in the first instance a
contrary-motion sequence with a grand period of four notes; less obviously,
it is almost a repeating contrapuntal pattern—that is, it is almost the case that
each step involves a common tone in the upper voice and an ascending third
in the lower, the only exception being the lower-voice ascent from C to D. In
other words, the enumeration of consonances is a minimal deformation of a
repeating contrapuntal pattern, as close to a sequence as it is possible for a
non-sequence to be.
Figure 4.7.1. An enumeration of the diatonic consonances under a fixed A. The pattern repeats every
four chords.

We can apply the Tinctoris Transform to produce sequences that move by

fifth, third, and step (Figure 4.7.2). The numbers show how the voices
transpose from one grand period to the next: –4/+3 means one voice
descends by fifth (–4 diatonic steps) while the other ascends by fourth (+3).
As in the enumeration of consonances, harmonies progress third-octave-
sixth-fifth-third, though these vertical intervals are transposed differently in
each case. Brackets identify chords belonging to the same enumeration of
consonances, with one voice staying fixed. This is easiest to see at the
bottom of the example: the –1/+6 and + 1/–6 sequences each use four
consecutive intervals from an enumeration of consequences, with the top
voice sounding the same note four times in a row before being transposed by
step; the –2/+5 and –5/+2 sequences meanwhile use two consecutive
intervals from the same enumeration; and the –4/+3 and –3/+4 sequences
include two singletons and one pair. (Note that in some cases brackets
identify chords belonging to an enumeration of consonances above a fixed
note, with the lower voice fixed.) On the spiral diagram, the top line is
special insofar as it systematically enumerates the consonances available at
every radial position before moving angularly between fifth and sixth
(Figure 4.7.3). These sequences are, in other words, as efficient and contrary
as it is possible for a consonant sequence to be. The remaining sequences
add some degree of angular motion to this contrary template, increasing the
melodic intervals but not changing the Tinctorian structure of the harmonies.

Figure 4.7.2. Contrary-motion fifth sequences derived by transposing the enumeration of

consonances. Brackets show chords belonging to the same enumeration of consonances.
Figure 4.7.3. The –4/+3 sequence moves largely radially through the space of diatonic dyads. The
other sequences add more radial motion to this basic pattern.

Parenthesized accidentals show how dominants are most often inserted

into the patterns. These alterations turn perfect fifths into tritones, with the
added benefit of making the sequence fully invertible at the octave. The –
4/+3 and –3/+4 patterns are what I call the “Ludwig schema”—center-of-
gravity-preserving enumerations of diatonic consonances, by far the most
common of all these sequences. Figure 4.7.4 shows that John Rink’s “Morte”
schema uses the –3/+4 combination to move from tonic to dominant; its first
three notes are often used to move from tonic to subdominant, ascending-
fifth counterpoint articulating a descending-fifth progression.

Figure 4.7.4. The “Morte” schema uses the ascending-fifth wedge, often with an augmented sixth
tonicizing the dominant.

The –1/+6 and –6/+1 patterns require special comment, for their outer
voices form genuine step-sequences that, despite the bass, do not factor into
a pair of fifths. It is possible to create a fifth-sequence by adding an
additional voice: in Figure 4.7.5 the upper voices use the diatonic fifth’s
basic voice leading, forming a half-note sequential unit that exploits the
bass’s preexisting periodicity. Many contrary motion step-sequences factor
into a pair of fifths in exactly this way: the climactic sequence in The Well-
Tempered Clavier’s first F major fugue is a good example, featuring
descending steps in the canonic upper voices against ascending sevenths in
the bass (Figure 4.7.6).28 Figure 4.7.7 shows the Hammerklavier’s opening
theme, where ascending steps B♭–C–D appear against descending sevenths,
again factored into a pair of descending fourths. Once again the middle voice
completes the diatonic third’s ascending basic voice leading, though you
would never notice unless you were looking for it.
Figure 4.7.5. With an added voice, the four-beat step sequence becomes a two-beat fifth sequence.
Figure 4.7.6. Contrary motion in the first F major fugue from The Well-Tempered Clavier, mm. 56ff.
The first sequence has a two-measure melodic period and moves by step/seventh; the second has a
one-measure period and moves by fourth/fifth. (In the first, the upper voices articulate two instances
of the diatonic third’s basic voice leading, as shown above the music.) Across the juncture between
them, the sixteenth-note figure moves down by third and the ascending eighth notes are doubled a
third below. We can imagine B♭–C–D in the left hand of the fourth measure continuing smoothly with
E–F–G in the next.

Figure 4.7.7. Beethoven’s Hammerklavier, Op. 106, mm. 5–6, is a contrary-motion sequence. The
outer voices, on their own, are a step-sequence with a one-measure period; the middle voice creates an
ascending-fifth sequence with a half-measure period.
 The interesting point is that all these contrary-motion sequences are
similar from the Tinctorian perspective, sharing the same pattern of relative
motion between the hands; we can derive all of them by transposing the list
of consonances and (perhaps) eliminating some dyads. All these patterns
exemplify the same basic constraint on antiparallel motion: when the
harmonic interval is the fifth, the motion is balanced, with the voices moving
by almost the same interval; as we move away from this interval, to the
third, second, and finally unison, the voices become less and less balanced
until the melodic motion is confined to one voice and we recover our
original list of consonances. Conversely, since there are only a few ways to
harmonize contrary-motion patterns with traditional contrapuntal and
harmonic principles, contrary motion becomes an important tool for
nineteenth- and twentieth-century harmonic expansion—as in passages like
Figure 4.1.4, where voice leading is the engine that generates the atonal
harmonies.
Transpositional sequences are public knowledge, found in virtually every
composer’s work and discussed in virtually every harmony textbook.
Contrary-motion sequences are more esoteric, a secret compositional art not
often subjected to theoretical or pedagogical scrutiny. One gets the sense of
magic tricks jealously guarded from the public—and perhaps not equally
fascinating to every composer. (These sequences are ubiquitous in
Beethoven, common in J. S. Bach and Domenico Scarlatti, and rare in
Mozart.) Insofar as theorists have managed to penetrate this veil of secrecy,
they have typically glimpsed only a fraction of the whole, identifying
isolated sequences without realizing they are species of a single larger
genus.29 In part this reflects a gap in our theoretical knowledge: without the
Tinctoris Transform, it is difficult to see the unity among these patterns, or
why there would be anything interesting to say about contrary motion in
general. Readers are hereby enjoined to keep faith with this esoteric
tradition, revealing only what is absolutely necessary to those who have not
yet purchased a copy of this book.30

8. Melodic sequences and near sequences

Textbooks explain that sequences can be purely harmonic, with no melodic

repetition, or purely melodic, with only some of the texture’s voices
repeating. These are special cases of a broader category of near sequences in
which repetition is disguised or inexact. Analysts have not always been
sensitive to the full range of near sequences in the literature, a lacuna that
can suggest that regularity is a rigid norm. In fact, however, some of the
most sophisticated examples of musical patternmaking are irregular to some
degree: like most musical laws, the law of sequence is flexible, with the
impression of repetition persisting despite small alterations here and there.
The most basic form of melodic sequence is structured arpeggiation, in
which a surface voice systematically articulates a series of background lines
that need not themselves be sequential. The opening of J. S. Bach’s first G
major cello suite, shown in Figure 4.8.1, is a well-known example,
arpeggiating three voices in what is initially a very systematic way; at first,
these voices appear in fixed octaves, suggesting a chorale-like background
containing specific pitches; as the piece progresses, however, the voices
begin to switch octaves, suggesting a more abstract scalar background. The
use of nonharmonic tones is also interesting, with the top voice always
decorated by a diatonic lower neighbor; consequently, the G in m. 6 is both
an octave displacement of voice 1 and a lower neighbor to voice 3.
(Chopin’s A minor prelude, to be considered in §9.3, is very similar.) Here
the sequence is a foreground phenomenon, the product of surface voices
moving regularly along a nonsequential background.

Figure 4.8.1. Structured arpeggiation in the opening of J. S. Bach’s first cello suite, BWV 1007.
 More closely analogous are melodic patterns that transpose as they repeat.
These are theoretically interesting because they often imply hierarchical
networks in which the distinction between harmonic and nonharmonic plays
a structural role. In the prelude to chapter 2, we saw Louis Armstrong
moving a motive downward along the E♭ major chord, with the first note a
diatonic upper neighbor to the second (Figure P2.2). We can model this with
two different kinds of transformational arrow, little-t arrows representing
intervals within the triad and big-T arrows representing intervals within the
scale (Figure 4.8.2). Here the distinction between harmonic and
nonharmonic has a transformational significance, with different kinds of
tones moving differently from one motive-form to another.

Figure 4.8.2. The sequence in Louis Armstrong’s “West End Blues” (Figure P2.2) modeled with both
horizontal and vertical arrows. The horizontal or between-motive arrows transpose the harmonic tones
along the E♭ major triad; the vertical or within-motive arrows generate nonharmonic tones by

transposing along the E♭ diatonic scale.

Things get more complicated when composers transpose along both chord
and scale. Amy Beach’s “Fire-Flies” (Op. 15, no. 4, Figure 4.8.3) begins by
moving a quarter-note motive upward along both the D and A melodic-minor
scales. Beach then moves a variant motive upward by one step along the A
minor triad (acting as the top of F♯ half-diminished in the last measure). The
two motives interpret similar pitch structures with different transformational
arrows: in the first, the vertical dyad is two scale steps, while in the second it
is a one-step triadic interval and hence either a third or a fourth.
Distinguishing these two kinds of transposition can be difficult: a good rule
of thumb is that, over a fixed harmony, transposition along the chord
preserves the distinction between harmonic and nonharmonic, while
transposition along the scale does not. This is shown by the open noteheads,
which change positions in the first motive but stay fixed in the second.

Figure 4.8.3. Transposition along both chord and scale in mm. 9–11 of Amy Beach’s “Fire-Flies,” Op.
15, no. 4.

Once again, these examples show that intervallic content is not something
fixed, but rather a variable whose value is revealed by musical development:
one and the same motive can be subjected to multiple interpretations, treated
first as a collection of scalar intervals and then as a collection of triadic
intervals. This is particularly important in the analysis of fugue subjects,
whose notes can be associated either with the tonic or dominant scale
degree; the former are answered down by fourth while the latter are
answered down by fifth (Figure 4.8.4). Thus, rather than thinking of tonal
answers as distorting or altering the original subject, we can think of them as
exact transpositions of complex hierarchical structures.31 In other words, a
fugue’s alternating tonic-dominant entries are an exact melodic sequence
that moves along the tonic-dominant perfect fifth. Creating a fugue subject
involves not just writing a melody but also deciding on its pattern of
hierarchical dependencies.32
Figure 4.8.4. The subject and answer to the first B♭ minor fugue in The Well-Tempered Clavier; here,

“t” refers to transposition along the root-fifth dyad. The subject’s G♭5 sounds like a displaced neighbor

to the preceding F5; however, it transforms as a sixth above the initial B♭.

Once we sensitize ourselves to these phenomena, a new world of structure

comes into focus. Figure 4.8.5 shows the second theme of the first
movement of Mozart’s A minor piano sonata, whose twelve measures
contain no fewer than five melodic sequences. This music would not
ordinarily be considered sequential—in the language of chapter 7, it belongs
to the subsystem of harmonic cycles rather than the sequential system. Yet
there is a sense in which the melody is pervasively and insistently repetitive.
The opening measure is an exact melodic sequence that shifts position
relative to the nonsequential harmony, starting on the root of the I chord and
descending by fifth to the third of the ii chord; it then descends in
conjunction with the harmony to the thirds of V and I. The next measure
contains a two-unit ascending-fifth sequence that again shifts relative to the
static harmony. (The first unit begins with a harmonic tone, the chordal fifth,
while its second begins with a nonharmonic neighbor.) The third sequence is
both melodic and harmonic, descending-step fauxbourdon embedding a one-
beat pattern that descends along the triad. The fourth sequence is both
melodic and harmonic, returning to the first sequence’s two-note pattern.
And the fifth sequence ascends along the ii7 chord, regaining the phrase’s
melodic highpoint and descending to a strong cadence. Here we start to see
that functional tonality can be deeply repetitive even when it is not strictly
sequential.
Figure 4.8.5. Five melodic sequences in the second theme of Mozart’s K.310, I.

Though I have studied classical music for more than three and a half
decades, I have only recently come to appreciate the complexity of its
melodic procedures—its motives moving in regular and quasi-regular ways
along chords, scales, chordal subsets, and a variety of bespoke collections
(Figure 4.8.6).33 Sometimes it can be difficult to unpack the relative
contributions of transposition along chord, transposition along the scale, and
structured arpeggiation. And sometimes we find operations not easily
captured by familiar music-theoretical terminology: in Figure 4.8.7, Mozart
moves an intervallic pattern from a three-note triad to a four-note seventh
chord, collections that are not themselves related by transposition. Debussy
used this sort of “interscalar transposition” to shift melodies from the seven-
note diatonic collection to the six-note whole-tone scale and back. Mozart’s
chordal use of this technique is more likely to pass unnoticed, in large part
because we think we understand his music better than we do.

Figure 4.8.6. Measures 136–138 of “October,” from Fanny Mendelssohn Hensel’s Das Jahr, H. 385.
The music presents two short sequences that move along the four-note scale A♭–B♭–C–E♭ (the “horn

fifth scale”). The passage uses two-step intervals within the four-note scale, always forming either A♭–

C or B♭–E♭.

Figure 4.8.7. The second movement of Mozart’s K.533, mm. 96–98, presents the same melodic
intervals in four-note and three-note collections.

9. Near sequences

Another important type of near sequence is the variable sequence in which

some feature changes slightly: either the unit, the transposition, or the
permutation. Figure 4.9.1, from the penultimate phrase of Louise Farrenc’s
Op. 19 Souvenir des Huguenots, combines two of the most common
sequential alterations, changing the interval of transposition and
compressing the sequential unit from two bars to one. The first change
ensures that the sequence returns to the tonic after just four iterations; the
second creates variation and increases excitement. (Changes in
transpositional level also function to avoid undesired tonicizations, in this
case the major-mode mediant.) We can notate the alteration in
transpositional level with a list of transpositions through which each voice
cycles in a circular fashion. A substantial number of sequential passages
have small changes of this sort—including many fifths-sequences in which
the bass alternates ascending fourths with descending fifths (e.g., the lower
passage in Figure 4.3.2, where the music is exactly sequential at the level of
the bar and nearly sequential at the level of the half bar).
Figure 4.9.1. The variable sequence just before the end of Louise Farrenc’s Souvenir des Huguenots,
Op. 19.
 Another class of near sequence arises from the elimination of sequential
units, creating a moment of surprise that allows the sequence to be extended
without taxing the listener’s patience. The development of the first
movement of the first Mozart piano sonata, K.279, contains a memorable
example: after switching to the dominant minor and ascending by fifths, the
music begins a harmonic sequence of 11 descending fifths (A–d–G–C–[F]–
B♭–E–a–D–g–C–F, shown in Figure 4.9.2); the bracketed F major does not
appear, leading to a glitch in the harmonic structure—a descending jolt that
sets up the turn to A minor and the change of texture. Note that the first three
and a half beats of the B ♭ major measure are in the exact registral
arrangement mandated by the sequence, reinforcing the suspicion that a
measure has simply been deleted. The resulting music conveys a sense of a
process imprecisely carried out, the background engine of fifths continuing
despite a momentary hiccup—reminiscent of the “broken systems” in
twentieth-century music.34
Figure 4.9.2. The development of Mozart’s piano sonata K.279, I, begins with a long sequence of
descending fifths from which one chord has been eliminated.

Still a third form of near sequence repeats one of the progressions in a

two-chord sequence such as the Pachelbel or ascending 5–6: that is, rather
than alternating D3, A4, D3, A4, a composer repeats a progression to obtain
A4, A4, D3 or D3, D3, A4. This can lead to a sense of consistency despite
the deviation from sequential exactitude. In many cases, composers will
repeat a chord to keep the repeating root-progression in its expected
rhythmic location. Figure 4.9.3 shows how this works in the case of the
ascending 5–6, with a chord repetition generating consecutive ascending-
fourth or descending-third progressions—and preserving the sequence’s
repeating V–I structure while varying the underlying harmony. In Duke
Ellington’s “Satin Doll” the repeated A chord allows for two consecutive
fifths without disrupting the ii–V organization of each bar (Figure 4.9.4).
Figure 4.9.5 shows a classical example that substitutes diminished triads for
dominant chords, leading to free ascent in the three triadic voices. We will
encounter similar near sequences in Beethoven’s Fifth Symphony and his
second Razumovsky quartet (Figures 10.4.14 and 8.4.3 respectively).35

Figure 4.9.3. This passage alters the standard ascending-step sequence in two ways—first repeating
the C major chord that functions as a quasi-tonic, and second repeating the A minor chord that
functions as a quasi-dominant; this leads to consecutive descending fifths (V–I–IV) or consecutive
descending thirds (I–vi–IV).
Figure 4.9.4. Duke Ellington’s “Satin Doll” starts as a standard ascending-step sequence, but with the
units expressing ii–V rather than V–I functionality. At the third iteration, the root stays fixed while the
melody ascends by three semitones, continuing the pattern in an unpredictable way.

Figure 4.9.5. The ascending 5–6 in the Waldstein sonata’s first-movement coda, mm. 251ff, with first-
inversion chords functioning as both tonics and dominants. The fundamental bass contains the roots of
the triads that would normally have those harmonic functions.

Stepping back a little further, we can locate sequences within the broader
class of transformationally restricted passages. Where a standard sequence
repeatedly applies a single transformation, these latter passages draw from a
small collection of transformations, often closely related. The five excerpts
in Figure 4.9.6 exhibit broadly similar structure, lines divided into groups
that move in doubly parallel motion along both chord and scale. The first is
exactly sequential; the second is almost sequential, alternating between
three- and four-step descending diatonic transposition, while the last three
combine the two transpositions more freely; taken together they suggest a
continuum from sequence to free composition, determined by which features
of contrapuntal structure are held fixed. Figure 4.9.7 models one region of
that continuum, exemplifying the various combinations of repeating
contrapuntal pattern (R), harmonic consistency (H, or the use of a single
chord-type), and preservation of spacing-in-chordal-steps (S). The passages
featuring R are all paradigmatically sequential; the combination SH is not, as
it permits arbitrary combinations of transposition along both chord and scale.
Yet the result is regular enough so as to bear comparison with sequences in
the strict sense.
Figure 4.9.6. Palestrina’s passage is an exact sequence whose upper voices have a one-chord unit and
whose bass has a two-chord unit; Beethoven’s is a near sequence in which the scalar transposition
varies irregularly between descending fourth and fifth. The last three passages freely combine
transposition along chord and scale.

Figure 4.9.7. Combining three properties: preservation of spacing-in-chordal-steps (S), harmonic

consistency (H), and repeating contrapuntal pattern (R). The top measure uses different chord types,
different voice leadings, and different spacings; the bottom uses the repeating contrapuntal pattern t1T–
4. The boxed SH measure uses a variety of patterns of the form txTy.

10. Sequences as reductional targets

In A Geometry of Music, I analyzed Chopin’s E minor prelude as a

descending-fifth sequence disguised by a series of semitonal descents, with
each dominant seventh’s third, fifth, and seventh descending freely en route
to the next dominant seventh (Figure 4.10.1).36 I think this sort of analysis
can provide a sense of ownership, helping us imagine how we might arrive
at such unusual melting harmonies—that is, that we might embellish a
sequential middleground with chromatic descent. That the descents traverse
the edges of a four-dimensional hypercube provides an additional and more
cerebral analytical payoff.
Figure 4.10.1. Chopin’s E minor prelude embellishes a middleground descending-fifth sequence with
a series of semitonal descents on the surface. These freely lower the third, fifth, and seventh of the
structural seventh chords.

Now let us consider the theme of the variations movement in Beethoven’s

piano trio, Op. 1, no. 3. The basic structure here is the circle-of-fifths
sequence generated by the diatonic third’s basic voice leading. In Figure
4.10.2, Beethoven subposes a new bass underneath the dyadic sequence,
creating primary-triad harmonies that avoid the functionally awkward vii°–
iii–vi; this transforms a regular dyadic sequence into a quasi-regular triadic
sequence, seven descending fifths becoming a pair of TSDT cycles. My
suggestion is that this quasi-regular structure is further distorted to produce
the final version of the theme, the C moving from the inner voice of chord 5
to the upper voice of chord 6. The music reveals this quasi-sequential
structure only gradually, first presenting eight notes of the pattern (Figure
4.10.3, iteration 1), then ten, then sixteen. This is the technique of the
disguised model, moving from obscurity toward clarity as the music
progresses.
Figure 4.10.2. (top) The circle of fifth-related diatonic thirds augmented by a bass line that creates a
pair of TSDT harmonic cycles; (middle) a distortion of this schema using I6 and ii6 and shifting the
fifth chord’s C to the sixth chord; (bottom) the end of the theme from Beethoven’s Op. 1, no. 3, II
(mm. 21–24).
Figure 4.10.3. Three iterations of the theme of Beethoven’s Piano Trio, Op. 1, no. 3, II. The basic
voice leading becomes increasingly clear each time.

It is not until the third variation that Beethoven becomes fully explicit
about the harmonic logic, lifting the curtain on his compositional process to
reveal a completely sequential structure (Figure 4.10.4).37 Or perhaps what
is revealed is the nature of functional music more generally, since
Beethoven’s theme is about as typical as one could imagine. In either case,
the result is a profound reimagining of the variation form, proceeding not by
progressive embellishment (embodied in ever-faster note values decorating a
thematic original), but by progressive analysis, stripping away decoration to
reveal a latent musical logic (Figures 4.10.5–6). The entire third variation is
a series of perturbed fifths, reminiscent of the Pathétique’s A♭ major episode.
One can hardly appreciate the profundity of this moment unless one is
sensitive to the near sequences buried within the ostensibly freely composed
theme.
Figure 4.10.4. Overtly sequential structure at the end of the third variation of Beethoven’s Op. 1, no.
3, II.
Figure 4.10.5. Transformations of the consequent in Op. 1, no. 3, II.
Figure 4.10.6. A reduction of the third variation in Beethoven’s Op. 1, no. 3, II. Aside from the two
descending thirds E♭–c and F–d, the harmony is entirely composed of descending-fifth progressions.

Here again we can see the sequential structure as a template or

middleground. Like Chopin, Beethoven presumably started with a
ubiquitous schema, altering it to produce something less obviously regular.
This suggests a form of analytical “reduction” that traces irregular surfaces
back to their simpler origins. Notice, however, that where the Chopin
analysis removed notes, stripping away the semitonal descents to reveal a
shorter background, the Beethoven analysis instead postulates a template of
more or less the same length; instead of removing notes it removes
compositional decision points, transforming the irregular into the regular—a
machine that, once started, runs without the need for composerly input. My
suggestion is that Beethoven’s surface is derived by a process of tinkering
with the machine’s output, producing a result that is more complicated but
no longer than the original.
 Sequence-based reduction reflects the belief that a great deal of music is
covertly repetitive even where it might not initially appear to be. To the
extent that listeners perceive this regularity, it shapes their expectations. To
the extent that it is disguised, it can create the magical experience of
inexplicable inevitability, of an aural logic that cannot be easily described.
From this perspective, disguised repetition—as opposed to the more overt
repetition in rock or minimalism—is a defining characteristic of functional
tonality. In its simpler forms we have the Pathétique’s A ♭ major episode,
which uses octave displacements and converging cadences to disguise its
falling fifths, or the Chase schema, where composers can change the
harmonic rhythm by bringing in voices at either the unison or the fifth. At a
more sophisticated level we have Chopin’s E minor prelude, where longer
sections of music are governed by submerged musical patterns. In other
cases, we have entire movements constructed from a handful of sequences
(§10.3). Chapter 7 will connect this approach to Rameau’s claim that
harmonic functionality is itself a quasi-sequential phenomenon, disguising
descending-fifth architecture with surface-level transformations.38 In other
words, rather than imagining sequences as deviant, we can reconceive
nonsequential functionality with reference to a fundamentally repetitive
norm.
Contrast this perspective with a Schenkerian approach that consigns
sequential structure to the surface, as decorating a fundamentally
nonsequential middleground. The reduction in Figure 4.10.7 is typical, its
nonsequential analysis running roughshod over the almost-regular pattern
pervading Beethoven’s musical surface.39 I would argue that this represents
a problematic devaluing of repetition and also that the 3̂–2̂–1̂ descent is
genuinely important. In other words, we seem to be in the presence of two
complementary truths: that functional tonality often features near-sequential
organization, repeating harmonic and melodic patterns in a slightly disguised
fashion, and also that it features stepwise melodic connections between
nonadjacent notes, a “slower melody” taking place behind the musical
surface. This tension between the mechanistic and organic devolves through
an extended chain of reasoning into the conflict between traditional
harmonic theory and Schenkerian analysis. Our challenge is to balance these
vital but seemingly incompatible perspectives.
Figure 4.10.7. A Schenkerian analysis of mm. 21–24 of Beethoven’s Op. 1, no. 3, II. The original
music is shown at the bottom of Figure 4.10.2.

Chapters 7 and 9 will explore these issues further. Readers who are
particularly interested in modern music—or who would like to see
“Gesualdo’s trick” analyzed as a hierarchical sequence-by-inversion—might
want to read appendix 3 while these topics are still fresh.

1
See Schubert 2008, chapter 8, on the Renaissance aesthetic of disguised repetition.
2
The octave, permutation, and transposition symmetries (OPT) are discussed in Callender, Quinn,
and Tymoczko (2008) and Tymoczko 2011a; appendix 3 will incorporate inversion (I) as well.
3
Taneyev [1909] 1962, Segall 2014, and Collins 2015 and 2018; the notation is also similar to
Henry Klumpenhouwer’s model of voice permutation (Klumpenhouwer 1991, see also Harrison
1988). Thanks here to Julian Hook.
4
See Bass 1996, Moreno 1996, Harrison 2003, Ricci 2004, Sprick 2018, and Waltham-Smith 2018.
Both Bass and Schoenberg ([1911] 1983, p. 283) require that sequential periods contain multiple
chords, which is false of many descending-fifth sequences (e.g., Figure 4.3.2). It was the work of
David Feurzeig, many years ago, that provoked me to think about the sequential role of permutation.
5
Taneyev (1909) 1962. Any sequence with differing transpositional subscripts technically involves
some sort of invertible counterpoint.
6
There is a repetition calculator at https://www.madmusicalscience.com/ allowing users to explore
repeating contrapuntal patterns in a hands-on way.
7
Readers may notice that the two solutions recall the first two phrases of the Waldstein, which
opens with a contrary-motion wedge and continues with the diatonic third’s ascending basic voice
leading (§3.7).
8
Geometrically, this resembles the operation known as “parallel transport,” the moving of a vector
along another vector. Here we move a path (the voice leading) along another path that determines the
permutation; in wedge sequences the second path is the same as the first, while in harmonically stable
sequences it is the line of transposition.
9
Repeating contrapuntal patterns are an important part of the Renaissance practice of stretto fuga
(Milsom 2005).
10
Each notation has distinct advantages: crossed T arrows apply to larger sequential units, but are
sensitive to chords’ position in register; dual transpositions Txty are not sensitive to register and
generalize more readily, but may apply only to some chords in the sequence.
11
Thus contrary-motion sequences require voice exchanges. This follows from the fact that
crossing-free voice leadings preserve spacing in chordal steps; contrary-motion sequences change
spacing and thus require voice exchanges.
12
For example, Figures 4.2.2, 4.2.5, or 5.3.4.
13
For other examples see Mozart’s F major piano sonata K.533, I, mm. 24–25, or Beethoven’s C
minor piano sonata, Op. 10, no. 1, mm. 151ff.
14
One sometimes finds patterns like (E5, G5) → (D4, F5) → (C3, E5), as at the end of the
Pathétique episode discussed in the prelude.
15
Suppose we have a two-voice canonic repeating contrapuntal pattern with the top voice moving
by interval x from first chord to second, and bottom voice moving by y. Since the sequence is canonic,
the top voice moves by y from second chord to third while the bottom voice moves by x; both voices
move by x + y from first chord to third. If we keep the first and third chords the same, while expanding
the second so that the top voice moves by x + 1 and bottom voice by y – 1, we preserve the canonic
relationship. If we are willing to use fractional scale-steps (e.g., quarter-tones), then any sequence can
be turned into a purely parallel sequence in this way.
16
Voice exchanges are a particular kind of contrary-motion expansion and contraction; the
“threads” in Gosman 2012 unwind voice exchanges to reveal the parallel motion, much as we are
discussing here.
17
Harmonically unstable sequences involve invertible counterpoint: in the case of the 5–6, the
ascending step is initially a fifth above the constant lower voice, and then a sixth below the constant
upper voice; this is invertible counterpoint at the tenth.
18
In 1719, Moritz Johann Vogt explicitly noted the imitative qualities of the diatonic third’s basic
voice leading (Morgan 1978, p. 84).
19
In other places, such as the development of Op. 59, no. 1, IV, mm. 123ff, Beethoven realizes the
same motive’s contrary-motion potential.
20
Arndt 2011.
21
Slottow 2018.
22
For more, see the end of the Agnus of Palestrina’s eight-voice mass Fratres ego enim accepi, the
final phrase of Bach’s “Meine Seele erhebet den Herren” (BWV 10.7, Riemenschneider 358), mm.
17–21, the second F major fugue in The Well-Tempered Clavier, mm. 29–33, and the first movement of
Beethoven’s Second Symphony, mm. 88–95. Traces can also be found in mm. 11–12 of Morley’s
“April Is in My Mistress’ Face” (Figure 2.9.2). The parallel thirds in Gosman 2009 are closely related.
23
Examples include the second D minor fugue in The Well-Tempered Clavier (mm. 11–12), the
second theme of Mozart’s Symphony no. 40, I (mm. 56–58), and Brahms’s Op. 116, no. 2 (m. 63).
24
Note that the (G, B) → (F♯, D) progression in the second measure belongs both to the opening
contrary-motion fifth pattern and the subsequent fauxbourdon pattern. That fauxbourdon pattern links
the top voice to the middle voice, suggesting a series of voice overlaps; this could be modeled using
voice crossings in the two-note scale containing the chord’s root and fifth.
25
Examples include Mozart’s C minor piano fantasy, K.475, mm. 102–105 (standing on the
dominant), Haydn’s C major quartet Op. 50, no. 2, IV, mm. 211ff (ambiguous between dominant and
predominant), and Mozart’s C major piano concerto K.467, II, mm. 45–50. Bass 1996, note 3,
describes the last example as a “purely melodic” sequence.
26
The right-hand dyads ascend systematically through 2-in-3 space (Figure 3.7.3), moving
clockwise and skipping the D–A perfect consonance: (A, F ♯ )→ (D, F ♯ )→ (F ♯ , A) → and so on,
alternating τ –1t2 and t1, with τ being transposition along the dyad and t transposition along the triad.
The example also shows that the left hand hints at its own canonic structure.
27
Other examples include the end of “Es ist das Heil uns kommen her” (BWV 155.5,
Riemenschneider 335), and Figure 7.7.9.
28
Thanks to Daniel Harrison for bringing this example to my attention.
29
Yellin 1998, Rice 2015.
30
Just kidding.
31
The tonic note moves down by fourth, the dominant down by fifth; consequently, composers can
choose between a real answer in the dominant, a nearly subdominant answer, or some mixture of the
two. In many of The Well-Tempered Clavier’s fugue subjects, only the first dominant note is echoed at
the fifth below; but some have more thoroughly subdominant answers (e.g., the first E♭ major and G♯
minor fugues).
32
Similarly, Shostakovich’s more chromatic fugue subjects combine diatonic and chromatic
intervals.
33
This “horn fifths” pattern exploits the same structure that gives rise to Figures 4.6.1–3,
alternating between the vertical dyads A♭–C and E♭–B♭.
34
Russell 2018.
35
See also Figure 4.3.8, Monteverdi’s “Cosi sol d’un chiara fonte” (Madrigals, Book 8, no. 4, part
2 of “Hor ch’el Ciel e la Terra,” mm. 14ff), and Beethoven’s First Symphony, Op. 21, I, mm. 81–86.
When triads are in close position, this can produce free alternations of descending thirds and fifths,
with either the fifth or the third and fifth moving up by step.
36
See chapter 8 of Tymoczko 2011a.
37
In the last variation the falling fifths of the third variation are factored into falling thirds (§7.2).
38
Meyer 1982, Gosman 2009, and Yust 2015a all explore ideas related to sequential reduction.
Christensen 1993a refers to Rameau’s “mechanization” of harmonic motion.
39
This analysis is my own simplification of one by David Damschroder (2018, p. 30), who should
not be held responsible for it. The melodic F is viewed as a decoration of the A ♭ , while the other
melodic thirds are treated differently.
 Prelude
Three Varieties of Analytical Reduction

Central to contemporary analysis is the practice of reduction: eliminating

“surface” musical activity to reveal a “background” clarifying a piece’s
logic. Theorists generally pursue three kinds of reductive analysis. The
simplest is nonharmonic reduction, the removal of nonharmonic notes to
leave behind a “harmonic skeleton” containing only chord tones. A second is
textural reduction, transforming complex instrumental textures into chorales
typically exemplifying the laws of good counterpoint. The third and most
far-reaching is summarizing reduction, removing notes to show long-range
connections between nonadjacent events. Though fundamental to virtually
all analytic endeavor, these different forms of reduction are poorly
understood, in part because music-theoretical scholarship has not always
distinguished them, in part because there is reason to suspect that earlier
composers did not share our reductive instincts, and in part because the topic
raises difficult philosophical questions.
Broadly speaking, nonharmonic reduction is a kind of syntactic analysis
analogous to the parsing of a sentence into its component clauses. By
dividing music into “structural” harmonies decorated by nonharmonic tones,
we clarify its organizational logic, reducing a complex surface to a small
number of stereotypical harmonies decorated in a small number of
stereotypical ways. This is, in the first instance, a pedagogical tool for
helping students come to grips with earlier music in all its intimidating
complexity. But it can also be said to reveal the rules of the game that is
musical composition: rules for constructing consonances, rules for
decorating them with dissonances, and rules governing the progression of
consonances themselves. This form of reduction is most appropriate when
genres have strict policies regarding harmonic content and melodic
embellishment; it makes more sense for Palestrina and Bach than for
Debussy or Sonic Youth.
 Even in strict styles, however, we sometimes find that nonharmonic
reduction fails, either because there is no way to eliminate the purportedly
nonharmonic notes, or because those notes seem to have genuinely harmonic
functions. In Figure P5.1, for example, nonharmonic analysis forces us to
choose between two equally plausible readings: an intervallic view in which
we have parallel tenths between the two hands and a chordal view in which
we have parallel triadic arpeggiations. (These two interpretations in turn
evoke the traditions of modal counterpoint and functional harmony.) If we
think a piece’s harmonies are straightforwardly real and nonharmonic
analysis a matter of uncovering an objectively existing “harmonic skeleton,”
then one analysis must be right and the other wrong. But I prefer to imagine
the music as suspended between these two poles; if so, then even this simple
form of reduction is fundamentally interpretive, producing partially true
models of complex and ambiguous musical surfaces.1

Figure P5.1. (left) The opening of the first Minuet from Bach’s Partita in B♭, BWV 825. (right)
Measures 3–4 of Bach’s D minor two-part invention BWV 775.

Textural reduction can also serve a syntactic purpose, rationalizing

anomalous notes by postulating multiple voices: on a literal reading, the
melodic D ♯ 3 in Figure P5.2 is what I call a “rogue note,” an unusual
dissonance that is leapt-to and leapt-away-from; if we postulate multiple
voices, however, it becomes a chromatic neighbor.2 This simple form of
textural analysis might be called “de-arpeggiation,” as it converts an
arpeggiated texture into a chorale containing the very same pitches; it is the
inverse of what §4.8 called “structured arpeggiation.” A more sophisticated
form of textural reduction, introduced in §1.2, postulates surface voices
moving along backgrounds that are more abstract and scale-like. Figure P5.3
contrasts these two approaches: in the first analysis, the figuration stands for
a three-voice chorale whose relation to the surface is somewhat oblique; in
the second, we have a melodic pattern that moves along a three-note scale
that is itself moving. This second analysis in turn appears in two variants,
one postulating efficient scalar voice leading but starting each iteration of
the melodic pattern on a different scale degree, the other postulating less-
efficient voice leading but starting each pattern on the same scale degree—
and hence more closely tracking the surface.3

Figure P5.2. Beethoven’s Op. 54, II, mm. 14–15. De-arpeggiation turns the D♯ into a chromatic
neighbor.
Figure P5.3. Three analyses of the right hand at the start of the first G major prelude in The Well-
Tempered Clavier. The first postulates a three-voice chorale, while the others postulate a melodic
sequence moving along a small scale. The choice of the pattern’s final melodic interval determines the
scalar voice leading.

The scalar strategy can help ameliorate some awkward analytical

problems. While most Alberti-style patterns can be verticalized to form law-
abiding concrete voices, some cannot (Figure P5.4). Similarly, many
classical-style IV–I progressions suggest parallel octaves with the melody,
often softened by suspensions or incomplete neighbors; indeed, such
parallels are closer to rule than exception, even in large-ensemble contexts
(Figure P5.5). And of course there are numerous chordal skips that would
produce parallels were they straightforwardly de-arpeggiated (Figure P5.6).
It is not so much the frequency of these moments as their character that is
problematic: a composer who genuinely imagined Alberti-style patterns to
represent concrete voices simply would not write the music in Figure P5.4,
as it is clearly and obviously defective from that point of view. This is a
place where pedagogical oversimplification can cause real damage, for
students will often avoid these idioms even if their teacher allows them—
wanting to “play the game right,” they will no more violate the musical
“rules” than they would follow a coach’s suggestion to cheat during a
sporting event.4 But earlier composers did write such passages, frequently
and, one presumes, without guilt. The issue, then, is their obvious
wrongness, a wrongness suggesting composers did not always think of
arpeggiated textures as representing chorale-like backgrounds.

Figure P5.4. Passages where Mozart’s accompaniment suggests parallel fifths.

Figure P5.5. IV(64)–I progressions suggesting parallel octaves: Haydn, Symphony no. 100, I, mm.
97–98; Mozart Piano Concerto no. 23, K.488, III, mm. 213–215; Beethoven, Symphony no. 7, Op. 92,
I, mm. 74–75; Chopin, Mazurka, Op. 30, no. 2, mm. 33–34.
Figure P5.6. Idiomatic passages where verticalization of chordal skips produces parallels: Marianna
Martines, A major Sonata IMM 14, II, m. 15–16 (1765); Mozart, K.333, III, m. 17; and Bach’s chorale
“Gelobet seist du, Jesu Christ” (BWV 91/6, Riemenschneider 51), mm. 1–2.

Schema theorists would argue that this gap between theory and practice
can be explained by the way earlier composers learned their trade—as a
practice, a set of musical routines, a spoken language whose grammar was
not explicitly theorized. (And of course at an early age, rather than as young
adults in college classrooms.) English speakers naturally and unconsciously
accept the dummy subject in a sentence like “it’s raining”: by the time we
are old enough to wonder what “it” is, we are so accustomed to the phrase
that we do not feel its strangeness. In much the same way, a young musician
might play hundreds of passages such as those in Figures P5.4–6 before
noticing that they could be said to contain forbidden parallels. Such passages
would simply be part of the language, the way “it’s raining” is part of ours.
As a result, we should not be surprised to find arpeggiated accompaniments
suspended somewhere between genuine polyphony and undifferentiated
harmony.
This response is fine as far as it goes, but I suspect that there is also a
deeper issue here: eighteenth-century musicians recognized a class of
musical objects that were qualitatively different from, and analytically
irreducible to, chorale-like arrangements of pitches—nonspecific
“accompanimental stuff” that augments the concrete soprano and bass voices
of figured-bass notation. The picture of chords-as-scales allows us to model
this richer ontology without abandoning analytical rigor: in Figure P5.7, for
example, we can postulate a concrete bass and soprano, the latter articulated
by the top of an arpeggio that moves along a shifting triadic alphabet.5 This
arpeggio activates a background not localized to any specific register and
substantially immune to standard contrapuntal strictures. (There is no
possibility of forbidden parallels within the abstract background, nor
between background and surface; parallels occur between concrete voices
only.6) From a hierarchical perspective there is nothing suspect about this
passage; but if we try to reduce it to specific pitches, then we can avoid
parallels only at the cost of radically inconsistent analysis.7
Figure P5.7. The start of the trio of Beethoven’s Op. 2, no. 3. Carl Schachter’s reduction postulates a
doubled A3 that is not motivated by the music; my alternative features a regular melodic sequence in a
shifting scalar background.

Summarizing reduction, is, paradoxically, both the most speculative and

least problematic form of reduction: here we boil a complex passage down
into something simpler, finding familiar musical patterns guiding longer
stretches of musical time. If nonharmonic reduction reveals the rules of the
musical game, then summarizing reduction provides conjectures about a
composer’s strategic ends, analogous to the suggestion that an expert chess
player intends to put pressure on a specific square. In this way it can provide
a hypothetical blueprint, an outline or plan that helps us understand how a
composition might have been made.8 Such analysis is inherently speculative,
requiring a careful alignment between composer and analyst—a good sense
of a composer’s tendencies and characteristic moves. One major
achievement of recent schema theory is to reveal a much wider range of
strategies, directing our attention to a host of common patterns such as the
“Prinner,” “Indugio,” and “Morte.”
Crucially, schemas are often occluded by diminution, with important
schematic notes hidden among a wealth of decorative figuration.
Schenkerian theory typically understands diminution as a matter of less
important notes representing more important notes “at a deeper structural
level,” a “standing for” that derives from our experience with the
nonharmonic system. This habit of thinking is so ingrained that it is
surprising to realize that it is superfluous. For consider a summary of a
movie: when we omit episodes extraneous to the main plot, we do not imply
that this material represents or prolongs the core narrative. Or imagine a
group of improvising actors beginning with the skeletal outline of a plot to
be embellished spontaneously.9 Or a skilled chess player pausing their
offense as they fight off an opponent’s threat. In all these cases, there is a
higher-level plan that contains within it a variety of subsidiary lower-level
events, which we can eliminate from our summaries without invoking the
notion of representation.10
From this point of view, summarizing reduction gives us something like a
paraphrase of a composer’s musical ends, akin to those we apply to sports
or games or movies. Though these reductions can be important analytical
tools, they are by their nature tentative: we should no more aspire to come
up with a singular correct summary of a piece than we should strive to
produce the perfect paraphrase of a movie. (For this reason I am skeptical of
attempts to automate or quantify the process of summarizing reduction,
which inherently requires high-level understanding.11) Nor is there any
reason to expect that our paraphrases will take some specific form—that
they will conform to the principles of species counterpoint, that bass lines
will always move from tonic to dominant, or that the melodic structure of an
entire piece will necessarily descend from a tonic-triad note to the root. Here
I think the music-theoretical community could benefit from more
informality: rather than conceiving of summarizing reduction as a syntactic
project analogous to nonharmonic reduction, it would be better to imagine it
as producing hypotheses about composers’ intentions. We will return to this
point in chapter 9.
 1
Ian Quinn first alerted me to the complexities of nonharmonic reduction.
2
Christoph Bernhard ([~1640] 1973) was one of the first theorists to distinguish surface and
abstract voices.
3
Thanks here to Ian Sewell. The scalar approach is both highly constrained and yet capable of
recreating some characteristic features of Schenkerian analysis.
4
Burstein 2020 tries to justify traditional pedagogical strictness, but I worry that his arguments
resort to unjustified ideology (e.g., that upward resolutions of the cadential six-four chord necessarily
represent the standard descending resolution, or that occasional inner-voice parallel fifths create a
perceptible weakening of voice independence).
5
The extreme notes of an arpeggiation pattern tend to be treated as belonging to a concrete voice
when they form the outer voice of a musical texture; thus Mozart would typically avoid parallels
between a concrete melody and the lowest note of the Alberti patterns in Figure P5.4, which represent
a concrete bass line; similarly, Beethoven would typically avoid parallels between a concrete bass line
and the highest notes of a pattern such as that in Figure P5.7. Once again, this recalls figured-bass
thinking, which often combines concrete soprano and bass with abstract harmonies.
6
In pitch-class space there is no distinction between parallel fifths and parallel fourths, and parallel
octaves occur only in the (extremely rare) situation where we want to postulate scale-like backgrounds
with scale steps of size 0. Surface voices move inside the pitch-class background, and cannot form
parallels with it.
7
Weber (1817–1821) 1846, p. 827, explicitly argues that arpeggiated lines are not actual parts but
only “imagined” parts, and hence are less strictly bound by the rule against parallel fifths (p. 836).
8
For similar approaches to reduction see Temperley 2011a, Cook 1989, Sewell 2021. The
Schenkerian analyses in Figures P5.3 and P5.7 are discussed by Sewell; the scale-based alternatives
are my own.
9
Gjerdingen (2007, chapter 1) cites the commedia dell’arte as a model for musical schemas. Reef
2020 provides a nice example of a Bachian template that sits somewhere between schema theory and
Schenkerian theory.
10
Yust (2015, 2018) questions the connection between representation and prolongation, though he
ends up much closer to standard Schenkerian practice than I do.
11
See, for example, Lerdahl 2001 and Kirlin 2014.
 5
Nonharmonic Tones

Western music has traditionally been a music of rules: rules about how
chords are constituted, rules constraining the relation between harmonic and
nonharmonic, rules governing the avoidance of parallels or antiparallels,
rules about which notes can be doubled, rules about how chords progress,
and countless others besides. The rules regulating nonharmonic tones are
perhaps the most important, in large part because they ensure a certain sort
of teachability—allowing us to mark our students right and wrong without
consulting our aesthetic preferences. They are also intimately connected to
the practice of reductive analysis, the replacing of nonharmonic tones with
their harmonic partners. In many styles this produces a harmonic skeleton
that can be said to carry the weight of the musical argument. The
nonharmonic system thus allows theorists to play the role of structural
engineers, distinguishing the load-bearing and ornamental walls of the
musical house.
All of which works much better in theory than in practice. For as we will
see, received views of nonharmonicity are empirically inadequate,
assimilating suspensions to decorations even though they resist our best
attempts at reduction. They are plagued by inconsistency, as when they treat
the permissibility of parallel fifths on different musical levels. And they are
aesthetically problematic, encouraging us to ignore the myriad ways in
which nonharmonic tones can contribute to fundamentally harmonic
narratives. We do not have to stray far from the Austro-Germanic canon to
find composers like Domenico Scarlatti blatantly disregarding traditional
rules—or styles in which the distinction between “chordal” and
“nonchordal” is inherently blurry.1 These problems have provoked some
theorists to protest against the very notion of nonharmonicity, searching for
alternatives in which harmonic and nonharmonic are on a more equal
footing.2 Indeed, this is one motivation for the recent turn to figured-bass
nomenclature, where there is a less-sharp distinction between the harmonic
and linear realms.3
One issue here is the theoretical tendency to overvalue rules at the
expense of concrete idioms, a Platonic streak running throughout the history
of Western musical thought. Beyond that, however, is a deeper and more
uncomfortable fact about the language itself, namely that composers
sometimes viewed the nonharmonic system with a degree of irony, writing
music in which purportedly nonharmonic notes had important harmonic
effects. The nonharmonic system was less like a system of moral
commandments than it was like the speed limit, often seen as a nuisance and
sometimes obeyed only grudgingly.4 (This is obviously true of Monteverdi,
Domenico Scarlatti, and Mahler, but I believe it holds for Palestrina and J. S.
Bach as well.) Like the speed limit, the rules were real enough, explaining
behavior that would otherwise remain inexplicable. But a deep
understanding of earlier music requires acknowledging the gap between the
theoretical precepts of the nonharmonic system and the meaning of the
music produced under its aegis. This leads to an awkward terminological
problem, for we should neither claim that “nonharmonic tones are harmonic”
nor assert its negation. Nonharmonicity, or perhaps nonharmonicity, is an
inherently unstable notion, both useful and misleading, explanatory and yet
false to the music it purports to describe.5
Misunderstanding this point sometimes leads pedagogues to discourage
the very techniques that enabled composers to express themselves freely.
Thus we forbid practices that were merely somewhat uncommon:
consecutive strong-beat parallels, parallels masked by suspension, and
antiparallel fifths in four or fewer voices.6 Haydn’s reported comments about
Kirnberger aptly express this point: “too cautious, too confining, too
everlastingly many infinitely tiny restrictions for a free spirit.”7 This
persnickitude is often associated with the abdication of aesthetic
responsibility, for by forbidding antiparallels we allow ourselves to avoid
questions about whether a particular passage is pleasing or stylistic: the
whole subject is simply off the table. An alternative would be to
deemphasize rules in favor of more directly musical questions, emphasizing
fluid creativity over legalistic niceties. Sometimes what looks like rigor is
really just a desire to make it through the day with our red pencils intact.
 1. The first practice and the SNAP system

The first great era of nonharmonic control was based on what I call “the
SNAP system”: suspensions, neighbors, anticipations, and passing tones.
Though often said to originate with Josquin, this prima prattica stabilized
somewhat later, in the music of Palestrina and contemporaries such as
Lassus and Byrd.8 Earlier polyphony features an array of incomplete
neighbors, upward-resolving suspensions, and the occasional inexplicable
moment (Figure 5.1.1). The Palestrina style largely eliminates these,
retaining the cambiata as the sole surviving incomplete neighbor. The
resulting music is astonishing in its systematicity, almost completely free of
the occasional glitches that mark most human productions. Among his other
achievements, Palestrina was one of the great proofreaders in human history.
Figure 5.1.1. Unusual nonharmonic tones. In mm. 106–107 of Josquin’s motet “Liber generationis,”
we have both an incomplete neighbor and an upward-resolving suspension (A–B on beat 4). In mm.
18–19 of Clemens’s “Concussum est Mare,” we have a nonharmonic tone that is both leapt-to and
leapt-away-from.

This music is highly susceptible to computational analysis—partly

because musical voices are directly embodied in the notation, partly because
it contains relatively few changes of tonal center, and partly because there is
little need for concepts such as the “nonharmonic consonance.”9
Consequently, it takes just a few hours of programming to produce a
reasonably accurate survey of nonharmonic usage in Palestrina’s work.10
The success of such simple algorithms testifies to the tight constraints
governing the vocabulary: virtually all the nonharmonic tones can be labeled
and removed algorithmically, leaving behind a triadic skeleton exactly as
described in textbooks. This is perhaps the closest that Western music has
ever come to the theorist’s dream of purely consonant harmony decorated by
clearly nonharmonic embellishment.
 Palestrina’s nonharmonic tones can be said to be decorative in the specific
sense that they rarely obscure the entire duration of a consonance. This
means, first, that each individual nonharmonic tone takes up only a part of a
consonance; and second, that nonharmonic tones rarely conspire so as to
obscure a consonance entirely. Consonances are thus real sonic objects
usually sounding at some point during their lifespan. In later music,
harmonies are often obscured in a way that suggests they are abstract entities
or conceptual templates (Figure 5.1.2).11 The main exception here is the
suspension, which even in Palestrina can occupy an entire sounding
consonance (Figure 5.1.3). This is a first hint that suspensions have a unique
role in the nonharmonic system.
Figure 5.1.2. In m. 51 of Élisabeth Jacquet de La Guerre’s Sonata no. 1, I, an upper-voice échappée
occurs as the suspension resolves, so that the chord never sounds. Chopin’s Mazurka Op. 6, no. 3, m.
89, presents the same échappée, only now lasting the entire duration of the dominant chord.
Figure 5.1.3. Suspensions in Palestrina lasting the entire length of a consonance: the Agnus from the
mass De Beata Marie Virginis (III), m. 23, and the Gloria from the mass Veni Sancte Spiritus, m. 3.

One advantage of computers is that they allow us to look at repertoires

cross- sectionally—for example showing us every instance of a certain
progression, or every nonharmonic tone that resists standard reduction. This
cross-sectional view can reveal regularities that would otherwise escape the
analytic eye. For instance, it turns out that Palestrina almost always uses
anticipations in the context of suspensions—to the point where it might
make more sense to speak of a “suspension/anticipation system” rather than
of anticipations as a freestanding class of nonharmonic tone.12 At the same
time, and in uncomfortable conflict with this point, decorated anticipations
form one of the style’s most ambiguous idioms: the top voice on the left
passage in Figure 5.1.4 is most naturally read as an anticipation decorated
recursively by a neighbor tone. Yet the same reading of the passage on the
right would imply parallel fifths between outer voices.13 Thus a single
melodic formula can play different structural roles. When both readings are
possible, they can lead to very different interpretations: the suspension in the
last measure of the left passage in Figure 5.1.4 can be understood to resolve
either on the second or the third beat of the measure, depending on whether
the first F♯ is treated as a chord tone or an anticipation. (For what it’s worth,
I prefer the latter interpretation.) And one occasionally finds passages
suspended between these two possibilities, submitting to no fully satisfactory
analysis (Figure 5.1.5).14
Figure 5.1.4. Different meanings for similar figures. In mm. 163–165 of the Credo of Palestrina’s
mass In duplicibus minoribus (II), neither the B nor the A is harmonic; it is sensible to identify a lower
neighbor A decorating a quarter-note anticipation B. In mm. 19–20 of the Sanctus of the mass De
Beata Marie Virginis (I), however, we need to read the A as a passing tone to avoid the parallel B–
A/E–D fifths.
Figure 5.1.5. An idiomatic passage in m. 9 of the Kyrie of Palestrina’s mass Dilexi quoniam.
Underneath I provide two analyses. In the first, there is a C64 chord and the bass B and A are passing;
in the second, there is no C64 and the A–G–A is a half-note anticipation decorated by a neighbor.

Computational analysis also forces us to confront practices that we might

prefer to ignore, including 64 chords that act as fully fledged harmonies or
passing and neighboring tones that cut across the binary division of the beat
(Figures 5.1.6–7).15 By far the most significant of these is the Renaissance
seventh, a sounding seventh chord, formed by suspension, that cannot be
reduced to a consonance—an idiom appearing as early as Dufay and used
throughout the Renaissance.16 Figure 5.1.8 shows a short passage containing
three examples in close succession. One would like to say that the tenor B♭
on beat 3 of the second measure is “just a suspension,” but replacing the
nonharmonic tone with its tone of resolution forms a second seventh chord,
now in first inversion rather than root position (i.e., C–E♭–G–A rather than

C–E ♭ –G–B ♭ ). The problem is the starred G, a step away from the
suspension’s resolution; it is approached and left by leap and hence cannot
be assimilated to any known species of Renaissance dissonance. We are
forced either to invent an entirely new nonharmonic tone solely to account
for this figure, or else admit that there are suspensions that cannot be
reduced to a consonant background.

Figure 5.1.6. In mm. 9–11 of the Benedictus of Palestrina’s mass Io mi son giovinetta (1570), the
fourth above the bass resolves a suspension. In mm. 5–6 of the Kyrie of the mass Ave regina
coelarum, the leapt-to A is harmonic, which suggests that the alto F is similarly harmonic; otherwise,
it has to stand for the following E.
Figure 5.1.7. “121” neighbor and passing tones that cut across the binary division of the beat,
sometimes called “fake suspensions,” from Palestrina’s Pope Marcellus Mass (Gloria, m. 9) and
Adrian Willaert’s motet O salutaris hostia (m. 36). This idiom often gives rise to a 64 chord,
occupying the first half of the nonharmonic tone’s duration.

Figure 5.1.8. Three irreducible seventh chords in mm. 72–75 of the Gloria of Palestrina’s mass Veni
Sancte Spiritus. The starred notes cannot be dissonances according to standard contrapuntal theory.

For all its strangeness, this idiom admits a fairly straightforward

explanation. Let us suppose, first, that the voices not sounding the dissonant
suspension tend to form a consonance, and second, that this consonance
tends not to contain the tone of resolution except in the case of the 9–8
suspension over a root-position triad. If the suspended note is a dissonant C,
this means that the remaining voices can sound either G or G–D (a 4–3
suspension), a root-position B♭ triad (the 9–8), or some subset of D–F–A, the
maximal consonance not containing C or B. It is this last option that
produces the Renaissance seventh: since the collection D–F–A is dissonant
with the suspension resolution B, the voices have to move as the suspension
resolves (Figure 5.1.9).17 Everything makes sense so long as we abandon the
idea that there is a definite consonance governing the entire texture at the
moment of suspension. It would be easy to abandon this assumption were we
talking about late-medieval pieces whose lines were composed successively.
The difficulty is that the assumption of a governing consonance works well
in virtually every other late-Renaissance context.

Figure 5.1.9. The Renaissance seventh’s standard resolutions. From a modern perspective, all involve
root progressions of a descending third or descending fifth.

Suspensions are also unique in being the only nonharmonic tones that
regularly license parallels at the level of the harmonic skeleton—both in the
Renaissance and later (Figure 5.1.10). This poses another obstacle to the
view that suspensions are merely decorative, as it seems incoherent to say
that the first passage in Figure 5.1.11 is grammatical, the second is
nongrammatical, and the first stands for, embellishes, or represents the
second. The challenge is exacerbated by the cadential idiom in Figure
5.1.12; here parallels are often said to be permissible because they occur
“only on the surface,” disappearing when we consider the harmonic
skeleton. But how can it be that some parallels are allowed because they
occur only on the surface while others are allowed because they occur only
in the harmonic skeleton?
Figure 5.1.10. Suspensions masking parallel fifths in the harmonic skeleton, in the Kyrie of
Palestrina’s mass Assumpta est Maria (m. 54), and Bach’s chorale Ich hab mein Sach Gott
heimgestellt (BWV 351, Riemenschneider 19, m. 7).

Figure 5.1.11. The progression on the left is common, while the progression on the right almost never
appears. Should we say that the former stands for the latter?
Figure 5.1.12. An idiom in which parallel fifths appear “on the surface.” Here the anticipated tonic
appears alongside 5̂–4̂–3̂ to create fifths. For an example, see Bach’s Chorale “Freuet euch, ihr
Christen” (BWV 40.8, Riemenschneider 8, m. 4).

The suspension’s ambiguous status, I suspect, is connected to its cadential

role as a signal for musica ficta and the leading tone. Equally important,
though less easily quantified, is its role in generating the tension released by
the cadence itself: suspensions are dissonant harmonic events and not mere
decorations. Thus, rather than being sprinkled throughout the phrase, as we
might expect from the example of fourth-species counterpoint, suspensions
tend to cluster near cadences (Figures 5.1.13–14). A passage like Figure
5.1.15 might look ionian to the modern eye; to a Renaissance musician, the
suspension chain would signal an impending cadence even before the
leading tone appears. Indeed, Renaissance sevenths tend to have cadential
function: the most common form resembles and evolves into the baroque
predominant seventh, while other forms occasionally produce sounding
dominant sevenths (Figure 5.1.16).
Figure 5.1.13. Nonharmonic tones in Palestrina’s masses.
Figure 5.1.14. Suspensions in Palestrina’s masses.
Figure 5.1.15. The second Agnus from Palestrina’s mass Assumpta est Maria, mm. 44–47. The
suspension chain in the last measure should lead us to expect a D major chord, which in fact does
appear.

Figure 5.1.16. Sounding dominant seventh chords in mm. 11–12 of the Sanctus of Palestrina’s mass
Aeterna Christi munera and m. 38 of the Agnus of the mass In semiduplicibus majoribus (II).

Fétis wrongly claimed that these dominant seventh chords originated at

the end of the sixteenth century, mistaking standard Renaissance practice for
a harbinger of the future.18 The paradox is that the Renaissance seventh is
simultaneously new and old—an idiom that antedates the notion of a
“governing triad” while also evolving into the seventh chords of the classical
era. Seventh chords are the inevitable result of attempting to reconcile the
earlier practice with the philosophical view that nonharmonic tones must
necessarily be decorative.19 If the suspension always stands for its tone of
resolution, then the B♭3 in m. 2 of Figure 5.1.8 cannot be a suspension, and
therefore must be harmonic: this syllogism is appropriate in J. S. Bach,
inappropriate in Josquin, and of indeterminate validity during much of the
intervening time. Which means that the idealized consonant style of the
counterpoint textbooks never actually existed: while Palestrina’s music can
almost always be understood as using nonharmonic tones to decorate
exclusively consonant harmonies, it retains idioms that stubbornly resist
nonharmonic reduction—inherited from the past and pointing toward the
future. It is as if Palestrina wrote just before and just after an imaginary
consonant era.
It is natural to contrast the embodied and intuitive grammar of rock
harmony with the more explicit precepts of the nonharmonic system,
codified in books from Tinctoris to the present. But we have just seen that
nonharmonic usage, even in Palestrina’s style, resists simple formalization:
the true grammar of the nonharmonic system, the regularity that can be
inferred from compositional practice, is inherently schematic, more complex
and interesting than the simplifications found in textbooks. The theoretical
tradition’s reluctance to acknowledge this complexity testifies to a
Platonistic conviction that “surface” dissonances have to decorate a
consonant harmonic skeleton. In reality, however, this belief is just an
approximately true generalization, a model we apply to music that was
composed by people who thought differently. It is a fairly good
approximation to be sure: like Newtonian mechanics, it should neither be
rejected out of hand nor accepted as the complete truth.20

2. Schoenberg’s critique

The most famous critique of nonharmonicity occurs in chapter 17 of

Schoenberg’s Harmonielehre:
 I come now to one of the weakest points of the old harmonic system, to the point where it
suddenly abandons its usual procedure and, as I said in Chapter I, is patched up with another
system, which is not a system, in order somehow to include the most familiar harmonic events.
It is remarkable that this point has not yet occurred to anyone: Harmony, its theory, its
pedagogy, is concerned with non-harmonic tones! But non-harmonic matters have just as little
place in a textbook of harmony as do non-medical matters in a textbook of medicine.
(Significantly, the word medizinfremd is not used.) Whatever belongs in such a textbook is
there precisely because it is not non-medical: if it were, it would not be there. The expression,
“non-harmonic” tones, I can interpret only to mean that a number of tones are declared
unsuitable, or under certain conditions unsuitable, for forming harmonies; that such tones,
because they intrinsically lack the ability to form harmonies, i.e. chords (Zusammenklänge),
are designated as having nothing to do with music and consequently are thrown out of the art
and out of its theory. [ . . . ] Either there is no such thing as non-harmonic tones, or they are not
non-harmonic.21

As written, Schoenberg’s complaint seems to be a purely verbal one that

could be resolved with a simple change of terminology: instead of
“nonharmonic tone” we could agree to use a more neutral term like “linearly
constrained tone” or “tone of the second classification.” In this respect
Schoenberg betrays a naïve conception of language, as if meaning was
determined by words themselves rather than our behavior—the same rigidity
that might lead someone to insist that a sea cucumber must be a vegetable, as
a simple matter of lexicography. Yet behind the verbal complaint we can
discern the deeper observation that nonharmonic tones range from what
might be called “merely connective,” serving primarily melodic aims, to
those whose purpose seems to be to create a specific sound. In these latter
cases, uncritical allegiance to “nonharmonicity” can keep us from
understanding a genuinely harmonic logic.
Figure 5.2.1 returns to the 64 chord in Palestrina’s mass Descendit
angelus Domini (Figure 3.4.6). The example provides two interpretations of
the chord, one recording the standard claim that suspensions represent their
tones of resolution; from this point of view, the second inversion triad is
“really” an A–F–C sonority whose third is decorated by an inessential
suspension. The issue with this reading is that the second-inversion triad
appears in the context of a common sequential pattern, two-step clockwise
motion on the circle of diatonic triads, producing ascending-fifth harmonies
in which root and third repeatedly descend; it is difficult to do justice to this
sequence without treating the 64 as a harmony. Another problem is that the
64–53 pattern is ubiquitous in Palestrina, accounting for almost 10% of his
suspensions (Figure 5.2.2). Meanwhile, 64 suspensions almost never resolve
to 63 chords, with 64–53 resolutions outnumbering 64–63 by a factor of
about 30 to 1. Thus, rather than being accidental or inessential, the 6–5
motion is virtually mandated by the presence of the 4–3. Once again, it
seems perverse to argue that something ubiquitous “stands for” something
that almost never occurs. Like the Renaissance seventh, the 64 is a
suspension with harmonic significance.

Figure 5.2.1. Two interpretations of the 64 chord in the Kyrie from Palestrina’s Descendit angelus
Domini, mm. 39–40.
Figure 5.2.2. The most common suspension figures in Palestrina’s masses, transposed so that the bass
of the first sonority is C.

There is an old-fashioned view that Renaissance music is fundamentally

linear: a matter of lines and intervals, with harmonies simply the byproducts
of melodic logic.22 This picture is consistent with the formalist vision,
promulgated by Babbitt and others, that musical styles are logical systems
with different axioms or postulates.23 Historically, Renaissance counterpoint
functioned as the first of a triad of disconnected styles, with functional
harmony and twelve-tone music its canonical successors. (We see traces of
this view in curricula that contrast the vertical approach of “tonal harmony”
with the supposedly horizontal approach of “modal counterpoint.”) Yet
Palestrina’s music is filled with idioms like the 64–53, which are harmonic
in the sense that they cannot be explained as the byproducts of independent
melodic lines obeying the prohibitions of standard counterpoint.24 Chapter 6
will argue that functional harmony began with precisely these sorts of
idioms, understood intuitively and not explicitly theorized; over decades
they gradually strengthened and grew more complex, metamorphosing into
the lawlike regularities of the eighteenth century.
One can make a similar point about the “ascending subdominant”—an
ascending-fifth progression from IV to I, where one of the voices moves
from 6̂ to 1̂ through a “merely passing” 7̂; this ascending passing tone can
create a sense of dominant functionality, particularly when reinforced by 4̂–3̂
to create a resolving tritone.25 Ascending subdominants are found
throughout the Renaissance and baroque and are particularly important in J.
S. Bach (Figure 5.2.3, §7.7).26 The analytical challenge is that they span the
gamut from harmonic to nonharmonic: some of the ascending leading tones
in Figure 5.2.4 require a Roman numeral, others support one only shakily,
and others do not call for one at all. To do justice to this continuum of
musical practice, we need to recognize that nonharmonic tones can
participate in harmonic narratives.
Figure 5.2.3. Ascending subdominants in a range of composers.
Figure 5.2.4. Ascending subdominants in Bach’s chorales. In “Ermuntre dich, mein schwacher Geist”
(BWV 248(2).12, Riemenschneider 9, m. 3), the leading tone supports a clear harmony. In
“Menschenkind, merk eben” (BWV 318, Riemenschneider 18, m. 10), the harmony is more unusual.
In “Es ist das Heil uns kommen her” (BWV 86.6, Riemenschneider 18, m. 9), the leading tone is
passing.

The previous section showed that some contrapuntal configurations resist

reduction, in the sense that there is simply no way to reduce them to a
consonant background. Here we are dealing with the more delicate fact that
even when we can “reduce out” the nonharmonic tones, we may have good
analytical reason not to do so. Sometimes a nonharmonic tone has harmonic
effects. And some progressions only appear when accompanied by
nonharmonic tones—such as the V2–I in Figure 1.5.1, which Bach used only
when passing tones disguise the unusual treatment of the bass seventh. In
still other cases, a variety of nonharmonic tones conspire to produce a single
sounding sonority, such as Chopin’s beloved dominant thirteenth: that so
many different contrapuntal configurations generate the same harmonic
object suggests that harmony is goal rather than byproduct (Figure 5.2.5).27
A similar point could be made about the Tristan chord: while it often can be
explained as the byproduct of nonharmonic tones, it is so ubiquitous that it is
reasonably considered harmonic (Figure 5.2.6). And I am sure that readers
can cite other moments where the nonharmonic system seem to be a means
for producing nontriadic ends: the F ♯ in Monteverdi’s “Tu se’ morta,”
Domenico Scarlatti’s countless dominant-sevenths-with-an-added-tonic, the
gently overlapping horn calls at the end of the first movement of
Beethoven’s “Lebewohl” sonata, and so on.

Figure 5.2.5. Many different nonharmonic tones conspiring to produce the same “dominant
thirteenth” sonority in Chopin’s mazurkas: Op. 6, no. 3, mm. 89–90; Op. 41, no. 1, mm. 41–44; Op.
41, no. 2, mm. 1–2; and Op. 41, no. 2, mm. 18–20.
Figure 5.2.6. Three appearances of the “Tristan chord” that could be analyzed as the byproduct of
nonharmonic tones.

All of which, I propose, shows that Schoenberg was fundamentally right:

a musically sensitive theory cannot treat nonharmonic tones as having
“nothing to do with music” or even as having nothing to do with harmony
itself. Sometimes tones which are properly classified as “nonharmonic,” at
least from the standpoint of textbook theory, produce chords that are clearly
“harmonic” in the broader compositional sense. (As Schoenberg put it,
“these are chords: not of the system, but of music.”28) What then is achieved
by describing these notes as “nonharmonic”? The safest answer, I think, is
that nonharmonic tones are melodically constrained in ways that harmonic
tones are not. Consider Palestrina’s 64 chords: even if we grant that they are
harmonically significant, it is nevertheless true that the fourth is almost
never doubled and almost always resolves downward—indeed the
downward resolution of the suspension is as close to an inviolable law of the
style as one can find. By contrast, the sixth of the 64, though usually moving
downward, is more flexible, available for doubling and sometimes moving
differently.29 (A similar point can be made about the Renaissance seventh,
which invariably resolves down by step.) The real significance of the
nonharmonic system lies in these linear constraints, which forbid certain
vertical configurations except in the context of specific melodic obligations.
 Schoenberg, of course, went farther than this, denying the existence of
“nonharmonicity” altogether. As a historical claim this is flatly
unconvincing, asking us to ignore unmistakable regularities in nonharmonic
usage. (The Monteverdi-Artusi controversy was about something, after
all!30) As an aesthetic proposal, however, it becomes more plausible: our
experience with the whole panoply of twentieth-century music—from
Debussy and Schoenberg to jazz and rock—has taught us a more flexible
attitude toward the harmonic domain, making those long centuries of
compositional obedience seem somewhat baffling in retrospect. How, we
might wonder, could a composer as inventive as Mozart have been so utterly
conventional when it came to dissonance treatment? Was the nonharmonic
system merely a convenient tool for writing music quickly? Did composers
invest its conventions with metaphysical force? In asking these questions,
we can find ourselves in an uncomfortable relation to the past, valuing its
music while questioning the conceptual framework intrinsic to its creation.
Certainly the nonharmonic system was key to an incredible tradition of well-
ordered composition. Our challenge, as inheritors of this tradition, is to
balance our love for this style with our sense that its underlying syntactical
principles have lost their force. Once, the grammar of nonharmonicity
helped justify the music itself; now the music is the only thing sustaining the
grammar.

3. Monteverdi’s “Ohimè”

Schoenberg argued that composers sometimes used the nonharmonic system

to introduce harmonically significant notes, expanding their expressive range
while paying lip service to traditional counterpoint rules. This is more or less
uncontroversial in the case of Monteverdi, a composer born less than fifty
years after Palestrina. Glancing through his scores, we recognize all manner
of unusual moments, including unprepared and unresolved sevenths, voices
leaping away from suspensions, pedal tones, leaping anticipations,
incomplete neighbors, upward-resolving suspensions, Brahmsian rhythmic
syncopations, parallel perfect intervals, and notes that seem to defy reduction
altogether (Figure 5.3.1). Nonharmonic tones occasionally conspire to
produce nontertian sonorities or even challenge the very notion of
“underlying harmony” (Figure 5.3.2).31 And yet for all this strangeness
Monteverdi was a man of his time, a composer working in the broad
tradition of the late Renaissance. Critics love to wax offended at his
“shrieking” dissonances and unprecedented text-painting; but to ears raised
on Metallica and Penderecki, these dissonances can pass unnoticed, being
less salient than the bold triadic language he shares with Marenzio and
Morley.32 Paradoxically, hyperbolic rhetoric functions to legitimize the
nonharmonic system, suggesting that even minor contrapuntal unorthodoxies
are shocking to the ear: this is Victorian primness gussied up in postmodern
garb.33

Figure 5.3.1. Some contrapuntal oddities in Monteverdi’s madrigals. In A the upper voices are easiest
to understand as rhythmically displaced passing tones rather than suspensions: the F5 in m. 1 is
dissonant and hence not a preparation. In B the lowest voice arrives at F♯ a quarter note before the rest
of the voices; this could be either a leapt-to anticipation or a weak-beat suspension of the E. In C the
top voice resolves the suspension D upward to E. Example D is a standard cadential figure occurring
under an upper-voice pedal D5. In E the C5 seventh is unprepared. In F the top voice leaps away from
the seventh (G5 down to C).
Figure 5.3.2. The nonharmonic tones in A conspire to produce fourth chords. In B the harmonies are
difficult to identify.

Let us therefore look at Monteverdi’s setting of Giovanni Battista

Guarini’s “Ohimè, se tanto amate,” published in 1602 in his fourth book of
madrigals. The piece presents a range of characteristically Monteverdian
devices, including unprepared sevenths, suspensions resolving by leap,
incomplete neighbors, extraordinary chromaticism, and a few notes that
seem to defy nonharmonic analysis altogether. Along the way, we will revisit
some contrapuntal techniques from the previous chapters—including a
wealth of sequences and near sequences exploiting the geometry of diatonic
dyads and triads.
I hear the piece in four sections: the first stretches to m. 19 and makes use
of the diatonic third’s basic voice leading; the second stretches to m. 38 and
is constructed from two 9.5 bar phrases; the third, mm. 39–43, lasts just five
measures but contains the central poetic and dramatic conceit; while the final
section, mm. 44–67, features two parallel twelve-bar phrases. The first
contrapuntal anomaly occurs in m. 2, where the falling-third “alas” motif
seems to articulate a V7 chord with an unprepared seventh (Figure 5.3.3). If
we imagine the V7 to start on beat 3 then the upper voices’ A–C are “rogue
notes”: nonharmonic tones not connected by step to any chord tone. (The
variant in mm. 3–4 compounds the strangeness by using E♮ where one might

expect E♭.) One interpretive option is to take them at face value, as leaping-
third anticipations not entirely foreign to the tradition (e.g., the first passage
in Figure 5.4.6). Another is to imagine a hocket with rhythmic and registral
displacement, analyzing the high-register A and C in m. 2 as incomplete
neighbors attached to the preceding low-register G and B ♭ (Figure 5.3.3).
This is how the figure later appears (e.g., top two voices of mm. 6–8, and
alto and tenor in mm. 9–11). This reading also reflects the text, sundering the
sighs into male and female (low and high), rather than allowing them to
sound together as a unified syntactical structure.34 In either case, the crucial
point is that nonharmonic reduction has become an interpretative rather than
mechanical process: as in late nineteenth-century music, it is genuinely
unclear how to reduce the surface to a well-behaved harmonic skeleton—
with none of the answers being completely convincing.

Figure 5.3.3. The beginning of “Ohimè,” along with a reading in which the incomplete neighbors
embellish their predecessors.

The next six measures (“se tanto amate di sentir dir “Ohimè” / “if you so
love to hear me say alas”) present a familiar ascending-fifth sequence that
develops the preceding bars both melodically and harmonically (Figure
5.3.4).35 Cantus, tenor, and alto sing a six-beat melodic canon that
incorporates the “alas” motif, with Quinto and alto doubling at the third
below; this canon is essentially the ascending version of the diatonic third’s
basic voice leading, split into different octaves and with the bass sounding
chord roots. The sequential unit’s third half-note is harmonically variable,
supporting a first-inversion triad with a bass that either sustains the
preceding bass note or descends by third.36 As in rock modality, these
ascending-fifth root progressions produce descending-step voice leading.37
One can admire the compositional technique in designing the incipit of the
canon to fit with the “Ohimè” continuation, so that what initially sounds like
development transmutes into something more like resolution or explanation.
Figure 5.3.4. “Ohimè,” mm. 5–19, with brackets marking the six-beat sequence. Underneath, an
analysis of the opening canon.

The next text fragment (“deh perché fate” / “then why do you”) features a
new motive that overlaps with the end of the canon; here the suspensions
resolve by leap.38 Figure 5.3.5 analyzes the passage as a distorted ascending-
third sequence. The outer-voice counterpoint here is essentially parallel, with
the top voice initially singing chordal fifths and then roots (“thinking within
the chord,” §3.7). Monteverdi obscures this structure in two ways: first by
reversing the order of the phrases in the sequential repeat, and second by
reaching a cadence at the end of the first unit. This makes the second unit
feel like both a continuation and a phrase beginning; our ears hear some of
the organization in Figure 5.3.5 but only obscurely. (The A cadence also ends
the ascending-fifth harmonic motion that began in m. 3, B ♭ –F–c–g–d–A
crossing motivic and phrasal boundaries.) The reordering transforms the
relatively standard deceptive progression A–B♭ into the more dramatic A–F
(“the major-third system”). The passage continues with “chi dice ‘Ohimè’
morire?” (“kill the one who says ‘Alas?’ ”), returning to the A major triad via
another major-third progression, with the cadential dominant extensively
decorated by incomplete neighbors, an unprepared seventh, and octave-
displaced passing motions—a free counterpoint of lines decorating a stable
but hazy harmony.
Figure 5.3.5. A hypothetical derivation of mm. 8–15. Question marks are suspensions resolved by
leap. The result is an ABBA palindrome.

The piece’s second large section starts with the upper two voices moving
in parallel thirds (Figure 5.3.6), freely harmonized by descending steps and
descending fifths, and shifting position from root-third to third-fifth. (The
underlying logic here is shared with the Pachelbel progression, which
features descending melodic voices and is codified as a specific piece of
functional vocabulary; in the Renaissance, the same basic relationships are
exploited more freely.) This quasi-sequence dissolves in a wonderful
reduction of the harmonic rhythm, the voices singing a slow, sad “doloroso”
on ascending scale fragments. The voices take turn ascending by fourth and
fifth, harmony secondary to linear motion until the final cadence on D. The
phrase repeats up a fourth and with a fuller texture; the last melodic ascent is
displaced upward by third, producing a ♭ 9 incomplete neighbor and an
unapologetic seventh chord in m. 37. It is interesting that the two phrases
both start on a root-position D triad: the first begins with the descending
second (d–d–C–F– . . .), the parallel thirds sounding root-third, while the
second begins with the descending fifth (D–g–F–B♭– . . .), the parallel voices
sounding third-fifth.39

Figure 5.3.6. “Ohimè,” mm. 20–38, along with a reduction of each phrase’s opening.

Figure 5.3.7 presents the poem’s “turn” (“but if, my love, you wish to let
me live and live for me . . .”), set in a lydian-inflected B♭ and one of just two
phrases in the piece that are not immediately repeated.40 The shift to a faster
tempo and brighter tonality effectively conveys the poem’s move toward
amatory hope. Harmonically it is the piece’s most straightforward music,
replete with functional progressions and organized around an ascending bass
line from B♭2 to B♭3, with the upper voice outlining a stepwise descent from
F5 to B♭4.41 Contrapuntally, it is noteworthy for the dissonances in mm. 41–

42, particularly the cluster G–A–B♭ and the semitone A–B♭.

Figure 5.3.7. “Ohimè,” mm. 39–43.

The final section delivers the punchline, the promise of endless satisfying
sighs shared between lovers (Figure 5.3.8). We begin with emphatic V–I–V
progressions in G and C minor, the second containing undisguised parallel
fifths. The music then returns to the opening “Ohimè,” now harmonized with
parallel first-inversion triads; these create deceptive-resolution pairs (e.g.,
D–E♭, C–d, B♭–c), leading to a b6–g cadence closer to Poulenc than standard
Renaissance harmony. (Here the opening bars’ octave-displaced hocketing
has been regularized to a single line.) The lower voices then echo this music,
alternating tonic and dominant dyads in G dorian and recalling some of the
analytical challenges of the opening (Figure 5.3.9).
Figure 5.3.8. “Ohimè,” final section.

Figure 5.3.9. In mm. 53–55, we can read harmonies changing in quarter notes, as in the surrounding
phrases, or syncopated half-note motion.

We then get a varied repeat of this twelve-bar phrase: the homophonic

declamation returns, replacing the V–I–V with I–V–I, and transforming the
fauxbourdon “down a third/up a step” progression into descending fifths
(mm. 60–61). The lower two voices articulate the basic voice leading and the
top voice doubles at the third. (The interchangeability of descending fifths
and “down a third up a step” fauxbourdon is characteristic of baroque
tonality, as we will see in chapter 7.) As before, the ambiguity of the opening
“Ohimè” seems to be resolved, with each note assigned its own chord.
Ambiguity returns, however, with the utterly baffling measure that follows.
My best guess at an analysis is shown in Figure 5.3.10, where I try to capture
the sense of half-note harmonic motion.42 Crucially, this requires turning
around the ubiquitous “Ohimè” figure, so that it has appoggiaturas rather
than escape tones: in almost every previous appearance, the unaccented
“Ohi” has been nonharmonic, with the accented “mè” the genuine harmony;
here it is difficult to make that reading work with the two inner voices. (This
reinterpretation has a nice poetic significance, mirroring the transformation
from sighs of sorrow to sighs of passion.) We are left with music that makes
sense to the ear—a descending scalar rush of notes congealing on the
cadential 64—while resisting technical analysis. The nonharmonic system
has momentarily lost its authority.

Figure 5.3.10. An interpretation of “Ohimè,” mm. 63–67.

This is the sort of passage that provoked the most famous police action in
Western music history, Giovanni Artusi’s attempt to enforce the authority of
the learned tradition. Much of the commentary around this incident centers
on Giulio Cesare Monteverdi’s suggestion that text-setting justifies a
seconda prattica that departs from strict counterpoint. To my mind, however,
that is something of a red herring. More fundamental is the attempt to shift
the debate from inviolable musical rules toward acts of aesthetic judgment.
The crucial point, in other words, is the suggestion that we need to consider
whether a given passage is aesthetically fitting, rather than whether it obeys
the time-independent rules of the prima prattica.43
To be sure, the particular judgment Monteverdi asks us to make is one that
compares musical effects to poetic texts, but put that aside for the moment;
focus instead on the general form of the question. “Do I find this particular
musical passage to be compelling?” That question can be asked without
reference to a text, precisely as we do when modern music takes a more
flexible attitude toward the nonharmonic realm. Here the Monteverdis can
start to seem uncannily contemporary, as if they had anticipated modernity
three centuries too early: for the fundamental freedom they desired—to have
counterpoint judged by its aesthetic effects rather than by an unquestionable
set of musical laws—was one that composers did not obtain until the early
twentieth century. By contrast, Artusi seems to think that aesthetic judgment
should come into play only after we have verified that the music conforms to
the timeless norms of the strict style. What makes Monteverdi radical is the
subjecting of the rules themselves to aesthetic evaluation: for each of his
unusual contrapuntal effects, we are supposed to ask, “do I appreciate this
passage?” regardless of whether it violates traditional norms.44
Construed like this, the debate starts to look like a draw at best.45
Historians have sometimes dismissed Artusi as an ineffectual conservative,
pointlessly resisting the inexorable force of musical progress. This may be
justified so long as we focus narrowly on the question of prima prattica
rules. But if we turn to the more general question “should nonharmonic
tones conform to some rigorous contrapuntal grammar, whatever it may be?”
then we have to admit that history gave an answer more Artusian than
Monteverdian. Yes, the classical tradition incorporated some of Monteverdi’s
specific innovations, but it subjected them to rules as inviolable as those of
the prima prattica: in this sense, Monteverdi was an evolutionary dead end,
an unheard avatar of compositional freedom. While it may be true that Artusi
had a fundamentally scholastic mind, it is also true that the notated tradition
retained a scholastic streak right up to the twentieth century. That
scholasticism still survives, for though we live in a Monteverdian world, we
tend to become Artusi the moment we step into our classrooms.
 4. The standardized second practice

Eighteenth-century composers embraced a number of new contrapuntal

devices: seventh chords, the cadential 64, incomplete neighbors, and
pedals.46 We will discuss the first two of these in chapter 7. Figure 5.4.1
classifies the incomplete neighbors according to metrical weight (accented
or unaccented) and direction of stepwise connection. Three of the four
combinations are common, and two have their own names; the fourth
possibility, an accented incomplete neighbor connected by step to its
predecessor, is extremely rare (Figure 5.4.2). These neighbors are interesting
insofar as the notes they represent, for the purposes of reduction, are not
necessarily the notes to which they are connected by step: in Figure 5.4.3,
for example, the bass B ♭ is a melodic decoration of the following C, but
would be replaced upon reduction by the preceding F.47 Like the other
nonharmonic tones, incomplete neighbors can also decorate each other
recursively, as in Figure 5.4.4, where an incomplete neighbor decorates an
appoggiatura.48

Figure 5.4.1. The four species of incomplete neighbor.

Figure 5.4.2. Two examples of the missing incomplete neighbor (accented and connected by step to
the preceding tone): Chopin’s Mazurka Op. 33, no. 1, mm. 23–24; and “The Streets of Laredo,”
sometimes considered traditional and sometimes attributed to Frank H. Maynard. In both cases the
note can be interpreted as an added sixth.

Figure 5.4.3. An incomplete neighbor in m. 5 of “Wie schön leuchtet der Morgenstern” (BWV 172.6,
Riemenschneider 323).
Figure 5.4.4. An incomplete neighbor to an incomplete neighbor in m. 6 of the second B♭ major fugue
in The Well-Tempered Clavier.

Pedal tones are more complicated, in part because they are both a
rhetorical effect and species of nonharmonic tone: one can create the sense
of a pedal by changing chords over a static bass note, even if that bass is
always part of the prevailing harmony. Formally, a pedal tone is any note
that is initially harmonic and sustained through one or more chord
changes.49 In practice, however, this definition is overbroad, encompassing
such phenomena as Monteverdi’s “suspensions that resolve by leap,” which
may not feel much like pedals. Actual pedal tones are more schematic,
almost always occurring on the tonic or dominant scale degree, and typically
returning to harmonic status before they move: to my knowledge, Mozart’s
piano sonatas contain no exceptions to these principles among their hundred
or so pedals, though J. S. Bach’s chorales do contain a few (Figure 5.4.5).50
Pedals are unique in the nonharmonic bestiary insofar as they are
irreducible: where the other nonharmonic tones can be said to embellish
some specific note, pedals do not; removing them requires recomposition
rather than reduction. Many textbooks dodge this issue by stipulating that the
rest of the music embellishes the pedal tone itself, but I think this conflates
the syntactic process of nonharmonic reduction with the very different
enterprise of summarizing reduction.
Figure 5.4.5. Three unusual pedal tones in the Bach chorales: “Jesu Leiden, Pein und Tod” (BWV
245.14 Riemenschneider 83, mm. 11–12), “Nun danket alle Gott” (BWV 252, Riemenschneider 330,
mm. 3–4), and “O Gott, du frommer Gott” (BWV 399, Riemenschneider 315, mm. 1–2).

By the late baroque, many composers had once again become formulaic in
their treatment of nonharmonic tones. J. S. Bach in particular can be
astoundingly systematic: of more than ten thousand nonharmonic tones in
his chorales, all but about ten are well behaved. Most of the anomalies,
illustrated in Figure 5.4.6, involve what might be called “arpeggiated
anticipations,” where an arpeggiating voice (always moving down by third,
usually from tonic to subdominant) reaches the next harmony an eighth note
early; this is reminiscent of the start of “Ohimè.” There are a few other
unusual moments, including a trio of weak-beat suspensions, the
nonstandard pedals in Figure 5.4.5, a couple dozen “121” passing and
neighboring tones, a few upward-resolving suspensions, and so forth. Here
Monteverdian anarchy has been domesticated, standardized into a “second
practice” whose rules are almost as inviolable as Palestrina’s. This system
remains in force throughout much of the eighteenth and nineteenth centuries,
with few composers coming anywhere close to Monteverdi’s radical
freedom.
Figure 5.4.6. Anomalous nonharmonic tones in the chorales: “Herzlich lieb hab ich dich, o Herr”
(BWV 245.40, Riemenschneider 107, mm. 14–15), “Meinen Jesum laß ich nicht, weil . . .” (BWV
154.8, Riemenschneider 152, mm. 10–11), “Christus, der uns selig macht” (BWV 245.15,
Riemenschneider 81, mm. 11–12).

This is a subtle point, so it is important to be clear. My claim is that the

vast majority of the nonharmonic tones in the Bach chorales can be labeled
according to traditional counterpoint rules: we find suspensions, neighboring
tones, passing tones, anticipations, and the occasional pedal, more or less in
line with textbook theory. There are, in other words, very few outright
obstacles to nonharmonic reduction. But this does not mean that
nonharmonic reduction is analytically unproblematic: one often finds
irreducible pedals (Figure 5.4.5), suspensions masking parallel fifths (Figure
5.1.10), and nonharmonic tones that play a vital role in the harmonic
argument (Figure 5.2.4).51 Nonharmonic reduction is theoretically feasible in
the sense that it is possible to program a computer to do it reasonably well.
But it does not follow that reduction is analytically or aesthetically
advisable. One might say that the syntax of the music allows for a reduction
that sometimes leads us to misunderstand the music’s semantics.52
This standardization imposes a powerful constraint on musical analysis:
given a passage of baroque or classical-style music, the theorist needs to find
an analysis that rationalizes both the melodic and harmonic behavior. That
is, the task is to identify harmonies such that (a) dissonances belong to some
recognized species of nonharmonic tone; and (b) harmonic successions obey
functional norms (§7.1). These requirements are highly restrictive, implying
for example that notes both leapt-to and leapt-away-from must necessarily
be harmonic.53 Thus it is rather remarkable that we can analyze Bach’s
chorales such that more than 95% of the chord progressions conform to a
standard “grammar” of functional harmony, while all but a handful of
nonharmonic tones behave recognizably.
We can put these constraints to computational work by using the
nonharmonic system to “score” harmonic interpretations of particular
musical passages based on how well-behaved its voices are; this is
accomplished by assigning penalties to nonharmonic tones, with higher
penalties for the less-common varieties.54 (One can apply this procedure
recursively, first by scoring all the single-harmony interpretations of a
measure, then all the interpretations with two half-measure harmonies, then
those with four quarter-measure harmonies, and so on.) Figure 5.4.7 shows
that this produces a reasonably good automatic analysis of the opening
measures of the first Mozart sonata; for the sonatas as a whole, it correctly
identifies the chord about 75% of the time, which is an excellent result for a
simple algorithm not specifically tuned to the style.55 Here we see one
practical use for the nonharmonic system—despite legitimate grounds for
skepticism about its metaphysical status.
Figure 5.4.7. A computer analysis of the opening of the first Mozart piano sonata. The analysis is
mostly accurate, with the main issue being the unnecessary chord changes in mm. 6–8.

5. A loophole

The nonharmonic system does not directly deal with sonic resultants, but
only with the linear behavior of dissonant voices—a technical loophole
licensing almost any imaginable dissonance.56 Beethoven’s Ninth Symphony
uses incomplete neighbors to sound a complete harmonic minor scale, the
same scandalous set that stomps away at the opening of Stravinsky’s “Dance
of the Adolescents” (Figure 5.5.1).57 This Schreckensfanfare is one of
several famous moments where Beethoven uses nonharmonic tones to create
unusual harmonies, joined by the Eroica’s “false horn entry,” and the end of
the first movement of the Lebewohl sonata. If such anomalies are rare in
earlier music, it is largely for reasons of taste: as a general rule, earlier
composers intended to create consonant surfaces in which nonharmonic
tones avoided extreme dissonances.58 These intentions start to change over
the course of the nineteenth century even while the musical syntax remains
the same. The risk is that we focus too much on the stable syntax at the
expense of the evolving aesthetics.59

Figure 5.5.1. The same seven-note sonority appears in the last movement of Beethoven’s Ninth
Symphony (m. 208, just before the first vocal entrance) and Stravinsky’s Rite of Spring.

Beethoven used this loophole sparingly, as a special effect; in the “Rondo-

Burlesque” from Mahler’s Ninth Symphony, it is pervasive. Viewed from a
distance, the music articulates clear harmonic zones that act in recognizable
ways—though to be sure, its harmonies are sometimes extended or altered.
But like a pointillistic painting, the notes get harder and harder to rationalize
the closer you look. Figure 5.5.2 reduces the opening forty-five measures to
a harmonic skeleton, trying to synthesize the musical chaos into an easily
playable piano reduction; it is my attempt to imagine the kind of sketch that
Mahler might have used as a basis for his orchestration. That this reduction
is a riot of late-Romantic harmonic invention is one of the main obstacles to
convincing analysis; in making my reduction, I felt a constant worry that
there might be a more logical alternative.
Figure 5.5.2a. A reduction of the opening 23 measures of the Rondo-Burlesque from Mahler’s Ninth
Symphony.
Figure 5.5.2b. A reduction of measures 24–46 of the Rondo-Burlesque.

The piece begins off tonic, an unharmonized melody outlining a C♯ half-

diminished seventh chord. We then hear the same melody transposed down a
major third and extended into a nearly complete scale, the melodic A–B–C
in m. 3 the sequential counterpart to the earlier C♯–D♯–E. After this six-bar
introduction, the music settles into a series of eight-bar phrases. In the first,
the unremarkable Roman numerals paper over several anomalies: the V–
(VI–i) in m. 9, for example, would normally occur with the VI on a strong
beat, as a variant of the iv64–i cadential decoration (§9.5); similarly, the
tonic bass arrives in m. 14, a bar before the cadence. Less unnerving but still
problematic is the pickup C–E in m. 9, with the C belonging neither to the
V/V in m. 9 or the V in m. 10.
The next eight-bar phrase presents two counterbalanced sequences. The
first ascends by fifth from D minor (with added sixth) to A minor (with
subposed F♯) to E minor to the dominant of B minor and then to B minor
itself. This is reminiscent of mm. 5–11 of Monteverdi’s “Ohimè,”
particularly since the ascending fifths sometimes move through the dominant
of the next chord to articulate the “ascending subdominant” schema. (Note
that Mahler expands the final unit’s harmonic rhythm so that e–vii°/b
occupies two measures rather than one.) The arrival on B minor, v of the
dominant, overlaps with a root-position fauxbourdon descent to a bare fifth
on the dominant E. Once again, the descending-step logic is clearer when we
postulate added-sixth chords, treating the second chord in m. 22 as F major
rather than D minor in first inversion. Note also the melody’s flexible
relation to the harmony, with B–C–D–C–B over D minor echoed as F♯–G–

A–G–F ♯ over E minor, the melody ascending by fifth while the harmony
moves by step. The impression is less of harmony as container for the
melody than of melody as an independent actor that creates tension by
shifting its relation to the harmonic structure.
The music now turns to the relative major, oscillating between I, V, and
various decorations of V (mm. 24ff). The general technique—extensive
chromatic activity above a dominant pedal—is familiar, here taken to
Mahlerian extremes. Measure 26 gives us a flurry of eighth notes, A7 leading
to G7, the most conventional of the common-tone dominant sevenths (§3.8).
The chromaticism increases as the melody rises: in m. 28 we have a second
flurry on another common-tone dominant, B7 over G.60 This leads to the
remarkable explosion of mm. 29–30, with the upper voices descending at
different speeds along the minor seventh G♯–B–D♯–F♯, each note acquiring
a semitonal upper neighbor.61 The result is gleaming and uncategorizable, as
deliciously painful as nitrous oxide in a dentist’s chair.
The second phrase echoes its predecessor, now with a chipper violin
melody and still circling around the “B7 over G” dominant (mm. 32ff). This
leads to a long sequence of descending triads that I hear pointing to E7, or
V/V in D minor. The expected cadence is elided with the start of the next
phrase, the only interruption of the four-bar phrasing since m. 7. This is a
fine example of Mahler’s ability to use purely instrumental forces to convey
complex emotional states like intrusion, digression, and hesitation. The
subsequent music once again invokes the ascending voice leading of the
common-tone diminished seventh: here we have a surprising E–G–B–D
moving to the expected E–G♯–B–D, with the G ascending to G♯ over the
course of two measures, the E minor seventh the distant cousin of the
standard common-tone diminished seventh E–G–A♯–C♯.62 This ascending
semitonal voice leading continues through the phrase, moving through F–A–
B( ♭ )–D, to F ♯ –A–B–D ♯ , seventh chords ratcheting upward underneath a
truncated opening motif.63 I found this passage completely unintelligible
until I became aware of the retrofunctional idioms in §3.8.64
So goes my account of the opening section’s harmony. On the page it may
seem straightforward, but it was formulated over countless hours spread over
many years, with many wrong turns and rethinkings along the way. I have no
expectation that readers will agree with it, and indeed Mahler’s surface
presents numerous details that argue against my reading.
1. The V/V in m. 8 occurs over an E–A fifth that does not belong to the
underlying harmony, creating a D–D♯–E cluster on the downbeat; I treat
the E as a reattacked pedal tone surviving from the preceding harmony,
though it appears in a new octave as part of an active bass line (Figure
5.5.3). As in an improvisation, the bass player simply arrives at the
harmony a bit late.
2. My analysis treats the wind parts in mm. 15–18 as largely independent
of the underlying harmony, the neighbor notes moving from weak to
strong eighth notes with every repetition of the four-note motive (Figure
5.5.4, top staff). Here they act as a kind of orchestrational thickening
hinting at pandiatonicism.
3. The suggestion of a stable V in mm. 23–26 is undercut by bass motion
that could be taken to represent chord changes (Figure 5.5.5).
4. The “e minor seventh” in m. 38 is actually a minor ninth, occurring over
a towering clash: C–C♯–D–E–F–A on beat 4 of m. 37 and D–E–F–G on
beat 1 of m. 38 (Figure 5.5.6). The D–E–F bass could suggest that E is
passing.

Figure 5.5.3. Measures 7–9 of the Rondo-Burlesque.

Figure 5.5.4. Measures 15–18 of the Rondo-Burlesque.

Figure 5.5.5. Measures 23–26 of the Rondo-Burlesque.

Figure 5.5.6. Measures 38–41 of the Rondo-Burlesque.

These are genuine issues, but I know of no better analysis. The problem is
that Mahler’s language is so contrapuntally and harmonically inventive as to
make objective analysis almost impossible; there are simply too many ways
to parse the extended chords, pervasive linear dissonances, and unusual
progressions.65 (As we will see in the prelude to chapter 7, nonharmonic
reduction typically requires some prior understanding of the music’s
harmonic vocabulary.) While I strongly suspect that Mahler imagined some
reductive template along the lines of Figure 5.5.2, that template is all but
unrecoverable. The upshot is that nonharmonic reduction is no longer
something broadly intersubjective or syntactical, but rather autobiographical
and interpretive, a record of one’s personal coming-to-terms with a
refractory and ambiguous musical surface. Here we have the familiar
modernist slippage between compositional process and musical result, albeit
in the context of functional tonality.
Once again, I find myself pulled in two directions. In some sense, the
basic theoretical picture of “skeleton with decoration” seems reasonably apt:
if one function of analysis is to tell us how we might make a piece, then I
find my reduction to be successful, for I can just about imagine turning
Figure 5.5.2 into Mahler’s actual score. (And when I listen to the piece, I
have a sense of tonal coherence that seems consistent with the reduction.) At
the same time, the very act of reduction slights Mahler’s frenzied
counterpoint, its lines careening as close to chaos as those of Nancarrow or
Schoenberg. If Western norms of dissonance treatment were produced by a
conversation between composer and theorist, one gets the sense that the
dialogue has broken down. This is virtually explicit in the movement’s
sardonic dedication “to my brothers in Apollo”: superficially, the music
proves that Mahler could write counterpoint; but the pervasive delirium is
the sign of something deeper—an ironic attitude toward syntax itself. If the
meaning of this music lies in its clashes, extended sonorities and vectors of
dissonance, then it represents a deliberate challenge to the very idea of
nonharmonic reduction.

6. After nonharmonicity

The opening of Shostakovich’s Ninth String Quartet is in a flexible E♭ major,

the violins snaking in and out of the key to create a marvelous sequence of
tensions (Figure 5.6.1). These lines have their own individual logic and
character, the first violin articulating a series of 015 and 016 trichords (the
two three-note sets containing both semitone and fifth) against the second
violin’s octatonic fragments, G–A ♭ –B ♭ –B–C ♯ –D and C–D–E ♭ –F–G ♭ ,

intersecting to saturate chromatic space from B♭3 to E♭4. The beauty of this
music, to my mind, lies in the skillful blending of multiple logics—one
nontonal and set-theoretical, another tonal and harmonic, a third
emphasizing a descending fifth from B♭4 to E♭4. The same combination can
be found in jazz, originating with the blues and theorized under the rubric
“playing outside” (Figure 5.6.2). Unlike their weightless atonal cousins,
these sets acquire differential significance as they move in and out of the
key, flashes of color like fish in a coral reef.
Figure 5.6.1. The opening of Shostakovich’s Ninth String Quartet.
Figure 5.6.2. The fourth chorus of McCoy Tyner’s solo on “Pursuance,” from A Love Supreme.
Brackets show moments where the melodic 035 sets align with the harmony.

Here we can speak about the “emancipation of the dissonance” in a very

precise sense, for notes outside the harmony are no longer required to
resolve to harmonic notes according to rigid formulas. Indeed, the very
distinction between “inside the harmony” and “outside the harmony”
becomes problematic. This emancipation is a feature of virtually all
twentieth-century music whether notated or not. (In popular-music
scholarship, it is known as “the melodic-harmonic divorce.”66) This
loosening of restrictions should be distinguished from a very different and
more radical view that denies or abrogates the distinction between
consonance and dissonance altogether.67 In Shostakovich’s music there is a
fundamental sense of “inside” and “outside,” of tension and release created
by shifts relative to the prevailing harmony.68 This more limited
emancipation of the dissonance begins with composers like Chopin, Mahler,
and Debussy, who realized that harmonies could acquire sixths and ninths
without losing their meaning. It was radicalized by Stravinsky, who allowed
the nonharmonic tones to become chromatic, sometimes acquiring their own
secondary centricity.69
This is the freedom for which Monteverdi advocated, realized only
centuries after his death. To evaluate this music, it is neither necessary nor
sufficient to consult a list of timeless rules; instead, we have to listen and
make an aesthetic judgment. The freedom was fueled by the near-
contemporaneous development of sound recording—a technology allowing
musicians to bypass notation and formal education (itself a significant force
in sustaining the nonharmonic system), communicating directly through
sound. Once again, the blues played a central role: in “Rising Sun Blues”
(Figure 2.10.1), the most striking nonharmonic tone is the sustained D ♮
against the A minor seventh, a note that neither belongs to the chord nor
obviously embellishes any chord tone, contributing to a five-note sonority
that is aurally hard to parse. Then there is the G♮ deliciously clashing against
the E major chord, essential to the blues but utterly foreign to the
nonharmonic system. Connected to this, and not apparent in the notation, is
all the delicate microdetail of Tom Ashley’s singing, the scoops and glides
that I have quantized into the semitonal grid. It is genuinely unclear whether
this music has a “harmonic skeleton” in the traditional sense, and whether it
admits of anything like nonharmonic reduction.
The most vigorous protest against this new style of musical thinking came
from Schenker, who like Artusi felt that the rules of the nonharmonic system
were the foundation of musical coherence itself. Schenker frequently insisted
on the connection between traditional counterpoint and his own analytical
techniques, and he certainly seemed to treat the abandonment of the former
as a threat to the latter.70 Supercharging Artusi’s scholasticism with an
infusion of Hegelian metaphysics, Schenker insisted that great music
necessarily permits an analysis rooted in traditional contrapuntal laws. Here
one can hardly avoid the thought that the nonharmonic system functioned as
a form of cultural power: power that distinguished learned from unlearned,
legitimate from illegitimate, and whose exercise ensured the viability of a
certain kind of analytical reduction. Its collapse, as Schenker sensed,
represented not just a change of musical fashion but a threat to an entire
species of intellectual authority.
Central to Schenker’s theory is the assertion that nonharmonic notes serve
no harmonic function whatsoever:
Between the dissonant passing note and the sustained note, therefore, no composite sound
exists. Anyone who, in disregard of this fact, posits a composite sound at the upbeat—between
the dissonance and the cantus firmus note—has not grasped the nature of dissonance, of the
passing note as strict counterpoint teaches it.71
This commitment could lead him into analytical contortions, as when he
asserted that the ear spontaneously decomposes the parallel thirds of Figure
5.6.3 into a series of staggered semitonal motions. One can sympathize with
Schenker’s predicament: a straightforward reading would have the sliding at
once nonharmonic, embellishments of the V–I motion, and harmonic,
forming major thirds whose sonic character is musically significant. One
wonders how Schenker would’ve dealt with the passage in Figure 5.6.4,
where Chopin slides an entire triad from dominant to tonic, chromatic
passing tones creating beautiful and harmonically significant flashes of
color. Or for that matter Figure 5.6.5, where chromatic thirds have become
the music’s primary harmonic engine. Committed to the absolute
nonharmonicity of the passing note, Schenker was poorly situated to theorize
such moments, and by extension the changing role of nonharmonicity over
the course of the nineteenth century.

Figure 5.6.3. Mozart’s Symphony no. 36, K.425, mm. 158ff, with Schenker’s hypothesized perception
of the chromatic ascent.
Figure 5.6.4. Chopin’s Nocturne, Op. 27, no. 2, mm. 32–33, in which complete triads slide
chromatically from dominant to tonic.

Figure 5.6.5. The end of Kabalevsky’s “Ditty” (Op. 30, no. 2), where parallel chromatic major thirds
create the harmonies.

In effect, Schenker took passing tones as the model for all dissonances,
and nonharmonic reduction as the model for reduction more generally.72
Here I have reversed his argument, suggesting that suspensions are
fundamentally less reducible than passing tones, and using that irreducibility
to propose that reduction is generally problematic. My claim is that the
nonharmonic system, rather than a means of prolongation, is a mechanism
for generating dissonances whose meaning and metaphysics is up for grabs
—with musicians and theorists sometimes disagreeing about their status. In
the bottom passage of Figure 5.2.4, it is vital to hear the “composite sound”
formed by the passing D♯ and the harmonic A; just as in Figure 5.2.1 it is
vital to hear the composite sound of the 64 formed by the suspension and the
consonant sixth. If this is right, then there is an inconsistency at the heart of
traditional musical syntax, a mismatch between the grammatical rules
composers followed and the harmonic meaning of the music they made.
One underappreciated feature of that grammar is its combining of
intervals from distinct alphabets (§4.8). In earlier music, nontriadic alphabets
usually produce decorative notes subordinate to those of the prevailing
harmony.73 Twentieth-century musicians sometimes combined alphabets
without this sort of subordination: Figure 5.6.6, from the end of Stravinsky’s
Rite of Spring, is similar to Louis Armstrong’s line in “West End Blues”
(Figure P2.2, §4.8), lower notes moving along the B♭ minor triad, with the
top voice moving along the C major triad, two or three semitones above its
lower neighbor; here, a complex transformational network creates a variety
of set classes sharing the same abstract structure. While the network is
reminiscent of the one we used to describe Armstrong’s phrase, it has a very
different meaning: the first note of Armstrong’s motive is an out-of-harmony
neighbor that resolves to a chord tone, whereas the top note of Stravinsky’s
chord is part of the harmony itself.
Figure 5.6.6. The Rite of Spring, upper voices at R196.

The combination of different intervals produces a harmonic variety not

easily captured with any available analytical language, a hidden musical
logic that tickles the ear. Figure 5.6.7 shows how we might use this
technique in tonal and atonal contexts, deploying hierarchical networks
whose structure is revealed by their transformations. Generalized in this way,
the techniques of the nonharmonic system provide an extraordinarily flexible
resource for twenty-first-century musicians. We can still take pleasure in the
way motives and melodies move in and out of the prevailing chord or scale,
and we can still enjoy the similar-yet-different configurations that result
from complex intervallic relationships. If we broaden our thinking as
suggested by the Prime Directive, the past starts to feel present again.
Figure 5.6.7. Two passages that transpose complex musical objects along both chord and scale. The
fourth note of the first motive is a diatonic step below the fifth note; the third note of the second
motive is a minor third above the fourth note.

1
Sutcliffe 2003, chapter 5.
2
See, for example, Schoenberg (1911) 1983, Narmour 1988, or Quinn 2018.
3
See Gjerdingen 2007 and Holtmeier 2007, who essentially invert Riemann’s argument that
figured-bass notation makes too little distinction between harmonic and nonharmonic (Riemann 1896,
p. 107).
4
In the course of a sensitive discussion, Weber (1817–1821) 1846, p. 830, protests against “the
furious persecution against every thing that has the name of fifths.”
5
For clarity, I will continue to use the standard term “nonharmonic” even though it occasionally
leads to paradoxical language.
6
Sometimes pedagogues forbid practices that are actually common, such as cadences in which
inner-voice leading tones do not resolve, or doubling the third of a first-inversion tonic. Some of these
superstrict prohibitions can be traced back to dogmas contemporaneous with the music in question,
suggesting a longstanding conflict between theoretical regularization and compositional heterogeneity.
See, for example, Vicentino’s prohibitions on simultaneous leaps in ensembles of various sizes
(Vicentino [1555] 1996, book 2, chapter 30, Arthur 2021).
7
Quoted in Mirka 2015, p. 165.
8
See, for example, Taruskin 2005, volume 2, chapters 14–16, which identifies Josquin as the
source of the “ars perfecta.” Monteverdi’s brother Giulio Cesare instead identifies Ockeghem (Strunk
1950, p. 408). Counterpoint textbooks tend to take Palestrina’s practice as representative of this
broader and more varied tradition.
 9
See the prelude to chapter 7. John Miller performed the extraordinary service of digitizing all of
Palestrina’s well-attested masses. See Sailor and Sigler 2017 for a quantitative study focusing on
dyadic pairs.
10
Unfortunately, further improvements take much more work; as with most programming projects,
a small amount of effort produces the lion’s share of the results.
11
This is one way of understanding Schoenberg’s statement that dissonances are “comprehended”
with reference to consonances (Schoenberg 1984, p. 259, and Tymoczko 2011a, p. 185): that the
nontriadic surface is understood in terms of a triadic object that never appears.
12
By my count, freestanding anticipations, not decorating suspensions and not themselves
decorated by other notes, comprise less than 1% of Palestrina’s dissonances—making them less
frequent than ascending neighbors. Freestanding anticipations are much more common in Josquin.
13
This figure sometimes appears in eighth notes (sixteenths in our notation), as in m. 10 of the
Benedictus of the mass Ave Regina coelorum.
14
Knud Jeppesen ([1946] 1970, p. 124) notes that accented third-species dissonances are
acceptable for Palestrina only when they form this slippery figure, and not in the context of
unidirectional melodic lines.
15
I call these “121 passing and neighboring tones”: three melodic notes with durations 1, 2, and 1,
with the nonharmonic tone having duration 2 and the harmonies progressing in a 2, 2 rhythm. Bach
uses the 121 neighboring and passing tones in BWV 393 (Riemenschneider 275) m. 3 and BWV 103.6
(Riemenschneider 120) m. 4.
16
See m. 47 of the Gloria of Dufay’s mass Ave Regina coelorum (three voices) and m. 25 of the
Credo of the mass Se la face ay pale (four voices). Dufay’s music contains many other moments that
resist nonharmonic reduction—for example, m. 7 of the Kyrie of the mass Ave Regina coelorum,
where an irreducible seventh is formed by passing motion. In Palestrina’s music, the idiom occurs at
least once per movement on average. Gauldin 1995a discusses this figure.
17
Guillotel-Nothmann 2018 connects suspension resolutions to the origins of root functionality.
18
Fétis (1840) 1994, pp. 32–33, and Christensen 2019. Christensen emphasizes the unpreparedness
of some of Monteverdi’s sevenths (p. 33 of Fétis [1840] 1994), whereas I am more focused on Fétis’s
apparent misunderstanding of the suspension idiom (p. 32 of same).
19
Compare Lerdahl (2020, p. 18): “once passing and neighboring tones and suspensions are
stripped away in Classical music, every chord is a triad.”
20
Numerous theorists have considered suspensions “essential” or “primary” dissonances, as
opposed to the less-essential, more decorative nonharmonic tones: Vincenzo Galilei (Palisca 1956, p.
88), Christoph Bernhard ([~1640] 1973, p. 79), Fux ([1725] 1971, p. 55), Fenaroli and other
partimento theorists (Sanguinetti 2012, pp. 103 and 125ff), and Jeppesen (1946) 1970. Even Rameau
often assigns a different fundamental bass to a suspension and its resolution (Christensen 1993a, pp.
123ff; this view is echoed in Schoenberg [1911] 1983, pp. 316–17). Others instead treat suspensions as
purely decorative, including Riemann (1896, p. 107), Prout (Figure 1.3.1), and Schenker ([1910] 1987,
I, pp. 261 and 266). Beach 1974 identifies Kirnberger as the progenitor of this second perspective.
Thanks here to David Cohen.
21
Schoenberg (1911) 1983, p. 309. The claim about medicine is outdated: health-care providers
have learned that they often need to consider non-medical matters such as economics, environment,
and sociology.
22
Jeppesen (1931) 1939, p. xi.
23
For an interpretation of functional tonality and twelve-tone music as formal systems, see Boretz
1970 and 1972, Westergaard 1975, and Dahlhaus 1990, pp. 61–62 (particularly the thought that there
may not exist additional “systems”).
 24
Note that in making this claim, I am using the term “harmonic” in a minimalist sense, essentially
as a label for residual preferences that cannot easily be explained by linear or melodic factors; I do not
presuppose inversional equivalence, root functionality, or any other sophisticated theoretical ideas.
Quinn (2018) offers an alternative definition of “harmonic” in terms of tonic-independent principles;
on this definition, the tendency for 64 to resolve to 53 would be a purely contrapuntal law—as would
a preference for descending-fifth progressions, if I read Quinn correctly.
25
Harrison 1994 considers harmonic function as arising from particular scale-degree motions;
while this is not how I always think of function (§7.1), I find it analytically useful here.
26
The ascending 7̂ passing tone is virtually mandatory when the 6̂–1̂ is in the bass, the IV6 chord
leaping to I only in the context of IV–IV6–I arpeggiations.
27
Narmour 1988 makes a similar argument.
28
Schoenberg (1911) 1983, p. 322.
29
Compare the Sanctus of Palestrina’s mass Assumpta est Maria, m. 31, or the Sanctus of the Pope
Marcellus Mass, m. 53.
30
Recall also that the harmonic/nonharmonic distinction has a transformational significance (§4.8).
31
Christoph Bernhard ([~1640] 1973, p. 90) associates many of these effects with improvisation,
an association echoed by Palisca (1994, pp. 54ff). The tradition of using nonharmonic tones to produce
quartal harmonies, often evoking folk music, continues in works such as Schubert’s C major string
quintet, D. 956, III, and Chopin’s Mazurka Op. 6, no. 2.
32
See, for example, Fétis (1840) 1994, p. 31, or the heightened language in McClary 2004, pp.
183ff (“shriek,” “grinding dissonances,” “shouting,” “BITCH!,” “transgression,” “bitter,”
“unpleasant,” “brutal assault”).
33
McClary’s extreme language is no doubt intended to get us to hear Monteverdi’s dissonances as
striking and outrageous, as his contemporaries might have. However, it is not obvious that ordinary
Renaissance listeners (as opposed to theorists) would have had such extreme reactions—perhaps the
fleeting dissonances would have eluded earlier ears precisely as they elude so many contemporary
listeners. Or perhaps Renaissance audiences were accustomed to more extreme levels of dissonance in
improvisational contexts, as suggested in Palisca 1994, chapter 3.
34
Octave-displaced melodies also appear in the bass voice at mm. 16, 25, and 35.
35
That is, the ascending B♭–F fifth of the previous phrase continues to the C–G–D–A of the next;
this is noted in Chafe 1992 (p. 92). Salzer 1983 reads the opening chords as a prolongation of G minor.
36
Chafe 1992 proposes that Monteverdi’s language is hexachordal and based on major and minor
triads (pp. 24–31). The sequence in mm. 6–12, however, suggests that it is fruitful to include the
diminished triad alongside major and minor: the b°6 in m. 6 plays a similar role to the g6 in m. 10;
these are both instances of a ubiquitous Renaissance voice-leading pattern, which can appear either in
authentic (vii°6–I) or plagal ( ♭ vii6–I) forms. To capture this similarity, it is useful to treat the
diminished triad as a genuine sonority, and hence to postulate a diatonic rather than hexachordal
background.
37
Salzer 1983 interprets the chords in the fifth-sequence as dominants of their predecessors rather
than “ascending subdominants” of their successors. I think this is an anachronistic flattening-out of a
three-dimensional Renaissance harmonic landscape.
38
To my ear, these figures have less the character of a true seventh, or even a suspension, than of a
syncopation or perhaps a resonance that has been sustained just slightly longer than the harmony it
belongs to—like a fleeting pedal tone.
39
This is the “chord repeats, pattern continues” technique discussed in §4.9.
40
See Theune 2007 on poetic turns.
 41
Salzer 1983 reads the phrase as an interruption form.
42
One might try a quarter-note reading where the first two beats are B♭maj7 followed by a°, but this
involves an unusual harmonic progression and a nonstandard resolution of the seventh.
43
On “fittingness” in aesthetics, see Danto 1981, p. 207, and Camp 2017. McClary is similarly
suspicious of what she calls the “ ‘the words made me do it’ defense” (2004, p. 182).
44
These two perspectives—one humanist and evolutionary, the other scholastic and timeless—map
onto broader worldviews whose conflict was playing out during these same decades. See Tomlinson
1990 and Wootton 2015.
45
Palisca 1994 makes a similar point, as does Chafe (1992, p. 5): “what distinguishes Monteverdi’s
music from both late Baroque rationality and the sixteenth-century style is the relative unpredictability
of his dissonances and rationalizations.”
46
Christoph Bernhard’s unpublished 1640 treatise contains one of the most interesting
seventeenth-century discussions of nonharmonicity; interestingly, it does not discuss pedal tones.
47
Christoph Bernhard describes this explicitly: “part of a note is cut off, so that this may be placed
in front of the following note in the degree immediately below” ([~1640] 1973, p. 108).
48
A modest degree of recursion is characteristic of the nonharmonic system: in Figure 5.1.4, for
example, we have an anticipation being decorated by a neighbor. One can also find neighbor tones
decorating passing tones and suspensions, and so on.
49
This minimal definition is found in Gauldin 1997. Other textbook authors such as Aldwell and
Schachter (2002) and Kostka and Payne (2003) place more restrictions on pedal tones. The bass G in
Figure 9.3.1 is unusual: if we want to consider the iiø7 to be harmonic then the chordal G is attacked a
measure too early; it is perhaps comparable to the bass A in m. 13 of Figure 5.5.2.
50
Pedal tones are the only species of nonharmonic tone largely limited to specific scale degrees.
Some genres, such as string quartets, are more lenient with pedal tones.
51
Bach’s keyboard music occasionally presents additional anomalies, including suspensions
resolving by leap.
52
Or, more neutrally, its significance; I use “semantics” in a general and metaphorical sense.
53
Or that they represent multiple voices as discussed in the prelude to this chapter.
54
That is, one might have a small penalty for the standard notes of the SNAP system (ensuring that
the computer prefers an analysis with no nonharmonic tones to one with some), a slightly higher
penalty for incomplete neighbors and pedal tones, and the highest penalty for “rogue notes” with no
standard explanation. The strategy was devised by my student Jeffrey Hodes in his undergraduate
thesis (Hodes 2012).
55
Assuming the key is correct.
56
This is something like the converse of the twelve-tone technique of creating consonant music by
hiding unwanted notes in an unobtrusive instrument.
57
Steve Taylor notes that the same chord appears at m. 327 of the first movement of Mahler’s
Second Symphony.
58
One sign of this intention is that nonharmonic tones generally form consonances with each other.
59
Meyer 1996 also emphasizes Romantic composers’ repurposing of earlier techniques.
60
Note that the “A7 over G” common-tone dominant resolves to G7 and is associated with the
diminished seventh C♯–E–G–B♭; the “B7 over G” resolves to C and is associated with C–D♯–F♯–A.
61
See Lewis 1983, pp. 148–50, for a polytonal interpretation involving the dominant of E minor.
My analysis evokes what jazz musicians call “upper structures,” chromatic, stacked-third extensions
of what is fundamentally V7 in C (cf. Levine 1989).
 62
It is also possible to identify the chord on m. 38 as a G7, with bass E an accented neighbor,
though that makes the underlying voice leading less clear.
63
The B/B♭ oscillation once again evokes the common-tone diminished seventh.
64
Note the three-note trumpet figure landing successively on G, D, and A.
65
Lewis 1983 contains a careful reading of the opening measures that disagrees with mine in
almost every detail.
66
Temperley 2007 and Nobile 2015. Christopher Doll modifies this to “melodic/harmonic
stratification” (2017).
67
Whether Schoenberg himself distinguished these two senses of “emancipation” is an interesting
question. As Hinton 2010 and Arndt 2011 point out, his description of emancipation changed over
time, from something anticipated to something that had occurred.
68
Exactly what determines this sense of “belonging” is a very complex matter; in popular music,
instrumentation plays a vital role, with guitars often playing pure triads while the vocalist provides
additional notes.
69
See Figure 1.1.4, which suggests a distorted B♭ minor. When the embellishing tones acquire their
own centricity, we have “polytonality.” Thus there is a hidden connection between the emancipation of
dissonance and polytonality.
70
This comes out most clearly in his various complaints over modernism, most notably “Further
consideration of the Urlinie: II.” The precise role of traditional contrapuntal principles, and their
specific applicability or inapplicability to various levels of reductive analysis, is a major point of
disagreement among Schenkerians.
71
Schenker (1926) 1996, p. 10. This point is echoed by Rothgeb (“the passing tone is . . . totally
without harmonic significance,” 1975, p. 268) and Lerdahl (2020, p. 18).
72
Schenker (1910) 1987, I, pp. 261ff, and Rothgeb 1975, pp. 273–74.
73
Fugue subjects are an exception, using different interval types to relate harmonic notes (§4.8).
 Prelude
Functional and Scale-Degree Analysis

Since there is a one-to-one correspondence between standard Roman

numerals and tertian structures over a bass, we can interpret Roman
numerals as referring to those structures. Thus a symbol like “I6” can be
taken to mean “the 63 sonority over scale degree 3” or “scale degrees 1, 3,
and 5 with 3̂ in the bass.” This is scale-degree analysis, a “thin” and
relatively theory-neutral record of the tertian sonorities in a piece. Pure
scale-degree analysis is a relatively modern phenomenon; the traditional or
functional approach uses harmonic symbols that are “thick” and theory-
laden—encoding the analyst’s phenomenological impression of a chord,
beliefs about its behavior, theory of its historical origins, or claims about its
“essence.”1 Thus a scale-degree theorist would label the starred chords in
Figure P6.1 as mediants while the function theorist might use V to record the
claim that they behave or sound like dominants. To a scale-degree analyst,
it’s just a fact that mediants sometimes appear where we would ordinarily
expect dominants; to a function theorist the label “III” betrays a
misunderstanding.
Figure P6.1. (top) The fifth phrase of Bach’s chorale “Jesu, meine Freude” (BWV 64.8,
Riemenschneider 138), where III+6 acts like a dominant; (bottom) the last measures of Robert
Schumann’s “Am Kamin” from Kinderszenen, along with John Rothgeb’s analysis. Above the music, I
show how the passage can be viewed as a “down by third, up by step” sequence (with mixed
chromatic and diatonic elements). On this reading the bracketed g6–a6 cadence repeats the bracketed
final g°–[a] sequential unit.
 The tradeoff here is between simplifying the process of chord-
identification and simplifying the grammar of harmonic progressions.
Rameau analyzed the deceptive cadence as a descending-fifth progression in
which the tonic chord is represented by the root, third, and sixth above the
fundamental bass; this yields a simple harmonic rule (“dominants descend
by fifth to the tonic”) at the expense of complicating the identification of the
fundamental bass (since either tonic or submediant can now serve as
fundamental bass for scale degrees 6, 1, and 3).2 In much the same way,
traditional theorists have often treated vii° as a dominant seventh missing its
root; once again this simplifies the grammar of progressions (“V goes to I”)
while complicating the process of chord identification (“V7 sometimes lacks
its root”). This same philosophical difference underwrites the longstanding
dispute about the proper labeling of the cadential six-four: to the scale-
degree theorist, I64 means “scale degrees 1, 3, and 5 with 5̂ in the bass,” and
it is just a fact that this chord sometimes occurs between predominant and
dominant. The function theorist prefers to say that ii6 progresses to a
dominant that is sometimes embellished by a 65–43 progression above the
bass. This gives an efficient grammar at the expense of an ambiguous
notation in which “V64” can refer to two different chords: scale degrees 1, 3,
and 5 with 5̂ in the bass and scale degrees 5, 7, and 2 with 2̂ in the bass.3
Discussions of this issue, when they do not immediately degenerate into
outright invective, often terminate in naïve philosophy. Many theorists have
claimed that labels like “I” and “I6” necessarily imply similarity, indicating
that both chords have the fundamental quality of tonicity or tonic-triadness.4
But this is not how language works: the sea cucumber is biologically an
animal despite its name. Philosophically, the scale-degree theorist is fully
justified in using Roman numerals to refer to collections of scale degrees:
from that perspective the relation between “I” and “I6” is a matter for
investigation rather than a presupposition of the notation. Roman numerals,
as used by the scale-degree theorist, are largely equivalent to figured-bass
symbols.
To be sure, the difference between these approaches is one of degree
rather than kind: in reality there is a continuum between almost completely
mechanistic analyses that label every sounding consonance, and highly
interpretive enterprises that provide almost no information about scale-
degree content at all (cf. the prelude to chapter 7). Nor can harmonic
analysis entirely avoid questions about whether sonorities are harmonic or
nonharmonic—questions that are extremely complex in the functional era
but arise in other styles as well (Figure P6.2). But we can acknowledge these
theoretical points while still hoping to separate, inasmuch as it is feasible,
the project of labeling harmonies from that of theorizing about their
meaning and role, using harmonic labels as a relatively neutral ground for
evaluating claims about syntactic behavior. From this point of view, the
behavior or origin or essence of a chord should have little influence on its
label.

Figure P6.2. (top) The second-inversion triad in m. 19 of the Agnus from Palestrina’s mass De Beata
Marie Virginis (II) (starred) is plausibly harmonic, occurring in a sequence of four ascending fifths (or
five, if we ignore the a6). (bottom) The second-inversion triad in m. 21 of the Agnus of the mass In
minoribus duplicibus (starred) has a much weaker claim to harmonicity, being produced by a passing
tone.
 In this book, I generally default to a fairly strict form of scale-degree
analysis. First, because it gives us a vocabulary for exploring how
collections like iii6 can gradually change their function (e.g., moving to IV in
the Renaissance, largely disappearing in the classical period, and acquiring
dominant function in the nineteenth century). Second, because it gives us a
common language in which to express disagreement or uncertainty about a
chord’s perceived function: for example, one listener may hear the cadential
64 as a dominant, while another may be more sensitive to its tonic qualities,
with neither being right or wrong.5 Scale-degree analysis provides a shared
vocabulary for discussing these disagreements, rather than encoding the
purportedly “correct” interpretation directly in its notation.6 And third,
because the scale-degree approach leads to greater methodological clarity.
For the scale-degree theorist, a term like “dominant” is roughly synonymous
with “V or vii°” and “dominants proceed to tonics” is a testable claim: we
simply check whether scale degrees 5, 7, and 2 typically proceed to scale
degrees 1, 3, and 5. This is something even a computer can do.
By contrast, functional analysis can sometimes come dangerously close to
circularity: at the extreme, labels like “dominant” are utterly divorced from
their scale-degree content, applied to any chord that is perceived (by the
analyst) as engendering a certain kind of expectation.7 Here the claim
“dominants progress to tonics” becomes definitional, free of any empirical
content whatsoever—an introspective report that some sonority arouses an
expectation of a tonic arrival.8 Here, harmonic analysis becomes a language
for talking about one’s experience—sometimes described as a matter of
“chord meaning.”
None of which is to imply that it is misguided: describing subjective
experience is a reasonable pursuit, and there is certainly something to be
gained from the thought that tonic function can be expressed by scale
degrees 1, 3, and either 5 or 6. Problems arise when one particular approach
is tacitly assumed to be the only option. Eminent scholars have claimed that
Roman numerals necessarily presuppose inversional equivalence, or that
terms like “dominant” assume the primacy of the bass voice.9 Similarly,
theorists sometimes complain about the “trivial transliteration of note-
content”—as if it were a mistake to aspire to somewhat objective
descriptions against which we can test our assertions.10 (Very often these
same theorists underestimate the degree of regularity to be found in local
harmonic progressions, regularity that becomes unmistakable when we stoop
to the level of “trivial transliteration.”) In rejecting scale-degree analysis,
theorists deprive themselves of a useful tool for charting the development of
harmonic functionality—indeed, virtually the only tool that can trace the
subtle ways in which configurations of scale degrees change their behavior
over the course of the sixteenth and seventeenth centuries.
My argument can be reformulated as a question. Suppose a musician is
interested in scale-degree content, desiring a relatively theory-neutral
account of the tertian harmonies in a piece. How should they go about this?
The two options are to create an entirely new language that is virtually
equivalent to, and intertranslatable with, familiar Roman numerals (for
instance, using “scale degrees 7 and 3 over 5̂” in place of “iii6”), or,
alternatively, to use familiar Roman numerals with a slightly different set of
implications. The first has massive costs in time and effort: when confronted
with these new terms, readers need to stop and think about what is meant,
whereas traditional Roman numerals can be immediately associated with
collections of scale degrees. The second runs the risk of unwanted
implications, but this can be forestalled by being clear about how one is
using words. One approach requires that readers internalize an entirely new
nomenclature, all in the fruitless attempt to stave off misunderstanding; the
other requires only that we trust our readers to be attentive, flexible, and
charitable enough to take us at our word.

1
See Rothgeb 1996.
2
Rameau (1722) 1971, pp. 71–73 (book 2, chapter 6).
3
This debate can be traced back at least to Weber (1817–1821) 1846, pp. 463–64; for more recent
discussions see Beach 1967 and Ninov 2016.
4
As Carl Schachter puts it, “there is no such ‘thing’ as a I chord in C major, but only an idea that
can find expression” in manifold ways; that single “idea” is presumably what is labeled by the Roman
numeral “I” (Schachter 1990, p. 166, quoted in Rothgeb 1996; my italics). Damschroder 2016, p. 5,
expresses a similar sentiment.
5
Rothgeb 1996 describes functional analysis as recording what a “perceptive listener” hears; such
perceptions are much more variable than scale-degree content.
6
In this respect my proposed reconceptualization of Roman numerals is consonant with the trend
toward the more theory-neutral labels of figured bass (Gjerdingen 2007 and Holtmeier 2007).
7
Some theorists divorce terms like “dominant” from scale-degree content altogether (Lerdahl
2001, Nobile 2016).
8
Some theorists use “subdominant” to refer to IV chords that progress plagally to the tonic rather
than authentically to V, so that “subdominant chords progress to tonics” becomes a matter of
definition; for the scale-degree theorist these same words constitute a falsifiable claim (Swinden 2005,
p. 253).
9
Dahlhaus 1990, p. 73. Speaking about the Renaissance cadence, Taruskin writes: “the progression
in the bass, from the fifth scale degree [ . . . ] to the final, is congruent with what we are accustomed to
calling a V–I or dominant–tonic progression. To call it that is to think of the motion of the lowest part
as the essential cadential approach, and to associate the gesture toward closure with the ‘dominant’
harmony” (Taruskin 2005, vol. 2, p. 496, my italics).
10
Rothgeb 1996, paragraph 11.
 6
The Origins of Functional Tonality

Scholars have situated the origin of functional tonality in the early 1400s, the
popular music of the early 1500s, Monteverdi’s works of the early 1600s,
Corelli and the 1670s, and even after J. S. Bach.1 To my mind this is like
trying to identify the precise moment in which dinosaurs evolved into birds:
functional harmony developed gradually over centuries, and there is no
single point at which the process began or ended.2 This is because the
transition was largely implicit, a matter of composers learning patterns and
idioms without necessarily recognizing them explicitly—embodied routines
that only later became the subject of explicit theoretical attention. To trace its
development, we need to look toward the musical knowledge implied by
music itself. Statistics are essential here, if only to point us in the right
direction.
Once we understand the development of tonal functionality as gradual and
implicit, we can incorporate a number of earlier insights. In what follows, I
will agree with Lowinsky that a simple form of protofunctionality appeared
in the early 1500s as an optional set of ionian-mode idioms emphasizing
root-position I, IV, and V.3 Over the next two centuries, these routines
colonized an increasing swath of music, ultimately coming to resemble
universal laws rather than genre-specific tendencies. During this time, they
also increased in complexity, moving beyond simple oscillations to longer
progressions featuring a wider variety of sonorities. Thus I also agree with
Fétis and Dahlhaus, who have drawn attention to the early 1600s as a time
when additional functional vocabulary was being standardized. The mid-
seventeenth-century enthusiasm for sequences represents another inflection
point, and a reason for highlighting Corelli.
Where I depart from earlier theorists is in rejecting a unified or essentialist
conception of functionality. The evolution of functional harmony, like the
evolution of birds, involved numerous separate processes occurring at
different times. One is the restriction of harmonic vocabulary to a small
number of chords and progressions. Another is modal homogenization, the
reduction of the available modes to just major and minor. The development
of inversional equivalence is a third. Broad acceptance of seventh chords and
second-inversion triads is a fourth. A fifth is the standardization of
modulatory pathways. Interest in sequential patterning is another. Yet others
concern melodic routines, from the leading tone’s increasing tendency to
resolve, to the development of the 3̂–2̂–1̂ and 5̂–4̂–3̂–2̂–1̂ templates that
underlie so many baroque and classical phrases. My guiding principle is that
these processes are largely independent, separable, and linked by no single
logic.4 As a result, we find some varieties of functional tonality that are
focused on primary triads, and others that make heavy use of seventh chords,
secondary diatonic triads, or sequences. No one key unlocks all the doors.
Again, many of these changes occurred in musical practice long before
they were noticed by theorists. To say this is not to imply that history is
guided by a hidden hand, or that it evolves toward greater artistic perfection,
but merely that it sometimes develops in a way that is akin to natural
language: from 800 to 1400, for example, English gradually lost its case
markers in a bottom-up process driven by the changing behavior of ordinary
speakers. To study this process by looking at medieval grammars, rather than
the records of linguistic behavior, is to get things backward—for the
grammarians neither noticed nor cared about these changes as they were
occurring.5 The development of harmonic functionality is analogous, a very
gradual regimentation of musical behavior that focuses musical attention on
a core set of harmonic and melodic routines. To explore it, we need to
consider the utterances of musical speakers, rather than the theories
surrounding those utterances. Paradoxical as it may sound, it is possible that
contemporary theorists, using contemporary tools, may end up
understanding some aspects of this grammar better than its native speakers.6

1. The logical structure of protofunctionality

A precondition for understanding the origin of functional harmony, and a

topic to which I as a nonhistorian can contribute, is a clear and principled
understanding of the phenomenon itself. We can start with simple counting.
In a typical major-key passage by J. S. Bach or Beethoven, more than half of
all sonorities will be either I, I6, or V(7). The flip side of this popularity is the
remarkable scarcity of sonorities like iii, iii6, and vi6, which in classical
music typically comprise less than 1% of all chords. Such frequencies are
sometimes known as zeroth-order properties, as they are context-
independent. They contrast with first-order properties that describe the
likelihood that one chord will progress to another: for example, the
probability that V will go to I. Theories of functionality often highlight first-
order properties—associating the experience of “dominantness” with that
chord’s tendency to move to the tonic.7 But to my mind functionality is first
and foremost a matter of chordal vocabulary.
Figure 6.1.1 presents the distribution of diatonic sonorities in the ionian
and major-mode music of a range of composers.8 Here we see harmonic
vocabulary gradually coming to focus on the tonic, dominant, and the two
main predominants, ii and IV, while the mediant grows so infrequent as to
become invisible. Particularly interesting is the gradual reduction of the
submediant, a narrowing that continues from J. S. Bach to Beethoven, as if
the process that led to functionality also persists during the functional era. It
is also interesting that IV appears to increase from Josquin to Palestrina,
only to decrease thereafter—suggesting a protofunctional practice
emphasizing I, IV, and V. These kinds of changes show that a harmonic
theory cannot be phrased solely in terms of rules like “chord X can go to
chord Y”: we also need to consider the basic preference for some chords
over others.
Figure 6.1.1. Chord roots in a range of composers, with size proportional to frequency.

We see these same tendencies when we look at specific pieces. Figure

6.1.2 shows the opening of the pavane of Dalza’s first Pavana alla
Venetiana, published in 1508 in the fourth volume of Petrucci’s lute
intabulations. The music is grouped into sixteen-bar periods whose
antecedents end with two bars of V and whose consequents end with a bar of
V and two bars of I. More than 95% of the piece is spent on the root-position
primary triads I, IV, and V; all but three chords are in root position. The only
substantial departures from functionality are the ♭VII6 in m. 9, the ♭VII in m.

27, and the occasional use of F♮ over the C major chord.9 One gets the sense
that each period embellishes a single harmonic template, rather like a
variations form—for instance, antecedent phrases always start with four bars
of I and end with two bars of V, but mm. 5–6 can contain either IV or I;
similarly, consequents always end with a bar of V and two bars of I, but the
preceding five measures are more flexible. There is also an eight-bar
“cadential interlude” after the third period, presenting two iterations of a IV–
V–I cadence, as if reinforcing the previous consequent’s close.10
Figure 6.1.2. The opening of Dalza’s first Pavana alla Venetiana.

The preference for fifth-related primary triads creates harmonic cycles

oscillating between the tonic and a contrasting chord. There is a subtle
geometrical logic underlying these cycles. The top system of Figure 6.1.3 is
an exact contrary-motion sequence, the bass dividing the octave in half while
the upper voices divide it into sixths. Underneath, I show that the major triad
exhibits this symmetry in an approximate way. The third staff shows how
this approximate symmetry manifests itself in diatonic space: the tritone F♯
major can be diatonicized in either descending or ascending fashion, to F
and G major respectively; the pattern achieves stepwise melodic motion by
using both quantizations to connect 5̂ to 7̂.11 The resulting diatonic structure
retains the traces of the chromatic symmetry: melodic steps balanced against
a bass that moves by approximately half an octave, cycling from tonic to
contrasting sonority and back.
Figure 6.1.3. Upper-voice steps against bass-line fifths.

This nearly sequential pattern is remarkable for a number of reasons. First,

it exhibits a feature central to Schenkerian theory, the bass acting as a fifth-
divider, nearly bisecting the octave into consonant intervals, while the
melody divides the third into nearly equal steps (a “step divider,” we might
say). Geometrically, the upper voices articulate a series of complete circles
in triadic space (i.e., Figure 3.4.1) while the bass divides the octave in half.
Second, it gives us an upper-voice analogue to the traditional “rule of the
octave”—associating melodic scale degrees with harmonic defaults. These
associations are apparent throughout the functional tradition, and particularly
important in its earlier stages.12 Third, the figure’s harmonic progressions
are closely related to the plagal and authentic cadences, allowing these well-
established idioms to appear in other parts of the phrase. Fourth, its
progressions are largely non-directional, and can occur either forward or
backward, with the exception of the relatively rare melodic 7̂–6̂. The figure,
in other words, shows how first-order tendencies could arise from the
conjunction of zeroth-order harmonic preferences (an emphasis on I, IV, and
V) and simple melodic constraints like the desire for stepwise motion. Fifth,
other than the completely parallel IV–V progression, the voice leadings on
Figure 6.1.3 have the form t2T–3 or –4, combining two-step transposition
along the chord with three- or four-step transposition along the scale. This
produces a sequential structure, the V–I–V–I (7̂–1̂–2̂–3̂) pattern repeating in
contrary motion at the interval of a fifth as I–IV–I–IV (3̂–4̂–5̂–6̂, Figure
6.1.4).13 Here we see the germ of several later schemas: 7̂–1̂–2̂–3̂ under V–
I–V–I is related to what Gjerdingen calls the “Fenaroli”; while 3̂–4̂–5̂–6̂
under I–IV–I–IV can be understood either as a transposed Fenaroli or a
reversed “Prinner.” (Similarly, 1̂–2̂–3̂ under I–V–I resembles Gjerdingen’s
“Do-Re-Mi.”) Our figure grounds these schemas in the basic geometry of
triadic counterpoint, suggesting that we will find them in any style that
makes use of the fifths-and-step defaults.
Figure 6.1.4. Sequential and schematic content of the fifths-and-steps arrangement.

Michaelis’s “Passando per una rezzola” is a brief frottola published in

Petrucci’s 1504 collection (Figure 6.1.5). Notated in cut-C mensuration, the
harmonies articulate a clear triple meter and three four-bar phrases—the first
ending on IV, the second with a weak cadence on I, and the third with a
stronger close on the tonic.14 All but two chords are root-position major
triads, and the overall harmonic progression is strikingly functional—indeed,
the sounding upper voice, considered together with the harmony, traces out a
5̂–6̂–5̂–4̂–3̂–2̂–1̂ melody harmonized almost entirely with the options in
Figure 6.1.3: the exceptions are the I6 chord replacing the root position I in
m. 6 (to be discussed shortly), and the two harmonizations of 4̂–3̂—the first
a quasi-cadential vii°6–I and the second a retrofunctional ♭ VII–I. The
combination of regular phrasing and familiar progressions anticipates not
just the strict functionality of the eighteenth and nineteenth centuries, but
also the looser functionality of our own time: add drums, change the words,
and you almost have a pop song.
Figure 6.1.5. The frottola “Passando per una rezzola” by Michaelis.

Figure 6.1.6 shows the opening of Bartolomeo Tromboncino’s “Se ben

hor non scopro el foco” from the same collection. (Tromboncino, “the little
trombonist,” would have been better nicknamed “Birichino” or perhaps
“Assassino.”) Once again we have regular phrases and a clear formal
structure. The first phrase ends with a Prinner variant substituting ♭ VII for
V6. The middle section, shown in Figure 6.1.7, is notable for its four-voice
canonic sequence; Figure 6.1.8 analyzes this sequence using a technique
from §4.3, applying rhythmic displacement and contrary-motion distortion to
antiparallel tenths.15 In the coda, the middle two voices ascend in canon to
form a series of F and B♭64 chords, delicately balanced between F and B♭
major, and exploiting the round in Figure 4.5.6.
Figure 6.1.6. The opening of Tromboncino’s frottola “Se ben hor non scopro el foco.”

Figure 6.1.7. The end of “Se ben hor non scopro el foco,” omitting the repeat of mm. 4–6 on Figure
6.1.6.
Figure 6.1.8. Deriving the four-voice canon: starting with an ascending-fifth sequence, every other
chord is contracted by contrary motion. Rhythmic displacement allows one voice to be duplicated at
the unison.

2. Similarities and differences

These pieces are so bewilderingly functional that they seem to challenge the
thesis of gradual evolution—suggesting perhaps that functional tonality did
spring forth fully formed in this early vernacular tradition. But on closer
inspection, it is clear that we are dealing with a distinctive protofunctional
dialect that both resembles and differs from later styles.
We can start with the resemblances.
First, these pieces seem to be harmonic in conception: the heavy use of
root-position triads, often in close position and related by “default” voice
leading, coupled with the deemphasizing of contrapuntal devices such as
imitation, all suggest that chords are genuine objects of compositional
focus.16 Unapologetic parallelism reinforces the centrality of harmony
(Figure 6.2.1).
Figure 6.2.1. Parallelism in Dalza’s second Pavana alla Venetiana (mm. 53–55).

Second, these pieces are dominated by primary triads, with root-position I,

IV, and V accounting for 70%–80% of all chords. These primary triads tend
to behave more or less as in later music: phrases frequently start with the
tonic, exploit weak contrasts with IV, ♭VII, or V (Figure 6.2.2), then move to
a more decisive contrast that features a robust dominant, which is either
resolved at the start of the next phrase (e.g., a half cadence, as in Figure
6.1.2, mm. 7–8), at the end of the phrase (full cadence, as in m. 15 of Figure
6.1.2), or by a surprise move to vi (deceptive progression, Figure 6.3.1).
Tonics are stable, dominants provide the most powerful contrast, and
subdominants provide a weaker contrast: in this sense, the basic pillars of
functional harmony are in place—though in a simple, oscillatory context that
typically cycles from tonic to contrasting sonority and back.17

Figure 6.2.2. Phrase structure in which a stronger V–I follows a weaker V–I (from the opening of
Dalza’s Pavana alla Ferrarese).

There is an interesting question about the role of IV: is it a subdominant

that moves to I, a predominant that moves to V, or a post-tonic to which I
tends to move? The conceptual challenge is that the zeroth-order and first-
order harmonic properties interact: we would expect music with a lot of I,
IV, and V chords to feature a lot of IV–I, IV–V, and I–IV motion. What we
need to know is, given the prevalence of IV, V, and I, are certain progressions
especially likely? We can explore this question by calculating the difference
between the overall likelihood of a chord (its zeroth-order probability) and
its (first-order) likelihood after a specific harmony: this tells us, for example,
whether V is particularly likely to occur after IV. Figure 6.2.3 shows that the
subdominant tends to act more like a predominant in Dalza while being more
neutral in the frottola, largely moving in accordance with its zeroth-order
probability.18 We will return to this idea later.

Figure 6.2.3. The “tendency” of IV–V and IV–I progressions. The number +19.9 indicates that the
likelihood of a V chord increases by +19.9 percentage points after a IV, relative to its overall
likelihood—or in other words, that IV chords “tend” to go to V chords. Negative values indicate
progressions that are suppressed.

Third, these pieces exhibit a precursor to inversional equivalence in the

form of a substantial zeroth-order preference for I6 over iii, despite the
otherwise heavy reliance on root-position triads. Figure 6.2.4 shows the
proportion of root-position and first-inversion triads appearing above each of
the bass degrees: for the primary degrees 1̂, 4̂, and 5̂, root positions occur
more than 90% of the time; for the secondary degrees 2̂ and 6̂ the percentage
is considerably lower; for 3̂, first inversion occurs more often than root
position (60% of the time in the frottola and about 85% of the time in
Dalza). One can hypothesize that this asymmetry reflects the aural similarity
between I and I6: perhaps, in this harmonically limited music, the I6 chord
simply sounds more consistent with the primary triads that dominate the rest
of the harmony.19 This zeroth-order preference for I6 also allows for
stepwise bass motion, particularly when combined with the idiomatic
Renaissance vii°6–I cadence. Figure 6.2.5 combines bass 4̂–3̂–2̂–1̂ with
upper-voice 6̂–5̂–4̂–3̂ to produce the basic voice leading of Gjerdingen’s
“Prinner,” as characteristic of the Renaissance as of the eighteenth century
(cf. Figures 6.1.5 and 6.5.1a, mm. 7–9).

Figure 6.2.4. The likelihood of root-position triads on the various ionian-mode scale degrees. The
mediant is, after the leading tone, the least likely to support a root-position triad, indicating a
preference for I6 over iii, even in genres that typically favor root-position triads.
Figure 6.2.5. The Prinner combines two standard ionian features, I6 as a default harmonization of 3̂
and the vii°6–I cadence.

In all these respects, I think it is appropriate to locate the origins of

functional harmony in this repertoire—some of the first music ever printed,
and some of the earliest popular music to survive.20 But we should also
recognize the important differences from the complex functionality of later
centuries.
First, mature functionality is a matter not just of what happens, but of
what does not happen, with its “harmonic grammar” underwritten by the
virtual absence of progressions like V–IV, ii–IV, and so on. This early music
exhibits harmonic preferences that evoke those of later functional music, but
there is no sense that nonfunctional progressions are forbidden: one can
easily find progressions like I–ii–I, or root-position dominants moving to
root-position subdominants, and so on. We are dealing with flexible
tendencies rather than rigid rules.
Second, there is no strong commitment to a single diatonic collection,
with ♭ VII common even in ionian contexts. The relevant conception of
mode, or tonal area, is one that draws freely on tonic, dominant,
subdominant, and subtonic. The ♭VII chord functions much as it does in our
own popular music, providing an alternate harmonization for 4̂ and 2̂.21
 Third, the harmonic routines are very simple. There is, for example, no
real sense that the ii chord is a predominant: it can of course play this role,
but in Dalza it more often moves to I than to V or vii°. Nor is there a sense
that the submediant has a characteristic role to play. We are dealing with the
fairly rudimentary set of options in Figure 6.2.6 rather than the more
complex routines of later music.

Figure 6.2.6. A harmonic map for simple functional music.

Related to this is the importance of IV: the two harmonic cycles I–IV–I
and I–V–I are, if not equal, then at least comparable in significance.
Furthermore, IV–I progressions sometimes have a subtle “dominant” quality
by virtue of an ascending 6̂–7̂–8̂ line. In other cases, I–IV progressions have
a tonicizing character, with tonic moving to subdominant so as to evoke
cadential formulae (e.g., the end of Figure 6.1.7). Figure 6.2.7 compares the
relative proportion of IV and V chords in a range of ionian and major-mode
music from the sixteenth century to the present. In some genres, IV is almost
as common as V, falling from about 45% in my sample of Lassus to
something more like 15% in Beethoven—a change due to both the declining
use of IV and the increasing preference for V.
Figure 6.2.7. The relative proportion of IV and V: 100% means only IV is used, 0% means only V is
used, and 50% represents an even balance between the two.

Fourth, though it is not apparent from the preceding musical examples, the
different modes feature different harmonic patterns. In dorian, for example,
the subdominant is often deemphasized relative to the subtonic, and pieces
often revolve around i, ♭ VII, III, and V.22 Here again we might think of
contemporary popular music, where major, minor, and mixolydian all
emphasize different chord progressions. Early protofunctionality is a mode-
bound option rather than an inviolable law governing all music—analogous
perhaps to the blues in 1960s rock, a package of musical moves that can be
adopted or avoided according to taste.
Fifth, inversional equivalence does not play a robust compositional role,
for the simple reason that the harmonic vocabulary is so restricted: there can
be no question about the equivalence of vi and vi6 in genres where vi6 hardly
appears. Instead we have a proliferation of root-position primary triads that
deemphasize the melodic function of the bass in favor of its harmonic role in
articulating roots. It follows that simple harmonic functionality does not
require inversional equivalence; the basic prerequisite for harmonic function
is harmonic limitation.
Sixth, the melodic tendency of noncadential leading tones is weak. In
functional music, outer-voice leading tones almost always move to the tonic.
In proto-functional music, noncadential leading tones are less constrained,
often progressing downward (Figure 6.2.8). This suggests that there are
important differences between cadential and ordinary dominants: leading-
tone resolution is a function of the cadential suspension rather than the
dominant as a harmonic entity or the leading tone as a melodic scale degree.

Figure 6.2.8. A descending leading tone in Dalza’s Tastar de Corde 1, mm. 85ff.

Seventh, there is no notion of modulation as an orderly progression

through secondary keys; the music rarely modulates, and sometimes does so
in ways that seem undirected or haphazard. Pieces sometimes begin in one
key and end in another.
None of which should dislodge the claim that the origins of functionality
can be found in this music. We can find evidence of protofunctional thinking
whether we focus on analytical details or zoom out to consider larger
statistical features of the genre as a whole. This protofunctional tendency can
be found throughout sixteenth-century homophony, in pieces such as
Willaert’s Villanesche (1545) or Goudimel’s 1564 setting of the Geneva
psalter—all of which place a heavy emphasis on root-position triads
supporting stepwise melodic motion. We can also find echoes of this
protofunctionality in vernacular styles from Sacred Harp polyphony to the
Carter family and three-chord rock-and-roll. Recognizing a distinct
protofunctional practice is important not just for understanding the origins of
functional harmony but also the nature of functionality itself.

3. Origin and meaning

Carl Dahlhaus has cautioned that we should not take these resemblances at
face value, since some of these pieces originated with a cantus-tenor duet to
which additional voices were later added.
Provided that the significance of a composition depends on the tradition out of which it arose
—provided therefore that a procedure’s origin determines its meaning—then a frottola like
Oimè et cor must be characterized as a two-voice composition with a supplemental bass.23

Tromboncino’s “Ah partiale e cruda morte!” is one of the pieces Dahlhaus

describes as originating with a two-voice framework (Figure 6.3.1). Like “Se
ben hor non scopro el foco,” it is a quasi-rounded binary form with a
sequential coda, largely emphasizing primary-triad harmony; throughout,
cantus and tenor form a syntactical two-voice pair complete with cadences
(Figure 6.3.2). But the three upper voices also tend to articulate close-
position triads, supported by a bass that mostly sounds roots. The music thus
seems to embody two forms of musical organization simultaneously.
Figure 6.3.1. Tromboncino’s frottola “Ah partiale e cruda morte!” I have eliminated a repeat of mm.
5–8.

Figure 6.3.2. The cantus/tenor duet, and close-position triads, in the first phrase of Tromboncino’s
“Ah partiale e cruda morte!”

A little thought reveals that this is not so mysterious, for as long as upper
voices form complete close-position major and minor triads, then cantus and
tenor will necessarily form an acceptable dyadic sequence of fifths and
sixths; conversely, as long as the tenor forms a third, fifth, or sixth with the
cantus, the alto can add the note that completes the close-position triad. (In
the opening of “Ah partiale e cruda morte!” the tenor is usually a third below
the cantus, falling to the fifth/sixth before the cadences.) Looking at
Tromboncino’s music, it is very difficult to know whether he began with a
two-voice framework or some analogue to OUCH theory that produced a
two-voice framework as byproduct. The deceptive cadence in the second-to-
last line provides some evidence in favor of the second option, for in the
cantus/tenor framework there is no deception and the passage repeats to
puzzling effect (Figure 6.3.3). This suggests that even if Tromboncino
started with a two-voice framework, he was already imagining the four-
voice structure. We should not put too much interpretive weight on these
questions of compositional method, as their importance tends to recede once
we recognize that different paths can lead to similar destination.
Figure 6.3.3. Close position upper voices in mm. 23–27 of Tromboncino’s “Ah partiale e cruda
morte!” The repeated cadences are attractive in the full arrangement, but oddly pedestrian in the
cantus/tenor duet.

Similarly, the origin of Renaissance cadential formulae lies in a number of

dyadic schemas, particularly the converging-voice idiom known as the
clausula vera. Dahlhaus and Bent observe that as late as 1532, Petro Aron
was describing Figure 6.3.4 as a cadence on E, not A. By that time, however,
there was abundant evidence that the meaning of these converging voices
was already in flux: (a) pieces ending with or without converging voices
(Figure 6.3.5); (b) cadential formulations where the converged-upon note is
not used as a tonal center (Figure 6.3.5); (c) a preference for the bass to
sound the converged-upon note, as revealed by the absence of cadences
where voices converge on the sixth above the bass (Figure 6.3.6); (d) the use
of deceptive-style cadences in contexts that suggest cadential evasion and
phrase elongation (Figure 6.3.1); and (e) theoretical lists of cadences that
lack converging voices.24 Further, the auditory system will lend salience to
the bass of a root-position triad through the phenomenon of “tonal fusion”—
a psychological effect that was as real in the Renaissance as it is today.25

Figure 6.3.4. A cadence in which voices converge on E, but which modern listeners will likely hear as
emphasizing A.
Figure 6.3.5. (top) The end of Josquin’s mass De beata virgine. The final V–I and IV–I progressions
lack the suspension formula and the final IV–I has no converging pair. (bottom) The end of the mass
Sub tuum presidium, attributed to Josquin. This is an E phrygian piece that ends with a cadence in
which upper voices converge on B, not used as a tonal center in the Renaissance.
Figure 6.3.6. Possible bass notes under a converged-upon C, along with the percentage of cadences
with that bass. The cadences on the sixth above the bass are very rare.

This divergence between theory and practice reminds us that earlier music
theorists are not “native informants” who provide authoritative information
about their musical culture, but rather theory builders who try to describe,
systematize, and explain that culture.26 Rather than taking their assertions at
face value, we should check their theories against the data provided by
music itself. Doing the history of theory without evaluating theoretical
claims is like doing the history of science without considering how
experiments turned out. It is also vital to bear in mind that earlier theorists
operated in an intellectual environment substantially different from our own:
in the early 1500s it was still possible to uphold Aristotle’s view that the
equatorial regions were uninhabitable, even though European sailors had
actually been there, or that heavy objects fall faster than lighter objects, even
though anyone could demonstrate the contrary.27 If Renaissance thinkers
upheld tradition over evidence in these matters, how sensitive should we
expect them to be to subtle changes in musical practice? The question is not
meant to be contemptuous but rather to acknowledge that early sixteenth-
century minds had their own ideas about the relative weight of evidence and
authority, that the habit of empirical thinking was still in its infancy, that
they did not have our tradition of accumulating intellectual progress, that
they lacked common terms for “discovery” and “innovation,” and that they
were largely unacquainted with probabilistic inference.28 Aron’s remarks
about the cadence were thus constrained by tradition in ways we may not
fully appreciate.29 All of which presents a fascinating topic for historians of
science, and a dangerous pitfall for unwary historicists. Music historians
have chided their “presentist” colleagues for unthinkingly projecting
contemporary concepts onto the music of the past, but they have been much
less sensitive to the equally erroneous assumption that earlier theorists
operated with contemporary standards of descriptive accuracy.30
One final example of the complex interplay between origin and meaning.
We have reason to believe that Renaissance composers had at least some
notion of inversional equivalence, for the repeating contrapuntal patterns in
chapter 4 are defined precisely by the fact that structurally analogous notes
move in similar ways: in the Josquin passage in Figure 4.2.2, root moves
down by step while third moves up by third, while in Figure 3.4.6 and mm.
20–24 of Figure 3.6.6, root and third move down by step while fifth stays
fixed. To construct these sequences a composer has to understand which
motions go in which voices, and this is tantamount to an implicit notion of
inversional equivalence. (It is difficult to repeatedly lower the root if you
cannot identify it!) Imitation thus provides one potential route toward
inversional equivalence, for in asking “how can I create a consonant three-
voice transposing round?” composers could easily be led to contrapuntal
patterns in which roots, thirds, and fifths moved in characteristic ways.
Does it follow that composers thought of inversionally related chords as
having the same musical function? Not at all, for it is entirely possible to
have an abstract understanding of concepts like “root,” “third,” and “fifth,”
while still believing that register is fundamental to a chord’s contextual
function and meaning. But at the same time, it seems likely that this weaker
version of inversional equivalence formed the foundation for the later and
more robust notion. (Furthermore, it is plausible that the early preference for
I6 over iii results from a perceived resemblance between E–G–C and C–E–
G.) The interesting question is how the version of inversional equivalence
implicit in sixteenth-century practice developed into the stronger form
explicit in the eighteenth century. This is a story of gradually shifting
musical meaning, written jointly by composers and theorists, and it begins in
weak and untheorized tendencies such as the favoring of I6.
 Against Dahlhaus’s carefully hedged historicism I would therefore
propose an alternative in which “origin” and “meaning” can come apart—
one in which composers follow their ears as they grope toward the new,
sometimes using earlier techniques to create music that exemplifies
innovative and untheorized structure. If we believe music can develop in the
way that natural language often does, if we believe that composers
sometimes develop novel techniques—sometimes without an abstract
understanding of their own musical practice, and with written theory trailing
behind—then we need something more sophisticated than Dahlhaus’s
identification of origin and meaning. And if Renaissance music exhibits
weaker versions of tendencies that strengthen throughout the sixteenth and
seventeenth centuries, then it is futile to make invidious contrasts between
approaches that “rely on common-practice tonality as a prism through which
to view early music” and “methods that respect the integrity and self-
sufficiency of the languages of early music.”31 Used cautiously,
contemporary theory and data-driven analysis can help us toward a
genuinely historical perspective, revealing as-yet-undiscovered regularities
in early music, and leading to a more nuanced view of functional harmony
itself. To label this project “anti-historicist” or “presentist” is implicitly to
reserve the term “history” for the study of verbal discourse. The history of
practice is part of the history of theory.

4. Harmony and polyphony

A basic question about the development of functionality is the role of genre:

was early functionality largely confined to secular and homophonic styles, or
does it also appear in the polyphonic tradition? Both answers, I think, have
an element of truth. Certainly, the earliest and most obviously functional
music is popular, secular, and homophonic, employing techniques that were
somewhat inimical to counterpoint. But it is also true that we can find
statistical evidence for increasing functionality throughout sixteenth-century
polyphony.
To frame our study of this issue, and to set our expectations for the sorts
of patterns we might find in analysis, I want to consider the raw triadic
progressions in a large sample of music by Ockeghem, Josquin, and
Palestrina. Here I have used a computer to record all the instances of a
complete vertical triad (major, minor, augmented, or diminished) followed
by a second vertical triad, categorized by the inversions of the chords and the
interval between their lowest notes. In performing this analysis, I permitted
one or two nontriadic sonorities between the triads, to account for passing or
other nonharmonic tones. The analysis was completely automatic and devoid
of human judgment, neutral with respect to mode and centricity; the only
question was whether the vertical intervals, sounding at any one time,
formed a triad. For this reason, it gives a relatively raw perspective on the
kinds of verticalities we might expect to find in sixteenth-century music,
unadulterated by human interpretation.
Figure 6.4.1 considers the zeroth-order chord distributions in the corpora,
showing a dramatic increase in root-position major triads from Ockeghem to
Palestrina; in fact, root-position major triads and second-inversion triads are
the only sonorities that increase between the two composers.32 Figure 6.4.2
records the five largest differences in first-order frequencies among the
repertoires, sorted by the difference between their likelihood in Ockeghem
and their likelihood in Palestrina. We see a substantial increase in
descending-fifth root progressions, tripling from 4% in Ockeghem to 13% in
Palestrina, while ascending fifths increase to a smaller degree, roughly
doubling from 4% to 9%.33 Meanwhile, there is an impressive decline in
motion between 53 and 63 sonorities over the same bass—with 6–5
progressions declining faster than 5–6. (In modern terminology: ascending-
third root progressions decline more quickly than descending-third root
progressions.) The last progression on the list, 64–53 over a fixed bass, was
discussed in §5.2: this is almost completely absent in Ockeghem, rising to a
respectable 4% of all progressions in Palestrina. To put that number in
perspective, 64–53 progressions in Palestrina are about as common as
descending-fifth progressions between complete root-position chords in
Ockeghem.
Figure 6.4.1. The prevalence of various sonorities in a broad selection of works by Ockeghem,
Josquin, and Palestrina. Numbers represent the percentage of complete triadic sonorities.

Figure 6.4.2. The progressions whose frequency changes most from Ockeghem to Palestrina. The bass
interval is given in parentheses with D5 = descending fifth or ascending fourth, 0 = no motion, and A5
= ascending fifth or descending fourth. Numbers are percentages of all two-chord complete-triad
progressions.
 Jeppesen writes that in Palestrina “the vertical, harmonic requirements
assume merely the exclusively consonant, full harmony of the chords, in
which modulatory relations [i.e., progressions] play only a small part.”34
This description of purely contrapuntal harmony-free music seems better
suited to Ockeghem, where the bass typically moves by unison or step—
exactly as we would expect if it were just another voice, not charged with
the dual responsibility of supporting meaningful vertical harmonies (Figure
6.4.3).35 By contrast Palestrina has a notable preference for major triads over
minor, for root-position over first-inversion chords, and for bass leaps by
fourth or fifth—leaps that almost invariably support root-position triads.36
To my mind, this suggests that harmonies are becoming objects of
compositional concern, and that the lowest voice is starting to embrace a
second role in supporting the overall harmony. This change from pure
counterpoint toward counterpoint-with-harmonic-patterning represents a
genuine difference within the prima prattica, to the point where one might
say that Ockeghem, Josquin, and Palestrina speak different musical dialects.
And this remains true even if these changes are largely implicit and
unselfconscious—a matter of internalized routines rather than explicit
thinking.
Figure 6.4.3. The most common progressions in Ockeghem are those in which the bass moves by
unison or step. Numbers are percentages of all two-chord complete-triad progressions. A2, D2, and 0
identify bass motion.

Figure 6.4.4 catalogues the total duration spent on complete triads in

ionian-mode pieces by Josquin and Victoria.37 As we move backward in
time, chord distributions flatten: Josquin spends less time on the tonic and
subdominant and more time on first-inversion triads; and Ockeghem’s
chord-distributions would be flatter still. Victoria’s distribution shows both a
telltale emphasis on IV and preference for I6 over iii. The bass degrees most
likely to support root-position triads are 1̂, 4̂, and 5̂, the roots of the three
primary triads, while 7̂ and 3̂ are most likely to support first-inversion
triads.38 This is strongly reminiscent of Dalza and the frottola, once again
suggesting a degree of harmonic thinking.
Figure 6.4.4. The frequency of complete triads in ionian music by Josquin and Victoria.

At the level of progressions, functionality can be said to involve both

forward-looking tendencies for chords to progress in specific ways (e.g., for
V to go to I) and backward-looking tendencies for chords to be approached
in certain ways (e.g., for I to be approached by V). Dominants tend to be
“constrained toward the future” while tonics are “constrained toward the
past”; this is, indeed, key to the dominant’s tension and the tonic’s release—
when we hear a dominant, we have a pretty good idea about what is coming,
while a tonic gives us comparatively little sense of what might happen.
Figure 6.4.5 looks at some forward- and backward-looking tendencies in
Palestrina’s C ionian mass movements.39 Notable here is the V6 chord,
which tends to be preceded (57%) and followed (55%) by tonic chords, its
neighboring role coming into focus. The vii°6 chord is similarly constrained
in both temporal directions: 60% of these chords move to C, and about 80%
of them are approached by ii (41%), IV (18%), or I6 (21%).
Figure 6.4.5. Chords that are most often approached and left in specific ways in Palestrina’s masses
(C-final movements, no signature).

One nice way of visualizing these harmonic tendencies is with a tendency

histogram, a graph that compares chord’s zeroth-order frequencies to their
frequencies either before or after a particular chord. Figure 6.4.6 presents
tendency histograms for the vii°6 chord in Palestrina’s ionian-mode mass
movements. When the dotted line lies above the solid line, as at the left of
each graph, that means that a particular progression happens more than its
zeroth-order probability: thus the graphs show a very strong tendency for
both ii→vii°6 and vii°6→I.40 When the two lines are nearly the same, as at
the center of each graph, that means that a chord’s frequency of appearance
is unchanged by the presence of the other chord; the second graph shows
that IV appears before vii°6 at about its typical rate, and hence that the
relatively large number of IV→vii°6 progressions is likely due to the
popularity of the IV chord itself. Finally, when the first-order line lies below
the second-order line, as at the right of each graph, that means the
progression is suppressed: there are many fewer I→vii°6 and vii°6→V
progressions than we would expect on the basis of zeroth-order probability
alone.
Figure 6.4.6. Tendency histograms for chords that precede and follow vii°6.

I find these graphs to be endlessly fascinating, encapsulating a wealth of

musical information I know only tacitly; I can easily while away the hours
exploring the tendencies in different repertoires.41 (Particularly interesting is
the divergence between raw frequency and tendency: some common
progressions, such as IV→I in Palestrina’s ionian music, seem to be
common largely because of the popularity of their individual chords.) Here
we can see a congealing of a substantial portion of the vii°6 chord’s standard
functional behavior—what we might call the “vii°6 subsystem.” Yet we also
see that the chord’s functional role is not completely settled: bass scale
degrees 1̂–2̂–3̂, for example, are typically harmonized by I–ii–I6 rather than
with a functional dominant; this is because I→vii°6 is suppressed and
vii°6→I is favored. In this music, vii°6 has a downward tendency.
Carl Dahlhaus and Margaret Bent both point to the vii°6–I progression as
demonstrating the inapplicability of functional-harmonic concepts to
Renaissance music: for them, functional theory requires assimilating the
vii°6 to an incomplete V7, but this assimilation is blocked by the Renaissance
habit of resolving the fourth scale degree upward.42 This argument reveals a
number of misunderstandings. First, functionally tonal composers often treat
vii° as an independent chord separate and distinct from V and V7. This is
most obvious in sequences, where we find patterns like I–IV–vii°–iii which
would become nonsequential if vii° were to be replaced with V or V7.43 But
it is also true in nonsequential contexts: in J. S. Bach’s chorales, for
example, the fifth of vii°6 is more likely to ascend than descend, whereas the
seventh of V7 almost invariably resolves down (Figure 6.4.7). Second, the
shift from vii°6 to V43 as the preferred dominant over 2̂, rather than being a
precondition for functional composition, is actually a change that occurs
during the functional era, as root-functional thinking increasingly takes hold
(Figure 6.4.8).44 Third, even after nineteenth-century composers came to
prefer V43 to vii°6, the later chord inherited the voice leading of its ancestor
—rather than resolving downward, the seventh often ascends (Figure 6.4.9).
Here V43 acts like vii°6, with its root a quasi-pedal tone and the chordal
“seventh” less a dissonance-relative-to-the-root than a consonance-relative-
to-the-bass. (Indeed, thoroughbass pedagogues explicitly describe the V43 as
an optional embellishment of vii°6.45) Thus, rather than establishing a gulf
between modal music and functional music, Dahlhaus and Bent isolate a
clear point of continuity between the two traditions—indeed, a case where
understanding early music can help clarify the difference between baroque
and classical dialects.

Figure 6.4.7. The five most common voice leadings for vii°6–I progressions in Bach’s chorales. Of
these, the majority involve 4̂ rising to 5̂, rather than resolving downward.

Figure 6.4.8. The frequency of the two most common dominant chords over the supertonic, vii°6 and
V43.
Figure 6.4.9. In the nineteenth century, V43 gradually comes to replace vii°6 as the primary dominant
over 2̂. However, it inherits its voice leading from the earlier chord.

To my mind, the moral is a general one: much of the perceived conflict

between harmonic theory and early-music analysis arises out of an overly
rigid understanding of functionality. Functional harmony is not a logical
system but an army of frozen idioms, and the story of its origin is one of
gradually congealing conventions, with its strict norms appearing first as
subtle preferences, sometimes invisible to the naked eye. Statistics are
indispensable here because they are humanity’s best tool for observing subtle
regularities. Of course, its evidence is provisional, particularly when we
crudely search through large amounts of music using computers; it goes
without saying that readers should focus on broad trends rather than precise
numbers. At the same time, however, crudity can be a kind of strength: for if
clearly protofunctional patterns appear in rudimentary statistics, and if these
same patterns also show up when we look closely at individual works, then
we have converging evidence that there was indeed a gradual turn toward
protofunctionality within sixteenth-century music.

5. The Pope Marcellus Kyrie

Let us therefore search for protofunctionality in a specific piece. It might

seem perverse to choose the Kyrie of Palestrina’s Pope Marcellus Mass
(~1562), since the movement shows no sign of the homophonic textures that
have been the focus of so much critical commentary. But the choice is
strategic in the following sense: if even the polyphonic Kyrie turns out to
have protofunctional tendencies, then this will help to loosen the association
between functionality and homophony.46 Furthermore, if extraversion and
populism are characteristic of the mass as a whole, and not just its
homophonic moments, this would be significant in light of the various
legends (still possibly true) about the music’s role in defending liturgical
polyphony.
I hear each of its three sections as having the same basic form: an
introductory idea leads to an accelerating and more intense second idea,
followed by an extended cadence. Figure 6.5.1 analyzes the opening
“Kyrie.” The music sounds its opening note three times before leaping up by
fourth and then stepping back downward. This motive appears four times in
the lowest voice, always generating similar harmonies: the opening note
supports a stable and ringing root-position triad (sometimes briefly touching
upon a first-inversion triad), the fourth leap is always accompanied by a pair
of root-position triads, and the descending steps are always harmonized in
parallel thirds leading to a cadence on the initial note, a version of
Gjerdingen’s “Prinner.” The preference for fourth-based harmonic structures,
rooted in primary triads, is critical for the overall effect, as we can see by
imagining contrapuntally plausible alternatives such as vi6–ii6–iii–ii–I.
Figure 6.5.1a. The opening of Palestrina’s Pope Marcellus Mass. Each cadence is marked by an
asterisk.
Figure 6.5.1b. The end of the first section of the Kyrie.

The bass sounds the theme twice in G mixolydian—which is to say, with

entries on the G–D fifth and cadences on G—and twice in C ionian, though
the pervasive F naturals may tempt modern ears into hearing the entire
opening in C.47 The music then accelerates, introducing the material in
Figure 6.5.2 while increasing the cadential energy. Here there are no fewer
than five irreducible seventh chords, all supporting cadences that survive
into the eighteenth century. (The up-by-step/down-by-third melody will
return in the final Kyrie.) The lower voices then pick up the falling thirds,
now on G rather than E, leading to the final cadence. Throughout, the
melodic and harmonic material is consistent with functional norms.
Figure 6.5.2. In mm. 15–18, overlapping “up by step, down by third” melodies create a one-measure
unit supported by a 4̂–5̂–1̂ bass.

The “Christe” opens with four statements of a three-voice, eight-beat

figure in which a first-inversion chord acts as a neighbor to a root-position
primary triad (Figure 6.5.3). These are grouped into fifth-related pairs C–F
and G–C, with the second adding chord-roots in the bass; this suggests an at-
least-tacit understanding of roots and inversional equivalence. Palestrina
then introduces a descending scalar motive that dominates the rest of the
section. This creates a series of three- and four-measure phrases each of
which ends with a 6̂–3̂–4̂–1̂ bass, articulating either an ascending-
subdominant or an evaded ii6–I cadence—an unusual progression that could
be a deceptive resolution in phrygian. The rolling, tumbling descent evokes
the Shepard-tone passacaglias of chapter 2. Figure 6.5.4 arranges the lines to
form a canon; in Palestrina’s music, these melodic fragments are divided
among the voices and displaced in register—as if the endless descending
sequence were an abstract background that was distorted on its way to the
musical surface.
Figure 6.5.3. The Christe from Palestrina’s Pope Marcellus Mass.

Figure 6.5.4. The Christe as a Shepard-tone passacaglia.

Figure 6.5.5 annotates the final Kyrie. The section opens with paired
voices in close canon, each articulating the descending form of the diatonic
third’s basic voice leading; because that voice leading produces a canon at
the fifth, the motive’s two transpositions sound almost the same notes
(Figure 6.5.6). The bass supports this fifth-based structure to create iconic
harmonic cycles, rather like Figure 4.10.2, with the canonic lines either the
upper or lower third of the governing triad. This is a wonderful example of
the inseparability of sixteenth-century harmony and counterpoint, the music
being at once harmonically significant and also generated by melodic
imitation. The next motive picks up the end of this figure, repeating entries
on C and G to create an accelerating (G, B) → (E, C) → (F, A) loop that lasts
until the end of the movement (Figure 6.5.7, cf. the lower-right pattern in
Figure 3.1.10). Measure 75 is of particular interest, with a IV64 chord
decorating the piece’s final V–I progression; that chord is itself embellished
with a irreducible seventh I have labeled vi65. Here the F, while dissonant
with the bass, serves both as resolution of the suspension G and as
preparation of a suspension that resolves to E—a second-inversion triad that
cannot be dismissed as merely nonharmonic.
Figure 6.5.5. The end of the Kyrie in the Pope Marcellus Mass. Each cadence is marked by an
asterisk.

Figure 6.5.6. The basic voice leading in the final Kyrie. The fifth-related entry pairs sound almost the
same notes.
Figure 6.5.7. Motivic entries at the end of the Kyrie of Palestrina’s Pope Marcellus Mass.

What is striking is the idiomatic character of the piece’s

protofunctionality: beyond the general fondness for primary triads and fifth
progressions, we find a relatively small set of schematic figures in which
harmony and melody are inextricably linked. In the opening Kyrie the “up
by fourth, down by step” motive gives rise to Prinner figures, IV–I6–ii–
vii°6–I, followed by an ascending-step/descending-third motive that
generates repeated cadential progressions. In the “Christe,” we have the two-
measure neighboring figure and the endlessly descending triads, while in the
last section it is dyadic voice leading and simple harmonic cycles.
Throughout, the close imitation creates a sense of near repetition; underneath
the ever-changing surface, we can hear something like a harmonic ground,
endlessly circulating through familiar harmonies. The overall impression is
less of functionality congealing from the top down (for example in the form
of greater attention to regular phrase structure or large-scale melodic
templates) than of bubbling up from the bottom—of an increasing focus on a
small number of voice-leading patterns. This music is neither purely linear
nor purely harmonic, but an inseparable amalgam of the two.
In all, about 85% of the movement’s progressions conform to functional
norms, with most of the exceptions falling into just two categories: 6–5
motions over a fixed bass, and progressions involving the mediant. These
numbers are characteristic of the Mass as a whole and higher than is typical
for Palestrina: in his other ionian-mode pieces, the proportion of functional
progressions is typically closer to 75% than to 85%. (In fully functional
music, the number would be closer to 95%.) By way of contrast, Figure 6.5.8
shows the close of a less-functional movement, the first Agnus Dei from the
mass In minoribus duplicibus.48 This music is suffused with nonfunctional
passages: prominent mediants; dominants that almost seem to be avoiding
the tonic; retrofunctional ascending fifths; a dramatic subtonic–tonic
progression; and a meandering lowest voice that seems to be avoiding the
low C3 from m. 17 to m. 29. It is almost completely devoid of the
subdominant emphasis of so much early functionality. Here we see another
side of Palestrina’s musicality, not so much “regressive” as learned or
introverted. The contrast between these movements reminds us that
Renaissance musicians were free to explore a vast harmonic territory, with
parameters like strength of centricity and degree of functionality varying
from one piece to another.
Figure 6.5.8. The end of the first Agnus Dei of Palestrina’s “In minoribus duplicibus.”

6. A broader perspective

I am proposing that there is a distinctively sixteenth-century form of

protofunctionality, evident as early as 1500 and particularly characteristic of
the “paratonal century” from 1550 to 1650. This ionian-mode style features:
(a) an emphasis on primary triads moving by fifth, articulating simple I–V–I
and I–IV–I cycles; (b) the favoring of I6 over iii as the default harmonization
of 3̂ in the bass; (c) the concomitant disfavoring of harmonies like vi6 and
iii6; (d) a wealth of simple predominant-dominant-tonic progressions in
which the predominant can be either ii, IV, or vi, with relatively little
preference among them; (e) the frequent use of vii°6–I, often preceded by I6,
IV, or ii; (f) precursors to later schemas like the “Fenaroli” and “Prinner”
rooted in the primary steps-and-fifths defaults of Figure 6.1.3; and (g) the
use of half, full, and deceptive cadences playing familiar roles. Not all these
features are equally prominent in every repertoire, let alone every piece, and
for many decades they coexist with nonfunctional modes of musical
organization. In other words, they are optional idioms rather than rigid rules.
Ionian mode provides the paradigm for two reasons. First, its major triads
occupy the positions of tonic, subdominant, and dominant, so that a
preference for major chords will reinforce the main functional poles. By
contrast, this same preference can have a centrifugal effect in dorian or
aeolian, as we will explore in chapter 8. The favoring of fifth-progressions
among C, F, and G major triads, regardless of which white-note mode is
being used, is one of the most challenging features of sixteenth-century
music: we are so used to a Copernican universe, in which chord-meaning is
relative to the tonic, that it is hard to imagine the absolute space of earlier
times, in which particular chords could have their own tonic-independent
tendencies. Second, the leading tone lies within the key signature, not
requiring explicit accidentals or implicit signaling via cadential
suspensions.49 Accidentals are thus a prerequisite for the transmission of
ionian routines to other modes: if alterations are primarily a matter of musica
ficta, they remain linked to specific cadential formulae, and hence cannot
proliferate throughout the phrase. It follows that modal homogenization
could scarcely occur before about 1550.
The decision to focus on local harmonic structure is motivated by my
belief that pieces can sound functional even while eschewing conventional
modulatory trajectories—John Coltrane’s Giant Steps, for example,
combines local ii–V–I organization with a key scheme that moves by major
third.50 (This is also consistent with psychological research that has called
into question listeners’ ability to track modulatory relations over even
moderate lengths of time.51) Furthermore, I believe many of the other
parameters that have been proposed as crucial to early functionality—for
instance, homophony or regular phrase structure—are somewhat secondary:
it is easy to come up with examples of polyphonic or irregularly phrased
music that are still strongly functional, from the Kyrie of the Pope Marcellus
Mass to the more functional fugues of Frescobaldi.52 The development of
functionality was a bottom-up process that began with local details and
gradually encompassed larger and larger timespans, ultimately including
phrase structure, modulatory patterns, and new formal templates.
To understand these changes, we need something between the raw
statistical data of §6.4 and the particular analyses of §6.5. To that end I have
created, over many years and with the help of more than a hundred
collaborators, a database of Roman numeral analyses matched to musical
scores, consisting of about a thousand pieces from Dufay to Brahms. Figure
6.6.1 shows the collection as it currently stands. In building it I have tried to
combine breadth and depth, analyzing a large number of pieces by
paradigmatic composers (Josquin, Palestrina, Monteverdi, J. S. Bach,
Mozart, Beethoven) and a smaller number of pieces by a wider range of
composers—with particular attention to the decades between Josquin and
Palestrina. For methodological details, consult appendix 4.
Figure 6.6.1. My corpus of hand-analyzed pieces; “major” refers to modes with a major third above
the final, “minor” to modes with a minor third above the final.

Figure 6.6.2 shows the most common harmonizations of descending scale

steps in three not-fully-functional corpora: Josquin’s ionian-mode pieces,
Goudimel’s 1564 setting of the Geneva Psalter, and a collection of
Monteverdi madrigals from Books 3, 4, and 5.53 The change from Josquin to
the later composers is quite striking, the mediant triad and supertonic
deemphasized in favor of primary triads. Again, one finds a similar primary-
triad emphasis in a wide range of quasi-functional music from Palestrina to
rock.
Figure 6.6.2. The most common harmonizations of descending scale steps in ionian-mode passages of
Josquin, Goudimel, and Monteverdi.

Figure 6.6.3 shows how composers treat different bass degrees—in

general, 1̂ and 5̂ are most likely to support root-position chords, and 3̂ and 7̂
least likely to use root position. Between these poles are the subdominant
scale degrees 2̂, 4̂, and 6̂, which can support either root-position or first-
inversion chords. Here we can see a slight rise in first-inversion chords as we
move toward the classical period, with ii6 eventually eclipsing IV as
principal subdominant. All of this suggests the congealing of the conventions
that were eventually to be codified as the seventeenth-century “rule of the
octave” (§7.5). These regularities start to appear in musical practice long
before they were described by theorists or pedagogues.
Figure 6.6.3. The percentage of first-inversion chords on each scale degree.

Figure 6.6.4 shows the proportion of diatonic triads found in a broad range
of music. Here I have anachronistically categorized chords by root,
considering I, I6, and I64 as forms of the same chord.54 What emerges is a
fairly dramatic increase in the number of V chords across the entire era.
There is also a clear emergence of IV as a privileged third sonority: up
through about Gombert, the subdominant chord is just another chord, no
different from ii, vi, or iii. After this point it starts to emerge as special, only
to be supplanted by the supertonic in the seventeenth century. Again we see
the prehistory of functionality moving through distinct phases, with the late
sixteenth century being relatively subdominant-focused, and the seventeenth
century being comparatively more oriented toward the supertonic. If this is
so, then we need to be careful not to generalize about functional harmony as
such. Instead, we will want to recognize the existence of multiple
functionalities, some featuring IV and others ii.55

Figure 6.6.4. Ionian-mode sonorities categorized by root.

Turning now to first-order properties, we see in Figure 6.6.5 that

dominants increasingly tend to progress to tonics.56 In much the same way, ii
increasingly tends to V, with its attraction-to-the-dominant eventually
matching the dominant’s attraction-to-the-tonic. It is interesting that vii°6
neither increases in frequency nor in its tendency to I—it has already settled
into its role as a dominant-functioned chord. Figure 6.6.6 shows the
likelihood that a dominant chord will be approached by IV, vi, and ii: in
much sixteenth-century music, there is rough parity among these options,
while in the seventeenth century, ii starts to supplant IV and the predominant
vi disappears almost entirely (Figure 6.6.7). Indeed, vi gradually adopts a
new role as a pre-predominant which usually moves to ii or IV prior to
moving on to the dominant: thus the harmonic cycles of harmonic
functionality gradually lengthen, from the simple oscillations of
protofunctionality (e.g., I–IV–I and I–V–I), to the standardization of the
four-chord I–ii–V–I, to the eventual inclusion of the vi chord in the five-
chord cycles I–vi–(ii or IV)–V–I. This process terminates with the baroque’s
seven-chord sequence of thirds, I–vi–IV–ii–vii°–V–I, to be discussed in
chapter 7.

Figure 6.6.5. The “tendency” of V→I and ii→V, defined for V→I as the likelihood of I following a V
minus the simple likelihood of V. This is the increase in the probability of one chord given that you
have just heard the other.
Figure 6.6.6. The relative popularity of the ii–V, IV–V, and vi–V progressions.

Figure 6.6.7. Percentage of all root-position vi (or vi7) chords that move directly to root-position V (or
V7).

Figure 6.6.8 shows the proportion of progressions that conform to the full-
fledged grammar of functional harmony described in §7.1. The picture here
is of a very gradual transition from modality to functional harmony: a gentle
and consistent rise from the early polyphony of Ockeghem and Josquin to
the music of Haydn, Mozart, and Beethoven, punctuated by outliers such as
Dalza, the frottolists, and Juan del Encina—and, in the other direction,
figures such as Castello. This data suggests it would be foolish to try to draw
a sharp line between modality and functionality; instead there is a very slow
evolution that takes place over several centuries, and which is particularly
dramatic in the century from 1550 to 1650. (As Jeppesen put it: “the entire
course of the development of music in the sixteenth and seventeenth
centuries was toward harmonic stability and tonal centralization.”57) We find
independent evidence for this process in individual pieces, crude
computational surveys, and large collections of human analyses. It is hard to
understand how scholars could once have favored a sudden or saltationist
perspective, for the data is as close to evolutionary as one could possibly
imagine.58

Figure 6.6.8. The proportion of progressions conforming to my grammar of functional tonality.

 7. “I cannot follow”

Marenzio’s “Ahi dispietata morte” (1585) sets a text written more than two
centuries earlier, one of a large group of sonnets lamenting the death of
Laura de Noves, Petrarch’s muse and unrequited love.59 In many madrigals
the specter of death is a metaphor for consummated desire or a gambit in the
game of sexual persuasion; here, Laura is really dead, and the poet is left to
reflect on grief, loneliness, and the consolations of memory. Besides being
intrinsically worthy of contemplation, Marenzio’s setting provides an
opportunity to review many of the topics we have been exploring.
Like many pieces of its time, it is composed of fragments that individually
conform to functional conventions, but are juxtaposed in surprising ways.
The piece begins and ends with E triads, and was likely conceived as E
centered. However, many of the E chords are major triads proceeding
directly to A: in this sense, they are “constrained toward the future” (§6.4,
§8.6), acting like dominants even though they may have been intended as
tonics; for this reason, I have analyzed the opening in an anachronistic A
minor. The piece opens with the pair of phrases in Figure 6.7.1, one
outlining E–a–E, the second answering with a–E–a. The soprano soars far
above the lower voices (an effect Monteverdi used in “Cruda Amarilli”), and
each voice sings a scale fragment, two ascending and two descending. The
cadence uses an irreducible seventh, introducing the suspension E with the
metrically weak 64 chord and blurring harmony and counterpoint into an
idiomatic whole.60
Figure 6.7.1. The opening of Marenzio’s “Ahi dispietata morte.”

The second phrase again features A minor scale fragments, but more
straightforwardly; I have analyzed the opening C–E dyad as A minor even
though the A does not arrive for two-and-a-half quarter notes. At the end of
the phrase, the upper voices prolong the word “vita,” overlapping with the
next phrase (“L’una m’à posto in doglia” / “The one plunged me into grief”).
Here we leave the A minor/E major orbit, the tenor/bass third A–C slipping
down to inaugurate a sequence of basic voice leadings, (G, B) → (E, C) →
(F♯, A), forming a harmonic cycle in G. The upper voices then repeat this
music up by fourth, a simple sequence grounded in imitation. As Figure
6.7.2 shows, the diatonic third’s basic voice leading continues throughout the
passage, shifting from (F♯, A) to (F♯, D) and continuing with thirds on G
and C; this superimposing of dyads is reminiscent of Figure P4.3. The dyads
lead to a deceptive cadence decorating vi with IV6, initiating a series of
fauxbourdon sixths and ending with a firm cadence on A.
Figure 6.7.2. The diatonic third’s descending-fifth basic voice leading in mm. 8–10.

The fifth phrase (“l’altra mi tèn”/“the other holds me here”) is another

imitative sequence, an ascending-third pattern that brings D minor to C
major (Figure 6.7.3). Figure 6.7.4 shows that it is a three-note, two-voice
canon at the distance of a quarter note, with the different iterations projected
into different voices. Here again imitation produces harmonic patterning, an
ascending-third sequence familiar in functional tonality. This sequence is
immediately balanced by a gentle alternation between I and IV64 whose
inner voices suggest a canonic A–G–F–E descent. This Prinner variant is a
widespread triadic schema rather than a localized eighteenth-century idiom.
The pattern will be the focus of the last section of the piece, and its
appearance here is a subtle bit of foreshadowing. We end with a wonderful
right angle, the bass ascending to E for a V–I cadence on an unexpected A
major.61
Figure 6.7.3. The second part of Marenzio’s “Ahi dispietata morte.”

Figure 6.7.4. The ascending-third sequence in mm. 14–16, leading to a IV64–I alternation.

The piece’s next phrase sets the words “et lei che se n’è gita seguir non
posso” / “and she who has gone I cannot follow.” This virtuosic music is
based on the Chase schema, with the underlying melodic pattern appearing
canonically eleven times as the music moves from A minor to C major
(§4.4). Canon here is a metaphor for following, with the notes portraying
what the words claim is impossible—a moment of negative text-painting
embodying the fantasy of mortality overcome. Figure 6.7.5 unpacks the
mechanics, rebarring the music to group the entries into ascending-step
triples (C–C–G, D–D–A, E–E–B). The final entries extend up by seventh
rather than fifth, leading to a cadence on C. This piece was my first
encounter with the Chase schema, and I remember being dumbfounded by
Marenzio’s contrapuntal and expressive virtuosity—admiration that was
later tempered by the realization that he was reusing a well-known piece of
Renaissance vocabulary.

Figure 6.7.5. The Chase schema in the seventh phrase of the madrigal.

At this point, more than halfway into the piece, we have had almost no
signs of E centricity, save an opening phrase that is easily heard in A minor.
Instead, the music has offered a trio of strong cadences on A (mm. 7, 13, 20)
and two less-marked turns to C (mm. 16 and 24). This is not to suggest that
it is impossible to hear the phrygian signals; on the contrary, I imagine that
cognoscenti would interpret the score as E centered. My point is that the
music does very little to establish E as a genuine, psychologically real
center; a casual listener from any culture would be hard pressed to single out
the E as the primary tone.
What is fascinating is that the second half of the piece starts to insist on a
recognizably phrygian modality (Figure 6.7.6).62 The contrapuntal virtuosity
of “I cannot follow” gives way to a homophonic passage on “ch’ella nol
consente” (“she will not let me,” a slightly odd blaming of the deceased for a
problem not of her making). The details of this little passage are fascinating:
it begins with what seems like functional music in G, ii–IV–V–i supporting
2̂–1̂–7̂–1̂ in the melody, but the cadential tonic is a minor chord that
progresses to A major, evoking a phrygian close.63 (The continuation, which
stays on A rather than moving to D, solidifies this impression.) The effect is
of an overlapping cadence, but without the standard suspension—a
surprising harmonic turn where the phrase seems to acquire an extra chord.

Figure 6.7.6. The third part of “Ahi dispietata morte.”

The next bit of music is transitional and tonally unstable. The lower two
voices sing “Ma pur ogni or presente” (“But in every moment”) in canon at
the fifth, using the basic voice leading to move through thirds and sixths on
D, G, C, and F. This is yet another example of the connection between the
basic voice leading and canon. The top three voices then vary this material in
singing “nel mezzo del meo cor” (“in the middle of my heart”), leading to a
cadence on a phrygian E that cannot be interpreted as a dominant. We
immediately turn to homophonic declamation with the C major “madonna
siede” (“my lady sits”), setting the stage for the final section.
That music, based on the words “et qual è la mia vita, ella sel vede” (“and
what my life is now, she sees”), proposes that the dead live on in our
remembrance (Figure 6.7.7). The texture is wonderfully suspended, slowing
the harmonic rhythm to luxuriate in a series of prolonged third-related
harmonies, enlivened by stepwise oscillation (e.g., the four A–G motions in
mm. 33–34, the four F–E motions in mm. 34–35, etc.).64 Here there are
almost no hints of (proto-)functional harmony. The basic contrapuntal
technique is shown in Figure 6.7.8; it is based on a four-note descending
figure that was first presented in mm. 11–12. Harmonically, the figure
appears in two forms: in one, the sustained voice creates oscillations
between fifth-related triads (“option 1” on the figure, used in the middle of
the phrase). The second and more modal version adds a second sustained
voice a third below the first, leading to a series of third-related triads, usually
arranged to form VI6–i progressions in which 6̂ acts as a descending
tendency tone (mm. 33–35, 41–43). To my ear, these VI6–i progressions
recall the vocabulary of minor-mode rock (e.g., Patti Smith’s “Ask the
Angels”).

Figure 6.7.7. The end of “Ahi dispietata morte.”

I have listed the appearances of the figure above the score, labeled by its
highest note, since the beginning is variable. The basic pattern can support a
unison canon at the distance of one half-note, but a third unison entry cannot
be introduced until three half-notes after the second—as at the end of the
piece, where we have four E entries in a row. However, it is possible to
accelerate the canon by bringing in an entry at the third (Figure 6.7.8). This
leads the music down by third from C to A for a phrygian cadence, the third
such cadence in less than ten measures; we then move downward to F and
back, but instead of continuing on from A to C, the soprano entry in m. 40 is
distorted to reach a high E. This opens onto a deeply spiritual ending and the
most extended phrygian passage in the piece—transforming the earlier E-as-
possible-dominant to an unmistakable E-as-phrygian-tonic. It is heartfelt and
expressive, music that is as alive today as it was more than four hundred
years ago: a poem of the mind in the act of making peace with death.

Figure 6.7.8. The contrapuntal logic behind the final passage of Marenzio’s “Ahi dispietata morte.”

But is it functionally tonal? I confess that the more I think about this
question, the more confused I become. It largely depends on whether we
think the glass is half empty or half full: clearly, Marenzio’s music is not
“functionally tonal” if we require that music treat the harmonic norms of the
eighteenth and nineteenth centuries as if they were inviolable musical laws.
But at the same time, it is clearly music where notions of tonic, dominant,
and subdominant, not to mention harmony, sequence, and deceptive cadence,
have some use. It is music that seems to move purposefully through key
areas, starting on an E-as-quasi-dominant, exploring the white-note A and C
areas, before introducing the phrygian flavors of its second half. If
“functionally tonal” means “music where the concepts of functional
harmony can be useful and enlightening,” then the music is functionally
tonal—despite the fact that it is in the phrygian mode.
Which should we emphasize: the differences from eighteenth-century
practice or the similarities? In some sense, the question is whether
functionality comes in degrees. To me, the affirmative answer is natural: this
music’s departures from later norms do not nullify the equally clear
continuities with that tradition. But in saying this I am conscious that I am
drawing on a twenty-first-century conception of functional harmony as
heterogeneous and largely conventional. (And perhaps more to the point: I
am drawing on my enculturation into popular styles that treat functionality
as an option rather than a requirement.) To a nineteenth-century musician
who had been taught to conceive of musical rules as natural laws,
comparable to the as-yet-unchallenged laws of Newtonian physics, things
might well seem different. Perhaps from that perspective, little departures
would have a much greater significance—as alienating and shocking as
minor departures from the laws of nature. Perhaps from that perspective,
genuine functionality would require that music rid itself of all errors and
fully embody the truth. That perspective may be gone at the level of our
conscious thinking, and yet still survive in our unconscious habits,
preventing us from acknowledging protofunctionality when we hear it.

1
Korte 1929, Georgiades 1937, and Besseler 1950 (early 1400s, all cited in Dahlhaus 1990, p. 83);
Lowinsky 1961 (early 1500s); Dahlhaus 1990 and Fétis (1840) 1994 (Monteverdi); Bukofzer 1947
(Corelli); Gauldin 1995b (after Bach).
2
This view is consonant with Christensen 1993b and Taruskin 2005, vol. 2, pp. 471–72. On avian
evolution, see Varricchio and Jackson 2016.
3
Lowinsky 1961.
4
Compare Taruskin’s suggestion that there is a unified logic based on fifths: “we [modern readers]
are therefore much more fully aware than anyone could have been at the time of the range of
implications the new cadential structure [i.e., the root-position V–I cadence] carried within it” (2005,
vol. 2, pp. 471–72, my italics).
5
In fact, there were relatively few such grammars; medieval grammarians were largely concerned
with Latin rather than “vulgar” language. In contrast to the standard practice of historical linguistics,
Bent explicitly recommends founding musical grammar on contemporaneous grammatical theorizing
(1998). This view is so foreign to historical linguistics that I am unable to find any linguists who have
considered it.
6
For a linguistic analogue see Labov 1975, p. 107: “this puts us in the somewhat embarrassing
position of knowing more about a speaker’s grammar than he does himself.”
7
The opposition between zeroth-order and first-order is broadly related to Dahlhaus’s opposition
between Riemannian and Rameauian conceptions: Riemann tends to emphasize zeroth-order chord
categories (“functions”); while Rameau tends to emphasize first-order progressions (Dahlhaus 1990,
Christensen 1993). Bobbitt (1955) and Meeùs (2000, 2018) emphasize first-order probabilities. There
are also second-order probabilities concerning the likelihood that a two-chord sequence such as ii6–V
will continue in a particular way, third-order probabilities governing the successors to three-chord
sequences, and so on.
8
This figure categorizes chords by root; one gets a very similar graph if one looks only at root-
position triads.
9
In general, the piece exhibits an intriguing and playful mix of scalar inflections. Since it was
published in tablature, it is unclouded by questions about musica ficta.
10
Dalza’s other Venetian Pavanes—which indeed feel more like variations on a single template
than distinct pieces—include similar cadential interludes.
11
For the sake of clarity, my realization includes parallel perfect intervals which would be avoided
in practice. The figure’s bottom level is very similar to a graph appearing in Rameau’s 1737
Génération harmonique (Christensen 1993a, p. 194). Unlike Rameau, I consider the sequence a
distorted quantization of an exact contrary-motion sequence (Yust 2015a). This recalls the strategy we
used to model rock harmony (Figure 2.1.7).
12
As the tradition develops, 4̂–3̂ is increasingly harmonized by V7–I. Other important possibilities,
not appearing on my simple model, include V harmonizing 5̂, IV harmonizing 1̂, and vii°6–I
harmonizing 4̂–3̂.
13
Including the octave repeat, the entire structure can be seen as a near sequence whose interval of
repetition changes from a fourth (B–E) to a fifth (E–B).
14
For a justification of my rebarring, see Boone 2000, pp. 36ff.
15
This same sequence appears in many other Renaissance pieces, including Monteverdi’s 1610
Vespro della Beata Vergine (Rotem and Schubert 2021).
16
Dahlhaus 1990, p. 288: “Many details [of the frottolas] cannot be explained without recourse to
the concept of a chord.”
17
“The I–IV, IV–I, and V–I fifth-relations in [Dalza’s] Pavanna alla ferrarese are so strikingly
marked that Lowinsky was unquestionably right to see in them something new” (Dahlhaus 1990, pp.
109–10).
18
This is a relatively small sample so things might change if we looked at more music.
19
Recall Gjerdingen’s remarks about I6, iii, and the rise of functional tonality (§1.5). It is also
possible that the preference for I6 reflects the influence of the mixolydian mode, where root-position
mediant triads are unavailable.
20
Compare McClary 2004, p. 15.
21
Compare Figure 6.1.5, m. 9; Figure 6.1.6, m. 4; Figure 2.4.1; Pinter 2019; and McClary 2004, p.
218.
22
For example, Dalza’s “Calata ala spagnole 1” and the “Caldibi Castigliano.”
23
Dahlhaus 1990, p. 281 (my italics). In a similar vein, Bent writes that “the failure to
acknowledge a discant-tenor duet . . . can invalidate an analysis” (1998).
24
Thanks here to David Cohen. Richard Taruskin (2005, vol. 2, pp. 465–72) makes a similar point.
 25
Parncutt 1988.
26
See Powers 1992a, pp. 43–4 for a similar view.
27
See Wootton 2015, pp. 72–73.
28
Hacking 1975.
29
Even in 1597 Thomas Morley was describing the consonance of the perfect fourth with a long
list of tenor-based precepts like “the fourth above the tenor is consonant only if the bass is a third or
fifth below the tenor,” paying homage to the dyadic tradition despite his own music’s manifestly
triadic and functional qualities.
30
Compare Bent 1998 and Dahlhaus’s remarks on the relation between V7 and vii°, discussed at
the end of the next section. Bent requires that “analytical tools be harmonized with early techniques
and vocabulary, to the extent that these can be recovered and extended, just as what is incompatible
with how they parsed their music should be avoided” (1998, p. 24).
31
Jesse Anne Owens, in the preface to Wiering 2001, p. ix.
32
Interestingly enough, this seems to reflect both an increasing preference for “major” modes and
an increasing preference for major triads within those modes. Major-mode pieces—interpreted as
“pieces ending on a major triad that is in the key signature, and which is the most-frequent chord in
the piece”—rise from a total of about 14% in Ockeghem, to 18% in Josquin, to 30% in Palestrina.
Meanwhile, the proportion of major triads among the consonances within a broad selection of minor-
mode pieces (i.e., pieces whose last root supports a minor triad in the key signature) changes from
22% in Ockeghem to 19% in Josquin to 29% in Palestrina; the corresponding numbers for ionian and
mixolydian are 24% for Ockeghem, 34% for Josquin, and 51% for Palestrina.
33
Some but not all of this is due to Ockeghem and Josquin’s greater preference for progressions
that move from a complete triad to an open fifth. In Palestrina, 27% of root-position major triads move
down by fifth to a root-position triad vs. only about 9% for minor triads, a difference that is too large
to be due to cadences alone.
34
Jeppesen (1931) 1939, p. xi.
35
The relative balance between major and minor, root position and first inversion, likewise
suggests that consonance is the primary harmonic consideration, as if Ockeghem had only weak
preferences among the different vertical possibilities.
36
Carl Dahlhaus claimed that Renaissance harmony is marked by “bass formulas” in which “the
difference between 53 and 63 sonorities over the same bass has little or no importance” (1990, p. 145):
in Palestrina, however, bass motion by fourth or fifth overwhelmingly tends to support root-position
triads. Meier (1988, p. 51) argues that the bass voice, along with the concept of the chord, becomes
increasingly important during the sixteenth century.
37
I used my computer to select one-flat pieces ending on an F triad and with F major as the most
common chord, and zero-flat pieces ending on C and with C major as the most common triad; I then
averaged the behavior in these two groups. Roig-Francolí 2018 considers some of Victoria’s proto-
baroque tendencies.
38
The preference for 63 chords over 7̂ is unsurprising, since the root-position triad is diminished.
39
That is, pieces with no key signature and which end on a C triad.
40
For clarity, I eliminate a number of chords whose zeroth-order tendency is very similar to their
tendency before or after vii°6.
41
This is a good place to reiterate that all my code and data are available online.
42
Dahlhaus (1990, p. 63) and Bent (1998, p. 44). Dahlhaus views the Renaissance “parallel
cadence” as combining 3–5 and 6–8 interval progressions, endowing the third above the bass with an
ascending tendency. The classical seventh-above-the-root, by contrast, has a descending tendency and
is rarely doubled.
 43
See Tymoczko 2011a, chapter 7.
44
V43 is awkward from an earlier perspective since the fourth above the bass is dissonant, yet its
tone of resolution, the third above the bass, is present in the chord; in Renaissance music it is very
rare, usually arising from a double suspension (i.e., D and F suspended against G, resolving downward
to C and E, with B forming the Renaissance seventh and moving up to C). I suspect composers like
Corelli and Bach still shared this point of view.
45
See Buelow 1992, p. 35, and Holtmeier 2007, p. 36. C. P. E. Bach writes: “to the chord of the
major sixth and the minor third, the perfect fourth is sometimes added without express indication”
([1762] 1949, p. 239). Thinking of V43 as an (optional) extension of vii°6 is very different from
thinking of vii°6 as a subset of V43.
46
Taruskin (2005, vol. 2, p. 655) articulates a common texture-based view: “the style of
Palestrina’s Kyrie does not differ especially from the ars perfecta idiom [i.e., the Renaissance
polyphonic style beginning with Josquin] with which we are familiar, because the Kyrie is a sparsely
texted, traditionally melismatic item where textual clarity was not of paramount concern. [ . . . ] It is in
the ‘talky’ movements of the Mass—the Gloria and the Credo—that the special post-Tridentine
qualities emerge.”
47
Jeppesen (1944–1945) 1975 agrees with this reading, arguing that the movement is in
mixolydian for the first eight measures and ionian thereafter.
48
This belongs to the same “tonal type” as the earlier Kyrie: it uses high clefs, has no signature,
and ends on a C triad (Powers 1981).
49
Lowinsky, by contrast, favors a double-mode theory where functional harmony arises
simultaneously in both ionian and aeolian (1961, pp. 8–10).
50
Stein 2002.
51
Cook 1987, Marvin and Brinkman 1999.
52
Long (2018, 2020) emphasizes the role of rhythm and phrasing in engendering harmonic
expectation. But it seems to me that the phenomena she discusses can appear in both modal and
functional contexts: regularly phrased eight-bar dance music, like irregularly phrased polyphony, can
be more or less functional. For this reason, I think we need an account of functionality that is
separable from issues of phrasing and rhythm, even if in actual historical practice these phenomena
were often intertwined.
53
For the Goudimel pieces, I have taken the preexisting melody as primary; it mostly appears in
the tenor but sometimes in the upper voice; for Monteverdi, I have used the upper voice, with no
implication that it is preexisting.
54
One gets a reasonably similar graph if one considers only root-position triads, or just first-
inversion chords.
55
Rock and jazz represent twentieth-century poles of this dichotomy: rock musicians favor IV and
jazz musicians ii.
56
Tendency here is the difference between the chord’s probability after a dominant and its overall
zeroth-order probability, as in Figure 6.4.6.
57
Jeppesen (1944–1945) 1975, p. 100.
58
Fétis (1840) 1994 (pp. 30ff) was an early advocate for the saltationist view. More recently,
Harold Powers (1981, p. 467) argued against the gradualist picture—in retrospect, placing too much
weight on theoretical statements, and not enough on musical practice. Long 2020 makes a similar
point.
59
Einstein 1948, p. 659, notes that this same text was set by Palestrina.
60
Note the similarity to Figure 5.1.5, which is in many ways its mixolydian counterpart.
 61
The third quarter of m. 18 is harmonically obscure, with my preferred reading being a IV64
whose root F is decorated by a suspension G.
62
Coluzzi 2015 explores modal flexibility (commixtio tonorum) in another Marenzio madrigal.
63
See §7.4 on the ii–IV progression, which is largely eliminated in later music. Here again we see
functional practice coming gradually into focus: in this music, the poles of tonic, subdominant, and
dominant are fairly clear, but many peripheral conventions are absent.
64
Newcomb 2007 describes this (perhaps uncharitably) as “chordal noodling.”
 Prelude
Could the Martians Understand Our Music?

Gottfried Weber argued that Western music was pervasively ambiguous, as

analysts have a choice about which notes to consider nonharmonic. The
fragments in Figure P7.1, for example, can be given either concise or
verbose readings depending on how one treats the labeled notes. Take a
moment to think about how you would analyze this music. Along the way
consider how the desire for consistency influences your judgments: for
example, the second and third progressions are loose retrogrades; should
they be analyzed so as to reflect this?

Figure P7.1. Possible analyses of four one-bar passages.

My answers are as follows. In the first progression, I choose the concise

reading, treating the melodic D as an ornament to a IV–I progression. In the
second, I favor the verbose reading because I–ii–vii°6–I6 is an important J. S.
Bach idiom, appearing once every three chorales or so.1 In the third, I
choose the concise I–vii°6–I6 interpretation. Meanwhile in the last, I reject
both concise and verbose options, treating the bass F as harmonic and the
tenor A as passing to obtain V–V2–I6. In each case I identify nonharmonic
tones to obtain what I consider the most plausible reading, avoiding
progressions I believe to be rare. (As Weber said, the ear chooses the “most
convenient, simple, easy, and suitable” explanation.2) This is justified by my
general conviction that functionally tonal music involves a small number of
characteristic progressions embellished by a small number of contrapuntal
formulae.
Ian Quinn has observed that this process raises a question of epistemic
circularity.3 All my readings avoid labeling ii–I progressions present on the
musical surface, justifying these decisions by claims about “what typically
happens.”4 Yet those justifications are themselves dependent on my readings
of specific passages like those in the example. Perhaps if I habitually labeled
ii–I progressions I would believe ii–I progressions to be more common,
which would in turn encourage me to find ii–I progressions elsewhere.
Weber said the ear chooses the simplest analysis: but what if our notion of
“simplicity” depends on prejudices not themselves justified by objective
data? Would intelligent Martians, if granted access to our music but not our
pedagogy, be able to do harmonic analysis?
The issue is that Western classical music combines two different syntaxes:
a contrapuntal syntax regulating dissonance and a harmonic syntax
governing the progression of “genuine harmonies.” The contrapuntal system
does not require any distinction between “genuine consonances” and
“consonances produced by nonharmonic embellishment”; it governs the
behavior of any and all dissonances. The harmonic system developed after
the contrapuntal system had stabilized; it does not govern consonances as
such, but only that subset qualifying as genuine. This does require that we
distinguish law-abiding harmonies from the merely “apparent” chords that
are the byproduct of nonharmonic motion. This is made all the more difficult
by composers’ tendency to ensure that simultaneous nonharmonic tones are
mutually consonant (§5.5).5 The avoidance of clashes inevitably leads to
“apparent chords” that do not participate in the harmonic syntax.
Incomplete sonorities are another important source of analytical
ambiguity.6 J. S. Bach’s “Jesu, du mein liebstes Leben” makes extensive use
of a one-measure contrapuntal fragment that occupies fully a quarter of the
piece (see mm. 2, 5, 6, 10, and 14 in Figure P7.2, with variants in mm. 3, 11,
and 13).7 Its third beat contains a bare third, consistent with a triad missing
either its root or fifth. What is interesting is that the piece seems to play with
these possibilities: in mm. 2, 6, and 14, it sounds like it is missing its fifth,
while in mm. 5 and 10, it sounds like it is missing its root.8 And “sounds”
really is accurate: to many listeners, not looking at the score, the B♭–D in m.
3, and the D–F in m. 5, really do sound, intuitively and pretheoretically, like
B ♭ major triads.9 It is almost as if the piece were trying to prove that the
same scale degrees can be given different meanings.10 One might say that
the ambiguity of its syntax is intrinsic to its semantics, the goal being to
create a miniature and multivalent musical kaleidoscope.11
Figure P7.2. “Jesu, du mein liebstes Leben” (BWV 356, Riemenschneider 243).

Sometimes we can experience these multiple interpretations

simultaneously. In Figure P7.3, Bach’s descending-fifth sequence arrives on
a (B ♭ , D) that, conceived as IV, continues the harmonic pattern. Yet this
measure is also the start of another descending-fifth sequence that
retrospectively suggests G minor. Here we face an uncomfortable choice
between analytical consistency and psychological accuracy: consistency
suggests that m. 11 be analyzed like the very similar m. 13, and that it
contain just a single chord, like most of the other measures in the piece. And
yet I have the overwhelming impression of arriving on a B ♭ that is

immediately reinterpreted as G minor. The B ♭ –D is what we might call a

pivot third, a chord that can be heard in two incompatible ways. My analysis
superimposes these hearings like the dual analyses of a pivot chord, even
though the passage does not modulate.12

Figure P7.3. Bach’s D minor two-part invention, BWV 775, mm. 7–14. I analyze m. 11 as ii and IV
simultaneously.

Key changes are a third source of ambiguity, as some listeners are

relatively quick to adopt a new tonal center while others retain the tonic even
through fairly long modulatory digressions. Common modulatory strategies
trade on this ambiguity in various ways. Standard pivot-chord modulations
present overlapping progressions that are each syntactical in different keys:
thus we experience the key change only after it has already happened (Figure
P7.4). (Even the comparatively unambiguous direct modulation can have a
similar effect, as listeners cannot know whether a foreign dominant seventh
is an applied chord or the start of a new key.) More unusual are what I call
garden-path modulations, where key areas are juxtaposed to create the
impression of an anomalous harmonic hiccup.13 Figure P7.5 presents a pair
of examples: the first produces a V–iii progression that sounds modal even
though everything before the D minor is syntactical in B ♭ major, while
everything after it is syntactical in F major.14 The passage from Grieg’s
Gade (Op. 57, no. 2) is more dramatic, creating an unmistakably mixolydian
cadence that is immediately and retrospectively demystified.15

Figure P7.4. Three modulatory strategies. In pivot-chord modulation, there is an overlap between
syntactical progressions in two different keys. In direct modulation, keys do not overlap, but the new
key begins with an unambiguous dominant. Garden-path modulations do not strongly signal the start
of the second key, creating a nonsyntactical progression across the key boundary.
Figure P7.5. Seemingly modal progressions occurring across key changes: Bach’s “In allen meinen
Taten” (BWV 13.6, Riemenschneider 103, mm. 3–4) and Grieg’s “Gade” (Op. 57, no. 2, mm. 99–
103).

Here Grieg seems to use functional harmony against itself, juxtaposing

key areas to create an effect of modality even without violating functionally
tonal laws. This is analogous to the way nineteenth-century composers used
nonharmonic tones to extend their harmonic vocabulary, writing music
where the supposedly “nonharmonic” notes create musically meaningful
sonorities. Such examples remind us that we need to consider syntax in the
context of aesthetics: when we discuss functional music, we often take it for
granted that composers wanted to create clear musical structures and law-
abiding harmonic progressions. But sometimes they wanted to write
dominant thirteenth chords or mixolydian subtonics. Where twentieth-
century composers were free to do exactly that, nineteenth-century
composers often maintained a degree of deniability, creating the impression
of nonfunctionality even while conforming to the prevailing system of
musical laws.
 These ambiguities greatly complicate the project of harmonic analysis.
Earlier I argued for an objective and nonpsychological labeling of scale
degrees; now I am talking about “how we hear” and making complex
intuitive judgments about nonharmonicity. It would seem that “analytical
objectivity” is strongly dependent on context: rather than mechanically
treating similar passages in exactly the same way, we have to balance local
details with a more general sense of what is to be expected in a genre, or in a
particular composer’s work. The challenge here is not so much subjectivity
as it is holism, the need to resolve ambiguities by drawing upon stylistic
knowledge external to the piece at hand.
What is needed, in other words, is the benign form of circular reasoning
that is sometimes called “hermeneutic.” For the interdependence of the
general and the specific can be unproblematic so long as there are enough
unambiguous passages to calibrate our expectations. In the case of the Bach
chorales, we can model this circularity with a two-stage process: we begin
by treating every consonance as a harmony, performing a raw analysis that
simulates the background expectations formed by a lifetime of learning. This
first-pass analysis is then used in a second pass that adjudicates among
allowable readings. Figure P7.6 shows how this works in two passages
discussed earlier. I began by programming a computer to identify keys on
the basis of scalar content, analyzing A harmonic minor passages in A minor,
C diatonic passages either in C major or A minor, and so on. I then counted
up all the chords and progressions to be found, treating every consonance as
a harmony. These raw counts served as input to a second stage that tried to
assign the most probable analysis to every progression (subject to the
requirement that all nonharmonic tones obey standard rules).16 This
approach reproduces all the readings I proposed in both Figure P7.1 and
P7.2, holistic computation capturing the subtle asymmetries of human
judgment.17
Figure P7.6. Counts of various progressions in an automatic analysis of J. S. Bach’s chorales, treating
every consonance as harmonic and using scale membership to identify keys. The top table shows that
IV→I6 is more common than either of the two-chord subprogressions of IV→ii6→I6; this is a reason
to declare the ii6 “merely passing.” By contrast, both two-chord subprogressions of ii→vii°6→I6 are
substantially more popular than ii→I6, as is the three-chord sequence itself. This is a reason to
consider the vii°6 harmonic.

Thus it may still be reasonable to aspire to a fairly objective labeling of

harmonies, subject to the following caveat: that the resolution of ambiguity
needs to be carried out on the basis of background knowledge. This caveat
greatly complicates the evaluation of harmonic theories: intuitively, one
might have thought that theory and analysis could be separated as shown on
the left of Figure P7.7, where a single theory-neutral analysis can be used to
evaluate many different potential syntaxes. Instead, what counts as the best
analysis may itself be theory-dependent, with the identification of
nonharmonic tones itself influenced by the harmonic syntax. This would
mean that we have to evaluate theory and analysis together as a single
package. What is surprising is that computational methods can capture this
holism, modeling the presuppositions essential to expert analytical judgment.
Intelligent Martians might not initially understand our music, but with
enough experience they could probably learn.

Figure P7.7. Two approaches to testing harmonic theories. On the left, different theories are evaluated
by a single objective analysis; on the right, different theories determine different optimal analyses,
with theory and analysis evaluated as a package.

Of course, the Martians are a metaphor: in reality, aliens would quite

likely lack the biological equipment to make sense of our music—even if
they heard air vibrations in the same range that we do, they would likely lack
the perception of octave equivalence, consonance and dissonance, our
enjoyment of rhythmic entrainment, and so on. Rather than being a universal
language, music is fundamentally tied to human biology and culture.18 Why
we put it on spaceships is beyond me.
 1
Compare Figure 3.1.7. Appendix 4 discusses Huron’s reading of this idiom.
2
Weber (1817–1821) 1846, p. 658, and Saslaw 1992.
3
Quinn’s skepticism lies behind the approach taken in Quinn 2018, which attempts to do without
the notion of nonharmonicity; he and I had a profitable and longstanding debate about these issues
over the course of many years, relatively little of which found its way into print. Temperley 2018 (p.
35, fn. 10) notes a similar circularity.
4
The holistic and skeptical philosopher W. V. O. Quine argued that theoretical terms were severely
underdetermined by empirical evidence: his (long-I) “Quinean skepticism” (1953) is in some respects
a generalized version of (short-I) Quinnian skepticism—which suggests that our harmonic and
contrapuntal beliefs are underdetermined by objective musical facts.
5
Weber (1817–1821) 1846, p. 677.
6
Weber (1817–1821) 1846, pp. 192–94.
7
Note the ascending subdominants in mm. 3, 11, and 13, the “slippery figure” (§5.1) in the tenor of
m. 8, here a decorated quarter-note anticipation, and the “Renaissance seventh” idiom in m. 15
(VImaj7–iv6).
8
The mediant is a common destination in minor and the F♮ and E♭ on beat 2 of m. 6 combine to
suggest V65/III; by contrast, the rarity of the major-mode mediant encourages us to hear a rootless
tonic in m. 10.
9
To test this, I asked two generous friends to label the chords in a recording of the chorale. Both
heard asymmetrically as described in the text. (As Weber would say, their ears really did choose the
simplest explanation!) See Goldenberg 2021 for more on incomplete chords.
10
Consider how mm. 2–3 are echoed in the relative major at mm. 5–6, so that m. 6 is functionally
analogous to m. 3, while also containing virtually the same notes as m. 2.
11
The same thing happens with larger sonorities as well: in mm. 42–44 of the F♯ minor fugue from

book 2 of The Well-Tempered Clavier, we hear a long series of descending fifths, E–a♯–D♯–g♯–c♯–

f♯–B–E–A, followed in m. 45 by the progression Dmaj7–b–c♯7–A, leading to another long sequence of

descending fifths, b–E–A–D–g♯–C♯–f♯. If we read the triads in m. 45 as rootless sevenths (Dmaj7–

g♯65–c♯7–f♯65), the entire passage becomes a descending-fifth progression.

12
Other pivot thirds can be found in m. 22 (A–C as F major and A diminished) and (arguably) m.
42 (F–A as both F major and D minor).
13
A “garden-path sentence” is one in which an apparent syntactical violation results from
misconstruing an ambiguous phrase: in “the old man the boat,” “old man” can be either adjective-noun
or noun-verb, with the former reading statistically more likely but producing a nonsyntactical
sentence.
14
A good number of Bach’s supposedly “modal” progressions occur across what can be
retrospectively understood as key boundaries. They present a challenge of segmentation, or the
placement of key boundaries; this is a tonal analogue to a familiar problem of atonal analysis (Hasty
1981).
15
Here the effect is reinforced by the thematic structure and phrase rhythm: the passage is a
rewriting of the opening four-measure theme, which uses a G♯ and ends with a standard V–I cadence.
16
When comparing progressions of different lengths, I compare the average frequency of their
two-chord subprogressions.
 17
Readers can consult the online materials for the other progressions. My approach is inspired by
Bayesian analysis, which emphasizes the importance of prior knowledge in interpreting new evidence
(Gelman et al. 2013). Ongoing work with the physicist Mark Newman uses hidden Markov models to
identify chords from raw notes and keys from chords; this almost completely assumption-free
approach manages to produce remarkably accurate analyses (cf. Nápoles López et al. 2019).
18
For musical universals see McDermott and Hauser 2005, Savage et al. 2015, and Jacoby et al.
2019. Alexander Rehding and Daniel Chua are insightful and entertaining in imagining potential alien
responses to the Voyager spacecraft’s “golden record” (2020).
 7
Functional Progressions

We now turn to the principles governing chord-to-chord progressions in

functional tonality, the tradition beginning roughly with Corelli, dominating
European music for almost two centuries, and surviving to this day as an
important set of harmonic options. That there are such rules is doubly a point
of contention: Schenker denied it, asserting that many apparently harmonic
events are in fact the byproducts of linear motion, while contemporary
practitioners of “microtheory”—enemies of abstraction and partisans of
idiom—have suggested that we can describe functionality using a small
collection of schemas. I will instead suggest that we can construct a simple,
approximately true, and heterogeneous grammar of chord-to-chord
possibilities, combining the three independent subsystems of harmonic
cycles, fauxbourdon, and sequence. Thus I argue that functional tonality
involves broadly grammar-like rules even while acknowledging that it is not
governed by a single principle.
These local harmonic constraints underwrite an aesthetic picture of
functional harmony as repetitive and nearly sequential, cycling through the
same small map of harmonic options and changing just enough so that the
ear stays intrigued. This mechanical quality is obvious in baroque music,
somewhat sublimated in the classical era, and returns, transformed, in the
music of Beethoven and Chopin. By the start of the twentieth century,
repetition and sequence are viewed with suspicion—encouraged, in part, by
a tendency to link “genius” with the organic. Perhaps as a result,
contemporary textbooks tend to deemphasize the repeating harmonic
structures at the core of this music—a repression that is aesthetically and
politically problematic, pushing high-art functionality away from more
manifestly cyclic styles, and drawing too sharp a line between the covert
repetition of the classical tradition and the unapologetic repetition of our
own time.
 1. A theory of harmonic cycles

Figure 7.1.1 shows the distribution of major-mode sonorities in J. S. Bach’s

chorales and Mozart’s piano sonatas. It is interesting that the submediant
appears almost exclusively in root position: the near disappearance of vi6,
combined with the absence of iii and iii6, ensures that bass notes 1̂ and 3̂ are
harmonized almost exclusively by tonic chords, 5̂ by either dominant or
tonic, and 7̂ by dominant (Figure 7.1.2). In other words, the norm is for the
bass to support tonic or dominant whenever it can, with 2̂, 4̂, and 6̂ providing
the bulk of the harmonic variation.1 The submediant can support the broadly
similar vi and IV6 chords, while the subdominant allows for both IV and ii6.
The supertonic is unique in permitting the functionally distinct sonorities ii
and vii°6 (or V43 in later styles). Bass strongly correlates with harmonic
function.

Figure 7.1.1. The frequency of diatonic chords in Bach and Mozart, combining triads and sevenths.

Figure 7.1.2. In functional tonality, bass degrees 1̂, 3̂, 5̂, and 7̂ typically support tonic and dominant,
while the remaining degrees support the other harmonies.

We have seen that the simplest forms of functionality are built around
elementary subdominant and dominant cycles: I–IV–I, I–V–I, and I–IV–V–I,
all articulated by root-position chords. As functional harmony develops, the
dominant cycle increases in complexity, coming to play a greater and greater
role, while the subdominant cycle remains static or even decreases in
prominence. Figure 7.1.3 provides a theory of harmonic cycles in baroque
and classical music, adapted from A Geometry of Music.2 Here I arrange the
diatonic triads in a chain of thirds descending from tonic to dominant: chords
are completely free to move rightward and vertically, but leftward motion
must proceed along one of the arrows. The size of each symbol corresponds
to its frequency. In a typical functional piece, more than 95% of the
progressions will conform to this model, particularly if we exclude sequence
and fauxbourdon; conversely, all the root progressions allowed by the model
are common.3 This two-way fit poses a challenge for those who would like
to replace Roman numerals with particular schemas, as one would need a
fine-tuned collection of idioms—no doubt varying with genre—to reproduce
the progressions on the map.

Figure 7.1.3. A model of functional harmony where rightward and vertical motion is unrestricted but
leftward motion follows the arrows.
 This model incorporates a variety of earlier approaches to functional
harmony. Inversionally equivalent chords are placed above one another,
meaning that they act similarly. Functional categories are encoded by
horizontal position, with neighboring sonorities (e.g., IV and ii, or vii° and
V) participating in similar progressions. The model also privileges certain
types of root progression: since the “strong” progressions (descending third,
descending fifth, and ascending step) move rightward by one, two, or three
steps, the model will permit more of these motions than their “weak”
counterparts (ascending third, ascending fifth, and descending step). Where
the simple protofunctionality of §6.1 was decidedly Riemannian, privileging
primary triads as locations in harmonic space, our more sophisticated model
is more Rameauian, highlighting a specific way of moving from tonic to
dominant.4
With just eleven chords, the map can be learned by example and without
explicit instruction. It is also consistent with figured-bass theory, as its
guiding principle is that bass lines move along an abstract chain of
descending thirds from 3̂ to 5̂, with 53 preceding 63 over the same bass, and
ascending thirds permissible only when they support 53→63 progressions
(e.g., from root position to first inversion, Figure 7.1.4). Interested readers
might try constructing alternative models that produce an approximately
root-functional output without making any overtly root-functional
assumptions: they will find, I suspect, that a thirds-based arrangement is
virtually unavoidable. And while the model includes a few progressions that
occur only in specific inversions, it can be approximately reformulated using
roots (Figure 7.1.5).
Figure 7.1.4. The descending-thirds arrangement as a figured-bass composer might conceive it, along
with a translation into contemporary chord symbols. The main numbers represent bass degrees; the
superscripts represent vertical configurations: 63 chords on 1̂ and 53 position chords on 7̂ are rare.

Figure 7.1.5. A root-functional version of the model, with some arrows requiring specific inversions.

I conceive of this map not as a set of universal laws but as first-order

defaults surrounded by a penumbra of second-order idioms and practices.
Some features of the penumbra remain fixed throughout the functional
tradition: for instance, mediants have a rare but conventional role as post-
tonic chords moving to vi or IV, as alternative dominants, and as substitutes
for the tonic (Figure 7.1.6).5 Similarly, V–viiø7 can support melodic 7̂–6̂–5̂
against bass 5̂–4̂–3̂, parallel tenths often decorated by octave displacement.6
In other cases we can trace the penumbra’s historical evolution, as harmonic
practice congeals over the course of the eighteenth century (leading to a
decrease of vi–V and nonsequential iii) and then expands throughout the
nineteenth (leading to an increasing use of ct°7 and iiø7 and the recovery of
iii). Figure 7.1.7 suggests that the functional tradition is like an hourglass,
with Haydn, Mozart, and Beethoven the point of maximal harmonic
regimentation.7

Figure 7.1.6. The functional flexibility of the mediant, which can act as a pre-predominant, a
dominant, and a tonic substitute.
Figure 7.1.7. Some salient major-mode differences between earlier and later functional dialects.

In some styles, harmonic vocabulary depends on mode: in rock, for

example, VI–VII–i is common in aeolian but rare in ionian, while I–V–IV–I
is common in ionian but rare in aeolian. Functional tonality exhibits an
unusual degree of modal symmetry, using virtually the same repertoire of
progressions in both major and minor. One distinctive feature of minor-mode
harmony is its greater attraction to the mediant—both as a modulatory
destination and as a tonicized chord within phrases. As a student, I was
puzzled by this asymmetry, wanting some principled justification for this
minor-mode behavior. I now think this desire was misplaced: rather than a
deviation from an established default, the minor mode never fully abandoned
its preexistent modal attraction toward the mediant, with III remaining the
primary harmonization of bass 3̂ until well into the seventeenth century
(§8.6). If this is right, then it shows the importance of understanding
functional harmony as a historically contingent development, rather than a
self-consistent logical system.
 I should emphasize that I am not suggesting that local harmonic
constraints are the only principles animating this music: on the contrary,
composers aim to produce compelling music while also satisfying the
harmonic grammar. The next chapters will consider a host of musical
principles not captured by simple chord-to-chord graphs: for instance, that
certain progressions will preferentially be found in specific parts of the
phrase, that melodies articulate stepwise connections between nonadjacent
notes, that phrases will sometimes be governed by the two-voice logic of the
imperfect system (§3.1), and so on. And of course, any collection of genre-
wide principles needs to be supplemented by composer-specific preferences.
Past a certain point, it is foolish to talk about functional harmony in general,
instead of a specific musician’s idiolect.

2. A more principled view

Figure 7.2.1 shows a simple but beautiful piece of musical mathematics, a

sequence of falling fifths perturbed so that it returns to its starting point after
seven notes, its final interval tempered to a diminished “near fifth” almost
but not quite the same as the others. Reinterpreted as a sequence of four-step
diatonic intervals, these falling fifths can in turn be halved to form a cycle of
descending thirds. Because the diatonic scale has seven notes, the bifurcation
can be repeated again and again, the thirds splitting into steps which can
themselves be factored (more abstractly) into descending fifths and
ascending fourths—a circular nesting of interval cycles each contained
within its predecessor, a musical ouroboros.8 Figure 7.2.2 shows a passage
containing four levels of the pattern, steps inside thirds inside fifths inside
steps.
Figure 7.2.1. Nested interval cycles, with each staff dividing the staff above it into two equal
intervals.

Figure 7.2.2. Nested interval cycles in The Well-Tempered Clavier’s first A major prelude.

Two aspects of this structure are particularly relevant. First, its layers can
be rearranged to produce the basic voice leadings for the primary sonorities
of functional harmony (Figure 7.2.3). Indeed, each layer of the ouroboros
can be extended to form a “maximally even” diatonic chord, with
overlapping collections linked by their basic voice leading.9 Second, the
tertian layer can be used to reformulate the third-based harmonic grammar:
rather than taking triads to be indivisible musical atoms, we can describe
music as moving along the descending-third chain of pitch classes as shown
in Figure 7.2.4. In other words, harmonic cycles often generate complete
chains of diatonic pitch classes, descending by third through two abstract
octaves.10 This descending chain of thirds will often continue across
harmonic cycles to produce an ongoing pitch-class sequence.

Figure 7.2.3. Basic voice leadings for the fifth, third, triad, and seventh chord, derived by rearranging
contiguous segments of the interval cycles on Figure 7.2.1.
Figure 7.2.4. The harmonic grammar expressed with pitch classes. Again, rightward moves are
unrestricted while leftward motion follows the arrows.

This in turn suggests a path from simpler protofunctionality to the more

sophisticated system of the eighteenth century. In chapter 3, we noted that
third-related triads and seventh chords are harmonically close insofar as they
share all but one of their notes, with the last moving by step. Functional
tonality exploited this relationship to expand the protofunctional I–IV–V
into the richer I–vi–IV–ii–vii°–V, the additional chords smoothing the
transitions between functional poles. (The unsmoothed progressions V–I and
IV–I remain primary signifiers of functionality.) The result is a system in
which nearly equal divisions of the octave play two separate roles: simple
root-position exchanges like I–IV–I and I–V–I divide the registral octave
nearly evenly, bass moving by approximately half octave while melody
moves by approximately one sixth of an octave (§6.1). Meanwhile,
paradigmatic four-chord progressions like I–IV–V–I and I–ii–V–I divide the
circle of diatonic triads into three nearly even segments, creating a
functional sequence of tonic, subdominant, and dominant, whose pitch
classes articulate a chain of descending thirds (§3.4).
These nested interval cycles allow composers to titrate the degree of
sequential repetition in their music. Some functionality is not at all
sequential on the surface, its regularity evident only at the abstract level of
pitch-class content; in simple I–IV–V–I protofunctionality, even this level of
sequential structure is fairly obscure. In other passages, however, we find
overtly sequential structure: a single voice articulating falling thirds, two
voices chaining the diatonic third’s basic voice leading, and a variety of
regular and quasi-regular three- and four-voice sequences. Figure 7.2.5
subposes a variety of bass notes underneath the diatonic third’s basic voice
leading; this is the schema that was discussed in §4.10. In effect, the staves
parse the chain of descending thirds differently: the two upper-staff voices
take turns playing successive notes from the descending-third chain,
combining to form a perfectly regular descending-fifth dyadic sequence; the
lower staff parses the third-chain as a sequence of triads whose largely
descending-fifth root progressions are perturbed by the occasional ascending
step or descending third. Here we start to understand how functional music
can move seamlessly between monodic, dyadic, and triadic logics.

Figure 7.2.5. Upper-voice regularity harmonized to form slightly irregular triads. The two staves
present the same sequence of descending thirds in slightly different ways.

These ideas can be used to interpret “Oh, Pretty Woman,” Roy Orbison’s
1964 ode to sexual objectification (Figure 7.2.6). The melody descends
through a single octave, moving downward by three thirds and a step;
underneath it, the harmony moves by third through two abstract octaves,
initially in parallel with the melody, but then inserting an E7 between D
major and A major. These descending thirds extend from the end of the first
statement through the beginning of the second, continuing the sequence
across the verse’s repetition. The bridge’s “down by third, up by step”
melody can be derived by omitting notes from the descending-third sequence
(e.g., F D b c E C a f, etc.); here it is supported by a pair of nonsequential
harmonic cycles—articulated in the electric guitar by a series of ascending
triadic voice leadings.11 Once again the pattern continues through the repeat,
the melody skipping only one note on the circle of thirds (or, if we view it as
an elided two-voice pattern, switching voices from bottom to top). Orbison’s
short song thus presents all three levels of the ouroboros: descending steps
on the melodic surface, embellishing descending thirds in the abstract
progression of pitch classes, mostly grouped into descending-fifth triadic
progressions, and supporting a descending-step melodic sequence in the
bridge.12 The beauty of this music is its multivalence: it is not based on fifths
or thirds or steps but rather on a delicate balance of all three, requiring only
the slightest compositional nudge to bring out one or another of these
aspects.

Figure 7.2.6. Roy Orbison’s “Oh, Pretty Woman.”

Such music is repetitive in two different ways. First, and most obviously,
it passes repeatedly through the same abstract set of harmonic possibilities.
Second, and more subtly, its phrases can be more or less sequential
depending on how a composer chooses to parse the abstract chain of
descending thirds. This sequential structure is latent in the grammar,
available for composers to exploit if they choose—rather than something put
into a particular piece, it can be drawn out of its generic fabric (cf. §4.3). At
some basic level, mature functionality simply is the descending sequence of
thirds—sometimes manifest on the musical surface and sometimes locked
away in the musical subconscious (Figure 7.2.7).13 The magic of this style
lies in the interaction between overt and covert, symmetry and asymmetry,
sequential melodic descent filtered through a nonsequential grammar.

Figure 7.2.7. One of the second themes of the first movement of Beethoven’s Op. 22, twice
descending by thirds from F to F.

All of which might start to suggest that functional tonality is a natural

language—deriving from diatonic geometry and resting ultimately on the
acoustic pillar of the perfect fifth. This is not exactly wrong: functional
tonality is a manifestation of one set of affordances given by the preexisting
geometry of musical possibility. But there are two important caveats. First,
the chain of reasoning we have been following is not one of necessity, but
rather possibility: having tempered a series of falling fifths, we are not
required to subdivide the fifths into thirds, nor to construct progressions by
grouping this sequence into triads. Nor need we employ fifth-based
modulations between nearby scales, nor limit ourselves to just the major and
minor modes, nor take any of the thousand other steps that jointly produced
functional harmony. Rather than grounding style directly in nature or
mathematics, the best we can hope for is to describe musical evolution as a
series of small but conventional leaps, some seizing upon possibilities that
may be suggested (but not directly required) by cultural, biological, or
mathematical phenomena.
Second, triadic functionality is not the only language that can make a
claim to naturalness. We have seen that there is something equally natural
about progressions formed from the ascending sequence of diatonic thirds—
which can be rearranged to form the stepwise descent central to so much
melodic behavior. These “retrofunctional” progressions—the very opposite
of those found in the functional tradition—are equally natural, and equally
suited to human music-making.14 Here we have two opposed musical
regimes, a tonally functional system exploiting descending-third harmony
and requiring special melodic treatment, and a modal regime exploiting
ascending-third harmony and giving rise to descending melodies. The
question is not whether one is “more natural” than the other, but how
musicians in different styles make thoughtful use of the possibilities
available to them.

3. Rameau and Bach

It is in this context that we can start to appreciate the virtues of Rameau’s

theory of harmony. Rameau proposed that functional music depends on the
repeating contrapuntal pattern shown in Figure 7.3.1, where root and third
stay fixed while fifth and seventh descend by step. This sequence can be
short-circuited by the double emploi, in which a single triad like F–A–C
stands for F major and d minor simultaneously. This leads to a blurring of
harmonies separated by diatonic third.15 Rameau was not always clear about
the circumstances under which this blurring was permitted, and his views
evolved over time, but he explicitly favored progressions by descending
third and ascending step, both of which can be generated by applying third
substitution to a descending-fifth progression.16 In the spirit of rational
reconstruction, let us stipulate that well-behaved Rameauian progressions
start with the tonic and move forward on the sequence of fifths to vii° or V,
either by descending fifth or using the double emploi. (See Figure 7.3.2,
where the stipulation is that there be no left-pointing arrows, so that the
double emploi shortens rather than lengthens the descending-fifth journey
from tonic to dominant.) What results are seven progressions using
descending fifths, descending thirds, and ascending steps: the full circle of
fifths, three four-chord progressions consisting of one descending third and
three descending fifths, and three progressions consisting of one ascending
step and two descending fifths. This reconstructed Rameauianism is both a
reasonable approximation to functional practice and extremely close to the
theory I have been expounding.
Figure 7.3.1. Rameau’s descending-fifth paradigm on the top staff with third substitution underneath.
Replacing the F major seventh with the D minor seventh produces an ascending-step progression;
replacing it with the A minor seventh produces a descending-third progression. Each substitution
preserves three common tones, shown here by open noteheads, with the fourth note moving by step.

Figure 7.3.2. A model of functional progressions inspired by Rameau. Chords can use third
substitution to move forward along the arrows to either bracketed chord, inserting either a descending
third or ascending step into the sequence of descending fifths (e.g., C–a–d–G or C–d–G). This graph
allows for seven progressions, all but one of which is common in functional music; all but one make a
single rightward pass through Figure 7.1.3.

This is because every progression allowed by the neo-Rameauian model

articulates a consecutive segment of the descending-third chain pitch classes
(Figure 7.3.3). So long as we are willing to follow Rameau in accepting
triads as four-note sonorities lacking either a sixth or a seventh, we can
analyze every rightward progression allowed by the descending-third model
in terms of Rameau’s paradigmatic progression. In other words: a theory
based on the descending-fifth sequence of seventh chords, whose roots and
sevenths are sometimes omitted, is necessarily going to resemble a theory
based on the descending-third sequence of triads, which can sometimes be
extended through the addition of sevenths. Figure 7.3.4 goes through the
derivation for one specific progression; a similar analysis could be given for
virtually every other functional progression, including most of the sequences
in §7.6.17 If there is a weakness of Rameau’s theory, it is its inability to deal
with simple interchanges like I–V–I and I–IV–I, treating them as elisions of
more complex progressions. My approach avoids this problem by tracing
Rameau’s seventh-chord paradigm to an even more basic object, the chain of
descending diatonic thirds.

Figure 7.3.3. Each of Rameau’s three basic progressions articulates a segment of the descending-third
chain of pitch classes.
Figure 7.3.4. Derivation of the I–vi–IV–ii–vii°–V–I progression from Rameau’s descending-fifth
paradigm. The boxed progressions can be interpreted as elided versions of the seventh-chord
progression on the bottom staff. Black noteheads on the bottom staff are missing from the associated
triads.

Rameau’s descending-fifth template runs through The Well-Tempered

Clavier’s first D major prelude like a murmuring cantus firmus, a musical
“core” that is extended to the three-measure fragment shown in Figure 7.3.5.
Bach employs two distortions of the melody’s tail, transposing it either
upward or downward by third; this introduces a descending third, locally I–
vi, into the otherwise-omnipresent fifths.18 While chords descend by fifth,
the keys move in the opposite direction, ascending by fifth from D major to
A major and then further up the ladder to E minor, B minor, and F♯ minor.
This pattern of distorted fifth-progressions is well-suited to a prelude, and
readers may enjoy improvising with it—or perhaps programming it into a
computer.19
Figure 7.3.5. (top) The basic sequential melody of the first D major prelude in The Well-Tempered
Clavier, with its implied chordal voice leading (bottom).

The result is vaguely sequential and yet never entirely predictable, an

unusually clear look at the regularity at the heart of tonal-functional syntax.
Indeed, Bach assembles his quasi-sequential phrases into a higher-level near-
sequence that would be completely regular were the first statement on E
(Figures 7.3.6–7); the perpetually descending stepwise bass flows naturally
from the sequence, loosely shadowing one of the upper voices, which
themselves articulate a baroque analogue to the “Shepard-tone-passacaglias”
of §2.5 (Figure 7.3.8). Here we see three levels of semi-regular harmonic
movement: chords descending mostly by fifth, keys largely modulating
upward by fifth, and larger phrases descending mostly by fifth again.

Figure 7.3.6. The fundamental harmonic progression of the prelude, with the descending scalar bass.
The bass line here derives from the A major statement.
Figure 7.3.7. The D major prelude makes four passes through this flowchart.
Figure 7.3.8. The prelude’s constantly descending bass line, with each beat representing a measure.

Though Bach was hardly a loyal Rameauian, the choices available in his
open-form prelude closely track the basic concepts of Rameau’s theory.
There are no ascending thirds or fifths in Figure 7.3.7, but rather a
preponderance of descending fifths (65%) and descending thirds (19%), with
a smaller number of ascending and descending steps (10% and 5%
respectively, the latter mostly moving from tonic to leading-tone seventh).20
For me, the root motions are salient: it feels as if the dancer is moving in a
particular way rather than visiting a collection of familiar places. In other
words, it exhibits a Rameauian functionality based on types of root
progression rather than a Riemannian functionality based on locations such
as “subdominant.” Yet the churning motion-oriented music also conforms to
the precepts of the thirds-based grammar, presenting no anomalous moves
and touching on a large fraction of the permitted progressions.
The near-equivalence between Rameau’s theory and my own is an
example of theoretical underdetermination, our ability to construct different
theories to account for one and the same set of observable facts. In physics
we have the Lagrangian, Hamiltonian, and vector formulations of Newtonian
mechanics. In music we have Rameau’s descending-fifth theory, my own
descending-thirds theory, and many others besides: theories based on idioms,
functional categories, linear relationships, and so on. The interesting thought
is that none of these theories is right or wrong; instead, they offer different
perspectives on a musical practice that was largely understood tacitly. In
choosing among them we must ask which gives the simplest and most
compelling explanation of the broadest range of musical behavior. Different
theorists will make different choices, reflecting preferences that are as much
aesthetic as scientific.
 4. Functional melody, functional harmony

Seventh chords are intrinsic to Rameau’s theory, not just in providing a tool
for analyzing descending-third progressions, but also in coordinating
descending-step melody with functionally tonal harmony. In reality,
however, seventh chords are not nearly so prevalent as Rameau’s theory
suggests. How then do real-life functional composers manage the tension
between triadic functionality and descending melody, given the tendency of
“strong” progressions to generate ascending melodic steps? That is, how do
functionally triadic composers negotiate the fact that efficient triadic voice
leading often leads to ascending melodies?
Here I think we can give five separate answers, each of which helps us
understand something different about the style. The first is that functional
composers use a wide range of root progressions, including “weak”
progressions that give rise to descending triadic voice leading. While it is
true that ascending thirds are generally rare, descending steps are reasonably
common, and ascending fifths ubiquitous: progressions like IV–I and I–V
are important in virtually all functional genres.21 Functional tonality does not
so much avoid ascending-fifth progressions per se as ascending fifths not
involving the tonic triad.
The second answer is that functional harmony involves a norm of larger
voice leading, expressed by larger motions on the triadic circle; in particular,
simple harmonic cycles like I–V–I and I–IV–I often make a complete circle
through the space (Figure 7.4.1). These larger voice leadings permit the
addition of passing tones that create brief seventh chords, turning
contrapuntal lemons into harmonic lemonade. Thus, insofar as functional
composers have to negotiate a conflict between harmonic and melodic
forces, this is largely a matter of avoiding obvious contrapuntal possibilities
—ascending-third progressions and ascending fifths not involving the tonic.
Practically speaking, the remaining options are more than sufficient to
produce effective music.
Figure 7.4.1. Functional tonality often travels the long way around the triadic circle.

Third, functional composers often write melodies that embed or disguise

the circle of descending thirds (e.g., Figures 7.2.2, 7.2.6–7, 9.1.1, 10.2.13).
These third-based melodies allow a harmonizing voice either to reorder the
thirds (producing the “Ludwig” schema) or move in parallel with them,
either a third above or a third below (Figure 7.4.2). Figure 7.4.3 shows Bach
using both techniques in the opening of The Well-Tempered Clavier’s first A♭
major fugue. 22
Figure 7.4.2. Harmonizing a descending-third melody with either contrary or parallel motion.

Figure 7.4.3. Measure 5 of the first A♭ major fugue in The Well-Tempered Clavier.

A fourth answer is that functional tonality coordinates melody and

harmony by expanding its harmonic vocabulary. The most common
technique is to use seventh chords to fill in the descending thirds that often
arise from larger motions on the triadic circle. Imagine, for example, that
you are a modal composer who has just used a V chord to harmonize
melodic 5̂; your melody is 5̂–4̂–3̂ and you are aiming to place a I chord
under 3̂. What chord goes under 4̂? The available triadic options are IV6,
♭ VII, vii°6, and ii (Figure 7.4.4). Functional music fuses the two chords of
the V–vii° progression into a dominant-seventh composite (Figures 7.4.5–6).
Virtually the same reasoning applies when the melody is 2̂–1̂–7̂ and the first
note supports a supertonic triad; here standard modal solutions include ii–
IV–V and ii–I6–V while common functional solutions are ii–ii7–V and ii–
I64–V: again, ii–ii7–V fuses ii and IV while I64 embellishes the dominant
(§9.5).23 Yet a third instance of this phenomenon has 8̂–7̂–6̂ supporting I–?–
IV; here the functional tradition retains both the modal solution I–iii–IV
(sometimes as I–III–IV) and the fifth progression I–Imaj7–IV.

Figure 7.4.4. Modal possibilities for melodic 5̂–4̂–3̂ over harmonic V–?–I.
Figure 7.4.5. Modal and functionally tonal approaches to similar musical problems.
Figure 7.4.6. Harmonizing descending melodic steps in the Bach chorales and Mozart piano sonatas.
The notation I(64) indicates that the tonic chord could be in any inversion including second, whereas
I64 requires second inversion.
 This suggests that functional-harmonic restrictions likely drove the
expansion from triads to seventh chords, partly as a way of recovering lost
melodic possibilities: scale-degree progressions such as 2̂–1̂–7̂, which had
indigenous modal harmonizations, gradually acquired functional alternatives
involving sevenths and I64, alternatives that allowed descending-fifth
progressions to harmonize a greater range of melodic possibilities.24 If this is
right, then it seems unlikely that we can derive functional rules from voice-
leading considerations, as some theorists have wanted.25 Instead something
like the opposite is true: at the most general level, melodic goals generate
nonfunctional triadic progressions; to counteract this tendency, composers
need to replace those triadic progressions with more complex alternatives
involving seventh chords and the cadential 64.
A final answer is that surface voices sometimes descend relative to an
ascending background. Leonard Cohen’s “Hallelujah” uses this strategy
alongside the others discussed in this section. The reduction in Figure 7.4.7
is largely determined by the melody, whose primary notes are shown with
open noteheads; these ascend by sixth from G4 before falling back to C5.26 I
analyze the final phrase as descending through the triadic background,
moving soprano–alto–tenor before cadencing on a C4 that is the same
abstract voice as the C5 that ends the second phrase, only now in a lower
octave.27 The strategy here is essentially the inverse of Neil Young’s
“Helpless”: there we found omnipresent descending voice leading giving
rise to modal harmony, with the vocal melody ascending by triadic step
against a descending triadic background; here we have omnipresent melodic
ascent, with the vocal melody descending through the voices from one verse
to the next. The two songs, with their two different harmonic languages, are
in effect mirror images, with “Helpless” descending while “Hallelujah” rises
toward the heavens.
Figure 7.4.7. An outline of Leonard Cohen’s “Hallelujah.” The numbers above the example show the
steps each progression takes along the circle of diatonic triads (Figure 7.4.1).

5. Fauxbourdon and linear idioms

Fauxbourdon begins in the fourteenth century and persists throughout the

functional era—parallel triadic motion beholden to no further harmonic
laws. As such it is Exhibit A for the thesis of continuity between modal and
functional practice, and a clear example of the multiple syntactical systems
comprising functional harmony.28 In the simplest cases it is decorative,
gliding from one syntactic harmony to the next as in Figure 7.5.1. Here we
can imagine removing the parallel chords to reveal a functionally
permissible residue. (In much the same way, popular musicians sometimes
use stepwise motion to embellish familiar progressions, as when IV–I
becomes IV–iii–ii–I.) In other cases, fauxbourdon serves a textural function,
creating a palette-cleansing wash long enough to interrupt the connection
between first chord and last.
Figure 7.5.1. Fanny Mendelssohn Hensel’s, Sechs Lieder, Op. 1, no. 2, mm. 24–27, contains
“decorative” fauxbourdon audibly connecting IV6 to V6.

One of fauxbourdon’s main roles is as the source for an idiom that

harmonizes descending bass lines, a schema that is perhaps the most
important exception to the general principles governing harmonic cycles.
The joyous opening of Bach’s Fifth Brandenburg Concerto uses the idiom to
move from the high D4 tonic down to the low D3; the bass line uses parallel
first-inversion chords with the exception of the root-position chord above 4̂
(Figure 7.5.2).29 Early theorists had devised “the rule of the octave” to
govern such situations, calling for an applied dominant on 6̂ and dominants
on 5̂ and 4̂ (Figure 7.5.3). But many functional composers preferred an
alternative that uses fauxbourdon to move through a weaker chord on 5̂ to a
metrically strong arrival on 4̂, which can then continue downward to I via
Gjerdingen’s “Prinner.”30
Figure 7.5.2. The opening of Bach’s Fifth Brandenburg Concerto.

Figure 7.5.3. Two approaches to harmonizing a descending scale: the traditional rule of the octave
(top) creates harmonic accents on metrically weak beats, while the fauxbourdon-based alternative
(bottom) does not.

This “fauxbourdon rule of the octave” (or “fauxbourdon ROTO”) admits

several variants: substituting IV for ii6, V2/IV for V6, and, most importantly,
I64 for iii6 (Figure 7.5.4). This last yields the I–V–IV–I chord progression of
Neil Young’s “Helpless,” recreating a ubiquitous modal progression within a
functional context—an idiom that is statistically common even while being
theoretically puzzling.31 And while classical composers often skirt past the
V–IV, grouping chords in pairs so that V6–IV6 occurs across a phrase
boundary, they occasionally milk it for its retrofunctional qualities—as in
Brahms’s G major violin sonata Op. 78 (Figure 7.5.5), whose melodic-
harmonic core would not sound out of place on a classic-rock playlist.32 The
resonance is real, with similar theoretical constraints giving rise to similar
solutions.33

Figure 7.5.4. The “fauxbourdon ROTO” and some common variants.

Figure 7.5.5. In the second theme of the first movement of Brahms’s G major violin sonata, Brahms
uses the fauxbourdon ROTO to create a strong sense of retrofunctionality.

The IV6–I64–ii6 idiom is sometimes reversed in the course of what might

be called “standing on the predominant”; usually, predominant chords on 6̂
and 4̂ are connected by I64, delaying the arrival of the cadential dominant
(Figure 7.5.6).34 In such contexts, theorists often invoke the doubly
problematic notion of a “passing 64 chord.” The first issue here is that we do
not see a widespread proliferation of “passing 64 chords” of which IV6–I64–
ii6 is just one example; instead, the vast majority of “passing 64s” are
instances of the fauxbourdon ROTO schema (Figure 7.5.7). This suggests we
are dealing with a specific idiom rather than a more general musical
technique. A second problem is that many purportedly “passing 64”
progressions are perfectly well formed from a root-functional perspective:
progressions such as I–V64–I6, IV6–I64–ii6, and V2–vi64–V43, like the
“neighboring” I–IV64–I and V–I64–V, are largely in accordance with Figure
7.1.3, and merely occur over an unusual bass. (This is particularly clear in
the case of I–V64–I6, where the dominant simply lacks its seventh.) It is not
obvious that these progressions have any special “passing” quality, as
opposed to simply evincing a greater tolerance for second-inversion triads.35

Figure 7.5.6. Standing on the predominant in the Symphony Op. 11, no. 2, I, mm. 108–115, by Joseph
Bologne, Chevalier de Saint-Georges.
Figure 7.5.7. “Passing 64” chords in Bach, Mozart, and Haydn. More than 90% involve tonic or
dominant.

Fauxbourdon also plays a crucial role in the development of sequences. In

Corelli’s Op. 1, a substantial proportion of the sequential passages use
fauxbourdon—usually first-inversion triads descending by step, but
occasionally root-position or ascending triads.36 The descending-step
passages are usually decorated with quasi-harmonic 7–6 suspensions that
give the impression of a chord change.37 In figured-bass notation, the
suspensions are represented using two figures, and perhaps for this reason,
the dissonant sevenths sometimes take on a harmonic character; when
accompanied by a perfect fifth, they become very difficult to analyze (Figure
7.5.8). Indeed, this idiom is a direct descendant of the Renaissance seventh
(Figure 7.5.9): read literally, roots alternately ascend by step and descend by
third, but, as with its Renaissance predecessor, the progression feels
suspended between the harmonic and nonharmonic realms. Even more
harmonic is the bottom staff of Figure 7.5.8, where the stepwise 7–6 descent
is embedded in a sequence of seventh chords descending by fifth. One of the
analytical challenges posed by baroque harmony is the seeming
interchangeability of these options, with sequences spanning the continuum
from “merely linear” fauxbourdon to “robustly harmonic” fifths, presenting
a full range of intermediate cases.38 And though these intermediate cases
largely vanish in the classical period, they occasionally resurface: in his Op.
17, no. 4 Mazurka (shown in Figure 9.3.6), Chopin updates the effect for the
chromatic era, creating a sequence in which all the 7–6 sevenths sound
harmonic, and half of them (G7 and E7) clearly are.39
Figure 7.5.8. A fauxbourdon sequence with 7–6 suspensions, along with two variants.

Figure 7.5.9. The 75–63 fauxbourdon sequence in mm. 34–37 of Monteverdi’s madrigal “Longe da te,
cor mio” (Book IV).

One final idiom shares fauxbourdon’s emphasis on parallel motion, two

voices moving in thirds against a fixed pedal point; its tonic and dominant
forms are shown in Figure 7.5.10, and an example appears in Figure 7.7.6.
The idiom is again delicately balanced between harmonic and nonharmonic,
mostly articulating well-behaved harmonies, but with the parallel thirds
suggesting “merely linear” motion. The D–F–C sonority is particularly
interesting since it seems to contravene the general prohibition on I–ii–I6
progressions.40 It feels uncomfortable to dismiss the chord as “merely
nonharmonic,” since it occupies the same rhythmic duration as the other
harmonies; the phrase does not sound like a I–I6–IV progression in which
the first chord lasts twice as long as the others. For this reason I call the
apparent “ii7” a pseudochord, neither harmonic nor nonharmonic but
suspended between those worlds.41

Figure 7.5.10. The “pseudochord” idiom in its tonic and dominant forms.

6. Sequences

Chapter 4 contains a general theory of repetition applicable to both modal

and tonal composition, and extensible to contemporary music as well.
Functional composers did not use all these options with equal frequency.
Figure 7.6.1 counts the sequences in Corelli’s Op. 1, Bach’s Well-Tempered
Clavier, and the Mozart piano sonatas, classified by the interval of
transposition from one grand period to the next.42 We see that sequences
typically move by small intervals and that descent is preferred to ascent; this
is the familiar preference for stepwise descending melody, but now
governing larger spans of musical time. From Corelli to Mozart there is a
substantial increase in descending-third motion, rising to almost a quarter of
the latter composer’s sequences. Fourth motion and ascending-third motion
are rare for all composers, generalizations that continue to hold throughout
the tradition. This is likely because of the large registral motions involved;
most so-called fifth sequences are actually step sequences—often using
permutation in the upper voices, with the bass alternating ascending fourths
and descending fifths.

Figure 7.6.1. Sequences by interval of transposition in Corelli, Bach, and Mozart. Numbers are
percentages in the work of each composer.

Sequences typically move through a diatonic space that may itself be

modulating. For a given diatonic interval of transposition, there are four
possible two-chord sequences.43 The four descending-step options are shown
in Figure 7.6.2, while Figure 7.6.3 counts them in our three corpora: in
Corelli, fauxbourdon is the most popular, supplanted by descending fifths in
the later styles. The down-a-third-up-by-step sequence is present in Bach but
largely disappears later; the last possibility, the “descending 6–5,” appears
throughout the Renaissance but is mostly absent from functional music.
Turning to ascending steps, the four possibilities are shown in Figure 7.6.4.
Fauxbourdon and the ascending 5–6 are by far the most common; the second
possibility, D2A3, contains two weak progressions and is all but unknown.
Ascending fifths are particularly characteristic of Bach—indeed, many
striking passages in The Well-Tempered Clavier are ascending-fifth
sequences (Figure 7.6.5). In other styles this sequence is rare but not
unheard-of, a second- or perhaps third-level default, usually lacking Bach’s
omnipresent tonicizations.
Figure 7.6.2. The four diatonic descending-step sequences with two or fewer chords per unit. Here I
neglect order and treat octave-related intervals as equivalent (e.g., ascending sixth and descending
third). Note that D3A2 and D4A3 both juxtapose 63 and 53 chords over a stepwise descending bass,
but with the vertical shapes in different orders.

Figure 7.6.3. Distribution of descending-step sequences in Corelli, Bach, and Mozart.

Figure 7.6.4. Distribution of ascending-step sequences in Corelli, Bach, and Mozart. Note the
unusually high percentage of D4A5 in Bach.
Figure 7.6.5. An ascending-step sequence in The Well-Tempered Clavier’s first C minor fugue, mm.
13–14; the two-chord unit is itself a near-contrary sequence moving its voices apart by octave.

Descending-third sequences use all four of the two-chord possibilities

shown in Figure 7.6.6: fauxbourdon creates descending-third sequences
when the melodic period is two chords long and pure descending-third
sequences can be understood as dividing the descending fifth in half (e.g.,
Figure 4.6.5). The Pachelbel pattern D4A2 comes in two varieties, one in
which sequential units act like I–V resolving deceptively, the other
containing vii°–I pairs; occasionally, composers combine the two (Figure
7.6.7). Finally, there is the “ascending third, descending fifth” (A3D5)
sequence, C–E–a–C–F . . ., which often uses seventh chords to create
stepwise descending voices (Figure 7.6.8).44 The popularity of all four
possibilities reflects the descending third’s importance: virtually any two-
chord pattern sounds functional when it repeats down by third. Ascending-
third sequences are considerably rarer, almost always using the D2A4 root
progression (Figure 7.6.9).45 This is the retrograde of the standard Pachelbel
progression, and hence supports thirds ascending by step (Figure 7.6.10).
True fourth- and fifth-sequences are rare, though we will find that A2D5
ascending-fifth sequences are characteristic of developments (§8.5).

Figure 7.6.6. Descending-third sequences in Corelli, Bach, and Mozart.

Figure 7.6.7. A mixed descending-third sequence from the first movement of Haydn’s Keyboard
Sonata, Hob. XVI/49, mm. 50–52. The ascending step is initially a deceptive V–vi progression, then a
dominant-tonic vii°–I; the third unit combines these possibilities, the deceptive bass G–A♭ appearing
one sequential unit too early, subposed underneath the authentic b°–c progression.

Figure 7.6.8. The use of seventh chords allows for a greater variety of stepwise descending thirds.
Figure 7.6.9. Ascending-third sequences in theory and practice.

Figure 7.6.10. A disguised ascending-third sequence at the opening of Pergolesi’s Stabat Mater, P. 77.
The stepwise upper voices are decorated by voice exchanges and octave displacement (E6→E5). The
sequence calls for ii° instead of iv on the second beat of m. 3.

This survey reveals an important higher-level harmonic regularity, with all

but a few common functional sequences featuring descending-fifth
progressions repeating at some regular interval. The main exceptions are
pure descending thirds, whose connection to harmonic cycles is clear, and
the Pachelbel sequence, where the engine of repetition is deceptive rather
than authentic.46 This favoring of descending fifths produces notable
asymmetries between ascent and decent. For example, the ascending 5–6 (or
D3A4) is common, while its retrograde, the descending 6–5 or (A3D4), is all
but unknown—even though both are equally useful from a linear point of
view, and even though the descending 6–5 was popular in the Renaissance.
The rarity of these contrapuntally plausible options means that sequences
cannot be merely linear, but rather have harmonic content.47
Many theorists have argued that sequential progressions have an
attenuated functional significance. Less common is the Rameauian
suggestion I have been making here, that nonsequential harmony is itself
quasi-sequential, an oblique and disguised reworking of the circle of
descending thirds. It is, to be sure, distinct from the sequential system—a
branching set of possibilities that allows for a number of different paths
through the same basic map, and one that is usually expressed only
harmonically. But these differences should not lead us to overlook the
system’s fundamental regularity: sequence and harmonic cycle are poles in a
continuum, rather than separate countries demarcated by a clear border.
This continuity is nicely illustrated by the second half of “Ich grolle
nicht,” the seventh song from Robert Schumann’s Dichterliebe, Op. 48
(Figure 7.6.11). We begin with a harmonic cycle that highlights its quasi-
sequential structure: the bass descending almost by thirds from tonic to
dominant, C–A–F–D–[B]–G, and articulating a I–IV–ii–V–I progression:
descending by fifth from tonic to predominant (I–IV), descending by fifth
from predominant to tonic (ii–V–I). The descending-third bass continues
into the following sequence, an A3D5 pattern whose units combine in pairs
to form a contrary-motion fifth-sequence. Figure 7.6.12 connects the figure
to Gesualdo’s trick, two voices moving in parallel with the other two
forming thirds alternately above and below this dyadic core; disregarding the
octave displacement, we can think of it as a single repeating transformation
operating at two different timescales.
Figure 7.6.11. An outline of the second half of Schumann’s “Ich grolle nicht” from Dichterliebe.
Figure 7.6.12. A derivation of the main sequence in “Ich grolle nicht.” At the top, the open-notehead
voices descend in parallel, with the other two voices creating seventh chords alternately above and
below; on the second line, these voices are placed in register as in §1.1. At the bottom, the actual
piece, where the third and voices ascend, and one of the descending voices is displaced by octave. The
arrows show that we can think of the entire sequence as a repeating contrapuntal pattern with one pair
of voices moving every half note (dotted arrows) while the other pair of voices moves every quarter
note (solid arrows).

This leads to a two-chord contrary-motion sequence whose outer voices

ascend by minor seventh and descend by major second, divided in half to
create ascending fourths and descending semitones—a Romantic-harmony
pair of subdominant–dominant progressions, with contrary motion producing
extended and unusual sonorities.48 We then hear a third contrary-motion
pattern that brings us closer to the world of harmonic cycles, a I–V7/vi–IV–
V near sequence.49 The music ends with ascending plagal progressions that
evoke the Fenaroli/Prinner round (Figure 4.5.6), though here in a three-voice
configuration, culminating in a gloriously sarcastic V–I cadence. Listening
to the piece I do not sense any grinding of gears as it passes between
harmonic cycles and sequence: the harmonic cycles have sequential
qualities, while the sequences articulate harmonic functions.
My purpose is not to deny the audible differences between the disguised
repetition of harmonic cycles and the more obvious repetition of the
sequential system. Rather, it is to emphasize that both use repetition in
complementary ways—and that they are embedded within a language that
deploys larger and equally repetitive formal, rhythmic, motivic, and phrasal
schemata. As a student, I inherited from my teachers a tendency to devalue
repetition in favor of constantly developing variation. Now that I am older, I
realize that repetition can be an affirmatively enjoyable musical good.50 The
magic of functional music lies in the way it balances identity and difference,
disguising its omnipresent repetition just enough to keep our ears on their
toes.

7. Bach the dualist

Schenkerians have warned that a focus on chord-to-chord regularities can

lead to impoverished analysis—as if too much attention to grammar would
tempt a literary critic to obsess over syntax rather than a novel’s inner
meaning. Though I am an aficionado of harmonic theory, I think the
criticism has merit: earlier composers internalized Roman-numeral norms so
deeply that they were virtually second nature, obeyed the way a writer
unthinking observes subject-verb-object ordering. As a result, Roman-
numeral analysis is often more correct than informative. The real question is
how innovative musical thinking manifests itself within the constraints of the
syntactic system. I agree with Schenker’s suggestion that this is often a
matter of voice leading: a single line moving stepwise so as to determine
harmonic trajectories (§2.1), a pair of voices articulating basic voice leadings
(§3.1), voices articulating long-range melodic connections between
nonadjacent notes (§9.1), and so on. The spiral diagrams blend Schenkerian
and traditional harmonic theory, allowing for both chord-to-chord laws and
idiomatic contrapuntal patterns arising from the geometry of musical
possibility.
C. P. E. Bach once wrote that the third of a major triad generally has a
“tendency to ascend.”51 When I first read this, I dismissed it as a sloppy
attempt at theorizing the leading tone—half convincing myself that he had
somehow forgotten the subdominant, whose third, I was taught, had a
descending tendency.52 But it is hard to feel entirely comfortable with the
hypothesis that a musician as gifted as C. P. E. Bach, dedicated enough to
write a figured-bass treatise, would somehow overlook one of the primary
diatonic triads. (I am, as readers know, prepared to think that earlier theorists
went wrong, but this seems like a step too far.) As I grew to appreciate the
role of ascending subdominants in baroque tonality, I began to wonder
whether 6̂ might have an ascending tendency in some styles—with the
diatonic third’s ascending basic voice leading acting as a counterweight to its
more familiar descending form. In other words, I started to wonder whether
C. P. E. Bach’s puzzling claim might be a reference to dyadic schemas such
as those in Figure 7.7.1.

Figure 7.7.1. The diatonic third’s descending and ascending basic voice leadings and their functional
uses.
 These patterns resonate with the music-theoretical tradition known as
“dualism,” which conceives of ascending fifths as inversions of descending
fifths, their “upside-down” counterparts. Dualist music theorists typically
operate in chromatic space, interpreting the major key’s dominant as the
chromatic inversion of the minor key’s subdominant; however, the two
forms of the basic voice leading can also be related by diatonic inversion
around the supertonic. From this point of view, the “ascending subdominant”
is the inversional counterpart of the V–I progression (Figure 7.7.2). But this
oversimplifies things, since both dominant and subdominant permit a
passing note that forms a resolving tritone: paradoxically IV–I is both a
musical antipode to V–I, and also an alternative expression of the same
functional logic. Could there be functional dialects where “dominant” is a
superordinate genus containing both descending- and ascending-fifth
species?53

Figure 7.7.2. The descending G–C voice leading is the diatonic inversion of the ascending F–C voice
leading; both permit a passing tone that creates a resolving tritone.

In J. S. Bach’s chorales, approximately a third of the root position IV–I

progressions incorporate an ascending leading tone, while in the Beethoven
string quartets, the ascending subdominant occurs just a handful of times. (It
is similarly rare in other classical-era repertoires.) While composer-to-
composer variation is hardly surprising, this suggests a drastically different
conception of a basic functional progression. Perhaps the picture of two
opposite functions derives from later musical dialects, rather than being
intrinsic to functional harmony itself. Perhaps we can recover the earlier
perspective by considering patterns such as those in Figure 7.7.3. Here,
horizontally related pairs are related by inversion, highlighting the contrast
between “right side up” and “upside down.” Vertically related pairs use the
two forms of the basic voice leading to express the same sequential logic,
moving by the same interval and showing that the two voice leadings can
function similarly.

Figure 7.7.3. Descending and ascending step- and third-sequences using the two kinds of voice
leading. The first two letters indicate the type of sequence, while the last indicates whether the
descending or ascending basic voice leading is used. Thus A2a indicates an ascending-step sequence
using the ascending form of the diatonic third’s basic voice leading.

To contemporary musicians, reared on the triadic logic of Riemann and

Rameau, this way of thinking might seem bizarre. But we will examine four
different pieces where J. S. Bach seems to thematize these dyadic
relationships, basing sections and even entire pieces on the contrast between
ascending and descending fifths. The resulting music is almost always
consistent with root-oriented triadic grammar; but it seems to arise from a
very different, dyadic thought process. In these pieces, Roman numerals stop
short of musical meaning.
(a) The first two-part invention, BWV 772. One of the clearest expressions
of Bachian dualism can be found in the first two-part invention, whose initial
sequence is a lightly disguised Pachelbel progression, moving downward by
third from tonic to V/V. Shortly before the end of the piece, the sequence is
inverted around B; though the new version sounds unimpeachably functional
it is very difficult to analyze, particularly if one imagines the inversion to
operate triadically—that is, turning the first version’s root-position triads
into second-inversion triads (Figure 7.7.4). The harmonic mystery dissipates
when we attend to the diatonic third’s basic voice leading: in the first
sequence, it tonicizes C major in the initial unit while becoming
nonharmonic afterward; in the second, it produces a pair of ascending-
subdominant progressions with perceptible dominant-tonic energy.54 (Note
that on my analysis several notes that were harmonic in the original become
nonharmonic in the inversion.) Here we have an almost-explicit
representation of the double relation between the two forms of the basic
voice leading, as both inversions and bearers of the same harmonic energy.
An analogous dualism can be heard in the piece’s two “alternating hands”
episodes, the first using the inverted theme to express the ascending-fifth
basic voice leading, the second alternating theme and inversion to outline the
descending-fifth basic voice leading (Figure 7.7.5). This succession of
tension-increasing ascending fifths followed by tension-releasing descent is
characteristic of Bach’s music.

Figure 7.7.4. Measures 3–4 of the first two-part invention (BWV 772), along with its inversion in
measures 19–20.
Figure 7.7.5. Two episodes from the first invention, the first using the inverted theme to ascend by
fifth; the second using theme and inversion to descend by fifths.

(b) The C ♯ major prelude from Book I of The Well-Tempered Clavier.

This prelude begins with a long ascending-fifth sequence based on the
“pseudochord” idiom and without any tonicization of new sequential units
(Figure 7.7.6); the ascending-fifth basic voice leading can be read from the
start of one unit to the next, though this is not obvious to the ear. These
ascending-fifth modulations ratchet up the tonal tension, ascending from C♯

to G♯ to d♯ to a♯, arriving at a descending-fifth sequence that releases the

accumulated energy. This second sequence is unusual insofar as its four-bar
unit modulates up by fifth and then down by step, so that it initially seems to
continue the opening sequence up to e♯ (Figure 7.7.7). Here the outer voices
clearly delineate the ascending basic voice leading (a fragment of pattern
A2a on Figure 7.7.3) while also hinting at the descending form. Where the
initial ladder-of-fifths ascent was weightless and independent of tonicization,
these last ascending-fifth motions have more energy, as if extra force were
required to keep the music moving in that direction. Descending fifths
eventually win out, falling past the tonic and down to the subdominant,
whereupon the initial music takes us from F ♯ to C ♯ to G ♯ for a long
dominant pedal. This pairing of ascending- and descending-fifth sequences
anticipates what I will call, in §8.5, the “up-and-down-the-ladder” schema; it
is also reminiscent of the D major prelude’s tonal plan, though that piece has
another trip up and down the ladder of fifths (Figure 7.7.8).55

Figure 7.7.6. The opening sequence of the first C♯ major prelude from The Well-Tempered Clavier.
Open noteheads show the diatonic third’s ascending basic voice leading.

Figure 7.7.7. The second sequence of the C♯ major prelude.

Figure 7.7.8. The tonal plans of the C♯ and D major preludes compared. Solid brackets indicate
ascending-fifth modulations; dotted brackets indicate descending-fifth progressions that produce
descending-step or descending-fifth modulations.

(c) “Ach, lieben Christen, seid getrost” (BWV 256, Riemenschneider 31).
The previous pieces contrast the two forms of the diatonic third’s basic voice
leading, but in different ways: the two-part invention presents the ascending
fifth as both the descending fifth’s inversional dual and functional
equivalent, while the C♯ major prelude uses ascending and descending fifths
as contrasting directions in an orderly modulatory path. “Ach, lieben
Christen, seid getrost,” a freestanding chorale based on an anonymous
sixteenth-century melody, gives us a third approach, moving quickly from
ascending-fifth progressions to descending-fifth cadences, as if compressing
the prelude’s harmonic narrative down to the phrasal level. Its mercurial
contrasts feel spontaneous and embodied, the product of internalized
keyboard routines rather than a series of chord changes we are intended to
follow.
The music begins on A minor but immediately switches to C major for the
rest of the phrase (Figure 7.7.9); this is a standard chorale gambit, inserting
the global tonic at the start of an opening phrase in a subsidiary key, as if
paying lip service to monotonality.56 The IV–I ascending subdominant
features the diatonic third’s basic voice leading, supporting a “merely
passing” note that “just happens” to create a resolving tritone. The phrase
ends with a cadential schema reused by four of the chorale’s five phrases,
with the inner-voice leading tone falling by third.57 The second phrase
returns from the relative major to the relative minor by a sequence of
ascending fifths, C–G–d–a. This is a favorite Bach schema, the diatonic
third’s basic voice leading allowing for the tonicization of each successive
fifth; it is closely related to the four-voice Chase-schema variant in §4.6.
Relative to that schema, the seemingly superfluous vii°/V is both standard
and expected—sandwiched inside the ascending basic voice leading (C5,
E4) → (B4, G4).

Figure 7.7.9. “Ach lieben Christen, seid getrost,” highlighting the diatonic third’s basic voice leading.

The next phrase cancels the Picardy third with Renaissance decisiveness,
shifting directly from the cadence on A major to the dominant of C. Once
again, we return to the relative minor by the ascending-fifth schema, C–G–
d–a, now extended to a half cadence in A minor. The next phrase modulates
to G major by descending fifths, arriving at a dominant that moves to G by
way of a IV6–viiø7–I progression. Many analysts would consider the IV6
“merely passing,” but the chorale contains three C6–f ♯ ø7–G progressions,
only one of which is preceded by D; it is a particular schema that uses the
ascending basic voice leading, not a generic process of prolongation.58 This
last phrase loosely echoes the second, beginning with two more repetitions
of the four-voice Chase schema. In the D–a progression, the thirds switch
roles so that each staff twice articulates the diatonic third’s basic voice
leading: (G, B) → (F♯, D) → (A, C) and (D, B) → (F♯, A) → (E, C).59 At
this point the fifths reverse direction, descending to the final cadence as in
m. 4.
Here, much more than in the previous pieces, one can start to feel the
inadequacy of triadic Roman-numeral theory. For while virtually everything
in the chorale can be made consistent with that theory, this requires a series
of analytical tricks—the delicate placement of key boundaries, the use of
secondary dominants to license unusual triadic progressions, and so on. Such
tricks will not completely erase the sense of mismatch between triadic
grammar and Bach’s practice: theory tells us to value descending-fifth
progressions and smooth progressions between keys, whereas Bach seems to
delight in ascending-fifth progressions that meander from one key to
another. There is no path to the music that starts from the precepts in a
standard harmony text. But once we focus on the basic voice leading, it
starts to clarify: for here we find a systematic contrast between the familiar
ascending-fifth schema, appearing in both explicit and disguised form, and
the descending-fifth logic of the cadence.
(d) The second B minor fugue from The Well-Tempered Clavier. The last
piece of The Well-Tempered Clavier paints a similar picture on a bigger
canvas, using the two forms of the basic voice leading to articulate large
formal regions. Its episodes all use the diatonic third’s basic voice leading,
with episodes 1, 3, and 7 ascending and the rest descending. The subject has
two parts, the first an inverted arch combining basic voice leadings in near-
contrary motion, the second a 6̂–5̂–4̂–3̂–2̂ descent. The piece presents two
harmonizations of this descent, each exploiting a different dyadic pattern
(Figure 7.7.10).60 The first uses sequence D3a on Figure 7.7.3 and leads to a
series of energetically charged ascending-fifth episodes while the second
uses seventh chords and the diatonic fifth’s basic voice leading; this
produces a variety of harmonizations spanning the gamut from fauxbourdon
to descending fifths.61 The music thus charts a path from unusual to
conventional, gradually relaxing in intensity over the course of its ninety-
nine measures—much as our earlier examples balanced ascending-fifth
tension with descending-fifth release.

Figure 7.7.10. The subject of the second B minor fugue from The Well-Tempered Clavier, along with
two harmonizations of the linear descent.

Formally, the piece is in three large sections: the first twenty-six measures
contain a single episode and state the subject on B, F♯, and B; this music is
dominated by an energetic, trilling countersubject and the diatonic third’s
ascending basic voice leading. The central second section lasts from m. 27 to
m. 80 and introduces the oscillatory, descending-fifth countersubject. Here
we have very little B minor, with two entries on F♯ and one each in D major,
A major, and E minor; this music is dominated by descending-fifth
progressions. The third section is a transposed and condensed repeat of the
second, supplemented by a cursory final cadence.
Figure 7.7.11 uses open noteheads to highlight two fundamental voice-
leading patterns: the contrary-motion near sequence harmonizing the
subject’s opening and the countersubject’s basic voice leading. The
abundance of white shows how completely the two schemas saturate the
music. The first episode warps the subject’s final measure into an episode
that exploits pattern A2a on Figure 7.7.3, moving along the chain of
ascending fifths from A major to its relative minor. The m. 26 arrival on C♯7
signals the end of the first section and the disappearance of the trilling
countersubject. Bach’s willingness to drop this motive contradicts
compositional nostrums about unity and motivic consistency: instead, the
piece simply moves on to something new.

Figure 7.7.11. An outline of the fugue’s first section.

The first half of the second section is outlined in Figure 7.7.12. In some
ways this can be considered a second beginning, the counterpoint finally
having achieved its final form. It is interesting that this new countersubject
appears only after the piece has left the tonic, as if reserving its home-key
debut for the recapitulatory third section. Bach’s countersubject has a
characteristic baroque flexibility, appearing once as fauxbourdon, three times
as descending fifths, and twice as the intermediate 75–63 sequence (Figure
7.7.13).62 One weakness of triadic theory is its inability to capture this
flexibility: fauxbourdon is purely linear, the descending-fifth sequence
paradigmatically harmonic, and the 75–63 D3A2 sequence is an
uncomfortable hybrid that cannot easily be assimilated to either category.
Like the Renaissance seventh, this eighteenth-century idiom challenges the
distinction between harmonic and nonharmonic, opening a gap between
Bach’s thinking and our own.

Figure 7.7.12. An outline of the first half of the fugue’s middle section.
Figure 7.7.13. The six appearances of the 7–6 pattern, transposed to B.

The fourth entry gives way to a descending-step episode whose bass has
the same oscillating character as the second countersubject, as if echoing the
new musical mood (see Figure 7.7.12, and pattern D2d on Figure 7.7.3). It is
a canonic sequence whose upper voices outline the descending form of the
diatonic third’s basic voice leading, highlighted here for the first time. The
following A major entry gives us a descending-fifth harmonization of the
second countersubject, a diatonic 65–42 sequence in sevenths. This is
followed by a return of the diatonic third’s ascending basic voice leading,
embedded in a highly chromatic sequence, with the voices outlining exactly
the same notes that featured in the earlier ascending-fifth episode. (Compare
the reduction of episode 3 in Figure 7.7.12 to that of episode 1 in Figure
7.7.11.) Like that earlier passage, the music ascends by fifths from A to f♯,
once again moving from major to relative minor. Discounting the
recapitulatory third section, this is the last ascending-fifth sequence in the
piece: having dominated the first half of the fugue, its nonstandard logic
gives way to descending-fifth convention.
 The rest of the piece has some wonderful moments: a joyous cascade with
outer voices in a lightly disguised canon (episode 4 on Figure 7.7.14); a brief
ascending-fourth sequence based on the first two measures of the theme,
creating the false impression of stretto (entry 8 on the same figure); and a
final descending-fifth episode in which the bass loops the end of the subject
under a florid upper-voice countermelody, with the outer voices decorating
the diatonic third’s descending basic voice leading (episode 5). With the
return to the tonic, the fugue settles into a recapitulation. The expressive
effect is somewhat unclear, the recapitulation functioning neither as a
triumphant return nor a counterweight to any lengthy tonal excursion;
instead, it is a slight tweaking of the degree-of-sameness in a relatively
homogeneous piece. The final close is perfunctory, the subject shrinking to
its first four notes and appearing in the lower voices before the entire Well-
Tempered Clavier careens to a halt on an unexpected Picardy third.
Figure 7.7.14. An outline of the second half of the middle section of the fugue.

***
If it is true that J. S. Bach’s music is often governed by a dualistic contrast
between ascending and descending fifths, then what consequences follow? A
blunt answer is that for all its accuracy, Roman-numeral thinking is often
boring, insensitive to the contrapuntal structures that give direction to its
atomic and independent harmonic cycles. Rather than mere conventions,
these contrapuntal structures reflect a preexisting musical geometry; thus,
though they may sometimes appear in historical sources or partimento
manuscripts, they can also be rediscovered independently by musicians as
they learn their way around the keyboard. This independent rediscovery is
part of the process of developing a distinctive compositional voice.
A sense for this contrapuntal logic is a prerequisite for enlightening
analytical reduction. Consider the opening of Petzold’s Minuet in G, not
composed by Bach but evidently valued by him. Every Schenkerian analysis
I know treats the opening melody as embellishing a neighboring 5̂–6̂–5̂
progression, the E–F♯–G ascent a superficial prolongation of the primary E
(Figure 7.7.15).63 The assumption seems to be that the subdominant’s third
has a tendency to descend, acting as a neighbor to 5̂. The preceding analyses
suggest that the progression is instead an expressively marked illustration of
C. P. E. Bach’s ascending tendency. On this hearing, the tonic and stable G–
A–B–C–D is answered by a more active ascending-subdominant C–D–E–
F♯–G, the C–B bass and E–F♯–G melody creating a subdominant-dominant
hybrid. Habits formed by Haydn and Mozart will not help us hear this music,
nor adjudicate among our interpretive options. Nor will it do to retreat to
subjectivity, declaring analysis the record of our own responses to
decontextualized notes: for interpretation should be cognizant of a
composer’s vocabulary—or at a minimum, not conducted in ignorance
thereof.

Figure 7.7.15. Carl Schachter’s reading of the opening of Petzold’s Minuet in G.

A similar point could be made about Figure 7.7.16, which at first glance
seems to juxtapose two unrelated syntactical units, IVmaj2–ii7–V7–I in the
dominant and a descending-step sequence that returns to the tonic. To a
traditional theorist, the progression from V to vii°43/ii is an inessential
juncture, a pair of words belonging to different sentences. To a Schenkerian,
the local progressions and modulations are superficial, generated by a deeper
linear process in which the tenth (C4, E5) descends by step through the
octave to (C3, E4). Those who have been sensitized to the diatonic third’s
basic voice leading, however, may notice the descending-step dyadic
sequence (G, B) → (F, D) and (F, A) → (E, C): this gives higher-order
direction to the harmonic cycles while continuing the modulatory journey up
the ladder of fifths (C major→G major→D minor→ . . .). Bach embeds the
dyadic sequence in a series of nonsequential triads, superimposing the first-
unit (F, D) with the second-unit (F, A), so that (G, B) is root-third of a G
major triad while (F, A) is third-fifth of D minor (c.f. Figure P4.3). This
dyadic logic joins two Roman-numeral cycles and cuts across the parallel
tenths of Schenker’s analysis. The compositional thinking here is
recognizably similar to that which produced the four-voice Chase schema in
“Ach, lieben Christen, seid getrost,” or the ascending-fifth sequences of the
B minor fugue; indeed, virtually the same progression occurs in the G major
cello suite, extending the ascending fifths past the supertonic to the relative
minor (Figure 7.1.17).64
Figure 7.7.16. A sequence of ascending basic voice leadings in mm. 8–15 of The Well-Tempered
Clavier’s first prelude, along with Schenker’s reading.
Figure 7.7.17. Measures 9–14 of the prelude to the G major cello suite, BWV 1007.

Should the ubiquity of these schemas lead us to postulate corresponding

structures in the minds of listeners, as Meyer, Gjerdingen, and Byros have
suggested? Here I think we should be cautious. It is plausible that an
experienced listener might judge the bottom staff of Figure 7.7.1 to “sound
like Bach”—just as enculturated audiences might say that a I–♭III–♭VII–IV–I
progression “sounds like rock.” But there is a huge leap from this kind of
tacit recognition to an explicit awareness of schemas as specific items of
musical vocabulary: “oh, that is an ascending subdominant pattern, oh, now
we are hearing an ascending-fifth sequence, oh, here’s a Prinner.” Anyone
who has taught ear training will know just how rare such awareness is, even
among talented musicians with substantial education. For this reason, I
suspect that the transmission of schemas occurs predominantly from
composer to composer, through explicit instruction, implicit learning, and
score study—augmented by a substantial helping of independent
rediscovery. Readers who are familiar with Bach might ask themselves how
much they thought about the diatonic third’s ascending basic voice leading
prior to this book. If the answer is “not at all,” then music is not a
straightforward transmission of information from composer to listener;
rather, it is a more complex mixture of implicit and explicit.

1
This preference for tonic and dominant is embedded in the traditional “rule of the octave,” to be
discussed shortly; the connection supports Holtmeier’s 2007 contention that the rule of the octave is
not simply a schema but a theory of functional tonality.
2
My theory is indebted to a long tradition of “harmonic maps” (Riemann 1896, Piston 1941,
McHose 1947, Kostka and Payne 2003).
3
Some bass lines are rare for linear reasons, such as ii6–V6 with its leaping tritone or V6–I6 with its
unresolved leading tone.
4
See Christensen 1993b, p. 97.
5
It is possible that the mediant’s harmonic flexibility helps explain its infrequency: by blurring
tonic and dominant, it undercuts the dichotomy at the heart of functional syntax. Nonsequential
mediants are most common at the beginning and end of the functional tradition, virtually disappearing
during the classical style.
6
See mm. 32–33 of Mozart’s Piano Sonata in C major, K.279, II, or m. 13 of his C minor sonata
K.457, I (with the leading tone in an inner voice); the tenths often evoke the Prinner schema.
7
Horn and Huron (2015) make a related point about modal homogenization, which occurs on an
even larger timescale.
8
John Clough (1979b and 1994) and Jason Yust (2009) have explored this structure as it relates to
prolongation (§9.1–§9.3). I will be more concerned with its role in generating harmonic and melodic
patterns, including sequences and near sequences. An unpublished paper by Yust, written when he was
a graduate student, anticipates several features of my approach.
9
Because the size of the diatonic scale is a prime number, every “maximally even” chord is a near
interval cycle and every interval cycle can be extended to form a maximally even chord. Since the
ouroboros contains fifths, thirds, and steps, it contains basic voice leadings for nearly even chords of
every size.
10
For example, I–IV–V7 forms the cycle (G–E–C)–(C–A–F)–(F–D–B–G), returning to its starting
note two octaves below. Very simple cycles such as I–V–I and I–IV–I skip a few notes on this chain of
thirds.
11
The A major triad at the end of the bridge sounds like C: V/ii, suggesting another repeat; instead,
it becomes the global tonic by compositional fiat.
12
Temperley 2018 notes the bridge’s sequential structure.
13
There are many well-known examples of descending-third melodies, including the opening of
the Hammerklavier’s first-movement development (Rosen 1971, pp. 407–13), the opening of
Brahms’s Fourth Symphony, and Brahms’s Op. 121, no. 3.
14
One could of course make an analogous point about many non-Western styles.
15
This blurring of third-related harmonies is already present in the 1722 Treatise on Harmony,
which derives the deceptive cadence by noting that C–E–A can replace C–E–G (Rameau [1722]
1971). In that book, C–E–A can have a fundamental bass of either A or C.
16
This is particularly clear if one examines the musical examples Rameau offers, which sometimes
add content to his explicit theorizing. These examples, like my own reconstruction, highlight the
centricity-defining role of I and V7.
17
Nathan Martin has referred to Rameau’s “astonishing ingenuity” in deriving common
progressions from his descending-fifth paradigm (Martin 2018). But if functional music tends to move
along the descending circle of thirds, then this ingenuity is easily explained; what makes it seem
astonishing is the nonobvious relation between Rameau’s descending-fifth paradigm and (what I take
to be) the more basic circle of descending thirds.
18
Since the second distortion undoes the first, quarter notes 8–10 are shifted up by third relative to
the underlying pattern. The second distortion avoids the tonic chord’s seventh, abandoning the leading
tone to create a brief moment of triadic arrival; my analysis includes this missing seventh in
parentheses.
 19
Readers can visit my https://www.madmusicalscience.com to hear artificially generated music
using Bach’s template. Those who find Bach’s figuration challenging can use another pattern, relaxing
whatever stylistic rules they want—the important thing is guiding the perpetually descending sevenths
through a range of keys.
20
Rameau himself might not have considered them true descending-step progressions, as he
considered vii° an incomplete V7.
21
The most common ascending-third progression is vi–I6, and the most common descending-step
progressions are vi–V and V–IV6. The functionally anomalous cadential six-four allows for vi–I64 and
ii–I64 (§9.5), while vii° (sometimes thought to be a metonym for V7) allows for I–vii°.
22
Compare the second B major fugue, which exploits the double counterpoint latent in the two
parallel options.
23
For modal ii(6)–IV–V progressions see Figures 6.5.5 (m. 56) and 6.7.6 (m. 25).
24
The ascending use of 4̂ in I–V43–I6 is also relevant here, as it allows I–V–I to harmonize 3̂–4̂–5̂.
25
See Headlam 2012 and Yust 2018, chapter 10, for expressions of this Schenkerian hope.
26
The only descents come from “weak” progressions in both more and less conventional forms (a
IV–I–V found in virtually all functional music, and I–vi–I and IV–vi–IV progressions more idiomatic
to rock) and slightly longer voice-leading paths (IV–V and V/vi–vi, the latter using a seventh). I hear
I–vi–I and IV–vi–IV as weaker versions of subdominant-tonic oscillation, with vi substituting for IV
in the first case and I in the second. These variants start to become popular in the middle of the
nineteenth century (§10.7).
27
This ascending voice leading continues through the repeat even while the vocal melody
disguises it, with the middle voice now sounding the melodic G4. This is illustrated, on my example,
by the two open noteheads in the final chord, one representing the last note of the first verse, the other
the first note of the second verse.
28
Traditional theorists rarely distinguished fauxbourdon from harmonic cycles: Rameau ([1722]
1971, p. 126), for example, adds a fundamental bass to fauxbourdon, while Weber ([1817–1821] 1846,
p. 457) includes fauxbourdon passages in his discussion of chord grammar. Later theorists such as
Schenker and Lerdahl also conceive of functional harmony as unified.
29
The contrary-motion outer voices are a common idiom; see Figures 3.2.5 (“fauxbourdon”) and
10.6.9, or Beethoven’s Op. 18, no. 5, IV, m. 59ff.
30
It is sobering to contemplate the divergence between this musical “rule of the octave” and the
alternative found in historical treatises—a theoretical construct that is perhaps more accurate as a
description of general functional principles than of musical practice (Christensen 1992, Holtmeier
2007; see Sanguinetti 2012, p. 20, who cites a rare discussion of the fauxbourdon ROTO in Valente).
Once again we find an embodied and untheorized musical knowledge, idioms learned by ear or by
score study, passed on from composer to composer without leaving clear theoretical traces.
31
Both Temperley (2011) and Rothgeb (1975) describe the fauxbourdon ROTO progression as
unusual.
32
Brahms frequently uses retrofunctional V–IV progressions in his second themes; often these
involve some version of the “fauxbourdon ROTO,” but sometimes they do not. For other examples,
see the second themes of the first movements of the G major string sextet, Op. 36, and the C minor
piano quartet, Op. 60.
33
Compare the fauxbourdon ROTO harmonization of 8̂–7̂–6̂–5̂ to the template in Figure 6.1.3.
34
My label is a play on the term “standing on the dominant,” coined by Erwin Ratz and
popularized by Caplin 1998; Gjerdingen’s “Indugio” schema (2007) is closely related. For similar
examples from Mozart, see K.279, I, mm. 25ff; K.283, III, mm. 33ff; and K.309, I, mm. 47ff.
 35
Even in the sixteenth century, Zarlino encouraged composers to be freer in their use of second-
inversion harmonies ([1558] 1968, p. 193).
36
Harrison 2003 explores this connection in detail; Allsop 1992 explores the prehistory of the
Corelli style.
37
William Caplin has noted that the “7” sonorities in these passages have a quasi-harmonic
character (1998, p. 31, footnote 29), and I have argued for the importance of the underlying down-a-
third-up-by-step root-pattern (2011a, chapter 7).
38
The “Crucifixus” from Bach’s B minor mass uses the 7–6 and 7–43 options almost
interchangeably (Biamonte 2012).
39
Examples can also be found in Brahms: Op. 76, no. 4, mm. 28–30 (discussed in Tymoczko
2011a, chapter 8), and Op. 119, no. 2, mm. 15–16.
40
One might think that this idiom counts as evidence against that prohibition, except that the vast
majority of apparent I–ii–I6 progressions occur in precisely this context, with the predominant
invariably lacking its fifth.
41
The second G major prelude in The Well-Tempered Clavier contains an extended example of the
idiom in which the parallel thirds exceed their usual bounds.
42
To generate this list, I programmed my computer to identify passages of melodic repetition; I
then analyzed each sequence by hand, eliminating the relatively small number that I considered not to
be true sequences. I also compared the results to my own analyses of individual pieces to check for
completeness (appendix 4).
43
Hook 2020 uses a similar system of categorization. Longer sequences can usually (but not
always) be reduced to just the first and last chord in the repeating unit.
44
This sequence appears commonly in Beatles songs, for instance “You Never Give Me Your
Money.”
45
Interestingly, Bach uses the A5D3 sequence, the inversion of the common D5A3, at least twice
in The Well-Tempered Clavier: in the first F major fugue (m. 13) and in the second C♯ minor fugue
(m. 23).
46
To these we can add fauxbourdon and the “down a third up a step” sequence, though the latter is
rare after the baroque.
47
See Riemann 1896, pp. 120ff, and Caplin 1998, p. 29, for the view that sequences are almost
entirely linear. Schenker went further, denying that sequences existed and speaking of linear
progressions instead; see Schenker (1935) 1979, Slottow 2018, and Sprick 2018. Weber (1817–1821)
1846 argues that sequential progressions are similar to those used by harmonic cycles (p. 430), though
somewhat looser (p. 432).
48
Note that here it is important to consider the melodic E5 as at least quasi-harmonic, as it
continues the sequential pattern of mediants-acting-like-dominants, even though it quickly resolves to
an orthodox V7.
49
This pattern uses both the ascending-third (I–V7/vi) and ascending-step patterns (V7/vi–IV–V) in
Figure 4.4.1.
50
Margulis 2013.
51
Bach (1762) 1949, p. 205.
52
Alternately, Nathan Martin has suggested to me that it may be an echo of the pre-Renaissance
idea that imperfect consonances resolve to perfect consonances.
53
Delair’s 1690 figured-bass treatise recommended the diatonic third’s ascending basic voice
leading for ascending-fifth bass lines (Nicolas 2019).
54
This analysis echoes some of Schoenberg’s observations about the piece (Neff 2012).
 55
The F major two-part invention has a very similar harmonic plan, modulating upward by fifths
I–V–ii–vi, descending by fifth-related chords vi–ii–V–V/IV–IV to reach the subdominant, and ending
by modulating back up by fifth.
56
Typically these “orphan tonics” also participate in the harmonic narrative of the opening phrase,
though sometimes in an unorthodox way (e.g., a vi–I6 progression).
57
To call this “non-normative” is to engage in speculative fiction; in reality, this is a standard
cadence that should be included in any realistic description of leading-tone behavior.
58
Burstein (2018, p. 6) writes that the IV6–I progression is “atypical,” but by my count I is the
second most common successor to IV6 in Bach’s chorales.
59
This could be described as a variable sequence with changing permutation.
60
The Well-Tempered Clavier’s second F♯ minor fugue also juxtaposes these harmonic strategies,
initially harmonizing 6̂–5̂–4̂–3̂–2̂ with the descending-third sequence before shifting to descending
fifths for the rest of the piece. In that fugue, however, the descending-third harmonization is fleeting
and momentary, with the bulk of the piece tracing out long sequences of descending fifths.
61
Compare the two harmonizations of the descending tenths in Figure 3.4.11.
62
These intermediate forms typically thwart the turn to the relative major that otherwise
characterizes the minor-mode entries, the III+maj7 a second-level Baroque dominant (e.g., A+maj7–f♯ in
m. 29 and G+maj7–e in m. 73). Compare Rings (2011, p. 152) on the evanescent nature of fugal key
areas.
63
See for example Proctor 2004 or Schachter 2016.
64
Note that Figure 7.7.11 contains the same pattern we have been considering, (E, G♯)→ (D, B)

and (D, F♯)→ (C♯, A), with (D, B) and (D, F♯) superimposed.
 Prelude
Chromatic or Diatonic?

A famous passage from the second movement of Beethoven’s Ninth

Symphony arranges eighteen descending-third progressions into four-chord
groups (Figure P8.1). The result is an intriguing mixture of diatonic and
chromatic, never establishing a clear tonic and wandering through tonal
space—yet spelling out a long series of fifth-related diatonic scales. My
analysis postulates a nested pair of repeating contrapuntal patterns, triads
moving inside diatonic scales that are themselves moving. This hierarchical
analysis helps simplify the description of the melody: if we conceive the
passage chromatically, we have to say that major and minor triads move
differently, the fifth of every minor triad ascending by semitone, and the fifth
of every major triad ascending by two semitones; if we postulate diatonic
scales we can instead say that the fifth of every triad ascends by scale step to
become the root of the next.1 Sometimes this scale step is a minor second,
sometimes it is a major second, but from a diatonic perspective it is always
one step. It goes without saying that a very large amount of music is
structured in exactly this way; my analysis simply assumes that Beethoven’s
passage exhibits the same sort of hierarchy found throughout his work.
Figure P8.1. Measures 143–176 of the scherzo of Beethoven’s Ninth Symphony, along with the
composite diatonic collections. Below the music I show the repeating contrapuntal patterns at the level
of both chord and scale. “T” represents diatonic transposition while T represents chromatic
transposition. I also include a neo-Riemannian analysis.

In this case, the diatonic scale’s notes are themselves descending

sequentially so as to create a series of descending-fifth modulations. When
the A4 in m. 2 moves to B♭ rather than B♮, it reveals that an abstract object, a

scale step or melodic slot, has moved down by semitone from B to B ♭ .

Because of these shifts, the chordal pattern never reaches the diminished
triad and the music does not arrive at a clear dominant, though to my ear it
still has a distinctly diatonic cast. These scalar progressions are somewhat
more goal directed than the triadic progressions: the chords move down by
third from C major to A major (spelled B♭♭), making a right-angle turn to V7
of E minor, while the keys move by fifth from C down to E where the music
remains.
 We can use our spiral diagrams to represent the motion of scales within
the twelve-tone chromatic collection (Figure P8.2). The recipe in §2.1 allows
us to calculate the descending basic voice leading: sliding downward from C
to F gives us T–7, revisiting our initial angular position four times for t4. The
resulting voice leading, T–7t4 or T5t–3, lowers the leading tone semitonally
while keeping every other note fixed. It follows that Beethoven’s passage
uses two hierarchically nested basic voice leadings: the chordal voice
leading takes a clockwise step in its spiral diagram, connecting third-related
triads by an ascending diatonic step; the scalar voice leading moves
counterclockwise in its diagram, connecting fifth-related scales by a
descending chromatic step (Figure P8.3). These are hierarchically self-
similar transformations occurring on two different structural levels.2
Figure P8.2. The spiral diagram for diatonic scales in chromatic space. Sliding clockwise along the
spiral from C to F passes seven points for T–7 while passing twelve o’clock four times for t4. The

combination T–7t4 shifts B to B♭ while keeping all other melodic slots fixed; this is the basic voice
leading.

Figure P8.3. Beethoven’s passage takes eighteen counterclockwise steps along the triadic spiral and
eight clockwise steps along the spiral of diatonic scales.

My account contrasts with one of the foundational analyses of neo-

Riemannian theory, which models the passage with two chromatic voice
leadings: the “R transform,” which either raises the fifth of a major triad by
two chromatic semitones, or lowers the root of a minor triad by the same
amount; and the “L transform,” which either raises the fifth of a minor triad
by semitone or lowers the root of a major triad by semitone.3 Thus where I
identify two hierarchically nested transformations, the neo-Riemannian
analysis has two nonhierarchical transformations in alternation, operating
directly on chromatic triads and unmediated by the diatonic collection.
Neither of these is correct as a matter of simple fact. The advantages of my
interpretation are, first, that it describes Beethoven’s sequence using
standard harmonic procedures, second, that it offers a unified explanation of
the triadic voice leadings, with each triad’s fifth moving up by scale step,
and third that it generalizes easily to any chord in any scale. The advantage
of the neo-Riemannian analysis is that it connects Beethoven’s procedures to
nineteenth-century chromaticism. The two analyses thus highlight
Beethoven’s position at the fulcrum between classical and Romantic eras.
From this perspective, scale degrees and pitch classes are fundamentally
analogous. Before writing this book, I thought of them as ontologically
different, scale degrees labeling movable positions in a collection while
pitch classes labeled notes. This conception went along with a relatively flat
view of musical hierarchy, in which scales of various sizes sat side-by-side,
each mapping their own scale degrees directly to pitch classes (Figure P8.4).
Surface voices, on this picture, move within scales, coming together to form
chords as byproducts. Having written this book, I have come to believe that
musical objects can participate in more complex hierarchies, with any
collection’s output available to serve as the input of the next. This gives us a
unified perspective on phenomena from modulation to subchordal voice
leading (e.g., the diatonic third’s basic voice leading, or the contrary-motion
patterns in Figures 4.7.1–3). This idea sounds simple when written on the
page but is extremely difficult to implement at the speed of music, whether
in improvisation or at the slower pace of notation—and all the more so when
one is exploring unfamiliar collections. Music theory can help musicians
internalize this abstract perspective, explaining how to replace specific
idioms with more general techniques applying to all levels of the collectional
hierarchy.
Figure P8.4. Flat and hierarchical models of chord and scale.

1
This is exactly the same reasoning that allows us to analyze “doe, a deer” as a single diatonic
pattern moving up by scale step (Figure P2.1).
2
These nested diagrams recall Figure 3.7.4 but represent different musical levels: where the earlier
figure depicted dyads moving inside scale-like triads, our new diagram represents chord-like triads
inside seven-note scales.
3
See Cohn 1997 for the analysis and appendix 3 for the neo-Riemannian transformations.
 8
Modulation

This chapter uses the spiral diagrams to model scales and modulation. We
begin with some conceptual issues surrounding key changes, including the
notion of “modulatory distance” and the meaning of enharmonic spelling in
a fixed chromatic universe. We then construct a circular model of key
relations, using it to describe several common modulatory schemas. This
leads to a reconsideration of the seventeenth-century process of modal
homogenization, by which other modes take on the characteristics of ionian.
Finally, we survey some features of scalar thinking in the twentieth and
twenty-first centuries. In many ways, the conceptual work has already been
done; the task is simply to apply familiar techniques to new musical
problems.

1. Two models of key distance

What does it mean for two modes to be “close”? One answer focuses on
modulations that minimize the motion of scale degrees, reducing the overall
sum of distances as measured from tonic to tonic, supertonic to supertonic,
and so on. From this point of view C major and G major are not close, since
their scale degrees are almost half an octave apart; instead, the most
proximate keys are the parallel major and minor—as well as semitonally
related keys such as C and D ♭ major (Figure 8.1.1). This notion of key
distance is most relevant when instruments do the same thing before and
after the modulation: one singer, for example, singing the same melody in
both old key and new. (If you can sing a tune in C major, you can probably
sing it in D♭ major whereas G major may be out of your range.) It underlies
modulatory practices such as “sidestepping” or “pumping-up,” familiar in
jazz and popular music but sometimes found in classical music as well.1

Figure 8.1.1. Scale-degree preserving mappings from C major to C minor and D♭ major. The open
notehead is the tonic.

A second answer focuses on modulations that minimize the motion of

pitch classes irrespective of their scale-degree roles. Here the nearest major
keys are related by fifth, while the nearest major/minor pairs are in the
“relative” relationship (Figure 8.1.2). This notion of key distance is implicit
in classical modulatory norms, and most useful when different material is
presented in the two keys—for example in the first and second themes of a
textbook sonata—or when composers have a wide variety of instruments at
their disposal. For in this case no problems arise from the nearly half-octave
scale-degree shift. This second approach links “key distance” and pitch-class
content, with nearby keys sharing a large number of common tones; the
earlier conception leads to a definition of “key distance” in which nearby
keys need not share many notes at all.
Figure 8.1.2. Scale-degree changing voice leadings from C major to G major and A minor.

Underlying both is the idea that scales can be modeled as collections of

musical voices, circularly ordered by pitch class and able to change both
pitch class and scale-degree role (§1.2, Figure 8.1.3).2 The two approaches
emphasize different kinds of motion on the spiral diagrams. For modally
matched keys, the scale-degree-preserving voice leadings are slides along
the spiral; for mode-changing modulations, we combine these slides with the
parallel modulation connecting keys sharing a tonic. What results is a model
in which abstract voices never change scale-degree roles (Figure 8.1.4).
The second notion of key distance permits voice leadings that do change
roles—represented by loops on the spiral diagram. For major keys, distance
is given by iterations of the diatonic scale’s basic voice leading, which either
lowers the leading tone (clockwise) or raises the fourth scale degree
(counterclockwise). The nearest major/minor pairs are now linked by the
relative relationship. The diagram in Figure 8.1.5 dates back to the
eighteenth-century work of Heinichen and Kellner; indeed, Heinichen’s
“Musicalische Circul” is one of the earliest geometrical representations of
musical structure, antedating the Tonnetz by almost a decade (Figure 8.1.6).
Figure 8.1.3. Each abstract voice is associated with a pitch class (bold), a scale degree, and a letter
name. The two circles represent C major (left) and G major (right).
Figure 8.1.4. A model of key distance based on the motion of scale degrees.
Figure 8.1.5. A model of key distance based on the motion of pitch classes.
Figure 8.1.6. Heinichen and Kellner’s circular models of key relations.

To get a feel for this second approach, it is useful to play the scales at the
piano with a fixed fingering; different scalar voice leadings will correspond
to different ways of shifting your fingers as you modulate from one key to
another, with each finger tracking an abstract voice.3 Suppose for example
you play a C major scale with the fingering shown at the top of Figure 8.1.7.
The simplest flatward modulation lowers the leading tone by semitone,
lowering the note played by the fourth finger of the right hand, and moving
the tonic from third left-hand finger to the right-hand thumb. A series of five
such modulations sends the tonic of the first key to the leading tone of the
last, but a series of seven sharpward modulations maps the tonic of C major
to the tonic of C♯ major. In both cases, the fingers embody the melodic slots,
taking on new scale-degree roles as you modulate. The size of the
modulation corresponds to the total distance moved by all fingers.

Figure 8.1.7. Two paths between the same points on the scale circle. (a) The 150° (five pie-slice)
clockwise path from C major to D♭ major. (b) The 210° (seven pie-slice) counterclockwise path from

C major to C♯ major. The first maps the tonic to the leading tone; the second tonic to tonic.

Two warnings. First, there is an inherent ambiguity in identifying scalar

voice leadings, as any change of scale can be nullified by an equal-and-
opposite change in the surface motion of the voices (appendix 1); the best
we can say is “if scales move like this, then surface voices move like that”—
or perhaps, “this choice of scalar voice leadings rationalizes the logic of the
surface voices better than the alternatives.” Second, while abstract voices
often trace specific paths through pitch-class space, scale degrees typically
do not. Consider a modulation from tonic to dominant: at the level of pitch
classes, it is easy to see why we might want to say “the fourth scale degree
has been raised by semitone to become the leading tone of the new scale”
(cf. Figure 1.2.5). By contrast, it is unclear whether the “tonic role” has
moved down by fourth or up by fifth—indeed, that can seem to be a
fundamentally misguided question. Changes in scale degree do not imply
determinate trajectories through musical space: instead, pitch classes simply
take on new scale-degree roles without occupying any intermediate points,
tonicity passing directly from one note to another.

2. Enharmonicism and loops in scale space

Consider a series of modulations, modeling scale degrees on the piano as

just described. Start with C diatonic and lower the second, third, sixth, and
seventh scale degrees to form A♭ diatonic; then repeat that modulation again,
lowering the new second, third, sixth, and seventh scale degrees to form E
(F♭) diatonic; then repeat the procedure one more time to return to the white

notes (D ♭ ♭ ). You will find that your left-hand third finger, which formerly
played the tonic, is now playing the leading tone (Figure 8.2.1). Contrast that
modulation with the series of key changes that ascends by fifth from C major
to E major and then returns to the white notes as above, by lowering the
second, third, sixth, and seventh scale degrees. Here each finger ends up
back on its initial note. We can say that the first sequence of modulations
forms a nontrivial voice leading from the C diatonic scale to itself, moving
each voice downward along the scale; the second produces a trivial voice
leading that leaves each voice exactly where it began.
Figure 8.2.1. Two modulatory sequences; solid arrows are descending major-third modulations that
lower the second, third, sixth, and seventh scale degrees; dashed arrows raise the fourth scale degree.

All of which is very clear on the spiral diagram in Figure 8.2.2. The first
sequence moves in a complete clockwise circle, taking one third of a turn
with each modulation, while the second takes four counterclockwise steps
before moving four steps clockwise; the different outcomes thus reflect the
geometrical fact that loops represent transpositions along the chord. Here,
however, the “chord” is what we would usually call a “scale,” a seven-note
diatonic collection embedded within the twelve-note chromatic aggregate.
Other than that, there is nothing unusual going on: if a series of modulations
makes a complete loop surrounding the center of scale space, then it will
change the scale-degree function of the abstract voices (here represented by
our seven fingers); if the modulations do not make such a circle, then the
abstract voices will return to their original scale-degree functions. The spiral
diagram allows us to see this as a specific manifestation of a more general
phenomenon that applies to both chords and scales. Enharmonic equivalence
is produced by loops in scale space, just as chordal inversion is produced by
loops in chord space. Indeed, enharmonicism just is chordal inversion at the
scalar level.4

Figure 8.2.2. The two modulatory sequences from Figure 8.2.1 graphed in 7-in-12 space.
 What is confusing is that notation normally but not always tracks the
underlying scalar voice leading, with the seven letter names A–G
corresponding to the seven abstract voices or melodic slots. Thus when we
lower the white-note scale’s leading tone we add a flat to B, indicating that
melodic slot B has moved down by semitone. Each one-step clockwise
motion on the circle adds a flat or subtracts a sharp; counterclockwise moves
add a sharp or subtract a flat. If we rigorously follow this rule, then
nontrivial modulatory sequences will connect different spellings of the same
scale: the modulations on the left of Figure 8.2.1 produce C major, A♭ major,

F♭ major, and D♭♭ major. The spelling does not indicate the presence of two

distinct harmonic objects, for in a fixed chromatic universe, C major and D♭♭
major are one and the same—acoustically, conceptually, and in every other
sense. Rather, the spelling indicates the presence of a nontrivial path, a loop
in voice-leading space that changes the voices’ scale-degree roles: in C
major, the C slot (or “voice” or finger) has tonic function, while in D ♭ ♭
major, the D slot is tonic.5 The modulatory sequence on the right of Figure
8.2.1 adds four sharps and then subtracts them, sending C major to C major,
with the identity of the spelling reflecting the triviality of the associated
voice leading. This is conceptually the same phenomenon we encountered in
Figure 1.2.2, where a series of background voice leadings made a complete
loop in chord space, so that a single pitch acquired a new scale-degree label;
in both cases, surface voices need to move in order to remain in the same
place.
However, there are also merely notational enharmonic changes that do not
correspond to any scalar voice leading. In Gottschalk’s Le Mancenillier (Op.
11, Figure 8.2.3), the music shifts from G ♯ minor to the parallel major

(spelled as A♭) and then down by fifth to C♯ minor; this entire passage then

repeats down by fourth, passing through D ♯ major (spelled as E ♭ ) before

returning to the initial G♯ minor. The underlying scale-degree logic is not at

all puzzling, and would be unremarkable if transposed into C minor or D
minor. In C♯ minor, however, Gottschalk’s sequence takes him through G♯

and D♯ major, both requiring double sharps. His flat-side respelling of these
keys is merely notational in the sense that it does not record any genuine
pitch-class voice leading: if we were listening to the piece, or watching a
pianist depress keys in slow motion, we would have no way of knowing that
the respelling had occurred. Similarly, if we were to model its scalar voice
leadings with our fingers there would be no difference between playing G♯

major and A ♭ major; the switch from sharps to flats has no observable
consequence. By contrast, genuine enharmonicism involves an observable
change in the melodic slots’ scale-degree roles.
Figure 8.2.3. Scalar voice leadings in Gottschalk’s Le Mancenillier, Op. 11. The flat-side respellings
are merely notational and do not represent scalar voice leading.

The contrapuntal significance of enharmonicism has been obscured by its

historical connection to issues of tuning and temperament. Absent a fixed
chromatic background, a sequence of twelve fifths will take us close, but not
exactly back, to our starting point—in a Pythagorean framework, for
instance, C major and D ♭ ♭ differ by about a quarter of a semitone.

(Geometrically we could represent this as in Figure 8.2.4, with B ♯ major

slightly to the left of C major, and A ♯ ♯ ♯ major to the left of that: fifth-
related diatonic collections are found approximately 34.4° around the circle
rather than 30°, so that a sequence of fifth modulations never returns to C
major.) These are genuinely different scales. When we work in a universe
with twelve fixed chromatic pitch classes enharmonic spelling acquires a
new meaning: instead of signifying a difference in musical identity it
signifies the presence of a nontrivial route from a chord to itself. And this is
true whether we are using equal or unequal temperament.
Figure 8.2.4. A Pythagorean space places B♯ and C major at distinct spatial locations.

The habit of thinking of enharmonic spelling as representing different

locations, rather than different paths between the same musical location,
persisted long after it was obviated by the advent of a fixed chromatic
universe. For example, both David Lewin and Richard Cohn have noted that
it is impossible to notate the progression in Figure 8.2.5 such that (a) each
voice leading is written in analogously (for instance as a major-third
progression rather than diminished fourth); and (b) the final chord is spelled
the same way as the first. From this Cohn draws the conclusion that
“classical methods of analysis” and “diatonically based models” are
inadequate.6 Cohn is correct that these sorts of modulations exploit
chromatic voice leading—and specifically the proximity of C major and E
major in 3-in-12 space (Figure 2.1.3). But it does not follow that we need to
abandon the very idea that chords can be embedded inside diatonic
collections, nor the hierarchical organization characteristic of so much
Western music. That might be true if one believed that diatonicism and
classical analysis necessarily presuppose something like a Pythagorean
picture of scales floating freely in a continuous and unquantized chromatic
universe. But in a fixed chromatic universe it is entirely possible to repeat
the two-tiered voice-leading schema on the right of Figure 8.2.5, operating
with chord and scale simultaneously: all we have to do is apply one voice-
leading schema at the level of the chord and another at the level of the scale.
What is new in these progressions is that chords take the lead, moving by
efficient chromatic voice leading and pulling scales alongside them.7

Figure 8.2.5. (left) A progression dividing the octave equally, which is awkward to spell; (right) a
two-leveled repeating contrapuntal pattern involving a three-note chord (stemmed) inside a seven-note
scale.

Indeed, the spiral diagrams show us exactly when a repeating contrapuntal

pattern can return each voice to its starting pitch: this will be possible only
when it moves purely radially, and hence can occur only when the relevant
circular space has multiple chords at the same angular position. For if the
repeating contrapuntal pattern takes a nonzero number of angular steps, then
further repetitions will always increase the total quantity of angular motion.
(It cannot change direction, since §4.2 showed that a repeating contrapuntal
pattern corresponds to a pattern of geometrical motion in a spiral diagram.)
The space of chromatic triads contains three chords at each filled radial
position, allowing us to find repeating voice-leading schemas that return
each voice to its starting point (§2.1). The existence of this pattern is a
simple consequence of the fact that 3, the size of the triad, divides 12, the
size of the chromatic scale. By contrast, any repeated modulatory move,
when embodied by the pianist’s hands, will necessarily return to its starting
scale in a new registral position. For exactly the same reason, there can be no
diatonic voice-leading schema connecting distinct triads that, when iterated,
returns each voice to its starting pitch.

3. Minor keys

So far we have constructed two models of key distance, one emphasizing the
proximity of parallel major and minor, the other emphasizing the proximity
of relative major and minor. Since both relationships are important, it would
be nice to be able to model them simultaneously. This led the eighteenth-
century theorist F. G. Vial to propose the structure in Figure 8.3.1; one of its
diagonal axes corresponds to motion along the circle of fifths while the other
alternates between parallel and relative modulations. Vial’s graph is an
accurate model of functional practice—which is to say, it represents the most
common modulations as short-distance geometrical paths. But it obscures
the circular geometry of enharmonic equivalence: after all, the two paths in
Figure 8.3.2 look similar, but only one returns its scale degrees to their initial
position. The model gives us no insight into the unified logic governing both
chord and scale.
Figure 8.3.1. F. G. Vial’s model of major and minor keys. The NE/SW diagonal represents motion
between fifth-related keys; the NW/SE diagonal alternates between relative and parallel modulations.
Figure 8.3.2. The two progressions in Figure 8.2.1 look abstractly similar in Vial’s space but produce
very different effects: the solid path moves every abstract voice down one scale degree while the
dashed path returns each voice to its starting position. This difference is not apparent in Vial’s
geometry.

For this reason, I favor an alternative that more closely resembles the
spiral diagrams. Figure 8.3.3 derives this model in three stages. The first
superimposes the circular spaces for the diatonic, acoustic (i.e., melodic-
minor ascending), and harmonic minor scales, using dotted lines to connect
the collections belonging to G and D minor; as in Figure 3.3.9, I represent
each spiral with a single circle whose arcs apply the basic voice leading. The
top-right circle improves on the model by duplicating the diatonic collection
and rearranging the scales’ radial positions: now the circular arcs represent
shifts between a single key’s scale-forms rather than applications of a basic
voice leading. The lowest graph reflects this same information more
compactly, no longer distinguishing the minor scale-forms. In all three
models, spatial extension allows minor keys to provide shortcuts between
otherwise-distant keys. This nonlocality arises not from deep features of
musical geometry but rather the convention of treating distinct scales as
syntactically equivalent, with changes between minor scale-forms not
counting as modulations even though they move through scale space.

Figure 8.3.3. Three models of key space. Top left: the superimposed diagrams for the diatonic,
harmonic minor, and melodic minor collections, moving outward. Top right: a version that rearranges
each scale’s radial positions while separating major and natural minor. In both of these graphs,
boldface type indicates natural minor, regular type is harmonic minor, and italic is melodic minor.
Bottom: a more compact diagram that does not distinguish the three specific minor-scale forms.

These models reconcile circular voice-leading logic with the insight that
parallel and relative keys are both closely related. Because A minor is
spatially extended, it can be close to C major, G major, D major, and A
major. The first and last are the relative and parallel relationships. The
proximity to G major reflects the single-semitone voice leading that turns a
G major tonic into an A melodic-minor leading tone.8 And though
modulations from D major to A minor are uncommon in D major pieces,
they are quite common in G major: indeed, the modulatory pathway I–V–ii
(G–D–a) is one of the most important in functional tonality. Thus the model
is reasonably accurate analytically, even though it is a two-dimensional
simplification of a vastly more complicated seven-dimensional space.9
It is useful to augment this model with information about the global tonic.
Figure 8.3.4 borrows a strategy from chapter 2, using darker colors to
represent keys whose tonic triads are contained in the C major scale; these
form a stack of fifths ascending from subdominant to mediant, F–C–G–d–a–
e, all clustered in a single region of circular space.10 This in turn shows that
enharmonicism requires chromaticism: because modulation typically uses
efficient scalar voice leadings, there is no obvious sharpward path from iii
(the sharpest of the diatonic keys) to IV (the flattest). Hence a sequence of
diatonic modulations will tend to cancel out, adding and subtracting the
same small collection of sharps and flats, and not requiring enharmonic
respelling. This means that composers who understand key areas diatonically
—limiting themselves to the triads contained in a single diatonic scale—will
not travel in complete circles through key space. It is only once key areas are
conceived chromatically, allowing for sharpward motion from mediant to
subdominant, that enharmonicism becomes commonplace. This is what I
take to be the truth behind Lewin and Cohn’s claim that such modulatory
sequences force us to abandon “diatonically based models”: not that we need
to abandon the basic picture of chords inside scales, but rather that we need
to consider modulatory destinations in chromatic space, rather than limiting
ourselves to keys whose tonics inhabit a single diatonic collection.
Figure 8.3.4. Circular key space, with key areas darkened to reflect proximity to C major.

With this model in hand, we see that the coordinated motion of

Beethoven’s Ninth, rather than being unusual, is in fact necessary for
modulatory sequences to generate repeating patterns of Roman numerals.
Consider Chopin’s D major prelude: here, the sequential units articulate I–
IV–V–[vi or I]–V/V–V progressions that ascend by fifth along the D major
scale, from D major to A major, E minor, B minor, and the dominant of B
minor (Figure 8.3.5).11 As in Beethoven’s Ninth, we have coordination
between levels. Beethoven’s passage is exceptional because each level uses
its basic voice leading, taking a single angular step in the relevant spiral
diagram. Chopin’s looser and more general coordination-between-levels is
commonplace—indeed, essential to modulatory sequences as we know them.
For if chords and keys move differently, then sequential units will move
relative to the underlying scale, leading to different tonal functions and
different Roman numerals.

Figure 8.3.5. Chopin’s D major prelude.

Chopin’s prelude ascends the ladder of fifths to F ♯ major, the point of

maximal distance from the tonic, before sliding chromatically to A7 and
returning to the opening; the repeat is largely faithful, except that the
penultimate chord, vii°7/iii, is reinterpreted as a common-tone diminished
seventh of D. These last progressions illustrate the difference between
chordal and scalar approaches to modulation: the initial key changes exploit
scalar proximity, moving by fifth between closely related key areas for a
very smooth modulatory effect; by contrast, the phrases end with
modulations that exploit the voice-leading properties of chords themselves—
at the end of the first phrase the root of the F ♯ minor chord moves
semitonally to form the G–A–C♯ subset of A7, while at the end of the second
phrase the vii°7/iii resolves chromatically to I. Here it is the chordal subsets
that are close: F♯–A–C♯ to G–A–C♯, and E♯–G♯–B–D to F♯–A–A–D. In A
Geometry of Music, I distinguished “scale-first” from “chord-first”
modulations in twentieth-century music; Chopin’s prelude exploits the same
mechanisms in a more traditional style.

4. Modulatory schemas

As functional tonality develops, its syntax starts to encompass larger and

larger stretches of musical time. In the Renaissance, chords and tonal centers
progress relatively unsystematically, with the nonharmonic system providing
a good portion of the harmonic energy: the oscillation between stability and
tension, “home” and “away,” is largely produced by the motion of non-chord
tones to chord tones. By the late sixteenth century, chord progressions
provide a higher level of motion away from home and back again. But it is
only toward the middle of the seventeenth century that we start to find a
grammar of modulatory destinations that is broadly analogous to its
grammar of chord-to-chord successions (Figure 8.4.1). A related
development, to be discussed in chapter 9, is the increasing use of regular
phrase templates such as “sentence” and “period.”

Figure 8.4.1. A schematic outline of standard modulations in baroque and classical music.

In their details, the grammars of chord and key are mirror images: chords
tend to progress from tonic to secondary destinations to dominant, while
keys move from tonic to dominant to secondary destinations. Chordal norms
are more stable than the modulatory norms: from 1600 to the present we can
find pieces deploying TSDT harmony in recognizably similar ways, even
while patterns of modulation change: for example, classical phrases often
end by tonicizing the dominant, but this modulation is rare in jazz or rock-
and-roll. Modulatory norms are also form-dependent in a way that chord
grammar is not: sonatas are structured as journeys from tonic to foreign keys
and back, while fugues and rondos permit more frequent returns to the tonic.
Beyond these very general conventions are some more specific
modulatory schemas. These typically occur in sonata-form developments,
where the majority of classical-style modulations are found. One of the
simplest is what I call the helicopter drop, a sudden and dramatic turn to a
distant key, followed by a gradual return to the tonic. The first-movement
development of Beethoven’s Pathétique starts by jolting the listener from G
minor to E minor, whereupon it descends by fifths back to the tonic (Figure
8.4.2).12 The first-movement development of the second Razumovsky
quartet is similar, moving through a circle of major-third related minor triads
(g–e♭–b) to arrive at a B minor that feels more distant than it actually is. The
remaining seventy measures pursue a single ascending-step near sequence
(Figure 8.4.3). Particularly interesting are the reinterpretations of the A and
A ♭ chords, marked by asterisks on the example: in both cases, sequential
logic dictates that the chords should be dominants moving up by fourth;
instead, they act as tonics moving down by third (§4.9). These
transformations extend the long sequential climb so that it twice ascends
from A to C before finally breaking free to rise to the tonic. (Note also the
two enharmonic respellings, signifying two loops in scale space and a
recapitulation in G♭♭♭ minor.) The embedding of diverse thematic material
within a single modulatory sequence is characteristic of Beethoven’s
practice.
Figure 8.4.2. The development of Beethoven’s Pathétique sonata, Op. 13, I. Enharmonic
reinterpretation of the diminished seventh takes us to a distant E minor, from which the music
descends by fifths back to a converging cadence and the tonic.

Figure 8.4.3. The development section of Beethoven’s Op. 59, no. 2, I.

Another important schema is what I call a paired-sequence design, a

balanced pair of sequences with the second largely undoing the work of the
first. The pivot is often rhetorically marked, a “point of farthest remove,”
where the music delays for a while—standing on the dominant, introducing a
new theme, or reaching a strong cadence. The first-movement development
of Haydn’s Symphony no. 62 is typical, beginning with an ascending
arpeggio that moves through the harmonies A7–F ♯ 7–[D]–b–G–E7–C ♯
(Figure 8.4.4). This is a near sequence by descending third, the fifth-
progression F♯ 7–b omitting a sequential step. The point of farthest remove
features an augmented sixth, a standing-on-the-dominant and a brief caesura;
a descending arpeggio then takes us down the ladder of fifths, C♯7–F♯7–B7–
E7–A7, discharging harmonic tension as the recapitulation arrives. (Note the
nice correspondence between key and motive, an ascending arpeggio for the
first sequence and a descending arpeggio for the second.) Compared to the
helicopter drop, the departure and return are more equal and balanced: where
the former takes a large leap on circular scale space, followed by a series of
short steps back, the latter involves a pair of broadly comparable motions.

Figure 8.4.4. A paired-sequence design in the development of Haydn’s Symphony no. 62, I.

The development of the first movement of Beethoven’s Op. 2, no. 1, uses

the same basic pattern (Figure 8.4.5). Here the initial sequence ascends by
step from the relative major A♭ to B♭ minor to C minor, both modulations
using the i–Ger65/ii–V/ii progression even while deploying different themes.
Instead of pausing on the antipodal point, Beethoven immediately retraces
his modulatory steps: we begin in C minor, still referencing the second
theme, and descend by fifths to A ♭ whereupon the modulation stops while
the chords continue, the music’s thematic content liquidated into generic
tonal material. If we imagine a pair of spiral diagrams, as in Figure P8.3, the
chords continue to move along their circle while the keys stop. The
introduction of E ♮ brings a standing-on-the-dominant that features a
characteristic contrary-motion pattern (§10.1).
Figure 8.4.5. A paired-sequence design in the development of Beethoven’s Op. 2, no. 1, I.
 5. Up and down the ladder

The most elaborate modulatory schema is a version of the paired-sequence

design that I call the up-and-down-the-ladder. Originating in the baroque,
this schema pairs ascending- and descending-fifth sequences: the ascending
fifths increase the tension and often employ the diatonic third’s ascending
basic voice leading; the descending fifths release the tension and return us to
equilibrium. This pattern becomes a central classical-era developmental
schema, surviving into the nineteenth century as a common modulatory
template.13 The “up-then-down” organization can be understood as a
byproduct of the linear ordering of diatonic keys, arranged in fifths from flat
to sharp (IV–I–V–ii–vi–iii). The keys at either end are not close by any
definition, and composers rarely modulate directly from one to the other.
Thus diatonic tonal journeys are typically bounded, requiring a there-and-
back-again structure, with up-then-down (or tension-then-relaxation) being
energetically preferable to down-then-up (or relaxation-then-tension).
In the classical period the schema most often appears in sonata-form
developments, where it acquires a few additional characteristics:

1. An early turn to the dominant’s parallel minor.

2. A two-step ascending-fifth sequence moving from v to ii to vi.
3. Continuation of the ascending fifths to a tonicization of V/vi, often
utilizing the augmented sixth and sometimes preceded by an 8̂–7̂–♮6̂–
♭6̂–5̂ bass descent.
4. An arrival on vi, the point of farthest remove for the development as a
whole; this can be either a thematic passage, a cadence, or standing on
the dominant.
5. Return to the tonic key in one of several ways:
a. a descending-fifth sequence (i.e., moving down the ladder);
b. a quick, surprising turn to the global dominant (e.g., V/vi–vi–V7);
c. direct motion from V/vi to the tonic.14

The ascending fifths ratchet up the tension, their sonic signature a sequence
of Sturm-und-Drang i–V7/V progressions; this leads to an energetic peak
that is discharged by more conventional harmonic movement. In larger
pieces the schema sometimes occupies only one part of the development.
 The appearance of V/vi as a point of farthest remove has sometimes been
identified as a characteristic of the early eighteenth century; however, the
up-and-down-the-ladder schema appears throughout classical music,
particularly in the relatively simple developments of Mozart’s piano
sonatas.15 Figures 8.5.1–3 provide some shorter examples from the sonatas
of Haydn, Mozart, and Beethoven. The development of the first movement
of K.332 is notable in that Mozart alters an expositional sequence to
conform to the ascending-fifth schema—an unusual rewriting that testifies to
the idiom’s gravitational pull (Figure 8.5.4). Figure 8.5.5 lists a few other
notable examples; we will encounter several more in chapter 10.

Figure 8.5.1. The development of the first movement of Haydn’s Divertimento in C major,
Hob.XVI:1.
Figure 8.5.2. The development of the first movement of Mozart’s K.332, I.
Figure 8.5.3. The development of the first movement of Beethoven’s Op. 49, no. 2.

Figure 8.5.4. In the development of K.332, I, Mozart changes a descending-fifth sequence into an
ascending-fifth sequence.
Figure 8.5.5. Some other developments that use the up-and-down-the-ladder schema.

Though largely indigenous to development sections, the up-and-down-the-

ladder can occasionally appear in transitions as well. The exposition of the
first movement of Beethoven’s C major piano sonata Op. 2, no. 3, is in three
parts (Hepokoski and Darcy’s “trimodular block”): the first medial caesura
leads to G minor and the music in Figure 8.5.6, modulating up the circle of
fifths to D minor and A minor, before returning down the ladder (a–D–g),
ending with an augmented sixth and standing on the dominant of G. It is
interesting to imagine this as displaced developmental rhetoric, a familiar
template that usually though not always occurs elsewhere in the form.
Similar moments occur in the recapitulation of Beethoven’s Op. 18, no. 2, I,
and the transition of Op. 59, no. 2, I.
Figure 8.5.6. The middle section of the exposition of Beethoven’s Op. 2, no. 3, I. Each unit of the
initial sequence expands the fauxbourdon ROTO, i–v6–iv6–i64–ii°6 becoming i–V2/V–V6–V65/iv–
iv6–i64–V6/V.

6. Modal homogenization and scalar voice leading

This theory of modulation can also help elucidate the phenomenon of modal
homogenization, by which the various Renaissance modes came to take on
the characteristics of ionian. This process maps ionian-mode procedures into
a new domain, implying something like a voice leading between scalar
collections—a “modulation,” as it were, taking place over the span of
decades rather than seconds. Our two notions of key distance correspond to
the two ways in which this can happen: in a scale-degree-preserving
transformation, the modal tonic stays fixed while other degrees change
position—as when the seventh degree is raised from a modal subtonic to a
functional leading tone (Figure 8.6.1). In a scale-degree-changing
transformation, we have a fixed vocabulary of pitches that change meanings,
centricity shifting from one note to another. This necessarily involves
ambiguity during the transition, the multiplicity of interpretive possibilities
reflecting a gradual change of scale-degree roles.16

Figure 8.6.1. Scale-degree-preserving and scale-degree-changing models of modal homogenization

for mixolydian. In the first, G mixolydian becomes G ionian; in the second it becomes C ionian.

Ambiguity about tonal centers is common in music that admits a wide

variety of modes, and particularly when musical styles are changing. Satie’s
third “Gnossienne” is a good example, its main motto potentially either iv–i
in E minor or i–v in a chromaticized A minor (Figure 8.6.2). Context
suggests it is meant to be heard in A minor, though the pull of the harmonic-
minor scale is strong enough to allow the more old-fashioned hearing. By
contrast, when Stravinsky uses the same mode at the start of The Firebird’s
“Infernal Dance,” the tonal center is unmistakable. In much the same way,
Laurie Anderson’s “O Superman” can be heard in A♭ major with contrasting
C minor; or in C minor with contrasting A ♭ . Here the ambiguity is not a
product of stylistic innovation, but of the role of rhythm, phrasing, and
expectation in producing popular-music centricity (§2.5).17

Figure 8.6.2. Satie’s third Gnossienne can be heard in either A minor or E minor.

Seventeenth-century modal homogenization is also bound up with a

deeper shift from what I have termed absolute musical space, in which
particular chords have their own tonic-independent tendencies, to the tonic-
relative (or “Copernican”) perspective of our own time. In Palestrina’s style,
for example, there are five tonal centers available in the white-note scale, C–
G–d–a–(e), with the last allowing only the weaker phrygian cadence.
Secondary emphasis groups these into modally matched pairs, affiliating C
with G (each solmized as “ut” in their respective hexachords) and D with A
(each solmized as “re”), with the E (or “mi”) tonality in its own category
(emphasizing A as a secondary center).18 Signs of this affiliation can be seen
in a wide variety of statistical features, including chord distributions and
cadential destinations (Figures 8.6.3–4).19 Thus we have a pair of authentic
modes emphasizing the fifth as a secondary scale degree, with the remainder
emphasizing the fourth.20 In other words, the major triads on C and G tend
to affiliate regardless of which is tonic: mixolydian G emphasizes C as
subdominant where ionian C emphasizes G as its dominant.
Figure 8.6.3. Common complete triads in Palestrina’s masses. In ionian and dorian, V is substantially
more common than IV; in mixolydian/aeolian, IV is somewhat more common than V. Modal identity
was determined by computer.
Figure 8.6.4. Cadential destinations in Palestrina’s zero-flat modes as determined by computer. Ionian
and dorian have many dominant cadences and relatively few subdominant cadences; mixolydian and
aeolian have fewer dominant cadences and more subdominant cadences. Note the general scarcity of
cadences on E and F.

By contrast, the major-minor system involves a double relativity: first, all

major keys, and all minor keys, emphasize the same secondary degrees;
second, the two modes are very similar to one another (e.g., both major and
minor emphasize the dominant). Gioseffo Zarlino foreshadowed this more
homogeneous conception when he argued that every Renaissance mode
should emphasize the fifth and third above its tonic, a view that was less an
accurate description of his contemporaries’ practice than a programmatic
statement about how music could be.21 This led him to suggest that
purportedly phrygian pieces had hypoaeolian elements: since a pure
phrygian would emphasize the fifth above the tonic, an E mode that
emphasized A necessarily had an aeolian quality.22 Zarlino’s analysis
suggests that late Renaissance modality may have been ambiguous even for
enculturated listeners: a listener who accepted plagal patterns of emphasis
might hear E-music-that-emphasized-A as being in E, whereas a listener who
demanded authentic emphasis might hear the same piece in A, even if it
happened to end on E.
With these issues in mind, let us consider the modes individually.
(a) Mixolydian. Since it is just a single accidental away from ionian, the
mixolydian mode can readily incorporate ionian routines by raising its
seventh degree; but since it tends to emphasize I and IV, it can also adopt
ionian qualities by exchanging primary and secondary tonal centers,
gradually evolving from “G music that also emphasizes C” to “C music that
emphasizes G.”23 These two possibilities can both be heard in the opening of
Scheidt’s 1624 setting of the mixolydian “Gelobet seist du, Jesu Christe”
(SSWV 135, Figure 8.6.5). The end of the first verse has a strong C major
quality, beginning an eighteen-measure passage with no F♯; the second verse
presents series of harmonic cycles in G major. It is likely that Scheidt
conceptualized the entire piece in G mixolydian; yet it contains substantial
passages that are reasonably heard in C major.24

Figure 8.6.5. The last phrases of the first two verses of Scheidt’s 1624 “Gelobet siest du, Jesu Christ.”
The first-verse phrase is easily heard in C, the second is more strongly in G.
 As a general rule, mixolydian tends to move toward ionian by way of
scale-degree-preserving mappings. Marenzio’s 1586 three-voice villanelle
“Vorria parlare e dire” is representative: F ♯ outnumbers F ♮ almost 5-to-1,
with the mixolydian seventh appearing only in mm. 4–5 (Figure 8.6.6). The
harmonic progressions can mostly be parsed into functionally plausible
cycles, though there are a few nonfunctional moments (e.g., IV–vi–V in mm.
2–3 and I–ii–I6 in mm. 13–14, the shift from D to A in m. 11). The most
striking is the V– ♭ VII–V in mm. 4–6, with its staggered parallelism
(reminiscent, perhaps, of pop-music power chords).25 Here we have a
mixolydian that could be described as substantially ionian, but with a strong
subtonic emphasis.

Figure 8.6.6. Marenzio’s “Vorria parlare e dire” (1586).

 (b) Phrygian. At the other end of the spectrum is phrygian, maximally
distant from ionian and the only mode not permitting an authentic cadence.
If we go back far enough, we can find phrygian music with a strong
secondary emphasis on the fifth scale degree, just as Zarlino described. The
phrygian of Hildegard of Bingen’s twelfth-century “Laus Trinitati” is not
entirely dissimilar from that of Stravinsky’s Symphony of Psalms, with both
emphasizing the tonic/fifth pair (Figure 8.6.7).26 Here there is little danger of
mistaking E for any sort of dominant; it is a genuine phrygian that
emphasizes its first and fifth scale degrees in a way that is analogous to
ionian.

Figure 8.6.7. Hildegard of Bingen’s “Laus Trinitati.”

By contrast, the secondary A emphasis of much phrygian-mode music,

from the mid-sixteenth century on, creates an ambiguity between phrygian
and hypoaeolian. Consider Johann Schein’s 1627 setting of “Aus tiefer
Noth,” shown as Figure 8.6.8.27 Our sense that the E has a dominant quality
is partly a matter of exposure to later music, but it is also a feature of the
music itself. Recall that dominants are typically constrained toward the
future while tonics are constrained toward the past (§6.4). Now ask how
Schein’s E chords behave: of the four E major chords in the piece, all but the
last proceed directly to A minor (first inversion in m. 1, root position in mm.
4 and 8). Furthermore, the E chord that starts the piece is audibly similar to
the G♯°6 chord that occurs at the end of that measure, a seeming dominant of
A. The only E minor chords are found in sequential passages, pairing with A
minor as ascending-fifth units: a–e–C–G in m. 6 (perhaps suggesting tonic-
dominant pairs ascending by third) and the (G)–a–e–F–C–d Pachelbel
sequence in the final phrase (perhaps suggesting tonic-dominant pairs with
deceptive resolutions). Meanwhile the piece’s E chords are approached
variously (from A minor and D minor), but never by a phrygian cadence.
The A minor chord is in many ways its mirror image: typically approached
by E (or g♯°6) but progressing in a number of different ways.
Figure 8.6.8. Johann Schein’s “Aus tiefer Noth.”

All of which suggests that even if E is being conceived as a tonic, it may

not be behaving as one: like a classical dominant, these “phrygian tonics”
are constrained in how they move, while the purportedly nontonic A minor
chord is constrained in how it is approached. In its statistical behavior, the
phrygian tonic is evolving into a minor-key dominant (Figure 8.6.9). In this
respect, phrygian is the opposite of mixolydian, evolving toward ionian by a
change of scale-degree roles.

Figure 8.6.9. Scale-degree-preserving and scale-degree-changing models of modal homogenization

for phrygian.
 (c) Aeolian and dorian. Between these two poles lie the aeolian and
dorian modes, which are somewhat slow to adopt the full suite of ionian-
mode conventions. As with mixolydian, the transition typically occurs in a
scale-degree-preserving way, with the aeolian and dorian tonics evolving
into minor-key tonics. But this process requires more extensive adjustments,
and the emphasis on the secondary major triads often pulls the music away
from its ostensible center.
Figure 8.6.10 shows a late modal piece, the Introit from Schütz’s St.
Matthew Passion (ca. 1665). This ostensibly G-based music emphasizes
major triads on C, F and B ♭ , all but effacing the sense of G as a genuine
center. After the initial i–iv–I progression, the music tonicizes C; this
initially leads me to hear C–F–B♭–F as I–IV–♭VII–IV in C mixolydian, with

the C centricity eventually overwritten by the cadence on B ♭ . The quasi-

palindromic piece then reverses direction with an ascending-fifth Chase
schema rising past G to D. (As in much Renaissance music, melodic
direction correlates with harmonic progression, ascending melodic steps
harmonized with descending fifths, while descending melodic steps are
harmonized with ascending fifths—here forming a phrasal up-and-down-the-
ladder analogous to those of classical developments.) The sequence’s final
fifth-progression, g–d, dovetails with a descending 6–5 that is not part of the
palindrome (Figure 8.6.11).28 All in all it is a wonderfully compressed
exemplar of baroque modality, elusive and familiar, modal and harmonic,
major and minor. Central to its mystery is the sense that the forces of minor-
mode centricity are weak, liable to being dislodged by the scale’s major
triads.
Figure 8.6.10. The Introit from Schütz’s St. Matthew Passion.

Figure 8.6.11. The palindrome hidden in Schütz’s piece.

This issue is subtle enough to deserve clarification. I am not arguing that

aeolian or dorian music is “really” in major, or that minor-mode centricity is
somehow illusory: Renaissance dorian and aeolian often place a perceptible
emphasis on minor-mode primary triads—i, iv, and v and their major
parallels—which are typically more prevalent than secondary triads such as
III, VII, and VI. Nevertheless, secondary major triads play a more significant
role than in later functional music, exerting a tonic-independent gravitational
pull (Figure 8.6.12). Indeed, for many composers, the mediant triad remains
the primary minor-mode harmonization of bass 3̂ until well into the
seventeenth century, being displaced by i6 long after I6 has become a major-
mode default (Figure 8.6.13); and of course, a fondness for the mediant is
one of the main modal asymmetries remaining in mature functionality
(§7.1). Ears accustomed to functional tonality will naturally tend to hear
modal music as “emphasizing” these secondary major triads, whereas it
would be more accurate to use a double negative: it deemphasizes them less
than J. S. Bach. Our sense of the music “digressing into major” is a
reflection of our having acclimated to the scarcity of secondary major triads
in later music.
Figure 8.6.12. Diatonic triads in the minor-mode passages of Bach’s chorales and in Palestrina’s mass
Spem in alium. Sonorities such as VI, III, VII are significantly more common in Palestrina than in
Bach.
Figure 8.6.13. The relative frequency of i6 and III in minor mode.

At the same time, there are some remarkably early examples of minor-
mode functionality. Morley’s “Leave, alas, this tormenting” is composed out
of short harmonic cycles which are unusual mainly in that they modulate
somewhat faster than we might expect (Figure 8.6.14). We begin with very
clear G minor, but immediately move to a stereotypical cadence in C minor,
only to be wrenched immediately into D minor before settling back into G
minor for eighteen measures. The III chord on the second system is a slight
but noteworthy inclination toward major. There are a few other moments
where Morley uses major triads in not-completely-functional ways (VI6–III6
on the third system, perhaps VII instead of v6 shortly thereafter, VII6 on the
bottom system), but these are exceptional; overall the music is strongly
functional, particularly when compared to pieces like Schütz’s St. Matthew
Passion.

Figure 8.6.14. The harmonies in Morley’s 1600 “Leave, Alas, This Tormenting.”

Future theorists will digitize and analyze a much larger quantity of

sixteenth- and seventeenth-century music, allowing them to tell the story of
modal homogenization in fine-grained detail. But I think the outlines are
already clear: the process of homogenization is comparatively
straightforward in the case of mixolydian, but more crooked and halting in
the case of the remaining modes. For a time, it is almost as if the composers
choose between major and modal rather than major and minor. During the
transition, we find functional and modal techniques existing side-by-side,
producing fascinating music whose basic principles are liquid and negotiable
—with highly functional pieces like Morley’s antedating highly modal
pieces like Schütz’s by the better part of a century. Historians may once have
looked upon the paratonal century from 1550 to 1650 as a time of confusion,
a rudimentary groping toward the absolute truths of classical harmony; now
it seems like a flexible, nondogmatic, and open period in which multiple
syntaxes overlap. Pieces differ primarily in the balance between the two
forces, and not in the fact of pluralism as such.

7. Generalized set theory

Grieg’s Elegie begins with an eight-bar parallel period showcasing the

composer’s fondness for chromatically descending bass lines (Figure 8.7.1).
The ear hears virtually the same music presented over B minor and what
sounds like G minor—harmonically distant yet surprisingly logical. The G
minor is a musical pun, for while B harmonic minor does not contain G
minor as a stack of thirds, it does contain its enharmonic equivalent as a
stack of scalar fourths, A♯–D–G. Grieg’s music thus manages to evoke two
incompatible hearings, a chromatic hearing according to which G minor and
B minor are related by transposition (reflecting the fact that they sound
alike), and a harmonic-minor hearing according to which they are not
(reflecting the fact that the collection of notes A ♯ –D–G sounds like it
belongs to B harmonic minor). It is not simply that we are forced to choose
between these two different interpretations but rather that both are equally
important for our understanding. Like a linguistic pun, it requires us to
process two incompatible meanings at one and the same time.29
Figure 8.7.1. The opening of Grieg’s Elegie, Op. 47, no. 7.

Grieg’s piece, like Satie’s third Gnossienne, represents an early example

of scalar exploration, treating nondiatonic scales as objects in their own right
rather than mere variants of the diatonic. I have argued that twentieth-
century music generalizes the traditional three-scale system to include the
scales in Figure 8.7.2. Four of these are the most-even seven-note collections
in twelve-tone equal-temperament: the diatonic, acoustic (or melodic-minor
ascending), harmonic minor, and harmonic major. Three are symmetrical:
the eight-note octatonic and the six-note whole-tone and hexatonic. These
scales are interesting for a number of reasons. They contain all and only the
“nonchromatic” sets (i.e., those that do not contain a chromatic cluster),
providing a way station between traditional diatonicism and complete
atonality. The scales are also linked by interesting voice-leading
relationships (Figure 8.7.3). Its seven-note collections have a systematic
structure when expressed on the circle of fifths, reflecting both their
intervallic makeup and the “heights” of their modes (Figures 8.7.4–5). And
the scales can be integrated with functional harmony via the principles of
“chord-scale compatibility” (Figure 8.7.6). We should not be surprised to
find these scales reappearing in a wide range of styles from impressionism
and jazz to minimalism and postminimalism.
Figure 8.7.2. Seven scales common in twentieth-century music. Solid arrows mean scales of that type
can be linked by single-semitone voice leading. Dashed arrows mean scales of that type share all but
one of the notes in the smaller collection.
Figure 8.7.3. The spiral diagrams for the diatonic, acoustic, and harmonic scales (bold, italic, and
plain text from center outward), with the symmetrical scales on the outside. Lines show single-
semitone voice leadings among the seven-note collections. Below, the single-semitone voice leadings
represented on a cube.

Figure 8.7.4. This table can be read in two ways. First, each column shows the notes of the column-
label scale as they appear on the circle of fifths: the diatonic scale is a compact stack of seven fifths,
the acoustic scale has a compact core of five fifths (a pentatonic scale) with two outliers, and so on.
Second, the table records the “brightness” of those scales’ modes, expressed as the sum of its intervals
above the tonic pitch; here each cell labels the tonic of a mode of the column-label scale, whose
“height” is given by the row label. See Figure 8.7.5.
Figure 8.7.5. Mode height as the sum of the intervals above the tonic.

Figure 8.7.6. The principles of chord-scale compatibility associate chords with scales containing
them; here, four common scales that accompany G–B–F, functioning as G7.

The spiral diagrams allow us to expand our horizons beyond this

particular collection, showing us the voice-leading possibilities of any chord
in any scale, and any scale in chromatic space. The diagrams also allow us to
consider more complex collectional hierarchies: thus, rather than using a ten-
or eleven-note scale directly (which often leads to chromatic saturation), we
can consider five-, six-, or seven-note collections moving inside these larger
collections, their chromatic intervals subtly distorting as they move around
the space. Figure 8.7.7 contains a triply nested chord progression I used in a
piece called Wheels within Wheels: the largest wheel (lowest layer) contains
seven-note scales moving inside an eleven-note scale (all the notes except
F♯); the middle wheel contains six-note chords moving inside those seven-
note scales; while the smallest wheel has four-note chords moving inside
those six-note scales. The music presents an ever-changing sequence whose
structure is aurally discernable even while remaining unpredictable. This
arrangement generalizes the nested hierarchy of Beethoven’s Ninth to an
additional layer, using three basic voice leadings simultaneously.

Figure 8.7.7. A triply nested sequence from Wheels within Wheels in which a seven-note scale
(bottom) iterates its ascending basic voice leading inside the eleven-note scale lacking F♯ (T8t–5).
Inside this scale, a six-note scale (middle) moves according to its ascending basic voice leading (T6t–
5); inside this, a four-note scale (top) moves according to the voice leading T3t–2. Chords with open
noteheads are generalizations of the melodic minor scale as discussed in the text.

Beyond that, the spiral diagrams reveal analogies between different chord-
and-scale domains. While the full picture is too complex to explain here, it is
worth surveying briefly. Suppose we have a spiral diagram with just one
chord at each angular position—that is to say, an n-note chord in an o-note
scale, with n not dividing o and the two numbers not sharing any common
factor: 3 in 7, 7 in 10, 5 in 12, and so on. In that case, the most-even n-note
chord will be a near interval-cycle, a circular ordering of notes all but one of
which are the same size, with the unusual interval one scale step larger or
smaller than the others. For example, the diatonic seventh chord E–G–B–D
is a near interval cycle, a stack of three two-step intervals (E–G, G–B, B–D)
with the fourth interval being just one step smaller (D–E, a step rather than a
third). Similarly, the diatonic scale is a stack of seven-semitone perfect fifths
(F–C, C–G, G–D, etc.) with the final interval (B–F) again just a chromatic
step too small. In this sense the seventh chord E–G–B–D is the structural
analogue of the C diatonic collection. Because both are near interval cycles,
we can form a generalized “circle of fifths” by shifting the position of the
unusual interval: with the diatonic seventh chord, we move D down by step
to C, connecting E–G–B–D to E–G–B–C, with the C major collection we
move B down by semitone to B♭, connecting C diatonic and F diatonic; this
is the descending basic voice leading for each collection. Thus we have both
harmonic and contrapuntal analogies between E–G–B–D and C diatonic:
harmonically, both are near interval cycles, contrapuntally both are linked by
a generalized circle of fifths.
To extend this analogy to other collections, we exploit the fact that we can
generate nearly even chords by scrambling the voice leadings on the
generalized circle of fifths.30 For example, the melodic minor scale is the
second-most even seven-note chord in twelve-tone equal temperament; it
can be generated by applying the voice leading E→E ♭ , which links the F

major scale to the B♭ major scale, one step too early on the circle of fifths:
that is, to C major rather than F major (Figure 8.7.8). The stack of fourths
D–E–G–A is the second-most even diatonic tetrachord; it can be generated
by applying the voice leading B→A one step too early on the circle of
diatonic thirds, to the E diatonic seventh chord rather than to the C major
seventh chord. In exactly the same way, we can obtain the third-most-even
seven-note scale, the harmonic major scale, by applying the voice leading
A→A ♭ two steps too early (i.e., once again to C major, rather than to B ♭
major).31 This procedure gives us a generalized recipe for producing chordal
analogues in different chord-and-scale domains. Once again, these
“analogues” will be similar both in their internal structure and their
contrapuntal relationships. The second-most even chord, for example, will
be “gapped” when expressed in terms of the interval that characterizes the
maximally even chord (Figure 8.7.9). The third-most even sonority will be
an inversionally related pair forming a doubly gapped stack of the
characteristic interval. And all of these objects will be linked by an
analogous system of single-step voice leadings. Thus one and the same
description can be applied to familiar seven-note scales, three- and four-note
diatonic collections, and countless other collections as well—a remarkable
system of analogies linking different chord-in-scale domains (Figures
8.7.10–12).

Figure 8.7.8. Scrambling the generalized circle of fifths to obtain analogous chords in different
musical domains.
Figure 8.7.9. When a spiral diagram has just one chord at every angular position, the four most-even
chords will be linked by the same single-step voice-leading relationships and will have the same
structure. Chord 1 is the most-even chord with a, b, c, connected by the basic voice leading (“the
generalized circle of fifths”). Chord 2 is the second most even chord, with its basic voice leading
linking 2a to 2b, and so on. Chord 3 comes in two inversionally related variants, 3a and 3A. This is the
abstract pattern shared by Figures 8.7.3–4 and 8.7.10–12.
Figure 8.7.10. The most-even triads in the diatonic scale.
Figure 8.7.11. The most-even four-note chords in the diatonic scale.
Figure 8.7.12. The most-even six-note chords in the eleven-note scale containing all the standard
chromatic pitch classes except B. For legibility, the circles have been straightened into lines.

These relationships hint at a music theory that has yet to be written, a

generalized set theory going beyond taxonomy to illuminate abstract
structures reappearing across different chord-and-scale domains. We have
not yet taught ourselves how to think in this way, much less to explore the
aesthetic significance of these ideas. But one way in which we can take
ownership of the functional tradition is to ask how its techniques might be
adapted to new harmonic realms. Earlier composers made knowing and
intelligent use of the quadruple hierarchy, moving voice in chord, chord in
scale, and scale in chromatic aggregate. Inspired by the Prime Directive, we
can adapt their techniques to arbitrary collections, using the spiral diagrams
to chart interesting paths through a new world of musical possibility.

1
Lewin 1984 and McCreless 1996; two familiar examples are the end of the scherzo of
Beethoven’s Hammerklavier Op. 106 and Miles Davis’s “So What.”
2
These pairings are the primary theoretical objects studied in Brinkman 1986, Agmon 1989, and
Rings 2011. These earlier theorists do not associate them with abstract musical voices as I do here; for
Agmon they are formal mathematical objects, and for Rings they are subjective perceptions.
3
Each finger plays a note that has a scale-degree role, corresponding to the three basic elements of
the model.
4
Yust 2018 pursues similar considerations using different mathematics; Tymoczko and Yust 2019
detail the close connections between voice-leading spaces and Yust’s Fourier models.
5
In isolation we cannot distinguish between C major and D ♭ ♭ major. But we can distinguish a

modulation taking C major to C major from one taking C major to D ♭ ♭ major. We could of course

rewrite these as modulating from D ♭ ♭ major to D ♭ ♭ major and from D ♭ ♭ major to E ♭ ♭ ♭ ♭ major,
respectively.
6
The words here come from Cohn 1996 and echo the sentiment in Lewin 1984, a very challenging
and obscure paper; Cohn 2012, pp. 2–3, reiterates the view. Waltham-Smith 2018 also seems to
endorse a similar perspective. Klumpenhouwer 2011 argues that we can avoid any sense of paradox if
we simply label keys chromatically.
7
As a keyboardist, J. S. Bach daily confronted an instrument in which the diatonic scale is a subset
of the twelve chromatic notes, and indeed celebrated this embedding in his most famous composition.
Were E ♭ minor and D ♯ minor different places then there would be little sense in the E ♭ minor/D ♯
minor prelude-and-fugue pair in Book I of The Well-Tempered Clavier.
8
Nowadays we tend to think that Kellner’s model puts the minor keys in the correct position, but
Heinichen’s model, which highlights the proximity of G major and A minor, has some benefits.
9
These models are consistent with, but more intuitive than, the three-dimensional lattice used in
§7.4 of Tymoczko 2011a. Some readers may prefer even simpler concentric models like Figure 8.1.5,
which is fine so long as one remembers that minor keys are “smeared.”
10
Compare Agmon 2013, building on observations made by the eighteenth-century musician
Joseph Riepel.
11
Note the IV–V7–I harmonization of the ascending basic voice leading for diatonic thirds (§3.1).
12
Other examples include the first-movement developments of Beethoven’s Op. 7, no. 1 (deceptive
cadence to ♭VI followed by a descending-third sequence to ♭II, skipping ahead one sequential step to
V), Op. 31, no. 2 (§10.3), and Op. 31, no. 3 (shift to VI followed by a descending fifth movement to
IV, which immediately ascends to the ii6 that starts the recapitulation.
13
Examples include The Well-Tempered Clavier’s first C ♯ and D major preludes, its second C
major fugue, Bach’s F major two-part invention, and the opening of the G major cello suite. Romantic
examples include Chopin’s D major prelude, discussed earlier, and the second song of Schumann’s
Dichterliebe.
 14
The nineteenth-century theorist Ernst Friedrich Richter cites Mozart’s K.533, I, an “up-and-
down-the-ladder,” as a paradigmatic development; his analysis closely tracks my description of the
schema (Richter 1852, pp. 32–34). Richter’s discussion is cited in Beach 1983, which provides a
Schenkerian discussion of developments that emphasize or close on V/vi. This option creates
something like the opposite of the helicopter drop: a gradual journey away from the tonic followed by
a quick and surprising return.
15
See Rosen 1988. In the online supplement to this book, I summarize the developments of all of
Mozart’s piano sonatas.
16
This ambiguity is facilitated by the fact that centricity in Renaissance polyphony is often fairly
subtle (Powers 1958).
17
On Shady Grove, Jerry Garcia and David Grisman perform the song “Wind and Rain” in a way
that I hear as strongly mixolydian; Crooked Still, on Crooked Still Live, plays a very similar version—
almost a transposition of Garcia and Grisman’s—that I nevertheless hear in ionian. Brittany Haas, who
played on Crooked Still’s recording, told me she initially heard the piece in mixolydian even while the
other musicians heard it in ionian.
18
These affiliations may reflect the presence of structurally similar pentachords, with the C
diatonic scale containing an 02457 pentachord on C and G, and an 02357 pentachord on D and A
(Judd 1992). These pairs are broadly consistent with those noted by Meier 1988 and Ceulemans 2017.
19
In determining the mode of a mass movement, I looked at key signature, final chord, and most
frequent chord. Thus my “mixolydian” pieces had no key signature with G major as the final and most
frequent chord. The frequency constraint was meant to eliminate ionian pieces that happened to end on
G. I also checked mode assignments by hand for plausibility.
20
The subdominant emphasis in mixolydian is noted by Kerman (1981, pp. 69–70) and Jeppesen
([1931] 1939, pp. 81–82); the subdominant emphasis in phrygian is noted by Zarlino ([1558] 1983,
chapters 20 and 30). Jeppesen (1931) 1939, pp. 59–82, is largely consistent with my own account.
Meier (1988, p. 135) notes Palestrina’s tendency to cadence on C even in mode 7 (mixolydian rather
than hypomixolydian).
21
See Christensen 1993b, who also emphasizes Zarlino’s classification in terms of the quality of
the third above the root.
22
Zarlino (1558) 1983, chapters 20 and 30.
23
Dorian is two accidentals away (F→F ♯ , C→C ♯ ), while aeolian is three. Phrygian is four
semitones from ionian if we fix the tonic (i.e., raise 2̂, 3̂, 6̂, and 7̂), but only three if we allow the tonic
to move (i.e., lower 1̂, 4̂, and 5̂).
24
Bach’s settings of the melody are similarly ambivalent: in BWV 91.6 (Riemenschneider 51) the
opening melodic note is a tonic; whereas in BWV248.28, it is a dominant (though I hear it gradually
acquiring tonicity as the chorale progresses). See Burns 1993 and 1995 for more.
25
Einstein 1948, chapter 8, contains numerous examples of early canzone villanesca with
extensive passages in parallel root-position triads; Einstein notes that these fifths “almost completely”
disappear in Marenzio (p. 586), though one remains in Figure 8.6.6.
26
The need for a term (like “phrygian mode”) to describe the similarities between Hildegard and
Stravinsky stands as a point against Powers’s suggestion that we abandon the term “mode” as a
musical descriptor, replacing it instead with “tonal type.” Hildegard and Stravinsky do not use the
same “tonal type,” but both their pieces can reasonably be said to be in phrygian.
27
Schein’s setting is likely the model for Bach’s.
28
The bass/alto canon begins in mm. 11–12 and continues across the sequential boundary.
Overlapping ascending fifths and the “descending 6–5” also occur in the Kyrie of Byrd’s Mass in
Three Voices (§3.5).
 29
In other words, the piece exploits the fact that the harmonic minor scale contains collections
related by chromatic but not scalar transposition, a property often referred to as “ambiguity”
(Rothenberg 1978 and Balzano 1980).
30
Tymoczko 2004 and 2011a. Junod et al. (2009) note some of these relationships in the 7-in-12
context.
31
To obtain the inversion of the harmonic minor scale, we do the same in the sharpward direction,
applying the displacement G→G♯ not to D major but to C major.
 Prelude
Hearing and Hearing-As

Music bumps against the limits of human possibility. It is easy to write

rhythms that performers cannot play, and possible to play rhythms that
listeners cannot understand. Composers regularly devise pitch systems that
elude even the most sensitive ear. The history of music is a tug-of-war
between the aspirational force of the musical imagination and the
gravitational pull of practical limitation, energized by the fact that audiences
generally do not like hard work. The result is genuine confusion about what
we hear: we may think we hear the organization in a complex piece of total
serialism, or the relationships described by set theory, or a sense of long-
term resolution in a sonata recapitulation, even while psychological
experiments suggest we are wrong. Music is unique among the arts, and rare
among human perceptual situations generally, in demanding skepticism
about first-person testimony.
Or perhaps it is not so unique. The history of science shows that
observations can be contaminated by expectations, and it is now standard
practice to ensure that experimentalists do not know what they are expected
to see. Judges and lawyers have come to understand that eyewitness
testimony can be surprisingly unreliable. For centuries, belief in the
supernatural led intelligent people to think they were witnessing miracles
and acts of witchcraft. An extensive literature on “implicit bias” suggests we
can be prejudiced even when we think we are not. Many oenophiles believe
they can distinguish cheap from expensive wine while failing to do so in
controlled circumstances. Perhaps the challenges of music perception are
just a specific manifestation of the more general problem of self-
knowledge.1
Here it is important to distinguish two issues. The first is that we can
sometimes think we are sensitive to musical features even when we
demonstrably are not—we might, for example, think we can hear sonatas
returning to the tonic key even while being vulnerable to clever tricks (for
instance, rewriting a piece so that it ends in a foreign key). This problem is
relatively tractable: to find out what I can hear, you can just run an
experiment. The second and more difficult issue is that we are sometimes
capable of passing empirical tests even without being able to hear the
relationship as such. In this case, experiments will not tell us what we need
to know.
Imagine a student with very good absolute pitch who is insensitive to
intervallic quality: asked to identify the interval D–F♯, they count on their

fingers—“let’s see, D to E♭, E♭ to E, E to F, F to F♯, that’s four semitones so

. . . um . . . a major third!” This student might pass simple or even advanced
ear-training tests, identifying intervals, chords, and progressions with a high
degree of accuracy. But we may legitimately feel they are missing
something. (They might encourage our suspicion by saying things like “I
don’t really experience a melody in one key as being similar to its
transposition into another.”) Rather than being directly sensitive to the
inherent sound of intervals, the way they directly perceive colors or smells,
the imaginary student calculates interval on the basis of absolute-pitch
information.2
Contrast this sort of calculational experience with direct sensitivity to
interval and chord quality as such. When I hear a dominant seventh chord, I
experience it as a perceptual whole: its character seems immediate, forceful,
and present. Often the sense of dominantness precedes any specific
awareness of the component pitches; I might know that I am hearing a
dominant before I realize what scale degree is in the melody; and I might not
be able to identify its inner voices at all. Furthermore, the perception of
dominantness is connected with a range of style-dependent expectations and
responses: in a classical context the chord will feel tense and full of
implication; in a blues it might sound like an energetic yet stable tonic.
These phenomena seem to be embedded in the perceptual experience rather
than generated by conscious calculation: I feel the tension as it were in the
chord itself.
Philosophers distinguish hearing as from hearing plus thinking: the
question is whether we hear a major third as a major third, or whether we
hear a pair of notes that allow us to think the thought “that is a major third.”3
Though not a common music-theoretical topic, it is arguably foundational to
the entire enterprise: for if we are committed to the idea that art is
experiential, then we have reason to be less concerned with what we can
calculate and more concerned with what we hear directly. This is why we
spend so much time teaching ear training rather than allowing students to use
scores or other technology.
The issue is relevant to questions about long-term key structure. For here
there is a fundamental and perhaps biological difference between two
different groups: the small number of absolute-pitch listeners who are
directly sensitive to keys, able to hear that a piece has returned to the tonic,
and the much larger group of relative-pitch listeners who have access to this
information only by way of explicit calculation. This inability to directly
hear keys should make us wary of statements like: “the most radical
transformation of a theme is from dissonance to resolution [ . . . ] this is
basically what happens in a sonata when the second group is recapitulated at
the end in the tonic.”4 And we should remain suspicious even if some
relative-pitch listeners are able to reliably identify the return of the tonic key
on the basis of rhetoric, instrumentation, score study, or conscious tracking
of modulations.5 For the experience of dissonance is felt and immediate in a
way that the calculation of keys is not. Indeed, it is reasonable to wonder
whether absolute-pitch listeners genuinely experience a recapitulation as a
release of tension, psychologically comparable to the resolution of a
dominant.6
All of which presents a profound challenge to music theory. We should
take seriously those experiments suggesting that our powers of long-range
musical perception are less impressive than theorists have traditionally
believed. We should be sensitive to the social and institutional forces that
encourage us to hide our limitations, exaggerating the power of our musical
ears. And we should be genuinely worried that so many music-theoretical
claims seem to be rooted in unverifiable introspection: that experts
confidently claim to hear vast stretches of music “representing” or
“prolonging” a single note; that they claim to hear a note in one measure
directly connecting to another fifty measures away; or that they claim to hear
a sense of return to the tonic after ten or more minutes. The danger is that
some substantial proportion of these claims might not be grounded in
anything real.
Yet at the same time we should recognize that theory can sometimes
change our experience.7 One of my favorite examples involves the track “In”
by the electronic musician Brothomstates (Lassi Nikko). At first listen, the
music sounds like a synthesized flute echoing in a gently lilting triple meter.
But as you tap along you invariably find yourself making small adjustments,
responding to subtle irregularities that bely the music’s placid and ambient
surface. Transcription is useless, yielding a jagged irregularity completely
foreign to one’s lived experience (Figure P9.1). Here we have two
perceptions both true and contradictory: a palpable sense of echo and the
unsteadiness of one’s tapping fingers. It is a perceptual impossibility,
comparable to those psychology experiments that convince people they are
seeing something completely red and completely green at one and the same
time.8

Figure P9.1. The opening of “In” from Brothomstates’ Claro.

The paradox evaporates once we attend to the distance between pitch-

repetitions, revealing a regular 3 + 2 pattern not at all obvious in standard
notation (Figure P9.2). What we are hearing, in other words, is an irregular
echo, a phenomenon unusual in both art and nature. The entire passage likely
results from just five notes played on a synthesizer, with the quintuple-meter
echoes forming a melodic composite both energetic and relaxed: intuition
filtered through algorithm to produce a psychologically compelling result.
This realization allows us to harmonize our seemingly incompatible
percepts: that the music sounds like it is produced by an echo, but that the
composite is rhythmically irregular. And this in turn can genuinely change
our experience, for instance by teaching our fingers when to tap, or changing
our sense of which notes belong together—helping us understand, for
example, that the C in mm. 5 and 6 is not just a neighbor but also an echo of
the C in the previous measures.

Figure P9.2. The music as the product of a 3 + 2 echo. Accented notes are played on the synthesizer;
the rest are produced by the echo.

I cannot offer any general theory of musical perception, nor any reliable
way to separate “hearing as” from “hearing plus thinking.” But a lifetime of
experience has convinced me of two facts. First, that it can be genuinely
difficult to know whether we are thinking something or experiencing it
directly—and hence that we can be drastically mistaken about what we hear.
(Indeed, I suspect that this confusion explains why so many ambitious
composers wasted so much energy manipulating inaudible pitch-
relationships.) And second, that theory can sometimes enrich our experience,
leading us to understand music in a way that genuinely increases our
enjoyment. The challenge for theory is to stretch our ears while remaining
realistic about human limitation. Done badly, it is like playing tennis without
a net, elevating blatantly implausible claims to the status of institutional
common sense. Done well, it is a delicate and intimate form of science that
increases the world’s supply of aesthetic value. No hard-and-fast rules
determine the boundary between these two activities, only taste and good
judgment.

1
Compare Brewer 2015 (“N-rays”), National Research Council 2014 (eyewitnesses), Thomas 1971
(the supernatural), Jost et al. 2009 (implicit bias), Goldstein et al. 2008 (wine).
2
Of course, most absolute-pitch listeners are directly sensitive to interval quality. I have, however,
met some who reported the sort of insensitivity discussed here.
3
The distinction originates with Wittgenstein’s (1953) distinction between “seeing as” and “seeing
plus thinking” which in turn was inspired by the earlier movement of Gestalt psychology. See
DeBellis 1999 for related discussion.
4
Rosen 1988, p. 197.
5
Repeated listening is not the answer, as it gives the listener cognitive access to information they
may not be able to perceive directly. Suppose a non-absolute-pitch listener learns a piece well enough
that they know what key they are in at any one time. This is compatible with their not having an
immediate and palpable sense of long-term tonal closure: they may know that a piece is returning to
the tonic, while having absolutely no sense of “dissonance being resolved.” In the same way, they
could memorize a twelve-tone piece accurately enough to identify, by ear, any divergences from the
score. But this tells us nothing about whether listeners are sensitive to twelve-tone structure in the
aesthetically relevant sense. Repeated experience allows us to think the thought “now I am back in the
home key” (or “now I am hearing the retrograde inversion of the row, starting on B ♭ ”), but this is
different from being sensitive to the relevant musical properties.
6
Cook 1987, Marvin and Brinkman 1999. It is entirely possible to notice that the music has
returned to the tonic key without any associated sense of release.
7
Davies 1983. Hansberry 2017 argues, correctly in my view, that scale-degree identity is to some
extent under our conscious control; indeed, I recall hearing about a new-music soprano who assigned
scale degrees to twelve-tone pieces, hearing them in relation to imaginary tonics.
8
Crane and Piantanida 1983.
 9
Heterogeneous Hierarchy

Music theory offers diverse tools for conceptualizing longer spans of

musical time: principles of motivic development, models of key relations,
theories of long-range melodic connection, and Formenlehre both old and
new. This chapter will explore these approaches in light of the theoretical
framework we have been developing. I begin by suggesting that the concept
of melodic strategy allows us to incorporate insights from Schenkerian
theory within the generally sequential framework of my harmonic grammar.
I then consider some connections between phrase structure and voice
leading.
The general picture I offer is one of heterogeneous hierarchy. Rather than
Schenker’s vision of many fundamentally similar musical levels, guided by a
unified logic and nested like the layers of an onion, I see diverse forms of
organization superimposed: a harmonic system directing the flow of chords,
a contrapuntal system overseeing the motion of surface voices (and largely
inherited from the Renaissance), a melodic system creating longer-range
linear coherence, a modulatory system governing the progression of keys, a
sequential system generating varied repetition, formal systems concerned
with phrasing and long-term development, and so on. To focus on one is in
no way to discount the others: after all, when we walk, our legs obey a local
alternation of left and right, and yet our journeys are nevertheless directed.
Nor is it disrespectful to say that these systems are relatively simple when
considered in isolation, for they come together to form a wonderfully
complex set of musical resources.1

1. Strategy and reduction

Rameau once wrote that it is almost impossible to give definite rules for
melody, as good taste plays a central role in its development.2 Schenker took
perhaps the single largest step toward unraveling this mystery when he
proposed that music typically involves a slower melody taking place beneath
the musical surface, with phrases often embellishing stepwise descents from
one tonic-triad note to another, and with bass lines moving from tonic to
dominant before leaping back to the tonic. This template is so ubiquitous
that examples are almost superfluous, but Figure 9.1.1 provides one anyway
—the opening phrase of Mozart’s G minor symphony K.550, embedding a
linear descent from D down to F♯, while reserving the strong root-position
dominant for the end of the phrase.

Figure 9.1.1. The opening of Mozart’s Symphony no. 40, K.550 features a linear descent from 5̂ to 7̂.

I am firmly convinced that this idea can help us understand a wide range
of music, and I regularly encourage young composers to listen for this
“slower melody” in styles both old and new. What is more challenging is
understanding its significance. Attention to melodic structure has sometimes
been set in opposition to the recognition of local harmonic patterns, as if the
two were somehow incompatible. To my mind, however, there is no conflict
at all: the art of functional composition consists in writing satisfying phrases
while also satisfying the constraints of the local harmonic grammar. This
sort of double organization is familiar from the world of sports and games.
In basketball, players execute higher-order aims (driving to the basket,
advancing the ball to the other end of the court, blocking an opponent’s
lanes) within the constraints given by the rules (no more than two steps per
dribble, no contact with opposing players, etc.). Similarly, chess players
pursue strategic goals (developing pieces, controlling the center) within local
constraints governing pieces’ motions. If the harmonic grammar provides the
rules of the game, then phrase-level melodic templates are the higher-level
goals.
What makes a melody directed? Schenker’s answer, which I endorse, is
that directed melodies often articulate coherent voice-leading structures.
Simple melodies, such as “Helpless,” have just one structural voice,
descending stepwise from one tonic-triad note to another. More complex
melodies, such as “Eight Days a Week,” dance between multiple voices of
an abstract background that itself moves in a coherent way—somewhat like
an Alberti bass, but less predictably. Still more complex is the technique of
moving contrary to a background voice leading, as when Beethoven ascends
along the predominantly descending Waldstein progression (§3.7). This
produces a dual melodic logic superimposing multiple directed narratives: in
the Waldstein, the two hands take separate paths from tonic to dominant,
bass descending by step while the melody ascends by thirds.
These facts imply that directed melodies will often articulate the doubly
parallel motion represented by our spiral diagrams—that is, the combination
of big-T transposition along the scale, and little-t transposition along the
chord. The nonharmonic system ensures that out-of-chord notes resolve by
step to in-chord notes; and registral limitations ensure that these within-
chord notes tend to outline close-position triads and seventh chords. The
resulting “background” provides melodic destinations, points of rest between
which the surface moves. (Again, I think of this as a kind of dancing, surface
voices leaping among background voices in an aesthetically pleasing way.)
Figure 9.1.2 analyzes the opening of Mozart’s Fortieth Symphony, where
hierarchical transposition is intrinsic to melodic directedness, the glue
binding harmony and melody together.

Figure 9.1.2. Dual transpositions in the opening of Mozart’s Symphony no. 40.

Mozart’s melody also exhibits another important technique, the

application of double transposition to specifically thematic objects with
distinctive rhythmic and registral profiles: the melody of mm. 2–5 is
transformed by double transposition into that of mm. 6–9. Figure 9.1.3
identifies the most common possibilities available to functional composers.
Though this technique is as old as functionality itself, it becomes
increasingly important as musical culture shifts from a polyphonic to a
homophonic default: double transposition of prominent thematic material is
relatively rare in The Well-Tempered Clavier, more common in C. P. E.
Bach, and ubiquitous in Mozart and Haydn. One sign of its importance is its
appearance in three of the eleven schemas in the appendix of Robert
Gjerdingen’s Music in the Galant Style (Figure 9.1.4).3
Figure 9.1.3. Thematic transformation by dual transposition. The starred transformation maps the
tonic-triad notes into a nontriadic subset of the dominant seventh.
Figure 9.1.4. (top) Three schemas that employ double transposition between tonic and dominant; the
labels above the solfege names refer to Figure 9.1.3. (bottom) The Do-Re-Mi in a melodic context.

The quadruple hierarchy thus underwrites a collectional analysis that is

more precise than its Schenkerian antecedents. Figure 9.1.5 shows that the
opening of Mozart’s piano sonata K.284 presents a pair of ascending
arpeggiations along the tonic triad, D–F ♯ –A. The introductory fanfare is a
two-octave near sequence of falling thirds that manages to sound iconic and
functional without implying any specific harmonies (§7.2). The first
arpeggiation is a near sequence moving the third D–F♯ along the scale D–

F♯–A; as in Figure 4.5.7, Mozart skips the perfect consonance to produce an

ascending sequence of thirds and sixths.4 The music then descends by thirds
to a tonic pedal, completing a four-measure sentence. The second
arpeggiation begins by harmonizing its melody with a contrary-motion
version of Gjerdingen’s “Quiescenza” (see Figures 3.1.10 and 4.7.4).5 This
sentence ends with a fauxbourdon ROTO variant replacing I64 with V43/IV,
leading to an outer-voice arrival on A. Here the large-scale strategic goal,
ascent to the fifth, is achieved using familiar dyadic and seventh-chord
voice-leading schemas—while also remaining consistent with functional-
harmonic norms. This is not so much a generic process of “arpeggiation,”
existing independent of and prior to the conventions of functional harmony,
but rather something more specific, a careful arrangement of schematic
building blocks that themselves exploit the fundamental geometry of musical
possibility.

Figure 9.1.5. An analysis of the opening of Mozart’s K.284, I.

Proceeding in this way I think we can have many of the benefits of

Schenkerian theory while jettisoning its problematic metaphysics. The key
is, first, to use the quadruple hierarchy to reconceive the notion of the
“background” in scale-theoretic terms, as a matter of surface voices moving
along abstract alphabets: subchord, chordal, scalar, and chromatic, each
nested within the next. Second, to reconceive reduction as a kind of
paraphrase, a speculative hypothesis about the longer strategic goals that
might animate a passage. We need not say less-important material
“represents” more-important material, any more than we say that omitted
material “represents” the events in a paraphrase. Instead, what is important is
simply that our reductions serve as plausible blueprints, their notes
connecting the way one scene in a movie might continue an earlier narrative.
A good blueprint is an analytical way station: simple enough that we can
imagine ourselves inventing it, and yet rich enough that we can imagine
expanding it into the finished piece.
If I am right, then a certain amount of musical hierarchy has an irreducibly
melodic character. It is easy to find evidence for long-range melodic
relationships, notes that seem to connect despite not being adjacent on the
musical surface. While I do not say that harmonies lack such hierarchy, I do
believe it is less common, in large part because functional music obeys a
strict set of local harmonic rules. Melody is locally free but hierarchically
structured, which is precisely why it is so hard to describe. Harmony is its
mirror image, largely nonhierarchical but locally constrained, and hence
more amenable to syntactic theorizing. Broadly speaking, Schenker was
right about melody but wrong about harmony.

2. Two models of the phrase

It follows that composers do not wander aimlessly through harmonic space,

moving from chord to chord without memory or thought of the future.6
Instead, they arrange outer voices to build satisfying melodies while
repeatedly passing through the same basic map of harmonic possibilities.
Different kinds of harmonic cycles will therefore tend to appear in different
parts of the phrase: early-phrase cycles will typically avoid melodic 2̂–1̂ in
favor of other scale degrees, just as the bass 5̂–1̂ will likely be reserved for
the final cadence (Figure 9.2.1). Some contemporary theorists describe the
early-phrase progressions as “prolongational,” as distinct from late-phrase
“cadential” progressions, but to the traditional harmonic theorist they are
essentially similar, alternative paths through a single harmonic map. The
difference between them is the result of large-scale strategy rather than
syntax.
Figure 9.2.1. Some common early-phrase and cadential progressions.

One way of putting this is that individual harmonic cycles have different
degrees of cadential weight, with phrases moving from less-conclusive to
more-conclusive over the course of a phrase. Figure 9.2.2 quantifies this
notion in a rough-and-ready way, considering whether a harmonic cycle
contains a predominant, ends with a root-position tonic-dominant
progression, and so on.7 Cadential weight, thus defined, generally increases
throughout a phrase, with the final cycle longer and stronger than its
predecessors. We can use this approach to construct a simple-but-effective
algorithm for harmonizing chorale-style melodies. Suppose we have a series
of melody notes n1, n2, and so on, to be harmonized by a series of chords C1,
C2, . . . Given a chord Ci harmonizing melodic note ni, we can choose a new
chord Ci + 1 by multiplying two probabilities, the (first order, phrase-
independent) probability that a particular chord follows chord Ci according
to a harmonic grammar and the (zeroth order) likelihood that that chord is
found at that point in the phrase. Figure 9.2.3 shows that this fairly simple
algorithm can produce a reasonably idiomatic output, comparable to the
efforts of a good second-semester undergraduate.8
Figure 9.2.2. (top) A simple model of cycle strength, assigning “strength points” to various features of
the cycle. (bottom) Modeled in this way, cycles increase in strength throughout the phrase in Bach's
chorales.
Figure 9.2.3. A computer-generated harmonization of the opening two phrases of the melody of
“Christus, der ist mein Leben” (BWV 281, Riemenschneider 6).

Chapter 6 argued that the earliest functional styles are so harmonically

limited that the question of inversional equivalence does not arise. We now
see why it is more important in more complex styles. Fundamentally,
inversional equivalence is a conceptual technology that synthesizes
harmonic and polyphonic ideals, allowing a wide variety of bass lines to
express the same small collection of chord progressions. The central claim of
traditional harmonic theory is not that inversion is irrelevant, but rather that
early-phrase and cadential progressions are similar, differing in degree rather
than kind. This is why earlier theorists often used the term “cadence” to refer
to harmonic cycles more generally.9 And this is why the full power of
traditional harmonic theory only appears when we interpret early-phrase and
cadential progressions as more and less conclusive journeys through a single
map of harmonic possibilities (Figure 9.2.4).
Figure 9.2.4. (top) Specific inversions are highly likely to be found at specific phrasal positions in J.
S. Bach’s chorales. (bottom) When we ignore inversions and group chords by functional category, we
find a much more even distribution.

All of which starts to narrow the gap between traditional harmonic theory
and the Schenkerian alternative, at least as concerns the structure of
individual phrases. For both parties can agree on the following:

P1. Phrases tend to begin and end with root-position tonic chords;
P2. Melodies tend to decorate stepwise descents from one note of the tonic
chord to another, perhaps after an initial ascent;
P3. Melodies often jump between the degrees of a series of abstract
scalelike chords that articulate close-position harmonies connected by
familiar voice leadings (§9.1);
P4. Bass lines tend to reserve the root-position dominant for the final
cadence.

These principles imply that traditional harmonic theorists will often find the
phrase-level structures emphasized by Schenkerian analysis. Similarly, they
will often be able to reinterpret orthodox Schenkerian graphs as highlighting
these features of functional phrases. From a practical point of view, the two
approaches are speaking a very similar language.10
However, they differ over subtle questions about the meaning and scope
of these features. Where traditional theory asserts that harmonic cycles are
concatenated sequentially, placed one after another like beads on a string,
Schenkerians argue that early-phrase progressions are nested inside a single
“structural” progression that lasts for the duration of the phrase (Figure
9.2.5). This view is described with admirable clarity in a pair of recent
textbooks, Steven Laitz’s The Complete Musician and Jane Clendinning and
Elizabeth West Marvin’s The Musician’s Guide to Theory and Analysis.11
Both use the term “the phrase model” to describe the recursive conception:
what makes the final dominant cadential, on this view, is not its rhythmic
role within the conventions of the eight-bar phrase, or its use of root-position
dominant chords and 2̂–1̂ melody, but rather the fact that it “connects”
syntactically to the initial tonic, with the earlier T–S–D–T cycles
hierarchically embedded within a higher-level “structural” progression.12
Harmonic analysis thus requires us to determine which chords belong to the
structural progression and which do not.
Figure 9.2.5. Steven Laitz’s interpretation of the opening of Haydn’s G minor sonata, Hob. XVI:44, II
(from The Complete Musician). Where a traditional theorist would see four successive harmonic
cycles, Laitz postulates that three of the cycles are embedded within the fourth. Above the staff, I note
the use of a schema from Figure 9.1.3 to relate tonic and dominant.

As a general rule, I do not find it useful to imagine chords lasting longer

than they do. At best this strikes me as a confusing metaphor for centricity, a
phenomenon that has much more to do with emphasis and expectation than
with hierarchy and persistence.13 (It may also reflect phrase-level
expectations that arise in certain functional subgenres, as we will discuss
shortly.) In other aspects of life, there is little temptation to connect stability
and representation: after all, when we travel we may feel unsettled, but we
do not say that our trips “represent” the state of being-at-home.
I suspect the desire to fuse these concepts results from two factors. The
first is a failure to consider alternative explanations for the phrase-level
linear structures so rightly emphasized in Schenkerian theory. The second is
a desire for a unified approach to musical structure in which phenomena
such as melodic directedness, centricity, harmonic function, phrase-level
expectations, and nonharmonicity can all be explained as manifestations of
the same basic principles. In Figure 9.2.6, for example, Schenkerian analysis
is presented as a later stage in a sequence of progressive reductions: the top
level removes nonharmonic tones to obtain a harmonic skeleton, while the
third level removes consonant chordal skips to obtain a melodic outline. Up
to this point, Schenkerian and non-Schenkerian theorists can agree. The next
levels, however, continue to remove harmonies in accordance with the
phrase model: the fourth reveals a single TD(T) harmonic cycle governing
the antecedent (resolving to the tonic in m. 5), while the fifth eliminates the
dominant to obtain a single TDT governing the whole phrase.14 The
approach thus extends nonharmonic reduction to larger time scales.

Figure 9.2.6. Lerdahl and Jackendoff’s five-stage reductional analysis of the opening of Mozart’s
K.331, I. Above the staves I add my own description of the reductional level. The first three are
compatible with traditional theory.

Deciding between these two conceptions of the phrase is perhaps the

single most pressing problem of contemporary music theory, as it directly
affects elementary pedagogy. Empirically, the question is what evidence
supports the claim that initial tonics “attach” to final dominants: in Figure
9.2.5, tonic chords are almost all unaccented, appearing only on the last
quarter note of every measure, yet the bulk of the phrase is taken to
embellish the opening tonic. Nor is it obvious, in the case of more flexibly
phrased music such as that in Figure 9.2.7, how to identify the initial tonic
that supposedly attaches to a given cadence; here Schenkerian analysis
seems to require fundamentally different mechanisms for the perception of
regularly and irregularly phrased music. There are also deeper
methodological questions about the foundations of the entire approach: in
chapter 5, I argued that nonharmonic tones do not necessarily “stand for” or
“prolong” harmonic tones; if even the simplest forms of reduction are
problematic, then there is all the more reason to be concerned about its
Schenkerian generalizations. Finally, there is the fundamental philosophical
question about hearing and hearing-as, for while it is possible to think “this
initial tonic connects to the final dominant,” it is much less clear that we
genuinely hear such connections. How could we even test whether the
recursive picture is true? Does “I feel the tonic as persisting” mean anything
more than “tonic chords are generally stable”?
Figure 9.2.7. Measures 39–48 of the first A minor fugue in The Well-Tempered Clavier. Here we have
a flexible musical texture without clearly articulated phrase breaks, making it difficult to determine
what initial C major chord should attach to the strong final cadence.

The upshot is that although Schenkerian theory has dominated American

music-theoretical discourse for more than half a century, there is little or no
solid evidence supporting it—and no historical evidence that earlier
composers in fact thought in this way.15 Thus as far as anyone knows, it may
be possible to hear functional harmony in two fundamentally different ways:
sequentially, as a sequence of harmonic cycles placed one after another, or
recursively, as a collection of “prolongational” cycles nested within one
another. It may be that these different ways of hearing are uncorrelated with
the quality of one’s musicianship—that one can be an excellent performer,
composer, or theorist and hear in either way. That would argue for a
pluralism in music theory, with different listeners perceiving a single piece
in fundamentally different ways. Another possibility, closer to Schenker’s
own view, is that both modes of hearing are available, but that sequential
hearing is a sign of inferior musicianship; if that is right, then there are
multiple ways to interact with music but some of them are objectively worse
than others. Still another possibility is that people generally hear in only one
of these ways, in which case many of us are wrong about what we hear.
After all I am convinced that I hear harmonic cycles sequentially, while
many of my friends and colleagues seem equally confident about the
hierarchical perspective.
That there can be pervasive disagreement over such a fundamental issue is
nothing short of alarming—as if linguists regularly disagreed about which
nouns attached to which verbs. And that these competing views have been
embedded in competing educational institutions is doubly concerning, for
depending on what conservatory one attends, or which textbook one reads,
one may encounter just one perspective, often presented as incontrovertible
and obvious.16 This is not so much an individual failing as a cultural issue,
evidence of an intellectual community insufficiently devoted to discovering
the truth. The problem, I suspect, is that Schenkerian theory appeals to a
powerful sense of how music should be: in this respect, it recalls the Platonic
faith that nonharmonic tones must be “merely decorative,” or that
recapitulations must create a sense of long-term resolution—convictions that
survive not because they are supported by evidence, but because they satisfy
intellectual and emotional needs.

3. Chopin and the Prime Directive

My view is that melody and harmony are heterogeneous, independent forces

that can be combined in a wide variety of ways. While it might be true that
certain kinds of melodies naturally suggest functional harmonization—for
example, those that embellish the chain of descending thirds—the same
cannot be said for melodies in general; indeed, a major theme of this book is
that stepwise descending melodies often generate retrofunctional harmonies
that contravene functional expectations. If that is right, then we should
expect to find composers employing one and the same melodic strategy to
produce a range of harmonic effects—for instance, using descending
melodic steps to move between diatonic modality, functional harmony, and
chromaticism.
This is exactly what happens in Chopin. At one end of the spectrum are
passages like the opening of the first etude, where stepwise descending voice
leading is fully compatible with the constraints of functional harmony
(Figure 9.3.1).17 At the other end are pieces like the A minor prelude, a
famously baffling miniature that avoids any intimations of A minor until its
final bars. Yet on closer inspection it is based on a very similar logic, its
voice leading disguised by the elaborate “nonharmonic tone machine” in
Figure 9.3.2. The background is a sequence of third-related triads e–G–b–D–
(F)–a, with each triad’s root descending by step to become the fifth of the
next (Figure 9.3.3). The G and D triads give rise to I64–V–(I) progressions
that continue the sinking counterpoint, with the latter inaugurating a series of
seventh chords linked by semitonal voice leading.18 These slithering
sevenths are utterly typical of Chopin and mid-nineteenth-century
chromaticism more generally; what is new here is the implied connection to
an earlier diatonic voice leading more characteristic of modality—another
contrapuntal tradition, adjacent to functionality, but utilizing a different
scalar and chordal repertoire.19 Ultimately, the entire prelude is built on
descending melodic steps: in the opening, diatonic steps connect mysterious
triads that just barely conform to functional norms; in the middle, chromatic
steps connect functionally anomalous sevenths (Figure 9.3.4). It is as if
Chopin were telling us that these two worlds were not so far apart as we
might have thought.
Figure 9.3.1. The opening of Chopin’s first etude uses stepwise descending voice leading, first with
diatonic triads and then with chromatic seventh chords.

Figure 9.3.2. Chopin’s structured arpeggiation embellishes three underlying voices, each of which
appears in two octaves. This is represented by the networks, which start in the center, move left, and
then move right.
Figure 9.3.3. Three increasingly detailed outlines of Chopin’s A minor prelude.
Figure 9.3.4. The A minor prelude takes short clockwise moves in two different geometrical spaces.

The first A minor mazurka (Op. 7, no. 2) is similar, interpolating VI6 into
iv64–i to create a strikingly modal opening, but proceeding onward to a
functional phrase with a clear 5̂–4̂–3̂–2̂–1̂ linear descent.20 Once again these
descending diatonic triads lead to nonfunctional sevenths, descending voice
leading now chromatic and futuristic rather than modal and nostalgic (Figure
9.3.5). The second A minor mazurka is similar but with the diatonic triads
giving way to chromatic trichords; the tetrachords are reserved until the
piece’s coda (Figures 9.3.6–7). An interesting wrinkle here is the scrambling
of the stepwise descents: rather than repeating the diatonic triad’s basic voice
leading, (A, C, F) → (A, C, E) → (G, C, E) → (G, B, E), Chopin gives us
(A, C, F) → (G, C, F) → (G, B, F) → (G, B, E), moving along the dotted
lines in Figure 8.7.10. This is the technique of Chopin’s E minor prelude,
now expressed with diatonic triads rather than chromatic sevenths.21 Here
we see something like an explicit awareness of the Prime Directive.
Figure 9.3.5. The A minor Mazurka, Op. 7, no. 2, opens with a series of short descending moves on
the triadic spiral, creating a iv–VI–i progression; the eight bars feature two loosely parallel iv–i–V–i
progressions in A and E minor. The contrasting middle section uses descending semitonal voice
leading between seventh chords.
Figure 9.3.6. The opening of the A minor Mazurka, Op. 17, no. 4. The move from A–C–F to G–B–E,
instead of taking place in the order Alto–Bass–Tenor, occurs in the order Bass–Tenor–Alto, producing
a series of nontriadic harmonies.

Figure 9.3.7. Chromatic seventh-chord descent in the coda of Op. 17, no. 4.

One final example allows for a more direct comparison with Schenker.
The first contrasting phrase of the Revolutionary Etude again features
descending stepwise voice leading embodied by triads and sevenths.
Schenker and I both identify the same initial voice leading, a chromatic 6–5
sequence using minor triads (Figure 9.3.8). But my analysis continues the
stepwise descent through the following dominant sevenths, while Schenker
instead follows the surface, inventing a 5–6 sequence to rationalize its voice
leading. His graph includes a provocative footnote:
all conventional concepts, such as functional harmony, modulation, etc., fail in regard to
measures 21–41. It would be wrong, for instance, to speak of a modulation to B-flat major, a-
flat minor, etc. The voice leading of the middleground alone offers the correct solution to the
problem.22

I agree that it would be perverse to describe this music as resulting from

standard tonal-functional strategies, as if Chopin began with the intention to
modulate to B♭ major, A♭ minor, D♭ minor, or to embody those keys with
unusual modal progressions. Instead, it manifests a fundamentally
contrapuntal logic that exploits stepwise descending voice leading, taking
short clockwise motions in a pair of spiral diagrams (Figure 9.3.9). This
descending background is further embellished by an ascending melodic
voice as in the opening of the Waldstein (§3.7).
Figure 9.3.8. (top) My analysis of mm. 28–35 of Chopin’s “Revolutionary” etude Op. 10, no. 12. The
bottom staff shows an idealized background voice leading, while the top staff shows the actual upper
voices, rising in thirds against the descending background. (bottom) Schenker’s analysis of the same
passage emphasizes a persisting C(♭).
Figure 9.3.9. The Revolutionary Etude takes short clockwise motions in the space of chromatic triads
(left) and chromatic seventh chords (right).

Yet Schenker and I disagree in our characterization of that logic. When

Schenker speaks of “middleground voice leading,” he is not referring to the
continuous process of chromatic descent; instead, he is claiming that the
music derives from or prolongs a persistent C (or C♭) shown by the arrows
and open noteheads on his reduction. For Schenker, the unusual surface is
essentially nonharmonic and decorative, with the real structure lying
elsewhere. For me the process is instead basic and irreducible: the relevant
middleground is a sequence rather than a static harmony; furthermore, I
conceive of this sequence as creating a series of abstract, scale-like objects
within which the surface voice ascends, rather than as a chorale made up of
specific pitches. Finally, I believe that the background does in fact intimate a
variety of functional phenomena: it is not “wrong” to hear modulations,
tonics, or dominants; it is just that these functional effects are guided by a
deeper logic.
When Schenker says that traditional concepts “fail,” he is, I think,
expressing a holistic conviction that his linear and prolongational approach
could replace the ideas of traditional theory. It was this belief that led him to
cast aspersions on such commonplace concepts as scale, sequence, motive,
modulation, and functional-harmonic convention.23 Recent Schenkerianism
tacitly softens his radicalism, presenting Schenkerian theory as an extension
of traditional music theory rather than a replacement for it. But this leads to
the methodological challenge of synthesizing two very different views of
musical structure: for how could one and the same chord be merely linear
and also a dominant obeying a local chord-to-chord grammar? How could
Chopin’s Revolutionary Etude simultaneously be the mere byproduct of
middleground voice leading and also modulate to A ♭ major? My approach
tries to ease this conflict by placing the two forms of explanation on
different conceptual levels, treating local harmonic constraints as rule-like
and syntactical, and larger linear relationships as strategic. For many
Schenkerians, tonality just is prolongation, and the passages we have just
considered are all functionally tonal. For me, Chopin’s varied harmonic
palette shows that style depends not just on musical goals, but also on the
chordal and scalar environment in which the goals are pursued. We see this
in the fact that Chopin can use a single melodic strategy to produce a wide
range of harmonic effects.

4. An expanded vocabulary of reductional targets

Schenkerian templates were originally presented as something like laws of

nature, rooted in the overtone series and applicable to all great music. By
reconceiving them as melodic strategies, we can open the door to other
forms of musical organization, including some that do not emphasize
stepwise descent.24
Consider the variations theme of Beethoven’s E major piano sonata, Op.
109 (Figure 9.4.1). Figure 9.4.2 reads each four-measure unit as a near
sequence by contrary motion, presenting thirds on E, B, E, and A(♯), with
each third accompanied by a voice exchange.25 The two four-bar units
combine to produce an eight-bar phrase featuring three distinct presentations
of its opening two-measure idea. On my reading, the first four-bar pattern
involves two octave displacements, one shifting G ♯ 3 to G ♯ 4 in m. 3, the
other displacing the repeated cadence to arrive on (B2, B4) instead of (B3,
B3). The second four-bar phrase has no displacements, the surface voices
switching hands to arrive at the expected cadence on (B2, B3).26 My
drawing perhaps expresses this more clearly than conventional notation.
Figure 9.4.3 shows that the theme is composed of two overlapping chains of
thirds: ascending thirds in the left hand, overlapping with descending thirds
in the right.

Figure 9.4.1. The theme of the variations movement of Beethoven’s Op. 109.
Figure 9.4.2. (left) My analysis of the opening of Beethoven’s theme. (right) A graphical
representation, with the vertical axis in scale steps and the horizontal axis in quarter notes.
Figure 9.4.3. The theme as a superimposition of two thirds cycles. Vertical dyads are connected by
either a basic voice leading (bvl+ or bvl–) or the two-step voice exchange (curved arrow).

Three pieces of evidence support this analysis. First, Beethoven makes the
contrary-motion design explicit in the piece’s third variation, where the
music articulates exactly the reduction I have proposed—the composer
helpfully analyzing his own music (Figure 9.4.4). This unveiling of covert
sequential structure is reminiscent of the Op. 1, no. 3 variations, which also
use the diatonic third’s basic voice leading (§4.10). Second, this kind of
thematic design, based on contrary motion and diatonic thirds, is
characteristic of Beethoven: we have already encountered examples in §3.2,
and will encounter many more in the next chapter. And third, the exploration
of dyadic logic continues throughout the theme: in mm. 9–10, we have a
curious descending-step sequence combining the two forms of the diatonic
third’s basic voice leading (Figure 9.4.5, compare pattern D2a on Figure
7.7.3). So the basic voice leading, which initially alternates with voice
exchanges, now appears more explicitly and without the voice exchanges.
Figure 9.4.4. The third variation of Op. 109 (bottom staves of each line) exposes the contrary-motion
structure of the theme (top two staves).
Figure 9.4.5. My analysis of the second half of Beethoven’s theme. Inward-pointing stems show the
descending form of the diatonic third’s basic voice leading; outward-pointing stems represent the
ascending form.

I chose this passage because I think it demonstrates the core Schenkerian

technique of analyzing complicated passages as decorations of simpler
templates. Standard Schenkerianism, however, with its single-minded focus
on stepwise melody—and its concomitant disdain for sequential structure—
is not well-equipped to analyze it, in large part because the music’s
vocabulary lies outside the Schenkerian catalogue of acceptable reductive
targets.27 Hence Forte and Gilbert promote the first measure’s weak F ♯ to
primary melodic status and ignore the theme’s notable close on 3̂ (Figure
9.4.6). This is a first example of what I take to be a broader incompatibility
between Beethoven’s lexicon and the generic expectations of Schenkerian
theory—and a reminder that we need to consider a composer’s individual
vocabulary when producing musical reductions.

Figure 9.4.6. Forte and Gilbert’s analysis of Beethoven’s theme.

The minuet from the Op. 22 piano sonata is broadly analogous to the Op.
109 theme, again emphasizing contrary motion and the diatonic third’s basic
voice leading (Figure 9.4.7). Here, however, it opens with an ascending-fifth
contrary motion pattern E♭–B♭–F, again decorated with voice exchanges but
now with each dyad preceded by its diatonic lower neighbor. Note that on
my analysis the initial melodic D belongs to the lower voice and continues in
the left hand. The underlying schema appears in many of Beethoven’s
compositions, including the preceding opus number (Figure 9.4.8); it is
easily missed if one is not expecting it.28

Figure 9.4.7. The Op. 22 minuet opens by decorating the ascending form of the diatonic third’s basic
voice leading.
Figure 9.4.8. The opening of the minuet of Beethoven’s First Symphony, Op. 21.

The sequential antecedent dissolves into a curiously static consequent that

sits on a V9 chord resolving to I without any linear descent. We then move to
a middle section that is a doubly distorted A3D5 near sequence: instead of
the purely sequential B♭–D7–g–B♭7–E♭–G7–c–E♭–a°— . . . , Beethoven gives

us B♭–D7–g–B♭ 7–[E♭]–G7–c–F7–B♭, omitting the E♭ and adjusting the final

unit upward by step so that it returns to the tonic after four iterations (Figure
9.4.9).29 There are also hints of an antiparallel contrary-motion sequence,
with the upper-voice thirds moving from B ♭ 5 to E ♭ 5 to A4 while the bass
roots ascend G2–C3–F3.30 The repeat brings us to a coda in which the
hanging G5 is doubly resolved: upward to B ♭ by way of the “ascending
subdominant,” and downward to D (harmonized by the tenor 4̂–3̂–2̂–1̂ to
generate a Prinner). Once again, the upper-voice melody never reaches the
tonic; insofar as there is a descent, it is in an inner voice, and postponed to
the coda rather than presented in the main body of the piece.
Figure 9.4.9. An analysis of the Op. 22 minuet.

Schenker’s analysis is largely insensitive to its sequential structure (Figure

9.4.10). The most obvious issue is his choice of primary note D5, the sixth
note of the upper voice’s eight-note melodic sequence: if one knows the
idiom, F is a more obvious choice. (Schematically, the initial D4 belongs to
the lower voice, even though Beethoven initially presents it in the right
hand.) Equally problematic is the interpretation of the middle section as a
generic descending-fifth progression D–g–c–F rather than a distorted near
sequence by descending third: not only does the latter reading engage more
musical details but it is more responsive to Beethoven’s sequential
vocabulary.31 The resulting analysis ends up flattening this highly personal
music into something generic.32
Figure 9.4.10. Schenker’s reading of Beethoven’s Op. 22 minuet.

All of which raises three methodological questions. The most important is

whether sequences are essential or merely superficial. Schenkerian analysis
is often presented as a process of removing notes, reducing music to a
nonsequential chorale-like background whose melodies invariably move
according to principles of good voice leading.33 I have argued that sequences
can also be legitimate targets for musical reduction, comparable to the
stepwise melodic connections that feature in Schenkerian analysis: for me,
the inner life of these pieces, their deepest semantics, consists in the
juxtaposition of schematic and sequential material. I take these sequences to
be atomic units, pieces of vocabulary comparable to English phrases like “it
is what it is.”34 Idioms are analytical end-states requiring no further
reduction.
Here, the two philosophies lead to very different perspectives on voice
exchanges. If the target of reduction is a chorale-like structure, then voice
exchanges must ultimately be reduced away. By contrast I think the familiar
picture of “surface” voice exchanges decorating stepwise melody, while
often useful, is misleading here: the point of Figure 9.4.2 is that the basic
voice leading and voice exchanges combine to produce a near contrary-
motion fifth sequence, with the voice exchange and the basic voice leading
producing very similar melodic intervals (§10.1). I claim this sequence is a
recurring schema important throughout functional music and crucial to
Beethoven’s language in particular—a unified idiom not representing
anything deeper. Indeed, the next chapter will argue that the schema is one
of a small number of contrary-motion patterns basic to functionally tonal
counterpoint. If I am right, then this is a case where the analytical habit of
eliminating voice exchanges can lead us astray.
A second question concerns the ontology of our reductional targets.
Schenkerian reduction produces a chorale-like output where specific pitches
are arranged in the manner of species counterpoint. I have instead proposed
a method that yields a scalar background, whose abstract “voices” are
melodic slots available in every octave (e.g., Figure P5.3 or P5.7). Thus my
analysis of the middle section of Chopin’s Revolutionary Etude featured an
ascending surface pattern within a descending harmonic field; Schenker’s
alternative features specific pitches which are projected into other octaves by
a process of register transfer. One could complain here that standard
Schenkerian analysis is insufficiently hierarchical, ignoring the ontological
difference between chord and scale—and flattening multileveled musical
structures in the process.
The last question concerns the degree of similarity to be expected among
functionally tonal pieces. I believe there are considerable differences from
composer to composer and genre to genre. Acknowledging these differences
might lead us to reconsider, for example, whether every functional piece
needs to articulate a linear descent to the tonic. Absent a preexisting
conviction about this matter, it is difficult to understand why any analyst
would identify a 3̂–2̂–1̂ descent in the two pieces we have just considered.
Here, Schenkerian analysis seems to suffer from the desire to make every
piece tell the very same story, an aspiration virtually explicit in the motto
“semper idem sed non eodem modo.”35 A better view, I think, describes
composers as deploying a range of strategies and schemas that vary with
time and place: rather than a single “language of tonality,” it paints a picture
of overlapping networks of context-dependent practice—an hourglass shape
with baroque tonality comparatively closer to the Renaissance, classical
practice more regimented and conventional, and more expansive in the
nineteenth century. Where Schenkerian metaphysics requires unity, the
melodic strategies allow for variability.

5. Simple harmonic hierarchy

Though I believe functional harmony obeys chord-to-chord laws, I do not

claim that its grammar is completely and utterly nonhierarchical. I have
already mentioned that fauxbourdon often embellishes functional
progressions, as when IV–iii–ii–I decorates IV–I (§7.5). Furthermore, I
agree with the standard view that the cadential six-four should be understood
hierarchically, with progressions like ii6–I64–V moving from a supertonic to
a dominant that is itself embellished by a second-inversion triad; in symbols
ii6–(I64⇒V), with (I64⇒V) indicating that the six-four is hierarchically
dependent on the following dominant. This perspective brings a host of
otherwise anomalous progressions under the umbrella of standard functional
principles; sequences such as V65/V–I64–V or vi–I64–V can be understood
to move from the initial chord to the final dominant by way of a subsidiary or
decorative 64.36 The view gains further support from the fact that second-
inversion triads can decorate the final tonic in cadential idioms like V–
IV64–I (Figure 9.5.1). It is more enlightening to view this as V–(IV64⇒I)
rather than a retrofunctional V–IV progression. Indeed, in earlier music we
find first-inversion triads functioning in this way (e.g., m. 12 of Figure
6.5.1).

Figure 9.5.1. While the standard cadential 64 chord decorates the dominant, an alternative decorates
the tonic: m. 30 of the first Agnus of Palestrina’s mass Quando lieta sperai; Chopin’s Mazurka, Op.
17, no. 3, mm. 39–40; the opening of the slow movement of Mozart’s A minor piano sonata, K.310,
where the G–B♭ suspensions embellish a deceptive resolution.
 Secondary dominants often require a similar approach. Consider the two
analyses in Figure 9.5.2. The first is not so much wrong as incomplete, for
while it correctly notes the presence of harmonic cycles in G major, D major,
and E minor, it implausibly postulates four separate key changes, as if the
mere presence of a dominant were sufficient to immediately cancel the
prevailing key. The better analysis depicts a multileveled structure—
identifying E minor both as the tonic of its own IV65–V65–i cycle and also
as vi in a higher-level G major cycle. This description recognizes the
psychological reality that very brief excursions to foreign keys do not fully
dislodge our memory of the home tonality: when we return to the tonic after
a short digression to E minor, we are capable of experiencing the entire
progression in a persisting G major. (It is not known whether listeners
experience the E minor as having any tonic quality, or whether we simply
accept V7/vi–vi as an idiom, perhaps a substitute for iii–vi.) We might say
that centricity is “sticky,” changing more slowly than harmonies themselves.
Musicians exploit this stickiness by briefly digressing from the home key, a
technique that is as old as secondary dominants and as recent as “playing
outside” (§5.6).37

Figure 9.5.2. Two analyses of the opening of Bach’s “Herr Jesu Christ, dich zu uns wend” (BWV 332,
Riemenschneider 136).

Rhythm, phrasing, and sequence can reinforce this hierarchical structure

in interesting ways. In Figure 9.5.3, for example, a two-unit sequence moves
from tonic to supertonic, embellishing each chord with its own dominant.38
The phrase structure reinforces the harmonic hierarchy: since mm. 9–16
mirror mm. 1–8, it is much easier to hear the entire unit moving upward
from I to ii, creating a large-scale I–ii–V progression decorated by lower-
level tonic-dominant oscillations. The third eight-bar unit is particularly
interesting, as it does not involve secondary dominants; instead it is an
example of “standing on the dominant,” in which hypermetrically weak
tonics may be felt to embellish rhythmically stable dominants. Here we have
harmonic hierarchy without secondary dominants or exact sequential
parallelism.39

Figure 9.5.3. The opening of the last movement of Haydn’s final piano sonata, Hob.XVI:52.
 Such examples clearly show that Western music can exhibit a limited
degree of hierarchical embedding, with subsidiary progressions decorating
larger patterns that can be felt to follow familiar harmonic trajectories. This
sort of recursive organization has generated significant interest among both
musicians and cognitive scientists, as it evokes the structure of natural
language: the hierarchy in Figure 9.5.2 seems analogous to that in a phrase
like “the kitten we got at the shelter has grown into an affectionate cat,”
where the sentence “we got the kitten at the shelter” is embedded inside “the
kitten has grown into an affectionate cat.” This suggests that music may
exploit the same cognitive capacities as natural language, with musical
“sentences” like I–V–I capable of being embedded inside one another to
form progressions like (I–V–I)–(vi–V/vi–vi)– . . . 40
This analogy has led some theorists to propose that secondary dominants
generally decorate the chords to which they apply.41 If so, it is wrong to hear
a direct relationship between a secondary dominant and the preceding chord:
instead, we should hear X–V7/Y–Y as a version of X–Y, with the second
harmony embellished by a V7/Y chord at a lower level, symbolically X–
(V7/Y⇒Y). (This is analogous to the way we connect “the kitten” but not
“the shelter” to the verb phrase “has grown”; it is the kitten rather than the
shelter that has grown.) Thus in the penultimate measure of Figure 9.5.2, the
meaningful progression is from G major to E minor, and not from G major to
A65. The G–A65 is like the misleading adjacency “the shelter has grown” in
the English sentence.
Once again, however, the complexity of musical practice thwarts simple
generalizations. For if secondary dominants invariably worked as just
described, then we would not find progressions of the form X–(V7/Y⇒Y)
unless we also found X–Y. Yet virtually any nonsyntactical progression X–Y
can be normalized by the insertion of the appropriate secondary dominant,
including V–IV, V–ii, I64–vi, and so on. Figure 9.5.4 provides a pair of
examples from the Bach chorales in which secondary dominants cancel the
implications of the preceding chord, allowing the music to rescind its
outstanding harmonic obligations.42 The paradox is that secondary
dominants are simultaneously a clear example of harmonic recursion and
also evidence for its limited scope: it is not much of an exaggeration to say
that any secondary dominant could signal either a momentary digression to a
foreign key, hierarchically embedded within a persisting tonic, or a shift to a
new key that simply abandons any existing implications. This suggests
harmonies have only weak implications for chords beyond their immediate
successors.

Figure 9.5.4. Secondary dominants licensing unusual harmonic progressions: “Wie schön leuchtet der
Morgenstern” (BWV 36.4, Riemenschneider 86) has a V–ii if the applied dominant is removed;
“Fröhlich soll mein Herze springen” (BWV 248.33, Riemenschneider 139) has a I64–vi without the
applied dominant.

To my mind, this difference underscores a fundamental feature of

functional harmony: where human language cannot be characterized in a
word-to-word fashion, a good deal of music does seem to exhibit chord-to-
chord structure. One can imagine linguistic utterances that thwart
expectations in a manner similar to the previous examples: “The beautiful
and . . . yesterday we went to the . . . and the kitten we got at the shelter has
grown into an affectionate . . .” The incomprehensibility we feel here, the
palpable frustration generated by unrealized implications, is unlike anything
one experiences in any of the musical examples.
The failure of hierarchical analysis, in even simple cases, speaks volumes.
In many ways, secondary dominants are comparable to nonharmonic tones:
both phenomena initially seem clearly hierarchical, yet turn out to involve
unexpected complexities. Nonharmonic tones do not always represent
harmonic partners, and secondary dominants do not always embellish their
successors. The issue, in both cases, is the limits of musical perception:
where we perceive language with almost crystalline accuracy, transforming
sound into meaning without error, we perceive music through a low-fidelity
haze. We are simply not accurate enough to be bothered by the fact that the
“Renaissance seventh” cannot be reduced to a consonant background, or that
the progression ii–V/IV–IV contravenes the principles of functional
harmony when the secondary dominant is eliminated. What is remarkable is
that music can move us so deeply despite our perceptual limitations.

6. The four-part phrase

Classical phrases often exhibit a quadruple grouping structure inherited from

Western European dance music.43 Figure 9.6.1 shows how four durational
units can be arranged to satisfy two constraints: that one unit repeats or
varies another, and that the final unit forms a cadence. When the third unit
recalls the first, we have an ABÁC or “period” structure—often further
organized into two loosely parallel halves, with a weak close after AB and a
stronger close after ÁC. When the recall is immediate, we have an AÁBC
“sentence,” again often articulated as (AÁ) and (BC). The third possibility
occurs when the third unit echoes the second for an ABB´C arrangement.
This structure has no standard name and is considerably less common than
the others, but it does occasionally appear (Figure 9.6.2); we might call it
“tail-development form,” because the phrase proceeds by developing the
“tail” of the initial four-measure idea. The fourth possibility, AÁÁ´C, is
reasonably common, appearing for example in Beethoven’s Op. 109
variations (Figure 9.4.1).44 This is a nice example of how abstract reasoning
can serve concrete analysis: rather than postulating sentence and period as
unexplained schemas, we can understand them as filling out an abstract
space of possibility—sensitizing our ears to new possibilities along the way.
Figure 9.6.1. Four ways to divide eight measures into four two-bar groups, with a cadence at the end,
and at least one group echoing another.
Figure 9.6.2. Four classical-style ABB´C phrases: A, Haydn’s Hob. XVI:8, IV; B, the opening of
Mozart’s violin sonata K.304; C, the opening of Beethoven’s Op. 31, no. 3; and D, the opening of
Beethoven’s Op. 79.

Though not unknown in earlier music, quadruply divided phrases become

popular in the late eighteenth century and remain so in our own time (Figure
9.6.3).45 Longer themes often nest these templates inside one another so that
similar structures appear on multiple temporal levels. The most common
approach is to embed an eight-bar sentence inside a sixteen-bar period
(Figure 9.6.4). In the rondo of his D major piano sonata K.311, Mozart takes
things one step further: the opening four bars are organized as a small period
that is continued by a four-bar sentence; together, they form an eight-bar
sentence that is embedded in a sixteen-bar period, triply nesting the standard
four-part templates (Figure 9.6.5).46

Figure 9.6.3. Grace VanderWaal’s 2017 “So Much More Than This” arranges the three units A, B, and
C to form an ABBC phrase, while Marshmello’s 2018 “Happier” arranges similar units to form an
ABAC period.
Figure 9.6.4. Sam Cooke’s 1961 “Cupid” nests two eight-bar sentences inside a sixteen-bar period.
The initial motive is transposed relative to the prevailing chord, changing harmonic notes to
nonharmonic notes and vice versa.
Figure 9.6.5. Triply nested groupings in Mozart’s K.311, III.

More complex construction can arise from elision, with one or more bars
shared by overlapping formal units. I hear the opening of the
Hammerklavier’s scherzo as a four-bar AA'A'A" that ends with a weak
cadence: heard on its own, these measures outline a weakly closed I–vi–IV–
V–I harmonic cycle, with the first three chords hierarchically decorated by
their own dominants (Figure 9.6.6). The final bar starts a second schema, a
four-bar “Ludwig” featuring outer-voice contrary motion and organized as
four-bar sentence; heard on its own, it would make a reasonably convincing
opening.47 To understand this music, we therefore need to recognize the
weak close in m. 4: whether we refer to it as a “cadence” or a form of
“noncadential closure” is immaterial; the important point is its role in
dividing the phrase into two overlapping halves, 4 + 4 somehow adding up
to 7.

Figure 9.6.6. The opening of the scherzo of Beethoven’s Hammerklavier, Op. 106. The first four
measures and the last four measures each sound like coherent four-bar phrases.

Here I have been using the terms “sentence” and “period” to refer to very
general grouping strategies that can exist on a variety of temporal levels.
William Caplin, our most influential theorist of phrase structure, instead
defines “sentence” and “period” as thematic patterns of what he calls
“formal functions,” largely determined by bass motion and always having a
length of eight “real” bars.48 This has, in my opinion, three problematic
consequences. First, it deemphasizes one of the most interesting features of
the classical style, the reuse of similar grouping strategies on different
temporal levels.49 Second, it deprives us of vocabulary for expressing
similarities between classical-style phrases and those in nonclassical styles
such as rock. Third, it directs our attention away from relatively
intersubjective features of musical structure, such as the contrast between
ABÁC and AÁBC grouping, and toward much more subjective judgments
about rhetorical role.
The second theme of the first movement of Beethoven’s Op. 31, no. 3,
illustrates the difficulties that can arise. From the perspective of melody,
motive, and grouping, Figure 9.6.7 is a period, an (AB)(ÁC) phrase with an
unmistakable pause on an unstable ii6. Caplin, however, does not recognize
it as such, taking its lightly embellished 3̂–4̂–5̂–1̂ bass to indicate that the
entire structure is “cadential,” lacking “initiating function” altogether.50 But
his writing also justifies another interpretation that highlights the left hand’s
ascending-step sequence. This would make the phrase functionally medial, a
continuation without proper beginning.51 Thus we could interpret one and
the same phrase as a standard period, a cadential structure lacking initiating
function, or a sequence expressing medial function.

Figure 9.6.7. Three features of the second theme of Beethoven’s Op. 31, no. 3, I, suggesting three
different formal interpretations. First, its rhythmic and motivic antecedent/consequent design; second,
its cadential bass line; third, its disguised sequential structure. This last feature may explain the
decision to begin with a first-inversion triad.

While I incline toward the first of these interpretations, I am much more

strongly committed to the view that they are all reasonable: motivically,
Beethoven has given us a period built from a largely sequential progression
over a bass line that usually signals a cadence. All these descriptions are
true, all are potentially available to listeners, and none need take priority
over the others. The advantage of thinner, more objective definitions of
“sentence” and “period”—based on grouping rather than a nebulous and
subjective concept of “function”—is that they give us relatively neutral
categories with which to express our differing perceptions: whatever we may
think about its formal function, we can agree that this theme is a period in
the sense of having a loosely parallel (AB)(ÁC) grouping structure, with a
weak close on ii6 answered by a stronger close on the tonic. To my mind,
that is reason to favor the simpler and more objective terminology.

7. Grouping, melody, harmony

Here, finally, we start to glimpse the full complexity of the classical style,
which obeys a local harmonic grammar while articulating longer-range
melodic connections and also grouping music into conventional four-unit
patterns, sometimes instantiated on multiple temporal levels, all occurring
inside scales that follow higher-level modulatory pathways and alongside
complex networks of motivic and thematic relationships which may
themselves make topical and generic references, while invoking the large-
scale conventions of forms such as the sonata. And that is completely
ignoring rhythm, orchestration, timbre, and any number of vitally important
phenomena. Music is difficult not so much because of the complexity of its
syntactical systems, but because it juggles so many simultaneously.
And though we can consider these systems separately, they sometimes
interact in complex and interesting ways. One example is when the local
harmonic grammar is overridden across phrase boundaries. In the first eight
bars of Figure 8.5.2, a V–ii progression occurs across mm. 4–5 of a sentence
form. Here the rhetorical structure ameliorates the harmonic hiccup: rather
than a baldly retrofunctional V–ii, I hear a first musical thought that ends on
the dominant followed by a new thought beginning on the supertonic. Such
examples show what is at stake in questions about the nature and location of
cadences; for whatever we call these I64–V progressions, we want to
recognize them as punctuations, important junctures that weaken the grip of
the local harmonic constraints.52
More common are those cases in which harmony supports grouping
structure. Chapter 6 noted that dominants and tonics are oriented oppositely
in time: dominants are approached freely, but constrained in how they move;
tonics move freely, but are approached in only a small number of ways
(§6.4). This differential orientation is fundamental to the logic of the period,
which begins with maximum harmonic stability, and hence maximal
uncertainty about the future, progresses to a high degree of harmonic
instability (and hence predictability), and then retraces its steps to achieve
repose.53 From this point of view, one might be tempted to consider Figure
9.7.1 an eight-measure period in which IV plays the role of the dominant,
appearing as the final chord of the first four-bar unit and leading to I at the
end of the second. This intuition motivates Drew Nobile to define
“dominant” and “tonic” in terms of the phrasal and entropic tendencies in
Figure 9.7.2. In effect, Nobile uses “dominant” to refer to a particular
position within the phrase, rather than within the more abstract harmonic
cycles that make up the phrase. I prefer the more traditional terminology, in
part because I prefer its linking of functional names to particular scale-
degree collections. Nevertheless, I feel the force of Nobile’s insight: there
are genres in which phrasal convention can lead us to hear a wider range of
harmonies as functioning like “dominants” in the broadest sense of the term
—unstable sonorities creating strong expectations of a coming tonic.54 Both
phrase structure and harmonic cycles can create tension and release.

Figure 9.7.1. Drew Nobile analyzes this passage from the Eagles’ “Take It Easy” as a period structure
with the subdominant chord playing the role of the dominant.
Figure 9.7.2. The energetic trajectory of an abstract period; it begins with low tension, moves to high
tension at the half cadence; and retraces its steps to end with low tension.

It is sentences, however, that present the most complex interaction

between the musical systems. We saw in §9.1 that classical composers use
dual transpositions Txty to transfer material from one harmony to another.
This gives rise to a fundamental sentence-form default whereby a melodic
fragment, harmonized by I–V, is answered by V–I.55 Here the demands of
thematic parallelism often lead to an interesting phenomenon: if the I–V
statement uses, say, descending voice leading (options A–B on Figure 9.1.3),
then the response will typically use descending voice leading from V to I
(options 1–4); and taken together the sentence will traverse a loop in chord
space, answering “V just below I” with “I just below V” (or alternatively,
answering “V just above I” with “I just above V”). The requirement of
motivic similarity thus ensures that motives will ascend by step along a
descending chordal background, producing scale steps at the surface.56
Chapters 6 and 7 argued that functional tonality makes particular use of I–
V–I voice leadings that travel a full circle in triadic space, rather than more
efficient voice leadings that often feature in modal counterpoint. We now see
that the conventions of sentence form endow these circular journeys with
rhetorical significance, the first half of the circle posing a musical question
that is answered by the second. Figure 9.7.3 analyzes the melody of the
second movement of Mozart’s K.282 as moving inside harmonies that act
like very small scales; its core claim is that the surface voices move contrary
to the underlying voice leading, ascending by chordal step as the chords
descend. Here we also have a clear example of a nonadjacent melodic
connection, the m. 1 D5 and the m. 3 C5 belonging to the same abstract
voice.
Figure 9.7.3. The opening of Mozart’s K.282, II.

This sort of analysis makes no reference to organicism, prolongation,

representation, verticalization, register transfer, obligatory register, or any
other controversial concepts of Schenkerian theory; instead, its assumptions
are those of standard scalar discourse, albeit at an unusual hierarchical level.
And though we are far from Schenker’s complete theory, we can see
glimmerings of a characteristic insight: that even when chords act like
abstract fields-of-play for concrete voices, they can still articulate interesting
contrapuntal structures. Schenker the musician was right to point to this
phenomenon, even if Schenker the metaphysician described it in ways that
may now seem problematic.

8. Beyond the phrase: hierarchy at the level of the piece

My belief is that large-scale form is produced by collage-like arrangements

of smaller sections whose differences in mood, mode, rhythm, and texture,
conjoined with thematic returns and contrasts, collectively form a coherent
and pleasingly dramatic experience.57 This play of energetic trajectories, of
similarity and contrast, is at the core of musical storytelling, equally present
in classical symphonies, free improvisation, and well-sequenced rock
albums. Its resistance to enlightening general description is one of the main
embarrassments of music theory—for the very features that most powerfully
stoke our musical interest, leading us to sit quietly for long stretches of time,
are also the most resistant to rational elucidation. Which is not to say that
they are beyond description, but rather that our words tend to add little to
what is obvious from simply listening.
This, I think, creates pressure to fill the void, to find something useful to
say about the nature of musical form. Earlier theorists often turned to
motive, proposing that nonobvious thematic connections contributed to a
sense of musical unity, explaining why some pieces are formally convincing
while others are not.58 This has always seemed bizarre to me, as it feels
obvious that a great piece can contain multiple unrelated themes or motives,
much as a great novel can contain many different episodes. Another option is
to follow Schenker in asserting that entire movements embellish the same
harmonic-contrapuntal templates that govern individual phrases. What
makes for large-scale continuity, on this account, is ultra-long-range
connection between pitches.
This approach is not obviously wrong in principle. Imagine for example
that classical sonata-forms tended to follow a first theme emphasizing 3̂ with
a second theme, in the dominant, that emphasizes 2̂ (locally 5̂). And imagine,
furthermore, that these emphases were strong enough that listeners,
performers, and analysts tended to agree about them, even without any
special training. Such a situation would suggest that composers were indeed
concerned with constructing movement-long trajectories of melodic
emphasis. (Whether listeners could hear these relationships, and whether
they were aesthetically significant, are of course further questions.) In actual
fact, however, the process of Schenkerian reduction is a good deal less
straightforward than this: constructing a long-term Schenkerian narrative
requires a host of nonobvious decisions about which notes are structural, and
these are often influenced by the desire to construct a satisfactory analysis.59
One reasonable conclusion is that functional music simply does not exhibit
intersubjectively identifiable trajectories of the sort Schenker postulated.
Any appearance to the contrary is made possible only because analysts put
their finger on the scale, gathering data with one eye toward making the
results vindicate the theory.
Putting that aside, there is a more important question about the aesthetic
relevance of large-scale pitch structure—about why anyone should care
whether a given piece contains the nonobvious relationships that
Schenkerian analysis purports to uncover. Imagine a long piece that exhibits
these kinds of long-range connections while presenting a satisfying sequence
of themes, moods, local pitch structures, and characters. Now suppose we
introduce small modulations that distort the pitch structure while preserving
the sequence of themes, moods, and characters. If the long-term pitch
relationships are aesthetically important, then these shifts should have a
significant impact on our appreciation for the overall work, yet it seems
implausible that they would: indeed, a non-absolute-pitch listener, not
intimately familiar with the piece, might not even notice them. How could
aesthetic value rest so squarely upon relationships with such little perceptual
import? Schenkerian theory seems caught between the empirically
implausible claim that we are sensitive to long-term prolongational structure
(without realizing it) and the equally implausible philosophical claim that
unheard relationships can be primary bearers of musical value.60 In this
respect, it is comparable to twelve-tone composition, two modernist
ideologies placing a heavy emphasis on relationships that are difficult if not
impossible to hear.
Many theorists therefore prefer to model long-term harmonic relationships
using keys and scales rather than Schenkerian linear connections.61 This is a
significant improvement, for it is indisputable that a large number of
functionally tonal pieces nest local chord progressions within formally
similar motion on the level of the scale. We have seen that the two processes
can be represented using the very same spiral diagrams, a hierarchical self-
similarity linking local musical details to progressions spanning entire
movements. Furthermore, scales and keys are empirically unproblematic to
the point where computers can identify them; by contrast, we have achieved
essentially no success at automating Schenkerian analysis.62 Thus at a purely
conceptual level, key relationships deliver uncontroversial long-term
hierarchical structure at least broadly analogous to that hypothesized by
Schenkerians.
Yet once again there are significant issues of perceptibility and aesthetic
relevance. Hepokoski and Darcy write that sonata-form recapitulations
deliver “the telos of the entire sonata—the point of essential structural
closure (ESC), the goal toward which the entire sonata-trajectory has been
aimed . . . the exposition’s structure of promise (presented there in the
dominant) finds here its goal and resting-point (in the tonic).”63 For Caplin,
recapitulation “functions to resolve the principal tonal and melodic processes
left incomplete in earlier sections” and “resolves [the] fundamental conflict
of tonalities when the subordinate theme is transposed back into the home
key.”64 When I read these sorts of claims, I try to imagine myself as a young
student who does not find long-term key relations to be particularly salient.
Would I worry that my perception was inadequate? Would that lead me to
contemplate leaving music? Or would I be tempted to misdescribe my own
perception, pretending to hear the relations that the experts told me to hear?
The language of “ultimate goal” and “resolution” suggests that listeners who
do not experience recapitulatory closure are in some sense defective.
There is also the worry that sonata-form recapitulations involve a degree
of repetition that is out of character with the dynamic qualities of the rest of
the form, particularly in those pieces with fairly literal recapitulations.65 In
shorter pieces, repetition can create a pleasurable rhythm of departure and
return; but as we consider larger timescales, the expressive purpose of
repetition becomes increasingly obscure—comparable, perhaps, to ending a
movie by repeating its first twenty minutes.66 Here we can start to sense a
conflict between the classical style’s twin inheritances of dance music and
opera, the one emphasizing sectionalized repetition, the other musical
storytelling.67 This conflict was aptly diagnosed by John Philip Sousa, one
of the first composers to regularly begin and end pieces in different keys:
 As a child I was brought up on band music. As I grew I noticed something about the marches
of that day—they did not climax. Speaking gastronomically, when they got through with the
ice cream they went back to the roast beef. And the beef had no new sauce on it, no new
flavor.68

Sousa bequeathed this attitude to ragtime composers like Joplin, yielding a

short-lived and uniquely American rebellion against recapitulation and long-
term tonal closure—a musical flexibility not heard since the sixteenth and
seventeenth centuries.69 This rebellion reflected the theoretical conviction
that long-term tonal closure is not necessary for the creation of satisfying
musical statements.
To be clear, Formenlehre theorists have made vitally important
contributions to recent theory, providing a wealth of new analytical concepts
that help sensitize us to classical-style conventions—including “medial
caesura,” “continuous exposition,” “trimodular block,” “standing on the
dominant,” “ ‘tight’ vs. ‘loose’ organization,” and the “one more time
technique.” This work is broadly schema-theoretical in character,
supplementing Gjerdingen’s more local additions to our analytic
vocabulary.70 Like Gjerdingen’s work, it has reoriented theory away from
long-term pitch-narratives, whether scalar or Schenkerian, and toward more
concrete and palpable gestures. One benefit of this shift is that it
deemphasizes justification, the project of explaining why “masterworks” are
great, in favor of a more constructive and concrete goal of informed
listening. If there is a difficulty here, it is the philosophical question of
whether it is important for the listener to consciously track these formal
markers and conventions.71
For a music theorist, learning to live without aesthetic reassurance is like
tightrope-walking without a net. It leaves us in a place where nothing
guarantees the aesthetic quality of the music we love: not motivic
consistency, not the Urlinie, and not the long-term resolution of modulatory
dissonance. But this is where we stand with respect to most of the arts.
Nothing guarantees the coherence of Guernica, or To the Lighthouse, or The
Blues Brothers other than our love for the work itself. The amazing thing is
that musical discourse has managed for so long to nurture the hope that there
might be something more—that we might be able to prove through reasoning
and analysis that a piece was satisfactory or coherent or organic or a
masterwork. These efforts have gone hand-in-hand with an unquestioned
conviction that the classical sonata is the paradigm of musical coherence,
with recapitulation providing a powerful sense of resolution or completion. I
do not expect to convince readers to share my doubts about any of this, but I
would hope to convince them that those doubts are reasonable, and that
students who share them are not defective but merely different. Otherwise,
we risk making agreement with orthodoxy a prerequisite for admission into
the discipline.

1
For similar views see Meyer 1996, p. 259, and Webster 2010 (“multivalence”). The approach also
resembles the “post-human organicism” of Watkins (2017) and Arndt (2019).
2
Rameau (1722) 1971, p. 155.
3
The technique seems to be genuinely schematic insofar as it is usually applied to phrase-initial
motives using tonic and dominant harmonies rather than motives more generally.
4
The upper-voice dyad moves by a loop (t1) and a slide (T1) on the 2-in-3 spiral diagram; the
three-note scale itself oscillates between D–F♯–A and C♯–E–G.
5
My analysis highlights the ascending basic voice leading (B, G) → (D, F ♯ ) to emphasize the
continuity with the following gesture.
6
Of course, no sensible person would imagine this to be so: traditional harmonic theory was never
meant as a recipe for producing beautiful music, any more than English grammar provides a paint-by-
numbers kit for producing beautiful poetry; the grammar plays a necessary but not sufficient role in
the production of compelling functional music. Critics of traditional harmonic theory sometimes treat
its limitations as fundamental flaws (Schenker [1930] 1997, pp. 1–9, Gjerdingen 1996 and 2015).
7
To be clear, this is only a very rough approximation of a complex issue; cadential weight is
affected by such factors as rhythm and temporal duration.
8
Here the algorithm chooses a voice leading Ci→Ci + 1 from the actual voice leadings in Bach’s
chorales.
9
Caplin 2004 rejects this term, but I think it captures the point that functional tonality extends the
harmonic logic of the cadence to the rest of the phrase. Caplin also tends to draw a sharp distinction
between “cadence” and “noncadence,” whereas I think the distinction is inherently fuzzy.
10
Sewell 2021 notes that Schoenberg provided simple melodic reductions along the lines I have
been describing.
11
Laitz 2008, Clendinning and Marvin 2016. Caplin 1998 has a similar view.
12
Though it derives from Schenker, the Laitz-Clendenning-Marvin phrase model is often presented
in starkly harmonic terms, shorn of the linear features fundamental to traditional Schenkerianism.
13
Tymoczko 2011a, chapter 5.
14
These later reductional stages tend to eliminate all the nontonic chords between the initial tonic
and the final cadence, regardless of the musical surface.
15
As recently as 2011, it was possible to publish an article in a prominent music-theory journal
broaching the question of how Schenkerian theory might, in principle, be tested (Temperley 2011a).
Morgan 1978 considers antecedents to Schenkerian analysis, all of which are compatible with the
version of traditional theory I articulate here; the distinctively Schenkerian focus on very long-range
pitch connections has little or no theoretical precedent.
 16
The fact that Schenkerian theory is associated with racist aesthetic opinions adds moral urgency
to these already pressing problems (Ewell 2020).
17
Note that the bass reaches 5̂ for the cadence while the melody hints at a 3̂–2̂(–1̂) descent,
returning to the third scale degree for the repeat. Carl Czerny ([1830] 1848, I, p. 92) has an alternative
voice-leading reduction that departs more radically from the musical surface.
18
These chords, A–d♯ ø43–d♯°43–d♯°65–d♯°65[ ♭ 3], transform A-as-dominant into A-as-tonic by
way of a French sixth that continues the ascending-third triadic sequence. Rings 2011 considers these
sorts of changes of scale-degree function.
19
For more on the “second practice,” see Kinderman and Krebs 1996 and Cohn 2012. The
particular idiom used here is the descending major-third version of the game described in §8.5 of
Tymoczko 2011a. My analysis has many points of overlap with that in Meyer 1996 (pp. 93–97); it
contrasts with those of Kramer (1990) and Subotnik (1991), who both read this piece as challenging
the very idea of analytical reconstruction. Hatten 2014 offers a sensitive discussion of the work’s
poetics.
20
The Op. 24, no. 4 mazurka similarly decomposes its final V–I cadence into a pair of descending
thirds in accordance with the logic of the 3-in-7 spiral diagram. Like the opening of Op. 7, no. 2, the
V–iii–I can likewise be understood either as a nod toward modality or as a drawn-out suspension.
21
See §4.10, Tymoczko 2011a.
22
Schenker (1933) 1969, p. 58.
23
See Schmalfeldt 1991 and Cohn 1992.
24
See Neumeyer 1987, Temperley 2006 and 2011, and Froebe 2015 for related views. It may be
that Schenkerian schemas have their origins in something more than arbitrary convention—for
example, the preference for descending stepwise melodies may be rooted in the physiology of
respiration (chapter 2).
25
Beethoven’s theme also recalls his earlier Op. 18, no. 5 minuet. Its I–V–I–IV progression would
be a falling fifth sequence were the initial chord F♯ instead of E. This same near sequence appears as a
chord progression in Figure 6.1.3 and as a succession of keys in Bach’s D major prelude (§7.3).
26
The location of the final cadence is interesting, and one could make the case that it belongs on
the downbeat, with both melody and bass delaying the arrival until beat 3.
27
As Robert Morgan put it, in Schenker’s theory, “stepwise motion forms the basis for all melodic
content” (Morgan 2014, p. 19).
28
It is also the diatonic inversion of the two-voice fauxbourdon framework at the start of the fifth
Brandenburg concerto (Figure 7.5.2). Yust 2015b refines Schenker’s reading of the piece.
29
The direct shift from B ♭ 7 to G7 is an example of what I have called “the minor-third system”
(Tymoczko 2011a). The switch from descending thirds to descending steps occurs in many Beethoven
pieces, including Op. 31, no. 3, I, mm. 68–70.
30
The middle section answers 5̂–4̂–3̂–2̂–1̂ in G minor with 6̂–5̂–4̂–3̂–2̂ in C minor, shifting the
rhythmic emphasis from tonic to dominant in preparation for the I64–V arrival.
31
Schenker’s reading ignores the B♭65 in the middle section, reducing it to a G minor triad.
32
Schenker’s invention of C7 in the opening is reminiscent of Dahlhaus’s denial of the autonomy
of vii°6 (§6.4).
33
“It should be evident now that the analytic procedure is one of reduction; details which are
subordinate with respect to larger patterns are gradually eliminated” (Forte 1959). Arndt (2018, p. 39)
argues that the focus on surface-to-depth reduction is a later accretion; Schenker proposed an idealist
theory of the Genius’s creative process, which proceeds from the background to foreground. In this
book, I am concerned with Schenker as he has been received rather than the historical figure.
34
Gjerdingen and Bourne 2015.
35
“Always the same but not in the same way,” Schenker’s motto, appearing at the front of Der
Tonwille and later works.
36
This is precisely why some theorists like the notation V64--5 3; my own preference is to notate
the scale degrees and trust that readers understand the idiom.
37
It is also clear that musical experience is often multileveled and hierarchical: we can hear a tonic
scale degree as decorating the third of a V chord that is itself unstable relative to the tonic. My
explanation of this phenomenon is that the tonic scale degree is unstable relative to the local harmony
while the local harmony is itself unstable relative to our functionally tonal harmonic expectations
(§7.1).
38
This example comes from Muns 2008.
39
Other examples include the opening of the B ♭ major prelude from the first book of The Well-
Tempered Clavier (featuring diatonic chords with a very weak dominant function), Chopin’s Mazurka
Op. 50, no. 3, mm. 53–61 (featuring secondary subdominants), and Figure 9.6.6 (secondary dominants
embellishing descending thirds). Salzer (1952) 1962 uses these brief moments of harmonic recursion
to argue against traditional harmonic theory.
40
Among those who have made this connection are Bernstein 1976, Lerdahl and Jackendoff 1983,
Katz and Pesetsky 2011, and Rohrmeier 2011 and 2020.
41
Brown 2005 (pp. 76ff) and Rohrmeier 2011 both make this argument, though they stop short of
explicitly demonstrating that their recursive principles actually account for the regularities in
functional pieces—or, what is more difficult, demonstrating that their recursive grammars rule out the
patterns we do not find in functional tonality. Rohrmeier 2020 seems to acknowledge that the
hierarchical approach does not expand our ability to describe what happens in actual pieces, merely
functioning as a theory of analytical intuitions about music.
42
For more examples, see Figures 10.4.7 (I–ii–IV–V upon removing the applied dominant) and
10.6.5 (V–ii6) and the end of the trio of Beethoven’s Op. 28, III, where D64, functioning locally as a
cadential 64, moves via secondary dominant to B minor and the end of the movement.
43
Rothstein 1989 emphasizes the classical style’s eight-bar norm.
44
Richards 2011 calls these “trifold sentences.”
45
Rothstein 1989, pp. 125–30, considers the origin of these phrase templates.
46
Lerdahl 2020, p. 89, makes the same point.
47
Note that this is the same ascending-fifth schema I identified in the Op. 22 minuet, on the same
pitches: IV–I–V in B♭ major (§10.1).
48
Caplin 2013 observes that we can analyze phrase structure in terms of motivic material, melodic
strategy, or harmony and bass. Temperley 2018, p. 90, favors a grouping-based strategy for reasons
similar to mine. Brody 2016 pursues the issue in a more general way, proposing “thematic design” and
“tonal structure,” as axes along which different theoretical approaches might disagree. In this section I
focus mostly on thematic design and motive; in the next I consider tonal structure.
49
Caplin 2009b allows for nested formal functions, but conceives them as generalized
“beginning/middle/end” patterns rather than the specific grouping strategies characteristic of sentence
and period.
50
See Caplin 1998, p. 111. Although Caplin in principle allows that cadential function is
determined by myriad factors, in practice he is almost exclusively focused on the bass: this leads him
to describe Beethoven’s entire theme as having cadential function although it is the start of the second
group. In this context, Richards argues that tonic chords are only “weak indicators of cadential
function” (2010, p. 26).
51
Caplin analyzes the sequential second theme of Beethoven’s Op. 2, no. 2, I as a “continuation”
(1998, pp. 112–13).
52
This provides another reason to reject Caplin’s strict division between “cadence” and
“noncadence.”
53
The association of tension with certainty about the future, and repose with uncertainty, may seem
counterintuitive; but recall that the tension is associated with the sense that the chord has to move in a
particular way. The uncertainty of repose is associated with the fact that the tonic has no analogous
obligations.
54
Nobile’s perspective derives from Lerdahl 2001, and recalls Long’s (2018, 2020) proposal that
we consider rhythm and phrasing as essential to functional tonality. These perspectives are closely
related to the ideas discussed in §9.2; indeed, one way of interpreting the Schenkerian notion of a
“structural dominant” is as “the dominant that occurs when we are expecting the end of the phrase.”
Here, phrasal role supersedes note-content as a determiner of function. This approach works best with
very regularly phrased music.
55
Caplin and Schoenberg present the opening of Beethoven’s Op. 2, no. 1, as a model sentence; to
my mind, however, it is unusual insofar as its basic idea has just one chord, giving it a more urgent
and developmental character than the typical Mozartian sentence (I–V, V–I). See BaileyShea 2004.
56
This is the origin of the “changing-note archetype” discussed in Gjerdingen 1988, later
rechristened the “Meyer” schema.
57
See Levinson 1997. Explaining what makes a piece “coherent and pleasingly dramatic” is no
easier than explaining what makes a good movie.
58
Reti 1951.
59
Marvin 2012–2013 points out that this challenge is particularly acute in the case of sonata-form
recapitulations. There is also the additional challenge of devising a Schenkerian analysis that is
compatible with traditional formal divisions (P. Smith 1994, C. Smith 1996, and Brody 2021).
60
This thought experiment is inspired by the experiment of Cook 1987, later repeated by Marvin
and Brinkman 1999. Gjerdingen 1999 expresses skepticism about some details of Cook’s experiment
even while endorsing his general conclusions.
61
Yust 2018 emphasizes keys within a largely Schenkerian framework, though this does not lead to
major changes in analytical practice.
62
Kirlin 2014.
63
Hepokoski and Darcy 2006, p. 232. On the same page they write: “The attaining of the ESC is
the most significant event within the sonata . . . It is here that the presence of the tonic becomes finally
secured as real rather than provisional.” In private conversation, Hepokoski stressed that he is
speaking of a compositional norm and making no claims about perception; nevertheless, he continues
to use the term “resolution” in this context (Hepokoski 2021, p. 12).
64
The first quote is from Caplin 1998, p. 161, continuing “and to provide symmetry and balance to
the overall form by restating the melodic-motivic material of the exposition.” The second is from
Caplin 1998, p. 163.
65
Scott Burnham speaks of “the embarrassment of trying to fit a large-scale repetition into a
dramatic narrative” (1995, p. 18). This embarrassing combination of drama and formulaic repetition is
echoed by Debussy’s alleged wisecrack about smoking a cigarette during classical-form developments
(cf. Rosen 2001, p. 117, suggesting the piece was by Beethoven, or Taruskin 2005, vol. 3, p. 22,
suggesting it was by Brahms). William Marvin (2012–2013) notes that an analogous problem arises
within the confines of Schenkerian theory.
66
Margulis 2013.
 67
Rothstein observes that the classical style’s origins in dance music and opera were emphasized
by Leonard Ratner and Charles Rosen respectively (1989, p. 130); to these, he adds popular music as a
third source.
68
The quotation comes from a 1918 interview with the Boston Post, quoted in Warfield 2011, p.
299.
69
In a fascinating 2017 article, Yoel Greenberg suggests that recapitulations arose from a
concatenation of two different practices: “the double return of the opening theme in the tonic in the
middle of the second half of a two-part form, and the thematic matching between the ends of the two
halves of two-part form.”
70
Byros 2015 also makes the connection between formal conventions and scale-degree schemas.
71
Hepokoski writes: “To feel this music musically—to identify with this music as music—is to
monitor the moment-to-moment tensions of the drives toward the succession of boundary cadences”
(2021, p. 8). I find it difficult to believe that typical eighteenth-century classical-music audiences
engaged in this sort of “monitoring.”
 Prelude
Why Beethoven?

Leonard Bernstein’s The Joy of Music opens with a question about

Beethoven’s preeminence. Acknowledging that the composer’s materials are
often simple—even simplistic—Bernstein wonders how it could be that
Beethoven is still the paradigmatic classical composer. (The assumption
seems to be that complexity is inherently valuable, foreclosing the
possibility that Beethoven’s music could be attractive because of its
simplicity, but we will put that aside.) Bernstein’s answer is form,
understood as the ability to write music that seems like the perfect or
inevitable continuation of what had come before; this leads to the inevitable
reflections on genius, artistic perfection, and the existence of God.1
The preschool motto “don’t yuck my yum” warns against criticizing
musical theodicies. But I think we can find a more practical approach to
Beethoven’s form. Figure P10.1 compares the density of attacks in the
exposition of the first movement of Beethoven’s Waldstein and the entire
first movement of Mozart’s first piano sonata. There are two notable
differences. First, the variation in Beethoven’s rhythm is more extreme than
Mozart’s, passing through a wider range of musical states; second, the cycles
of change are slower in Beethoven, with rhythm helping to demarcate formal
regions. Indeed, the Waldstein could hardly be more obvious about this, with
the first theme moving from eighths to sixteenths and the second theme
accelerating from half notes to triplet eighths to sixteenths to a trill. Rhythm
not only differentiates the exposition’s main formal zones, but also marks the
progression within them. It is hard to think of a previous composer who
habitually manipulated such large spans of musical time.
Figure P10.1. Attack density in the Waldstein’s first-movement exposition and the first movement of
Mozart’s first piano sonata.

Much the same point could be made about the use of register in the first
movement of the Tempest (Figure P10.2). Once again we see Beethoven’s
music oscillating dramatically, where a typical classical piece would stay in
a narrower range. And once again we see Beethoven using a basic musical
parameter in a form-defining way, most notably by constructing a long
registral arch linking transition and second theme—eliding the medial
caesura that typically separates those two formal regions. Registral
separation also seems to be a significant parameter, with the hands
expanding to the edges of Beethoven’s piano in the middle of the second
theme; this then initiates a grand contrary-motion sequence that returns to a
more normal configuration.2
Figure P10.2. Registral position of the two hands, averaged over a four-beat window, in the first
movement of the Tempest sonata.

This interest in extremes evokes the opposition between beauty and

sublimity. Eighteenth-century philosophers associated beauty with the
pleasant and well-formed: flowers, landscapes, small birds. The sublime was
a form of aesthetic pleasure associated with the overwhelming, awesome, or
incomprehensible: storms and mountain ranges, the ruins of a great city, the
idea of distant galaxies, Shelley’s Ozymandias. In music, beauty was
associated with the decorous: elegant major-key melodies stated at a
moderate tempo and dynamic.3 Even during his lifetime Beethoven was
recognized as the great composer of the sublime—wild and almost out-of-
control, exploiting unprecedented contrasts to create extremes of emotional
response.4 In this respect, his formal innovations are precisely not a matter
of “inevitability,” at least insofar as “inevitable” is kin to “appropriate” or
“tasteful”: instead they are closer to provocations, a willingness to tolerate
excesses other composers avoided—and in some cases, to embrace an
obviousness that might seem naïve.
This gives us a vantage from which to think about Beethoven’s oft-noted
anti-melodic impulse. Where Mozart and Haydn frequently oriented their
music around clear and singable tunes, Beethoven had a fondness for
material that was almost deliberately generic. Consider Figures P4.2, 3.2.4,
or 7.2.7. Here we have something like sub- or proto-music, unrefined
nuggets of musical ore. Similarly, the second theme of the Waldstein is
deliberately featureless, singable but not worth singing, and harmonized by a
stock Pachelbel progression (Figure P10.3). For me, the frequent banality of
Beethoven’s themes is inseparable from his focus on long-term formal
processes, as if he deliberately subordinated local interest to long-term
narrative.5 (In much the same way Beethoven’s harmony often consists of
simple tonic and dominant alternations, with many fewer secondary diatonic
triads than in J. S. Bach or Brahms.) Twenty-first-century listeners are
familiar with the gestural music of Xenakis and Penderecki, who rejected
serial pitch manipulations in favor of striking trajectories of force and line.
Strange though it may seem, I hear Beethoven as a kindred spirit, a
composer emphasizing larger shapes over local ornament and detail. Indeed,
there exists an atonal recomposition of the Waldstein’s exposition that
preserves a remarkable amount of its formal clarity and expressive intensity,
being surprisingly enjoyable in its own right.6 Atonal Mozart would just
sound awful.

Figure P10.3. The second theme of the Waldstein’s first movement.

 All of which gives us reason to suspect that Beethoven’s aesthetic
concerns may differ from those of his predecessors—and in particular, that
he was interested in loosening the formal schemas governing the classical
style.7 Hepokoski and Darcy orient their sonata theory around the perception
of a “medial caesura,” the articulating pause that conventionally precedes
second themes. Yet Beethoven’s tendency to avoid this gesture can be seen
as early as his first C minor sonata, where a passage of “standing on the
dominant” leads without pause to the second theme (Figure P10.4).8 And
while the Waldstein contains a pause between transition and second theme,
its “filled caesura” articulates a rhythmic deceleration linking the sixteenth-
note transition to the quarter- and half-note second theme. Here the fact of
the caesura takes a back seat to its function in carrying out the form-defining
deceleration (Figure P10.5).9 The danger is that we will interpret Beethoven
as more beholden to formal convention than he actually was: paradoxically
he is simultaneously eccentric and archetypal, an extreme outlier and the
classical composer par excellence.

Figure P10.4. Measures 44–66 of Beethoven’s Op. 10, no. 1, I, which move from transition to second
theme with no medial caesura. Hepokoski and Darcy describe this moment as a “caesura fill” even
though it could also be a “standing on the dominant” (albeit at a quiet dynamic); had the music
proceeded to a pause before its second theme, this passage almost certainly would have been identified
as such.
Figure P10.5. The Waldstein’s “filled medial caesura” features rhythmic deceleration.

Formenlehre theorists would no doubt respond that I am hearing

ahistorically, failing to grapple with sonata form as an a priori norm
governing the expectations of historically situated listeners. My response is
that we have very little evidence that sonata form in fact functioned this way,
particularly for listeners who were not professional musicians.10 No doubt
there were some who enjoyed tracking a piece’s progress through its
preordained stations, just as some contemporary audio engineers consciously
attend to the mixing and production of songs on the radio. But it seems
unlikely that this mode of listening was either a default or an ideal, or that
early nineteenth-century musical culture prioritized the technical signposts
of sonata form over and above its affective sequence of moods and gestures.
And even if that could be shown for the more conventional music of the late
eighteenth century, it is not at all clear that the conclusion would generalize
to Beethoven, a self-conscious revolutionary whose music evinces a lifelong
interest in new modes of musical continuity.
Thus I would argue that there is a conflict between Beethoven’s nascent
Romanticism, his compositional self-reliance, and recent approaches to
musical form. Caplin, Hepokoski, and Darcy all seem to want theories that
apply equally to Haydn, Mozart, and Beethoven, but Beethoven may have
been more ironic, more historically conscious, and more iconoclastic than
his predecessors.11 And so we find him starting pieces in medias res, writing
the Heilige Dankgesang, putting second themes in unusual keys, writing
periods without root-position cadences, omitting the medial caesura,
expanding the nonharmonic domain, and generally exploring alternatives to
inherited convention. To read him as a classical composer, a courtier
elegantly retracing a preordained sequence of sonata-form dance steps, is to
neglect those aesthetic qualities that can still inspire us, two hundred years
later. For his attitude seems to have been that music is fundamentally under
our control, that we can change the rules when we want to—that we can, for
want of a better phrase, take ownership of style itself. The result is music
with genuine philosophical content, implicitly humanist where so much
theory inclines toward scholasticism.

1
Bernstein 1956.
2
We would find a similar contrast were we to consider dynamics, with Beethoven’s ranging from
pianissimo to fortissimo rather than the forte-and-piano opposition of many classical sonatas.
3
Edmund Burke (1757) 1999, Part 3, chapter 25: “the beautiful in music will not bear that loudness
and strength of sounds, which may be used to raise other passions; nor notes which are shrill, or harsh,
or deep; it agrees best with such as are clear, even, smooth, and weak. [In addition,] great variety, and
quick transitions from one measure or tone to another, are contrary to the genius of the beautiful in
music. Such transitions often excite mirth, or other sudden and tumultuous passions; but not that
sinking, that melting, that languor, which is the characteristical effect of the beautiful as it regards
every sense.”
4
Hoffmann 1989.
5
Dahlhaus makes a similar point (1991, p. 83).
6
Lalite et al. 2009.
7
Dahlhaus 1991, Schmalfeldt 1995 and 2011.
8
Similar points can be made about the second theme of the first movements of the Eroica (where
there is no pause at all between the beat 1 dominant arrival and the beat 2 start of the main theme) and
the Tempest, to be considered subsequently. See Richards 2013a.
9
Compare the medial caesura in the first movement of the Fifth Symphony, which expands and
transforms the “knock of fate” motive into the start of the second theme (§10.4): here the caesura is
again doing formal work over and above its conventional function.
 10
In fact, we have some evidence that this is not the case, as contemporaneous “programmatic”
readings do not emphasize the medial caesura (e.g., the “Eroica” program in Marx [1859] 1997; see
Burnham 1995).
11
The subtitle of Caplin 1998 is “A Theory of Formal Functions for the Instrumental Music of
Haydn, Mozart, and Beethoven,” while Hepokoski and Darcy introduce their book by writing “from
one perspective the Elements is a research report, the product of our analyses of hundreds of
individual movements by Haydn, Mozart, Beethoven, and many surrounding composers of the time”
(2006, p. v).
 10
Beethoven Theorist

As an undergraduate, I took an ill-fated independent study with David

Lewin, a dazzling speculative theorist and a fine practical musician. My goal
was to write Bach-style fugues, but Lewin thought it would be more fun to
try to imitate Beethoven, a composer I had never really understood.
Predictably, my efforts were hopeless: while I had some minimal sense of
how to impersonate Bach, I was completely stymied when it came to
Beethoven. Over and over, I wrote overcomplicated, histrionic, and entirely
unmusical exercises; each week, Lewin would look at my work and say
something like “it’s really not about harmony, you know”—invariably
pointing to some marvelous passage consisting almost entirely of tonic and
dominant. Afterward, I would go home and once again fail to put simple
chords together in a way that sounded even remotely interesting. My
problem was that I had been made to understand what Beethoven’s music
wasn’t, but not what it was. What makes it work if not harmony?1
Almost three decades later I have the beginnings of an answer: sequences,
schemas, and patterns in general, and a specific set of contrary-motion
idioms in particular. As the first classical composer to grow up playing J. S.
Bach, Beethoven seems to have inherited the baroque love for mechanism—
to which he added his own distinctive emphasis on contrary-motion
counterpoint and primary-triad harmony, all filtered through a uniquely
forceful musical personality. Like Corelli, Beethoven was exquisitely
sensitive to the nearly sequential character of basic functional procedures,
which need only the slightest nudge to become manifestly repetitive. The
resulting play of near and exact repetition poses a challenge for theories that
deemphasize sequence or reject its harmonic significance, as patterns are at
the core rather than the periphery of Beethoven’s language.2
It is just here that we are liable to be distracted by the stereotype of
Beethoven as a composer of overwhelming force, whether divinely inspired
or oppressively patriarchal.3 For Beethoven was an intellectual who took
pleasure in the artful juxtaposition of schematic content—a composer who
poured his admittedly torrential energy through a sieve of utterly standard
devices. What is surprising, perhaps, is that this sort of intellectual play can
coexist with unprecedented rhetorical power. Humans like to put each
another into boxes, imagining that excellence in one domain implies
limitation in another, and it can be difficult to accept that one and the same
figure is both highly cerebral and highly expressive—as when the
supposedly dumb jock turns out to have a talent for mathematics, or when
Scott Burnham’s “Beethoven Hero” turns out to be Dmitri Tymoczko’s
Beethoven theorist.
This chapter will try to recover Beethoven the pattern-maker, highlighting
his ability to spin compelling music out of commonplace material. My
fundamental concern will be to explore how Beethoven built a personal
vocabulary from ready-to-hand material—drawing a life’s worth of music
out of theoretical relationships that other composers might dismiss as trivial.
After two introductory sections, I consider three famous opening
movements: the Tempest, the Fifth Symphony, and the Op. 28 “Pastorale.” I
then turn to Schubert’s Quartettsatz and the prelude to Wagner’s Lohengrin,
two pieces that pursue Beethovenian arguments amid the dissolution of the
classical consensus. In Schubert this is largely a matter of continuing
Beethoven’s quest for greater formal flexibility, but in Wagner it comes to
affect the musical details as well. The five analyses can be read in any order,
and readers will benefit if they study the pieces ahead of time.
One thing I will not emphasize is Beethoven’s tendency to reuse motivic
material, recycling the same ideas, sometimes in nonobvious ways,
throughout a piece. This is partly because I find motivic unity to be
aesthetically uninteresting: to me, it does not matter whether some
wonderful new theme bears a cryptic and nonobvious relationship to some
other music occurring elsewhere in a piece; what matters instead is whether
that new theme fits with what has come before, contributing to a satisfying
energetic and expressive trajectory.4 And partly it is because I am less
interested in motives than in the structures they decorate: when I imitate or
improvise classical-style music, motive feels superficial, a way of stamping
an almost illusory distinctiveness onto a preexisting vocabulary. Reusing a
motive feels neither challenging nor virtuous, whereas constructing a
satisfying melodic-harmonic core feels much more difficult. I worry that
Beethoven’s obsession with motivic unity distracts from his deeper and more
elusive virtues—primarily his ability to construct large-scale statements out
of familiar building blocks. What fascinates me are the features that recur
from piece to piece, rather than the motives distinguishing one piece from
another.

1. Meet the Ludwig

If a chord divides the octave very evenly, then it can be linked by efficient
voice leading to its transpositions. If its notes are clustered close together,
then its voice exchanges will be small. The minor third is interesting insofar
as it is balanced between these two extremes, being halfway between the
unison and the tritone—or in other words exactly one fourth of an octave.
This means that the most efficient voice leading connecting it to its tritone
transposition is exactly the same size as its smaller voice exchange. Two
contrary-motion voices, when restricted to this interval class, will alternate
between these two possibilities: (E ♭ 4, C5) → (C4, E ♭ 5) → (A3, F ♯ 5) →

(F ♯ 3, A5) → (E ♭ 3, C6) and so on.5 In Figure 10.1.1, Fanny Mendelssohn

Hensel uses this pattern in an exact contrary-motion sequence in which the
outer voices always move by minor third.6
Figure 10.1.1. Measure 17 of Fanny Mendelssohn Hensel’s “April,” from Das Jahr, H. 385, presents
a contrary-motion sequence in which outer voices move by minor third to form similar harmonies
(minor thirds or major sixths, 3 or 9 = 12 – 3 semitones). On the right, the schema’s diatonic
quantization.

The diatonic third, being approximately one quarter of an octave, inherits

this symmetry in an approximate way, allowing for sequences in which one
voice sounds thirds while the other alternates thirds and steps. Figure 10.1.2
embellishes this dyadic template using contrary motion, lower neighbors,
and a mixture of the two. In the descending-fifth sequence, one of the lower
neighbors creates note-repetition and is rarely used; in the ascending-fifth
version, both lower neighbors are common. The resulting patterns are
ubiquitous throughout Beethoven’s music, appearing in almost every
movement and producing a characteristically Beethovenian sound. To honor
schema theory’s penchant for cutesy names, I will refer to them collectively
as the Ludwig. A significant portion of Beethoven’s intellectuality involves
these patterns; when combined with parallel motion, the Ludwig is not so
much an idiom as a conceptual scheme, a very general way of approaching
tonal-functional counterpoint.
Figure 10.1.2. The Ludwig embellishes the basic contrary-motion pattern by adding stepwise contrary
motion and lower neighbors.

In development sections, this pattern is often used to produce overtly

sequential passages.7 More interesting are the many thematic statements that
embed it covertly. The simplest Ludwigs create V–I–IV or IV–I–V
progressions as in Figure 10.1.3. Here contrary motion endows primary-triad
harmony with new contrapuntal depth, replacing its fifths-and-steps defaults
with novel outer-voice possibilities (Figure 10.1.4). This fusion of
contrapuntal fifth-progressions, deriving from dyadic voice-leading
geometry, and harmonic fifth-progressions, deriving from functional
convention, is the Ludwig’s essential feature. More complex Ludwigs can
act as V–I–vii°/V or as V/IV–IV–vii°, replacing Gjerdingen’s oscillatory
“Quiescenza” with a more energetic expansion (Figure 10.1.5).8 The second
theme of the first movement of Beethoven’s first piano sonata uses a hybrid
descending-fifth Ludwig to expand the registral range by three octaves; this
culminates in a second contrary-motion passage returning the hands to close
position (Figure 10.1.6). The Pathétique sonata uses the schema at the start
of its first two movements, giving the second movement the character of a
major-mode variation on the opening allegro (Figure 10.1.7).9
Figure 10.1.3. Primary-triad Ludwigs: the opening of the string quartet Op. 18, no. 3, and the second
theme of the Appassionata sonata’s first movement, Op. 57, mm. 36–40.
Figure 10.1.4. The Ludwig as an alternative to the fifths-and-steps pattern of Figure 6.1.3. Here “D”
refers to a generic dominant (V or vii°).
Figure 10.1.5. The Ludwig as a variant of Gjerdingen’s “Quiescenza.”
Figure 10.1.6. The Ludwig schema in the second theme of Op. 2, no. 1, I, mm. 25–32. On the middle
line, the basic thirds-and-sixths structure; beneath that, the embellishing chords. The schema appears
twice so that the hands expand by two octaves.
Figure 10.1.7. The Ludwig in the first two movements of the Pathétique Sonata, Op. 13.

It is also common to find nonsequential Ludwigs that combine an

ascending fourth with an ascending fifth (Figure 10.1.8). Here it can be
useful to distinguish octave-focused Ludwig units where the octave or
unison is the primary harmony from third-focused units that emphasize
imperfect consonances (Figure 10.1.9). Figure 10.1.10 shows a contrary-
motion theme from the end of the exposition of the D major sonata, Op. 10,
no. 3, I; it is a hybrid, ascending-fifth Ludwig, moving from V/IV to IV to I,
whose first unit is octave-focused. Analysts who approach this passage
chordally will miss the significance of the “merely passing” B–G dyad,
which is essential to the schema but not the Roman-numeral harmony.10 The
Rondo of Op. 31, no. 1, opens with the lower-neighbor version of the same
pattern, again treating some schematic dyads as nonharmonic; it progresses
to a version of the schema that recalls the “Morte” (Figures 10.1.11–12).

Figure 10.1.8. Octave-repeating Ludwigs often found in thematic contexts.

Figure 10.1.9. Third- and octave-focused Ludwig units on the primary triads. Asterisks mark
emphasized harmonies.

Figure 10.1.10. The ascending-fifth Ludwig near the end of the exposition of the first movement of
Beethoven’s Op. 10, no. 3, mm. 93–96.
Figure 10.1.11. The ascending-fifth Ludwig in the opening of Op. 31, no. 1, III. Here the voice
leading ascends while the chords descend.

Figure 10.1.12. Another contrary-motion pattern in Op. 31, no. 1, IV, mm. 28–31. The D–F♯ dyad
appears one step higher than in the standard schema.

If there is a historical precedent here, it is likely Clementi, in many ways

the progenitor of Beethoven’s style. Like Beethoven, Clementi
deemphasized tunes in favor of registrally expansive gestures, and like
Beethoven he favored simple tonic-dominant alternation enriched with
secondary dominants. Thematic and sequential Ludwigs can be found
throughout Clementi’s early sonatas, which can sometimes sound uncannily
like the work of his more famous successor. Indeed, Beethoven is said to
have remarked that those who study Clementi make themselves acquainted
with the techniques of Mozart, but not the converse—as if the Ludwig were
a secret that could be acquired only in that composer’s work.11 If Clementi’s
themes are less strongly etched than Beethoven’s, his larger forms less
clearly defined, then this may reflect the difference between someone who
blazes a trail and someone who walks along afterward.

2. From schema to flow

It is also possible to understand the Ludwig as a more flexible network of

compositional options. Figure 10.2.1 augments the basic third-to-sixth unit
by providing a choice about whether to move by descending fourth or fifth.
This new flowchart represents the Ludwig not as a fixed idiom, a
precomposed chunk of music waiting to be plugged into a composition, but
as something malleable: not so much “the Ludwig schema” as “Ludwig-
ing,” a particular kind of musical flow. This figure contains all those
consonances that are invertible at the octave—or, equivalently, those which
can be expanded by third to produce another consonance.12 This means not
just that the hands can be swapped but also that any passage of two-voice
Ludwiging can have either of its voices doubled at the third. Figure 10.2.2
expands the chart to include the fifth: when the sequence is contracting, the
fifth comes before the third at the beginning of each unit; when it is
expanding, it comes after the third at the end of each unit. In some sense
Ludwiging is just contrary motion itself, exploiting the fifth-based geometry
of diatonic dyads (cf. Figure 3.3.8).
Figure 10.2.1. A flowchart for invertible counterpoint at the octave. If two notes share a stem, choose
one.
Figure 10.2.2. Replacing tritones with perfect fifths in the Ludwig’s basic unit.

Earlier, I described the second theme of Beethoven’s Op. 2, no. 1, as a V–

I–vii°/V descending-fifth Ludwig repeating at the octave. We can now say
that the underlying dyadic progression in Figure 10.1.6, E ♭ –A ♭ –D–A ♭ ,
makes four passes through the Ludwig network, comprising a single near
sequence rather than two exact sequences stuck together. The same basic
strategy can be found in a huge range of passages, from the opening of the
second piano sonata to the Op. 109 variations theme (Figures 10.2.3–4).
What makes these passages so characteristically Beethovenian is their
unidirectionality, harmonic and thematic repetition continuing an ongoing
process of registral expansion.13 Other composers are more likely to change
direction: for example, in mm. 11–18 of the Courante from Élisabeth Jacquet
de La Guerre’s Suite in D minor (1687), we have five Ludwig units whose
top voice alternates between descending and ascending, interleaved with
descending-fifth sequences built on the third’s basic voice leading (Figure
10.2.5).14

Figure 10.2.3. Measures 25–31 of Op. 2, no. 2, I, present five successive Ludwig units, with the left
hand octave-displaced.

Figure 10.2.4. The Op. 109 variations theme contains eight consecutive units (§9.4).
Figure 10.2.5. Measures 11–18 of the Courante from Élisabeth Jacquet de La Guerre’s Suite in D
minor (Pièces de clavecin, book 1, 1687) present five Ludwig units that alternate directions.

Contrary-motion Ludwiging can be augmented by the parallel-motion

options shown in Figure 10.2.6; together, these provide a simple but
comprehensive set of outer-voice defaults for functionally tonal music-
making. Here we have essentially adopted the Tinctorian perspective from
the prelude to chapter 3, analyzing functional counterpoint not in terms of
independent voices, but rather parallel and contrary patterns, isolating those
contrapuntal possibilities consistent with the harmonic grammar. And while
the resulting idioms have a schematic character, they are more general than
the independent atoms that usually go by that name, being closer a complete
list than an arbitrary set of building blocks.15
Figure 10.2.6. Functionally tonal parallel-motion schemas.

The interesting question is how the Ludwig can be both fundamental to

functional counterpoint while also characteristic of Beethoven’s music. If the
Ludwig reflects the basic geometry of contrapuntal possibility, then why is it
not equally ubiquitous in Mozart or Haydn or Monteverdi?
I have three answers that can be arranged in order of increasing interest.
The first is that the Ludwig was part of Beethoven’s improvisational
practice, a favorite lick or riff that he drew upon when needed. The Ludwig
can thus be understood as one specific manifestation of his more general
fondness for contrary-motion patterns (Figures 10.2.7–9). Though these
patterns are often simple, they occasionally achieve extraordinary
complexity, as in Figure 10.2.10, which combines free 5–6 ascent in the left
hand with a melody that descends through thirds and sixths. By contrast, a
composer like Mozart was more partial to outer-voice parallelism.
Figure 10.2.7. A contrary-motion wedge in Beethoven’s E minor piano sonata Op. 90, I, mm. 105–
107.

Figure 10.2.8. A passage from Op. 2, no 3, IV, mm. 127–133, that begins with contrary motion along
the F major triad and progresses to an ascending-fifth Ludwig, itself moving by near-contrary motion
along the diatonic scale.
Figure 10.2.9. (top) The contrary-motion pattern from Figure 1.5.6 can be doubled at the third and
extended through a complete octave to produce a series of mostly functional harmonies. (bottom)
Beethoven uses this strategy in the opening of Op. 130, I, mm. 18–19.
Figure 10.2.10. The second theme of Op. 2, no. 1, IV, mm. 34–42, combines free 5–6 ascent in the left
hand with a right-hand contrary-motion melody that descends through thirds and sixths.

Slightly more interesting is the idea that the Ludwig functions as a tool for
producing the dramatic gestural waves characteristic of Beethoven’s style.
The repeating passage from mm. 57–60 in the first movement of the Eroica
is exemplary, part of a massive sonic accretion using register (expanding by
two octaves), dynamics (full-orchestra crescendo), rhythm (quarter notes to
repeating eighth notes and ultimately sixteenths), and orchestration
(trumpets and timpani, Figure 10.2.11).16 From this perspective the Ludwig
is a pitch-domain analogue of the crescendo.
Figure 10.2.11. Measures 57–60 of the Eroica first movement, containing three consecutive Ludwig
units with the last two linked across voices.

The third answer is that Beethoven was theoretically fascinated by the

quasi-sequential qualities of basic functional idioms. I have at various points
mentioned his tendency to make music out of “unrefined tonal ore.”17 It
seems likely that this tendency in turn reflects an intellectual fascination
with phenomena other composers took for granted—leading him to fashion
music that highlighted the beauty of basic musical mathematics. Figure
10.2.12, for example, seems to announce that the I–V–I progression can be a
contrary-motion near sequence, dividing the octave almost in half. Similarly,
the Hammerklavier’s second theme is a four-unit contrary-motion Ludwig
whose fifth-progressions (a) exploit the melodic third-cycle sounded by one
of the Ludwig’s voices; (b) dramatize the underlying mechanisms of
functional harmony, raising its descending thirds to melodic salience, and (c)
insert a descending-third progression to create Rameau’s double emploi, all
without disrupting the melodic thirds (Figure 10.2.13).18 Such passages are
both elementary and extraordinary, utter trivialities that happen to clarify the
fundamental logic of functional harmony. Beethoven’s incessant Ludwiging
may therefore represent a kind of musical understanding, an implicit
knowledge that teeters on the edge of the explicit. If so, it is another
manifestation of a fundamentally intellectual musicianship, an obsessive
circling around theoretical ideas expressed with an unusual compositional
clarity.
Figure 10.2.12. Tonic-dominant-tonic progressions as contrary-motion, near-fifth sequences in mm.
73–76 of Op. 10, no. 2, I.

Figure 10.2.13. One of the second themes in the Hammerklavier’s first movement (mm. 66–68) is a
four-unit Ludwig that connects smoothly back to itself.

3. The Tempest

The first movement of the Tempest is one of Beethoven’s most characteristic

productions—all violence and gesture, largely amelodic, and almost entirely
sequential in its harmony. The reference to Shakespeare is icing on the cake,
connecting two giants of Western culture through a pair of their most iconic
productions. But if the piece manifests the violent and aggressive Beethoven,
then it also hints at a subtler figure—the humorist who repeats musical
fragments beyond what is reasonable, or deliberately lingers on seemingly
wrong notes, or builds extended movements out of quirkily simplistic
material. This is the composer who responded to his brother’s “Johann
Beethoven, land owner” with a sarcastic “Ludwig Beethoven, brain
owner.”19
There is a moment in the recapitulation that epitomizes this more complex
figure: at m. 189, the ascending-fourth transposition would bring the music
past the edge of Beethoven’s piano (F6). Rather than transposing down by
octave, Beethoven fixes D6 as an upper-voice pedal, creating unusual
dissonances as the notes crunch against the piano’s edge (Figure 10.3.1).20
Here the music seems to put itself into question, as if asking about the source
of its power. Sublimity is about exceeding limits, the storm destroying the
harbor’s sea walls or the musician pounding the keys so hard that “the piano
must break.”21 Here, however, the edge of Beethoven’s keyboard seems
impervious to his raging, a real-life barrier easily capable of withstanding the
force that human fingers can bring to bear. How is this failure in any way
sublime?

Figure 10.3.1. The upper-voice pedal in mm. 190–192 of the Tempest.

The answer requires a detour into eighteenth-century aesthetics. While

most philosophers identified the sublime with overpowering external forces,
Immanuel Kant instead associated it with the human drive to transcend our
own finitude—the internal capability that allows us, for any actual
cataclysm, to imagine one even greater. This is the impulse that lets the
toddler ask for a number larger than any you can mention, revealing a
concept of infinity greater than anything instantiated in nature. Kant’s
suggestion was that external forces can sometimes lead us to recognize our
inner resources: contemplation of an external object such as “the starry skies
above” sometimes reveals our physical limits (“annihilates as it were my
importance as an animal creature”), which in turn provokes us to recognize
our “infinite” rational nature (“infinitely elevates my worth as an
intelligence”).22 For Kant it was the human spirit that was truly sublime and
awe-inspiring, rather than external objects or events: monumentality and
excess are the false and illusory sublime.23 True sublimity instead looks
elsewhere, negating the finite and external in favor of a transcendent
interiority. This suggests that art may be able to achieve sublimity by
undermining itself.
I like to imagine that Beethoven fashioned his upper-voice pedal as a
reflection of this more spiritual view. For in this moment the edge of the
piano, symbolized by the pedal, annihilates the music’s physical aspect and
with it the notion that sound could reach beyond itself—its audible warpings
a subtle and humorous jab at Beethoven’s heroic aspirations. But at the same
time, it slyly hints at an esoteric message, a negative theology in which
musical failure points beyond music itself. Indeed, Beethoven alluded to
Kant’s discussion of sublimity in his conversation books, and had on his
writing desk a famously self-defeating inscription from an Egyptian temple
—“I am all that is, that was, and that will be. No mortal man has raised my
veil”—that Kant singled out as “most sublime.”24 Whether Beethoven got
this attitude directly from Kant or whether the two figures independently
shared a similar outlook is relatively unimportant: what matters is that
Beethoven’s sublimity lies not just in extremity and extravagance, volume
and length, but also in a kind of self-conscious reflection. Beyond the
tempestuous Beethoven lies a more thoughtful figure.
(a) Exposition. The exposition is a locus of intense disagreement and a
test for recent ideas about form: some analysts put the main theme at m. 1,
others at m. 21; some hear a second theme at m. 41, while others argue that
the piece has a continuous exposition with no second theme; some identify a
closing theme at m. 75, while others consider that to be a second theme. Part
of the issue is that the piece lacks familiar cadential markers—a well-defined
medial caesura or a definitive cadence before the closing section. Instead,
motive starts to supplant cadence as a determiner of form: the opening
phrase, shown in Figure 10.3.2, presents three figures I call “arpeggio,”
“cascade” (sometimes called “sigh”), and “turn.”25 As many analysts have
noticed, these three figures return throughout the exposition in something
like the way a development rotates through the exposition’s themes (Figure
10.3.3). Particularly interesting are the transition and second theme, which
combine to form a single registral arc (Figure P10.2), moving from arpeggio
(mm. 21–40) to cascade (mm. 42–54) to turn (mm. 55–63), as if expanding
the narrative of the opening.26

Figure 10.3.2. The opening of the Tempest uses a Ludwig to articulate three basic motives.
Figure 10.3.3. The arpeggio, cascade, and turn motives as they appear in the exposition. Each tick
represents one measure.

Yet for all the novelty, it is possible to hear a four-part exposition

containing five large paragraphs: a loosely periodic main theme at m. 1, a
sequential transition at m. 21, a second group at m. 42 (with another theme
starting at m. 55), and a closing section at m. 75.27 To be sure, there is a
sense in which the formal identity of these sections clarifies only gradually:
the opening music initially sounds like it could be an improvised
introduction, but quickly turns into a thematic allegro, while the transition
initially sounds like it could be a thematic statement, but quickly starts
modulating.28 This “becoming clear only in retrospect” continues in various
ways throughout the exposition. If we are alive to Beethoven’s playful
humor, then we will not be shocked by these misalignments between rhetoric
and formal function.29
I understand the opening twenty-one measures as a large period, two
quasi-parallel phrases each beginning with a slow arpeggio, moving to the
cascade, and ending with a cadence.30 The antecedent articulates a Ludwig
that continues across the phrase boundary, perhaps motivating the shift from
d: V to F: V (Figure 10.3.2). The music continues with an ascending-step
sequence whose last unit is repeated and embellished (Figure 10.3.4). That
is, the V65/V–V appears in two slightly disguised forms: first as N6–vii°,
then as vii°7/V–(i64⇒V7), joined by a semitonally ascending diminished
seventh.31 Melodramatic figuration leads to a powerful cadence and the start
of the transition. This begins the third motivic rotation, the ascending
arpeggio figure impassioned and in-tempo, and answered by a metricized
turn.

Figure 10.3.4. The consequent (mm. 7–20) as a disguised ascending-step sequence.

The transitional sequence deploys the technique I have called the

disguised model, altering its initial unit so that the sequential structure
clarifies as the music progresses (another instance of “becoming clear only
in retrospect”). The sequence uses fauxbourdon triads to ascend by fifth, the
phrasing accelerating from four bars to two to one (Figure 10.3.5).32 It ends
with an elongated vii°7/V, discharged on the dominant and ushering in the
second theme. This elongation replaces the medial caesura, acting as a
temporary suspension before the music snaps back into action: instead of
pausing on V before starting a theme on I, Beethoven pauses on vii°7/V
before starting a theme on V. Energetically, it lacks the quality of exhalation,
feeling more like a climax than a rest—and indeed it is near the high point of
the registral arc encompassing both transition and second theme (Figure
P10.2).

Figure 10.3.5. The transition (mm. 21–40) as a disguised ascending-fifth sequence using fauxbourdon
triads.

That second theme features the exposition’s only nonsequential harmony,

a V–i–V–N6–vii°–i6 progression that leads to a harmonized version of the
turn (Figure 10.3.6).33 Formally, it is a sixteen-bar sentence that arrives at a
weak cadence in m. 56; this cadence is immediately repeated as if we were
hearing a phrase extension or “postcadential echo.” However, its third
repetition completes a sentence with a more decisive cadence, leading me to
experience a pair of overlapping phrases, with the two-bar cadence of the
first theme simultaneously the initial presentation of the second theme—the
same technique used in the Hammerklavier scherzo (cf. Figure 9.6.6). This
sense of overlap is reinforced by a repetition that forms a loosely parallel 8 +
12 period. Thus, the first theme never quite ends while the second never
quite begins, and what initially sounds like a cadence turns out, in retrospect,
to be the start of something new.34
Figure 10.3.6. The second theme (mm. 41–57) can be read as a sixteen-bar sentence with a weak
cadence.

The theme’s consequent is tonally subtle. Caplin describes it as an

example of “invertible counterpoint,” the melodic turn A–B ♭ –A–G ♯ –A
moving from melody to bass. But since the reregistration would create
parallel fifths, the counterpoint is not truly invertible; Beethoven thus
changes rhythm and register, introducing a C♯ that allows us to hear (C♯, E)
→ (D, F) as a tonicization of D minor (Figure 10.3.7).35 Here the turn, just
heard as 1̂–♭2̂–1̂–7̂–1̂, potentially reacquires its original scale degrees (5̂–6̂–
5̂–♯4̂–5̂). For me, this music is highly unstable between A and D: where the
thematic repetition suggests an A tonic, harmonic syntax suggests D minor (a
hearing reinforced by the convention of recalling the opening key
somewhere near the end of the second group).36 The magic lies in the
availability of both interpretations. It is the rare classical-style passage that
can be said to be in two keys at once.

Figure 10.3.7. Since registral inversion produces parallel fifths (middle), Beethoven alters the motive
(bottom).

The repeat expands the register by transposing along the dyad, first by t1
then by t2, the right hand constantly ascending but always in a slightly
different way (Figure 10.3.8). The continuation reinterprets vii°7 of A minor
as vii°7 in the relative major, contracting the register with a descending-third
contrary-motion sequence that restarts the left hand’s eighth-note motion;
Beethoven liked this effect so much he reused it in the first-movement
development of his Second Symphony.37 We arrive, finally, on a two-
measure Ludwig fragment repeated no fewer than five times. (Here I take
the outer voices to be primary, A–B–C–E against C–B–A–G♯, with the latter
voice doubled at the third.) Its fragmentary nature gives this music the
character of a closing theme; however, both the quiet dynamic and the
dominant-pedal bass undercut the cadential arrival at m. 74.38

Figure 10.3.8. (top) Registral expansion by transposition along the dyad (mm. 63–69); (bottom) the
contrary-motion descending-third sequence returning the hands to their original alignment (mm. 69–
73).

Figure 10.3.9 summarizes the exposition as a series of six schematic

processes, each featuring an ascending bass line—a remorseless ratchet that
continues throughout the movement.39 That line is harmonized with contrary
motion at the beginning and end of the exposition, and with parallel motion
in the middle. Each process is in my view a unity, an internalized musical
pattern or sequence or schema of the sort that could guide improvisation—
and which the Tempest knits into a convincing musical argument. These
schematic processes lie underneath the motivic relationships that are readily
available to the ear, endowing the music with a deeper continuity.40
Ubiquitous motives, stock musical patterns, avoidance of straightforward
melody, and extremes of dynamics and register: that is Beethoven’s style in a
nutshell.

Figure 10.3.9. The exposition as a series of parallel and contrary sequences.

We can reasonably wonder whether we need any further reduction beyond

this. It would be quixotic to suggest that the piece would be aesthetically
defective if it possessed no further unity, or that six familiar processes—
easily summarized on just a few staves—were inherently too much for
ninety measures of music. Yet a good deal of analytical effort seems to be
driven by just this belief. Schmalfeldt, for example, provides the
Schenkerian reduction in Figure 10.3.10. The picture is that the white
noteheads stay in force throughout the intervening music, being singled out
by the ear despite their inclusion inside sequential processes. This “singling
out” has been taken to explain something aesthetically important about the
piece, its “organic coherence” or status as a “masterwork.” To me this all
seems implausible. My schematic reduction is simple enough that I can
imagine thinking it up on my own, and yet complex enough that I can almost
imagine recomposing the movement on that basis. By contrast it is
inconceivable that I would get anywhere starting from Figure 10.3.10.
Sequence-based reduction provides a sense of ownership that I do not get
from Schenkerian analysis, allowing me to understand how actual human
beings might find their way to music as complex and impressive as the
Tempest.

Figure 10.3.10. Janet Schmalfeldt’s analysis of the movement.

(b) Development and recapitulation. The development is a helicopter

drop, the rolled chords taking us far from the tonic (D6–f♯°7–F♯64), with the
six-four bypassing the dominant and progressing directly to a root-position
tonic. The compensating sequence follows the transition so faithfully that it
sounds more like repetition than a development proper: few other classical-
style developments are so completely based on what is arguably a
“transitional” sequence. Its ending is perturbed so as to move by fourth
rather than fifth, turning the entire pattern into a lightly embellished
ascending 5–6: ([f♯⇐C♯43⇒f♯ 6]–b)–(G6–C)–(A6–D). This leads to a half
cadence on A by way of an augmented sixth, followed by twelve measures of
standing on the dominant, during which the melody echoes the second
theme’s E–F–D–E turn (Figure 10.3.11).41 Finally, almost-parallel triads
return us to the opening; as in mm. 55–56, the tonic is surrounded by
semitonal neighbors.

Figure 10.3.11. The development, mm. 93–143.

The recapitulation is one of the more inventive in Beethoven’s middle

period. The slow arpeggio gives rise to an expressive lament, pedal down, its
final F poignantly suggesting D minor; this leads to the cascade, turn, and
cadence exactly as before.42 The next arpeggio is followed by a loosely
parallel lament in F minor; to facilitate comparison, Figure 10.3.12
transposes both melodies into the same key. The implied C7–f progression
expands to a new chromatic 5–6 sequence that evokes the development: C7–
f is followed by C ♯ –f ♯ and D–g.43 At this point the harmonic sequence

could connect smoothly with the E♭ 6–c♯°43–g♯°7 of the exposition but this

would upset the two-chord grouping; Beethoven instead omits the E ♭ and
proceeds directly to c♯° (Figure 10.3.13). Thus we can trace the ascending-

step sequence throughout the movement: E–a, F♯–b, G–C, A–d, [B–e], C–f,

C♯–f♯, D–g, E–A, lacking only the tonicization of E.44 We could say that the
recapitulatory repeat simultaneously continues the ongoing (if difficult-to-
perceive) sequential process.

Figure 10.3.12. The recapitulation’s two arioso laments, with the second transposed to D minor for
comparison.

Figure 10.3.13. The ascending-step sequence running throughout the movement.

From there the recapitulation largely repeats the exposition, with the
exception of the upper-voice pedal discussed earlier. The parallelism lasts
until the retransition, where repeated octaves lead to a spooky coda,
compressing the closing theme to a single measure, a fitting depiction of the
storm’s aftermath.45 For all its originality, the movement is almost entirely
schematic, composed of a handful of sequences stitched together with a few
moments of i–V–i functionality. Beethoven arrays these materials in the
service of a compelling formal trajectory, the large-scale registral, motivic,
and harmonic shapes portraying a musical storm that is sublime in every
sense of the term—overwhelming and rigorous music that escapes the
gravitational pull of the ordinary to achieve something unprecedented.

4. The Fifth Symphony

The first movement of the Fifth Symphony is in many ways cousin to the
Tempest: largely sequential, with a middle-register minor tonic, interrupting
its recapitulation for a doloroso aria, and opening with the juxtaposition of
heterogeneous material.46 The Tempest’s arpeggio is replaced by a fortissimo
“Knock of Fate,” appearing at the start of the form’s major sections:
exposition, transition, second theme, development, recapitulation, second
theme recapitulation, and, in the coda, the final thematic statement. (It is
missing from the recapitulation’s transition, replaced by the oboe lament.)
Both pieces are unusually taut in their overall form, suffused with motivic
relationships and exemplifying the analyst’s dream of unity. But where the
Tempest’s sequences are obvious, the Fifth’s are covert and disguised.
(a) Exposition. Schenker and Tovey both warn against taking the main
motive to be the first measure’s G–G–G–E ♭ ; for them the central theme is
this figure combined with its repetition down by one diatonic step, the
“down by third, up by step” melody so central to functional logic (§7.2).47 I
am not so certain. The opening measure articulates a systematic pattern of
chordal intervals that can be traced throughout the first phrase—a disguised
patterning, no doubt produced by the compositional subconscious, that lends
the music an audible if hard-to-describe consistency (Figure 10.4.1). This
pattern of chordal intervals becomes increasingly explicit over the course of
the movement, as if Beethoven were gradually revealing the underlying
structure (Figure 10.4.2).48
Figure 10.4.1. Measures 7–18 as a distorted arpeggio, cycling through three scale degrees that are
themselves moving.
Figure 10.4.2. Several appearances of the (0, 0, –1, –1) interval sequence.

Harmonically, the piece opens with an obsessive focus on tonic and

dominant: the only other harmony in the first thirty-eight measures is the
Italian sixth at the end of the first phrase. This harmonic limitation is offset
by rhythmic variety, the piece exploring many different four-measure
arrangements of tonic and dominant (Figure 10.4.3).49 The second phrase is
an “accelerating double sentence” whose continuation is itself a smaller
sentence: we start with a four-bar presentation that is repeated exactly; the
continuation is then constructed as a sentence with a two-bar unit, whose
continuation uses a one-bar unit (Figure 10.4.4).50 The continuation’s
Ludwig recalls the Pathétique’s allegro opening, departing from the four-bar
hypermeter, and presenting the first sustained series of quickly changing
harmonies. Schematic content balances the harmonic acceleration, the
complexity of the onrushing chords ameliorated by their idiomatic nature.

Figure 10.4.3. The symphony’s four-bar phrases distribute two harmonies in many different ways.
Figure 10.4.4. The second large phrase (mm. 26–44) as an accelerating double sentence.

We then hear three cascading phrases linked by a twisty maze of contrasts

and resemblances: phrases 1 and 3 use the same intervals along the chord,
phrases 2 and 3 use seventh chords, while phrases 1 and 2 add an extra
descending step to the third sequential iteration, perturbing their patterns in
analogous ways (Figure 10.4.5).51 These relationships are too fast to process
in real time, though the constant changes help keep the music fresh. The
modulation, such as it is, consists in the unexpected resolution of c: vii°7/V
to E♭: V, the F♯ reconceived as G♭.52 The following horn call expands the
Fate motive so that thirds become fifths, changing minor to relative major
without transposition: it does not feel like a medial caesura, in large part
because of the impression of being cut off in mid-utterance—the pause more
unexplained than preparatory, innovative rhetoric replacing familiar gesture.
As in the Waldstein, the caesura’s novel function seems more important than
its conventional ancestry.

Figure 10.4.5. Intervallic patterns in the three descending arpeggios.

At this point there is an interesting question of hypermeter. Both Tovey

and Schenker continue to read four-bar groupings through the transition, so
that the second theme starts on the last beat of the hypermeasure (Figure
10.4.6). This puts the tonic arrival on hyperbeat 3 rather than hyperbeat 1.
Riemann and Imbrie instead start the second theme on the second measure of
a four-bar group; on this hearing the second theme’s sequential units all have
a weak-weak-weak-strong design that mirrors the rhythm of the opening
motive. This correspondence is underscored by the basses, whose statement
of that motive arrives on the accented bar. For performers, the question is
whether to emphasize the theme’s second bar (encouraging the
Tovey/Schenker hearing), or the fourth-bar tonic arrival (encouraging the
Riemann/Imbrie alternative).53
Figure 10.4.6. Hypermeter at the start of the second theme (mm. 56–66).

The second theme is a long ascending-step near sequence (Figure 10.4.7).

As in the Tempest transition, Beethoven uses a disguised model, creating a
chainlike continuity wherein each unit of the sequence has a direct link to its
predecessor: the sequence’s first statement reworks the pitch classes of the
horn call; the second statement uses the same harmonic progression as the
first; and the third unit uses the same melody as the second.54 This third unit
appears a step too high, skipping the mediant and moving directly to IV,
whereupon it gets stuck on various secondary dominants before arriving at
V65.55 The ascending bass continues through the next phrase, a rushing
release of pent-up energy, its descending melodic waterfall counterbalancing
the sinuous ascent of the preceding sequence. This music uses a hybrid
descending-fifth Ludwig, (D, B♭)→ (E♭, G) → (G, E♭)→ (A♭, C), here acting
as V–I–IV and embellished with appoggiaturas; these are linked across the
repeat to form a continuous passage of Ludwiging (Figure 10.4.8). We arrive
at an exultant closing passage restoring the major third G–E♭ to its default

role as 3̂–1̂ in E♭ major. The exposition’s entire second half is thus a single
musical gesture, a steadily rising scalar bass beginning in stillness and
moving toward climactic affirmation (Figure 10.4.9). As Bernstein said,
there is something inevitable here, a palpably new focus on long spans of
musical time. This is not so much a matter of divine inspiration or
compositional genius, but rather musical thought: unusually long paragraphs
formed from entirely schematic material.

Figure 10.4.7. The second theme (mm. 71–94) as an ascending-step near sequence.
Figure 10.4.8. The Ludwig at the end of the second theme (mm. 96–102). The repeat creates four
units of continuous Ludwiging.

Figure 10.4.9. The ascending scalar bass in the second theme.

(b) Development. The exposition transformed G–E♭ from 5̂–3̂ in C minor

to 3̂–1̂ in E ♭ . The development responds with something like the opposite

gesture: what initially sounds like 5̂–3̂ in E♭ is undercut by a low D♭–C that
turns it into 4̂–2̂ in F minor. This begins a sixty-five-measure “up-and-down-
the-ladder,” ascending by fifth from F to C to G, standing on the dominant,
and descending by fifth to C and F. This leads to a mysterious fauxbourdon
passage, a transfigured second theme ascending from F minor to B♭ minor

before losing energy, crawling upward through the tritone G ♭ minor

(respelled as F ♯ ), crabbed and chromatic, only barely reaching E ♭ ♭ major

(notated as D) before grinding to a halt just when the dominant comes into
view (Figure 10.4.10).

Figure 10.4.10. A harmonic outline of the development.

The ascent up the ladder of the fifths takes place by way of a grand
contrary-motion gesture. Figure 10.4.11 provides a reduction, the top staff
cascading downward by six octaves from C7 to C1 while the lower ascends
from A ♭ 2 to E ♭ 4; at each stage the top staff descends by an octave and a
fourth, or ten diatonic steps, while the lower voice ascends by fifth. The
voice leading uses a descending-fifth Ludwig that moves from (A♭, C) to (D,
F) to (G, B), but the initial dyad is the upper third of F minor rather than the
lower third of A♭ major; this converts the A♭–D progression of descending-
fifth dyads into an f–d° progression of descending-third triads.56 Once again
dyadic logic augments the triadic logic of Roman numerals.
Figure 10.4.11. The two contrary-motion sequences in the first half of the development.

The standing on the dominant begins with a familiar retrofunctional move

from vii° to vii°7/V and then to i64. The descent down the ladder of fifths,
shown in Figure 10.4.12, is less complicated, as befits its role in reducing the
musical tension; it also contains the piece’s first syncopation (!), tying the
lower strings’ incipit across the barline. This music brings back the second
theme’s version of the Fate gesture, juxtaposing it with another contrary-
motion passage, descending lower-register arpeggios against rising upper-
register tritones; since each part moves by two steps along the seventh chord,
the downbeat interval class is always the same. There is a bit of contrapuntal
trickery across the repeat, with the arpeggiated tritone moving down by
semitone while the rest of the music moves by fourth and fifth. This leads to
a reimagining of the second theme and the movement’s first use of
fauxbourdon (Figure 10.4.10, last measure).57 The sequence recalls the
Tempest’s development, fauxbourdon ascending by fourth before becoming a
chromatic 5–6.
Figure 10.4.12. The descending-fifth, contrary-motion sequence in the middle of the development.

Once again, my analytical reductions have sequences and schemas as their

targets, reducing the music to a collection of nested patterns—contrary-
motion ascending- and descending-fifth sequences arranged as an up-and-
down-the-ladder pair. Figure 10.4.13 provides Schenker’s more note-based
analysis: at the highest level, he connects the A♭ that starts the development
to the G that is the first downbeat of the descending-fifth “down-the-ladder”
sequence; at the intermediate level he constructs the descending scale shown
in the example’s middle system; and at the most detailed level, he connects
the first Ludwig’s initial upper-voice A♭ to the lower-voice G that starts the
second unit. I have a number of technical reservations about these
reductions: in particular, many of Schenker’s choices seem arbitrary,
motivated less by internal musical logic than by an external mandate to force
the music to confirm to his preconceived reductional templates; this is most
obvious in his insensitivity to the Ludwig’s schematic unity, repeatedly
forcing it into a stepwise mold. Most important, however, is the question of
analytical goals. Schenker’s aim, it seems to me, is to show that Beethoven
is a superhuman genius, and hence better than you. Mine is to demystify
Beethoven by showing how you might have written this music yourself. I try
to do this by providing reductions that are simple enough so that one could
imagine inventing them, while being complex enough to serve as blueprints
that could be used to reconstruct the final piece. Schenker’s note-based
reductions are too far from the surface to serve this purpose—both because
they are insufficiently reflective of the piece’s details and because there are
countless pieces to be written embedding their notes.

Figure 10.4.13. Three of Schenker’s development sketches: (top) his long-range sketch; (middle) a
more detailed sketch; (bottom) his interpretation of the contrary-motion Ludwig.

(c) Recapitulation and coda. Oboe solo aside, the first half of the
recapitulation is fairly literal. In part this is because the transition requires
only a single change of harmony, vii°7/V moving to V rather than V/III. The
major difference is timbral: pizzicato lower strings and an increased role for
the winds, leading to the lament that replaces Fate’s second knock.58 The
second theme is transformed by the “pattern continues, chord repeats”
technique of §4.9; this takes I to IV and an expanded “standing on the
dominant” using ♭ 6̂ and ♮ 6̂ (Figure 10.4.14). Amusingly, the second theme’s

second sequential unit, melodic B ♭ –C–D ♭ –C with harmonic C7–F, is the

same in both exposition and recapitulation. These changes have the effect of
keeping the entire second theme area more squarely in the orbit of C major,
delaying the arrival of C minor until the coda.

Figure 10.4.14. The recapitulation’s version of the second theme (mm. 319ff).

The coda is most notable for what it does not do: in place of schematic
cleverness and voice-leading trickery, we have raw compositional force,
often featuring a stripped-down texture of octaves or dyads that evoke the
Knock of Fate. The overall impression is of accumulating energy: six
overlapping phrases grouped into three paragraphs, outlining large scalar
trajectories and building to a climax surpassing anything heard so far. Like
the Waldstein, it presents a long process of rhythmic change, beginning with
roughly fifty measures of mostly eighth-note motion, slowing to quarter
notes for the next fifty measures, arriving on a singular half-note at the start
of the fifth phrase, and reinstating eighth notes for the final thirty bars.
Motivically, it separates the opening motive’s pitch and rhythm, highlighting
its pitch content in the second phrase and focusing on rhythm in the
remainder.59
The coda begins with pulsing eighth-note harmony, strings and winds
alternately intoning the weak-weak-weak-strong rhythmic motto; the two
Ludwig units loosely recall the transition (Figure 10.4.15, compare mm. 34ff
of Figure 10.4.4).60 These arrive on a second-inversion triad that overlaps
with the start of the second phrase, simultaneously a cadential 64 and a
tonic-functioning beginning. The lower voice combines the Knock of Fate
pitches with the rhythm of the second-theme horn call, abandoning its
cadence in favor of a natural-minor descent in tenths, the counterpoint
contracting to bare dyads. Once again, we bypass the cadence, scale-
fragments instead changing direction and taking up the WWWS rhythmic
motto; offbeat eighths add a third voice that converts the tenths into tonic-
dominant harmony, as if the language were gradually rebuilding itself. One
has the feeling of massive forces reversing direction, an ocean liner slowly
coming about, with phrase 2 descending from A5 to D4 and phrases 3 and 4
ascending back to C5.
Figure 10.4.15. The coda’s first three phrases, mm. 374–422.

Phrase 4 returns to the coda’s opening, extending the Ludwig and

reinforcing its relation to the transition (Figure 10.4.16). The cadential
arrival is the coda’s sole half-note measure, the gigantic musical mechanism
simultaneously slowing and gathering force. Phrase 5 is an unusual example
of what might be called “development by reorganization,” using virtually the
same material as phrase 4, but organized downbeat-to-upbeat rather than
upbeat-to-downbeat. Eighth notes return for the final section, reuniting the
motive’s pitch and rhythm (Figure 10.4.17). The music of mm. 6–9 appears
twice, repeated in such a way as to clarify the theme’s arpeggiated nature;
the oboe’s A♭5–G5 helps us hear the repeat as a continuation. The repeat also
has the effect of withholding the dominant for the final cadence.
Figure 10.4.16. The coda’s fourth and fifth phrases, mm. 423–468.
Figure 10.4.17. The end of the coda, mm. 469–502.

All in all, the coda feels somewhat beyond analysis, devoid of the non-
obvious relationships that make verbal commentary worth reading. Here
Beethoven seems to be well on his way to a music of gesture, made from
contrasts of direction, grouping, and density rather than complexities of
harmony and melody. Though it uses familiar scales and harmonies, its
aesthetics feel genuinely novel, somewhere between the worlds of Haydn
and Penderecki. It is interesting that this music completes such a taut and
intricate movement, as if Beethoven’s expressive energy could no longer be
contained within the idioms of the classical style. The coda breaks its
generic bonds, bursting forth in a frenzy of unstructured energy.
 5. The “Pastorale” sonata, Op. 28

The first movement of the “Pastorale” sonata shows Beethoven exploring

similar ideas in a more placid frame of mind. Its most distinctive feature is
its rhythm of repetition: almost every phrase is played twice, with the repeat
sometimes extended. The piece thus suggests its own formal analysis,
articulating structural divisions through repetitive regularity.61 A second
point of interest is the emphasis on parallel motion, with fauxbourdon
playing a starring role—augmented occasionally by other sorts of
parallelism, and sometimes in dialogue with contrary motion. While many of
Beethoven’s movements are dominated by sequences or harmonic cycles, the
“Pastorale” is the rare piece that is largely devoted to the third system of
functional tonality. The combination of repetition and fauxbourdon
contributes to its relaxed lack of urgency.
(a) Exposition. I parse the exposition as being in six large phrases, with
the first two forming the first group, the third the transition, the fourth and
fifth the second theme, and the sixth a closing theme. The first phrase begins
with descending-third fauxbourdon, IV–ii–vii°–V over a tonic pedal,
replacing the suspensions of the standard 7–6 sequence with accented
passing tones (Figures 10.5.1–2). Having articulated almost the entire chain
of thirds from tonic to dominant, the inverted arch ascends by horn fifths to
recapture lost registral space. The lowest voice is displaced downward by
octave for the horn call, the unusual spacing setting up the repeat’s “thinking
within the chord.” Here the tenor moves upward as the upper voices
descend, transforming parallel into contrary motion—a process that will
recur throughout the movement (Figure 10.5.3).
Figure 10.5.1. The opening of the Op. 28 piano sonata.

Figure 10.5.2. The descending 7–6 compared to Beethoven’s descending-third fauxbourdon sequence.
Figure 10.5.3. Thinking within the chord in the repeat of the Op. 28 theme.

The second phrase is also fauxbourdon-based, a variant of the “standing

on the predominant” idiom with three voices moving back and forth between
ii6 and IV6 while the pedal shifts to the tenor (Figure 10.5.4). I hear this
music as responding to the horn fifths of the opening—a kind of “tail
development” whereby the second phrase picks up on the ending of the first:
Figure 10.5.5 tries to capture the relationship, but the analysis feels pedantic,
like pinning a butterfly to a board. The transitional third phrase is also
fauxbourdon-based, with three voices descending against a tenor pedal. We
hear a four-chord pattern first in A and then in E, moving twice up the ladder
of fifths to tonicize V/V. The vi65–V6–V43/V–V progression crosses
fauxbourdon with the Prinner, its two transpositions connecting to form a
descending scale—repetition that is again continuation (Figures 10.5.6–7).
This whole pattern is then repeated with an oscillating neighbor-tone
decoration—leading to a gradual weakening of the tonic, so that the first
presentation of b65–A6–E43–A is clearly in D while its repeat can be heard
in A. The final cadence is repeated while the right-hand scale descends to an
abrupt halt on E4, once again lending the medial caesura a novel expressive
valence: first, because the harmonically unsupported pause has the character
of an interruption, as if a performer had somehow gotten lost; and second
because the subsequent E ♯ seems to resume the melodic thread, the
purported medial caesura occurring in the middle of a continuous melodic
statement.
Figure 10.5.4. The second phrase (mm. 21–39) is based on the fauxbourdon ROTO schema.

Figure 10.5.5. The second phrase loosely echoes the first-phrase horn call.
Figure 10.5.6. The fauxbourdon-based transition, mm. 40–62.
Figure 10.5.7. The transition as a continuous process, the parallel first-inversion chords continuing
through the repeat.

Compounding this effect is a serious case of tonal indecision, the music

vacillating between A major and F♯ minor for almost thirty measures before
finally settling down in A (Figure 10.5.8).62 The second theme begins with
right-hand ascending steps against left-hand descending fifths, stating the
defaults in Figures 6.1.3–4.63 It is the bass that makes the music interesting,
its insistently descending fifths easy to miss if you are not paying attention.
But if you are, you can almost hear Beethoven making the music-theoretical
point that functional harmony encodes a contrary-motion sequence as one of
its basic idioms.64 This contrary-motion antecedent leads to a parallel motion
consequent, reversing the process heard at the movement’s opening: there,
the parallel-motion fauxbourdon theme was answered by a contrary-motion
variant; here a contrary-motion antecedent is answered by parallel-motion
consequent. The half-cadence C ♯ 7 slides to E7, with both sonorities being
decorated by lower neighbors to produce common-tone decorations;
unusually, they are half-diminished and minor sevenths rather than fully
diminished sevenths. This static phrase extension is one of the largest
nonrepeated passages in the movement, a treading water before the A major
arrival in m. 91.
Figure 10.5.8. The off-tonic beginning of the second theme, mm. 63–91.

That arrival occurs in the middle of a continuous forty-eight-measure

sweep extending from m. 61 to m. 108—or even longer, for listeners who
share my sense that the m. 61 E♯ continues the preceding music. Once again,
we have a sequence playing a thematic role: in this case, a standard A3D5
with parallel tenths in the outer voices, fauxbourdon parallelism having been
reduced from three voices to two (Figure 10.5.9). The sequence descends by
two thirds, moving from A to f ♯ to D before reversing direction with two
statements of the ascending-fifth Ludwig (Figure 10.5.10).65 Once again
parallel is answered by contrary, here within a single phrase. The second
statement shadows the original a third higher, descending by third and then
using the descending-fifth Ludwig to arrive at the same chord that completed
the first statement. It is fascinating to see Beethoven deliberately juxtaposing
two versions of the Ludwig, as if he considered them variants of the same
basic pattern. This restatement also tells us something about Beethoven’s
conception of musical identity, for this theme is not a specific melody but
something more general—a texture, a shape, a sequence of schemas. Here
Beethoven seems to declare that he is a composer of patterns rather than
melodies.
Figure 10.5.9. The second theme’s loosely parallel A major phrases, superimposed.
Figure 10.5.10. The two continuations of the second theme’s A major phrases.

The brief cadential passage then takes us to a cadence on A and a closing

theme that displaces the rhythm of the fourth phrase (Figure 10.5.8) by a
quarter note.66 Figure 10.5.11 shows that the left-hand figuration embeds the
horn fifths of the opening theme, a relationship that relatively few pianists
choose to emphasize; the right hand almost duplicates this melody at the
octave, leading to Gjerdingen’s Quiescenza and a brief tonicization of D
major; the closing theme thus contains both thematic and tonal references to
the opening. This music—repeated an octave below, with the melody
doubled at the original pitch level—takes us to the end of the exposition and
a descending quarter-note line. The horn fifths will be central to the
development, in both its opening-theme and closing-theme variants.
Figure 10.5.11. The closing theme is a loose eight-bar period with syncopated melody.

The six large expositional phrases are numbered on Figures 10.5.1–11.

Each is in some way sequential or schematic: the first a descending-third
fauxbourdon sequence moving to the horn-fifth ascent, embellished
contrapuntally in the repeat; the second, a version of the “fauxbourdon
ROTO”; the third, another fauxbourdon variant; the fourth, a contrary-
motion sequence basic to functional tonality; the fifth, a descending-third
sequence (like the opening theme) followed by a Ludwig; and the sixth,
horn-fifths followed by the Quiescenza. Half of these phrases move from
parallel to contrary or vice versa. There is no extended passage of
nonsequential, nonschematic, and melodic music to be found in the
movement. The ubiquity of patterning, in such a relaxed piece, reminds us
that schema and sequence are central features of Beethoven’s vocabulary,
compatible with a wide range of expressive moods.
 (b) Development and recapitulation. The development is perhaps the
largest “up and down the ladder” in the literature, featuring contrary motion
throughout (Figure 10.5.12). It begins two fifths below its usual starting
point: the schematic shift from dominant major to dominant minor, which
usually takes place across the exposition/development boundary, here occurs
within the development and on the subdominant. From there the music
ascends through five fifth-related minor keys, arriving at a V/vi that lasts for
thirty-eight measures (!), enlivened by a variety of rounds and contrary-
motion patterns.67 This leads to a very brief fragment of the closing theme—
once in B major and once in a B minor that moves directly to V7. Like the
exposition, the development presents a series of immediately repeating
statements. Here they shrink as the main theme is liquidated: we begin by
reducing the ten-measure main theme to its four-measure “horn fifths”
conclusion, arranged in eight-measure pairs, contracting to its last two
measures in m. 199, a single measure in m. 207, and ultimately a single note
in m. 240. Figure 10.5.13 shows the start of this process. The final
reappearance of the closing theme cements its connection to the opening, the
left-hand horn fifths more salient after a development largely devoted to that
motive.

Figure 10.5.12. An outline of the development.

Figure 10.5.13. Repetition in the first part of the development.

Unlike the previous pieces, the recapitulation involves only minor changes
to the exposition: inserting eighth notes into the first statement of the main
theme (mm. 279–280), adding an extra flourish in mm. 304–307, and
transposing mm. 316–319 so as to remain in the tonic. The coda returns to
the main theme and adds a brief cadence. It is difficult for me to perceive
this as anything other than perfunctory, for it neither contributes to what I
consider a genuine musical narrative, nor resolves anything I hear as long-
term tonal dissonance. In this respect it contrasts with the more varied
recapitulations of the Tempest and the Fifth Symphony. How is it that
Beethoven could be simultaneously so radical along some dimensions and so
conventional along others? What marked the difference, for him, between
unassailable features of sonata form, and those that were available for
experimentation? How did he understand the relation between large-scale
repetition and musical storytelling?

6. Schubert’s Quartettsatz

These three analyses should convey a sense of the intelligent play I hear in
Beethoven’s music. While his pieces are not always so pattern-based, nor so
heavily Ludwig-focused, these movements reflect an important strand of his
thinking—one that I think has been obscured by a fixation on genius,
heroism, divine inspiration, the Idee, musical unity, and the (Burkean)
sublime. I now want to consider how Beethoven’s techniques are
transformed in two pieces composed in his shadow, Schubert’s Quartettsatz
and the prelude to Wagner’s Lohengrin. I hear both composers as attempting
to match Beethoven’s expressive clarity while expanding his musical
vocabulary—Schubert in the direction of greater lyricism and formal
flexibility, and Wagner through a more thoroughgoing rejection of schematic
composition. If Beethoven’s language is one of limitation, a redeploying of
musical resources toward broader stretches of musical time, then both
Schubert and Wagner can be understood as trying to enrich their local
vocabulary while preserving Beethoven’s command of large-scale
architecture.
The Quartettsatz’s exposition is divided into three parts: we begin with
three phrases in C minor, the last of which moves to A ♭ ; we then hear a
lyrical second theme that leads to a two-part transition occurring after the
second theme. The transition begins with an ascending-step sequence whose
dominant sevenths never resolve, arriving at a G dominant that is V7 of the
global tonic—as if we were treading water tonally. We then hear a short,
repeated phrase that transforms this dominant into a tonic, leading to an
authentic cadence and a trio of themes in G major. This three-part structure
is in uneasy dialogue with the gestural targets of the classical-era sonata,
blurring, rearranging, and otherwise problematizing events such as
transition, medial caesura, and the “essential expository cadence.”68
Schubert’s combination of schematic detail and large-scale freedom has
prompted a great deal of criticism, but to a twenty-first-century musician it
suggests a genuine theoretical insight, that formal conventions and large-
scale key-relations are less significant than one might think.69 In this respect,
Schubert continues a process that begins with Beethoven’s more limited
problematizing of sonata-form conventions.
(a) Exposition. The opening measures present three basic elements: the
tremolo texture, the eighth-note neighboring figure, and the four-bar lament
bass (Figure 10.6.1). Both the motive and its development have a static
character, intrinsically accompanimental and always appearing on the same
pitch classes. We open with a four-voice canonic round expanding from
middle C to the quartet’s full four-and-a-half-octave range, low C2 to high
G6. Harmonically it articulates a third-focused tonic-subdominant Ludwig,
with the substitution of i64 for V43/iv evoking the fauxbourdon ROTO
(Figure 10.6.2). To the attentive analyst, the weak-beat i64 signals an
interruption, the unaccented second-inversion triad syntactically passing, yet
appearing at the end of each two-bar unit. Thus we could in principle infer
that the canon, when it breaks, must proceed to something other than V.
Eventually, the music satisfies this expectation with a dramatic Neapolitan,
the Ludwig continuing to the dominant through a voice exchange. In the
next six-bar phrase, the motive, such as it is, appears for the first time in an
upper voice, repeating just once before reaching a cadence with an unaltered
predominant. This is a softer and less angry sound, a notable contrast with
the previous Neapolitan.
Figure 10.6.1. The opening of Schubert’s Quartettsatz.
Figure 10.6.2. The opening fuses Ludwig and fauxbourdon ROTO.

The third phrase is shown in Figure 10.6.3. The ♭2̂ appears as part of the

cello ostinato, hinting at F minor and foreshadowing the coming A♭ major;

the cadence, when it arrives, reinterprets the Neapolitan to swerve to the
submediant.70 At this point a familiar Schubert emerges, a soaring melody
against a simple accompaniment in the lower strings—the movement’s
emotional center, and more sustained lyricism than in any of pieces we have
considered so far. If this is a second theme, it is preceded by no real
transition: instead, the accompaniment sublimates the neighboring motive,
the second violin E♭–F–E♭ inverting the initial C–B♭–C. The effect is of a
cinematic dissolve, our attention drawn by an upper-register melody so that
the motive gradually fades from consciousness, foreground receding into
background and ultimately beyond the horizon of our attention. It is as if the
motive has finally accepted its intrinsically accompanimental character.
Figure 10.6.3. The transition from C minor to A♭ major, mm. 13–26.

Figure 10.6.4 details the resemblances among the opening theme, the A♭
major theme, and the fauxbourdon ROTO schema: the opening substitutes
V2/IV for V6, bypassing the schema’s distinctive V6–IV6 progression, while
the second theme subposes new bass notes to avoid the equally characteristic
IV6–I64–ii6. The second-theme melody ascends upward against this
descending background, an arpeggiation that continues when it repeats at the
octave. This pattern of octave restatement will continue throughout the
exposition, which repeats almost as much as the first movement of
Beethoven’s Op. 28. The cadential phrase-extension provides a classical-
music take on the Renaissance ascending-fifth sequence, secondary
dominant concealing the V–ii retrofunctionality and minor tonic reinforcing
the descending voice leading (Figure 10.6.5).71 Its repetition is thus a
continuation, a single sequence of ascending fifths perturbed by third
substitutions.72
Figure 10.6.4. (top) Comparison of the C minor and A♭ major themes, and the fauxbourdon ROTO.

(bottom) The A♭ major theme ascends along a descending background.

Figure 10.6.5. An analysis of the A♭ major theme (mm. 27–60).

The long-delayed cadence arrives as the tremolos return and the mode
changes to minor, a shocking reversion to the bravura opening—with the
neighbor-note motive again in the foreground (Figure 10.6.6). This music
has the character of a sonata-form transition, and indeed it is possible to
imagine moving directly from the E♭7 in m. 26 to the A♭ minor in m. 61—an
alternative that would be rhetorically if not tonally more straightforward than
Schubert’s actual composition. As in Beethoven, texture, mode, and gesture
serve a thematic function, recalling the opening even in the context of
considerable variation.73 Figure 10.6.7 proposes a stepwise descending
background voice leading against which the surface moves; this effect of
continuous descent disguised by registral displacement is similar to the
Shepard-tone passacaglias considered earlier.74 We arrive on the dominant in
C minor, as if all the preceding music had gotten us exactly nowhere; I hear
this as Schubert’s take on the medial caesura, a reference that is clearer in
the recapitulation, where the texture contracts to a single voice (Figure
10.6.16).75
Figure 10.6.6. The beginning of the transition, mm. 61–80.
Figure 10.6.7. Descending chromatic voice leading in the transition.

Stasis is averted at the last minute by two measures in which chords

change on every eighth, the movement’s quickest harmonies (Figure 10.6.8).
Here we have the same effect found in the transition of the first movement of
Beethoven’s Op. 28, a modulating phrase that is immediately repeated. This
makes the repeat sound very different from the original: what was III of C
minor gradually becomes VI of G minor, a transformation that does not so
much occur in musical space as in interpretive space, our perceptions
revolving around the fixed point of an exact repeat.
Figure 10.6.8. The end of the transition, mm. 81–92.

The form then becomes a bit mysterious. To my ear, the preceding phrase
is transitional rather than thematic, motivically linked to the (faux) medial
caesura by the melodic F♯–G–A♭–G, and beginning on III of C minor. But
the phrase ends with a perfect authentic cadence and the next phrase, as
Aaron Grant notes, could be heard as closing: a six-measure sentence
containing a full statement of the fauxbourdon ROTO, bringing to
completion a motif present throughout the exposition (Figure 10.6.9).76 The
reference here is not simply one of motivic content, but of schematic
identity: both the C minor opening and the A ♭ major melody hint at the
fauxbourdon ROTO without stating it clearly, the former descending 8̂–7̂–6̂–
5̂–4̂, and latter 8̂–7̂–6̂, and continuing the schema’s harmony even further.77
In the G major theme, the schema emerges complete and undisguised,
Schubert finally coming clean about his material.78 The melody, meanwhile,
ascends in contrary-motion tenths and sixths against the bass, as in the
opening of the Fifth Brandenburg Concerto (c.f. Figure 7.5.2 and the
fauxbourdon passage in Figure 3.2.5).
Figure 10.6.9. The first G major theme, mm. 99–104. The melody alternates sixths and tenths above
the descending bass.

The first G major theme starts on G4 and repeats up an octave; the next
theme begins on the higher-octave G5, repeating at a stratospheric G6 (with
the cello instead switching to pizzicato). The descending octave moves from
bass to melody, sliding chromatically down from tonic to dominant, and then
diatonically down to tonic (Figure 10.6.10). Figure 10.6.11 shows that this
melody transports the preceding bass line’s intervals into the chromatic
scale, a subtle continuity that might initially escape the ear. The eight-
measure sentence is based on the Waldstein/Natural Woman progression,
descending (I–V)–(♭VII–IV)–(♭VI–♭III). Normally, this ♭III would proceed to
V, as in Morley’s “April Is in My Mistress’ Face”; here, Schubert drops
down to a striking root-position Neapolitan, as if extending the sequence
further—a dramatic continuation that transforms the Waldstein schema into
something novel.79 There follows a straightforward closing melody, a six-
measure sentence ascending by octave and repeating in the higher register,
the cello’s double pedal both instantiating the omnipresent three-note motive
and introducing an idea that will become important in the development
(Figure 10.6.12). This is the exposition’s third six-measure sentence.80
Figure 10.6.10. The second G major theme, mm. 113–124, along with its underlying voice leading.
Numbers show 90° arcs on the circle of chromatic triads (Figure 2.1.3).

Figure 10.6.11. The two G major themes present the same intervals in different scales.
Figure 10.6.12. The closing theme (mm. 125–130) as a six-measure sentence.

(b) Development and recapitulation. I hear the development in three large

sections, a static D♭ major center flanked by two sequential passages (Figure
10.6.13). We start with an ascending-step near sequence that omits the
mediant; formally, it fuses the opening tremolo, the coda to the lyrical
theme, and the cello’s closing ostinato.81 The phrasing, shown in Figure
10.6.14, is subtle: initially the V–I progressions decorate the sequential units,
repeating twice within the A♭ major section; in the second unit, we hear the

expected V–i progression in B♭ minor—but instead of repeating, the upper

voices ascend by third while the bass slides down by step. This means that
the second V–I progression, which repeats the previous measure from a
rhythmic/motivic perspective, has a different musical function: rather than
embellishing the tonic of an already established sequential unit, it is the third
unit of the harmonic sequence. This sort of variation is emblematic of
functional tonality, subtle irregularity enlivening a fundamentally repetitive
substrate.

Figure 10.6.13. An outline of the development.

Figure 10.6.14. In the first development theme the sequential unit contracts so that its V–I progression
changes function.

The following melody picks up the cello pedal, reversing the process at
the beginning of the piece: where the transition from C minor to A♭ major
steered the neighbor figure from foreground to the background, the cello’s
accompanimental pattern here moves from background to foreground. The
octave displacement also continues the exposition’s registral play: where the
exposition repeated themes in a higher octave, this theme contains its own
higher-octave repeat. The harmony exploits the minor-third system to arrive
at B♭ minor, linking A♭7 to F65 by semitonal voice leading across the repeat;
its repeated iiø43–V7 progressions evoke the transition’s iv6–V65 units
(Figure 10.6.7).
The third section contains what would be a textbook “up-and-down-the-
ladder” were the piece in B ♭ major: melodies descending chromatically as
the harmony rises i–v–ii–vi to a V/vi that is approached by augmented sixth.
Like the opening of the exposition, the development goes nowhere tonally,
starting with the closing theme’s G major and ending on the dominant of G
minor. This V/vi slides deceptively to the false tonic B ♭ major and the
recapitulation of the lyrical theme.
Of course, the piece is not in B♭ major and the recapitulation begins in the
wrong key. One could take this as a sign of Schubert’s flexible attitude
toward modulation, or perhaps even compositional frivolity: perhaps, having
found himself in the midst of a B♭ major up-and-down-the-ladder, he simply
followed its local promptings, trusting that he could make his way back to
the tonic.82 A more colorful possibility is that the key represents a confession
of musical theft, for the lyrical melody may derive from one of Mozart’s
more languid and proto-Schubertian concoctions, the adagio of the piano
sonata K.332.83 Figure 10.6.15 shows that the two phrases outline almost the
same contrapuntal substrate, both subposing a novel bass under the
fauxbourdon ROTO’s I64. Mozart then returns to the schema, giving us ii6
and I6 right where we expect. These last two chords are preceded with
pseudo-dominant root-position mediants, which I have labeled III+/ii and iii;
Schubert’s theme replaces these with standard-issue dominants. Perhaps,
having borrowed Mozart’s B♭ opening for an A♭ second theme, Schubert felt

compelled to restore it to its proper place—presenting it as a B ♭ major

beginning to make clear what he had done. As in a good mystery novel, the
confession is hidden, the B♭ major beginning the recapitulation rather than
the piece as a whole.
Figure 10.6.15. Comparison of Mozart and Schubert’s themes.

In any case, the theme immediately shifts to E♭ major for its repeat. While

close to C minor, this E♭ major will not take us smoothly back to the tonic:

the A ♭ major theme was a semitone above the exposition’s G major

destination, while the recapitulation’s E♭ is a minor third above the desired C

major arrival. The transition is therefore altered so that it ascends by just one
major second rather than two (B♭ to C65 rather than E♭ to F65 to G65, see
Figure 10.6.16); this takes us to C-as-dominant which is then transformed
into C-as-tonic for a repeat of the exposition’s final section, ending with the
opening round.84
Figure 10.6.16. The medial-caesura gesture in the recapitulation’s transition, mm. 241–244.

Looking back at this analysis, I am struck by three things. The first is its
schematic familiarity. The main harmonic and contrapuntal devices all
appear in earlier chapters: the Ludwig, the fauxbourdon ROTO, descending
chromatic sequences, the ascending-step near sequence that skips the
mediant, and the “up and down the ladder.” If I were ever to teach a course
on “composing like Beethoven,” or to advise a student struggling with that
task, I could imagine using this piece as a model; for like much of
Beethoven’s music, it illustrates the power of avoiding certain kinds of
invention. It is interesting not because of its material, but because of the
expressive use to which it puts that material. Schubert’s comfort with
Beethoven’s vocabulary is all the more impressive for the fact that it was
largely untheorized: he did not learn the Ludwig or the fauxbourdon ROTO
from a textbook; instead, he likely discovered them directly in music.
The second feature is the omnipresence of stepwise descending voice
leading. Figure 10.6.17 graphs several of the piece’s schemas in the relevant
spiral diagrams.85 The similarity of procedure is clear, with all these patterns
moving by small clockwise steps. Particularly interesting is the relation
between the diatonic fauxbourdon ROTO and the chromatic Waldstein
sequence, a geometrical analogy that Schubert highlights with a clear
thematic parallelism (Figure 10.6.11). Here we see something like the Prime
Directive, an awareness that musical patterns can be transported from one
chord-and-scale environment to another. We are not yet in the world of
Chopin or Wagner: rather than setting sail onto on the open seas of
chromatic counterpoint, Schubert prefers to stick closer to the shores. But we
can start to see the glimmerings of a new and more general understanding of
musical possibility.
Figure 10.6.17. Four passages from the Quartettsatz modeled on spiral diagrams. On the top left, the
basic “fauxbourdon ROTO” pattern C–e°–F6–C64, used at multiple points in the movement; top right,
the two chromatic triadic sequences, the Waldstein C–G6–B♭–F6 (second theme) and the ascending-
fifth C–G–D–A (development, dotted line). On the bottom, the seventh chord sequence C7–c7–D2,
used in the transition.

Finally, there is the piece’s deconstruction of classical form, with its three
separate tonal areas, its two secondary themes, its oddly placed medial
caesura, and its rearranged recapitulation. In this respect it continues
Beethoven’s efforts to loosen the formal conventions of earlier classical
music. Schubert’s music conveys the sense of sonata form as abstracted and
transformed: tonally stable thematic zones interspersed with transitions, any
of which might end with a medial caesura. Directed modulatory architecture
is giving way to contrasts of character and mood, the piece presenting not an
orderly march from tonic to dominant and back, but a meandering tour
through tonal and expressive space. For me this is liberating, suggesting that
Schubert was a kindred spirit: someone who did not feel that large-scale key
relations were central to musical coherence, someone who understood sonata
form as optional and conventional, rather than a musical law of nature. The
piece, in other words, makes implicit but genuinely theoretical claims about
the nature of musical form.

7. The prelude to Lohengrin

I hear the Lohengrin prelude as a slow-motion, otherworldly fugue—the

theme stated in A major, E major, and A major in successively lower
registers, with the earlier instruments adding countermelodies above each
new entry; in this respect it is like an imitative piece opening top-middle-
bottom, descending through pitch space—though more dramatic in its
registral plummet and with a polyphonic “theme” instead of a monophonic
line. Orchestration is the music’s lifeblood, with timbre as important as pitch
and rhythm: altissimo divisi strings, briefly joined by high winds, dissolving
to solo string harmonics. As the music progresses, timbre develops in
concert with pitch, each thematic entry assigned to a new instrumental
family (strings, winds, and horns), arriving finally at a loosely thematic brass
fanfare; this then leads to a giant Beethovenian wedge retracing the earlier
registral descent, now mirrored by an ascending bass line, and subliming
back into the stratosphere for a nine-measure coda.
 Liszt and Wagner both describe the prelude as a slowly waxing vision of
the Grail, followed by its evanescence.86 This description resonates not just
with the registral and timbral trajectory, but also with two of the piece’s
potential inspirations: the first is the “Dawn” section of Félicien David’s
1844 symphonic ode The Desert, which begins with altissimo A major string
chords and a prominent E–F ♯ –E motive (Figure 10.7.1); the second is
Palestrina’s Stabat Mater, which Wagner arranged around the time he was
beginning the orchestration of Lohengrin.87 Like the prelude, Palestrina’s
double-chorus motet juxtaposes homophonic passages almost as if they were
individual voices, generating a similar sense of sacred calm.

Figure 10.7.1. The opening of the “Dawn” section of Félicien David’s The Desert.

It is just here that the music touches on familiar concerns. For if

Schubert’s Quartettsatz is a cubist rearranging of sonata form, its individual
phrases recognizable though sometimes out-of-order, then Wagner’s piece
represents a far greater fracturing of classical language—reaching inside the
phrase to its schematic details. The prelude begins with three parallel stanzas
of approximately sixteen measures: each opens with a root-position I–vi–I, a
progression that is hardly used from 1650 to 1850, but familiar in a good
deal of other music (e.g., “Oh, Pretty Woman” or “Hallelujah”). From there
each stanza progresses to a short descending passage that hints at the
fauxbourdon ROTO, I–V6–IV6, but moving to V7 rather than the expected
I64; at tempo, the V–IV has the unmistakable character of a harmonic
retrogression.88 The music then restates this descent, continuing to a V6/vi
that inaugurates a descending-fifth sequence, its dominant thwarted by
vii°7/ii and a chromaticized converging cadence (Figure 10.7.2). Idiom here
has been subjected to radical reinterpretation: where earlier composers might
have imagined classical harmony as a universal musical language, Wagner
understands it as historically situated.

Figure 10.7.2. Measures 1–12 of the Lohengrin prelude. The first three stanzas all begin with this
material.
 After this shared introduction, the stanzas proceed freely, in something
like the way the successive voices in Renaissance polyphony diverge after
stating a motive. The first is in some ways the most hermetic: what sounds
like a repeat of the I–vi–I is interrupted by a descending chromatic 6–5
sequence, f ♯ –C ♯ –E–B (Figure 10.7.3). The A major context makes this
sound more like free-floating chromaticism than the typical descent from
tonic to dominant—a perception reinforced by the tonicization of E major.
Having arrived on B, we descend by fifth to A major, leading to an
unexpected phrygian cadence on G♯. Outer voices then descend in staggered
tenths, moving unsystematically through third-related harmonies to arrive at
E major halfway through the bar.

Figure 10.7.3. Measures 13–20, the end of the prelude’s first stanza.

It is easy to overlook how novel this harmony is, particularly when we are
thinking about Wagner’s later chromaticism. For though functional and
diatonic, the prelude is worlds apart from the schematic language of the
classical style: the proliferation of secondary triads decenters tonic and
dominant, and the music “wanders” even when it remains diatonic. Wagner
tends to disguise his schemas, as when he occludes the converging cadence
with semitonal motion or presents the Waldstein progression on unusual
scale degrees. One can draw a contrast with those Romantic composers who
present their wandering moments as departures from a tonic-dominant
background. Already in Lohengrin, wandering is becoming the norm.
The second stanza begins as the first, with the winds sounding the theme
in the lower register while the strings add a countermelody that, for me, is
largely textural (Figure 10.7.4). The second half of the phrase, shown in
Figure 10.7.5, uses generalized fauxbourdon to reach a cadence on F. The
countermelodies ascend by third as the harmony descends, as in the
Waldstein (Figure 10.7.6). The next four measures juxtapose phrygian and
authentic cadences on E, modality and functionality alternating as if they
were equally plausible options.89 The last dominant returns to E by way of
vi, hinting at the I–vi–I motto before descending by thirds B–G♯–[E]–c♯–A–

f ♯ . The relation between this music and the end of the first stanza
exemplifies Wagner’s innovative conception of musical identity: the two
passages seem to traverse similar terrain without quite being variations,
passing through the same stations in their own individual way. This formal
flexibility again recalls the Renaissance.
Figure 10.7.4. Measures 21–27, the start of the second stanza.
Figure 10.7.5. Measures 28–36, the end of the second stanza.

Figure 10.7.6. Descending-step harmonies and ascending-third melodies in Lohengrin and

Beethoven’s Waldstein.

The third stanza largely repeats the second, though in a new register and
with more powerful orchestration: the A major theme enters in the horns,
bassoons, and lower strings, centered around middle C and supported by the
full force of the winds and strings (Figure 10.7.7). An extra ♭ II in m. 49
resolves as an ascending subdominant to an F that moves via the major-third
system to A7, the brass entering fortissimo for the arrival on D (Figure
10.7.8). The following music can be heard as a decorated I–vi–I, D–
(b⇐F ♯ ⇒b)–(A⇒D), leading to two iterations of the converging cadence
suggesting different keys—the first beginning in D and the second ending in
A, with no definitive point of modulation in between. (As in the Quartettsatz
or Beethoven’s Op. 28, a repeated modulation can be heard in two different
ways.) This takes us to a strong dominant whose deceptive resolution
overlaps with the start of the final large phrase.90
Figure 10.7.7. Measures 37–50, the third stanza. This music largely repeats the second stanza.
Figure 10.7.8. The fourth stanza, mm. 51–57.

That music features a giant contrary-motion wedge, more than five

octaves from low-bass F♯1 to violin A6, converging over nine measures to
less than an octave (Figure 10.7.9). Programmatically, this represents the
evaporation of the Grail, ecstatic insight gradually contracting in a poignant
release of musical energy.91 Music-theoretically, it represents the evolution
of contrary motion from a specific schematic gesture, carefully arranged to
be consistent with functional laws, to a more general tool for generating
novel harmonic states—the ancestor of early-modern moments such as
Figure 10.7.10. On close inspection we can find traces of a familiar
descending-third schema: at the beginning, ascending steps in an inner voice
against descending thirds and fourths in the upper voice (compare Figures
10.3.8 and 10.2.10), overlapping with descending melodic steps and fifths
against ascending steps in the bass, and finally dissolving into a purely
harmonic sequence that alternates root-position and first-inversion chords.
These are linked by a nonfunctional C♯–E hinge connecting the F♯ minor
beginning to an A major continuation that features deceptive rather than
authentic resolutions. The arrival at m. 67 is the prelude’s first conclusive
cadence, prompting a return to the altissimo register and one last version of
the I–vi motto, now normalized into a I–vi–IV–I progression (Figure
10.7.11).

Figure 10.7.9. The contrary-motion wedge of mm. 58–72, along with an analysis.
Figure 10.7.10. The death of Kastchei from Stravinsky’s Firebird (R195). The lower voices feature
the repeating contrapuntal pattern shown in Figure 4.2.2, while the upper voices descend
chromatically and unsystematically, producing a wide range of sonorities.

Figure 10.7.11. The end of the prelude, mm. 73–75.

Christopher Reynolds rightly associates this music with Beethoven, and

there is indeed a clear resonance with many of the passages we have
considered.92 But where Reynolds traces the music to the Ninth Symphony
in particular, I would instead point to Beethoven’s lifelong fascination with
contrary motion—the Ludwig as a manifestation of a broader compositional
habit rather than any one particular piece.93 The main difference is that
Wagner’s language is much less schematic than Beethoven’s: like many
nineteenth-century composers, he combines Beethoven’s formal clarity with
a more varied harmonic vocabulary drawing on a much wider range of
historical influences. The atoms of the Lohengrin prelude are not the
ubiquitous I–V–I progressions of the Fifth Symphony, nor even the more
complex sequential formulations we have been examining, but something
looser and less definite. Sequences and contrary motion become tools for
achieving harmonic novelty.94
 A good deal of critical commentary has centered on the question of
Lohengrin’s relation to Wagner’s later music. At first hearing they are quite
different: Lohengrin more triadic and more familiarly functional, its
modulations less frequent and its key areas more stable. If we think of
“progress” as a vector pointing from Tristan to The Book of the Hanging
Gardens, then Lohengrin is clearly “not yet” progressive. But it is possible
to hear the piece as progressive in another sense, exploring a flexible fusion
of modality and functionality found in our own time. Lohengrin, in other
words, intimates a diatonic progressivism whose explorations occur within
familiar scales rather than outside of them. Here we come full circle,
returning to something like the freedom of the modal era—a largely diatonic
language in which function contends with other modes of musical
organization.
Yet despite these changes, the basic materials of Wagner’s prelude are not
unfamiliar. Consider some of the idioms it shares with the Quartettsatz: the
fauxbourdon ROTO, contrary-motion sequences, the Waldstein schema, the
major-third system, and so on. These patterns can be found wherever
intelligent composers use triadic materials, from Renaissance modality to
contemporary popular music. This is not so much a matter of historical
influence but of constraints inherent in musical space: as we rise to higher
levels of abstraction, styles start to converge, with different genres making
slightly different use of the same basic musical options.
In the four centuries from Ockeghem to Brahms, these possibilities are
constrained and then loosened. Initially the narrowing is harmonic, as the
modal freedom of the early sixteenth century is gradually restricted to a
smaller set of functionally tonal progressions. In the eighteenth century, we
have a gestural narrowing that increasingly focuses on specific schemata:
sentence and period, converging cadence, standing on the dominant, sonata
form, and all the rest. Beethoven is at the very center of this process,
simultaneously the apotheosis of classical condensation and the initiator of
the Romantic countermovement—the point of rest at the top of the arc. The
Quartettsatz was written at the beginning of the expansion, the music’s local
moves familiar even while its sonata-form architecture is novel. The
Lohengrin prelude is expansive even at the level of local detail, its schemas
distorted and its progressions announcing a new harmonic freedom. A
Hegelian would describe it as combining the thesis of local freedom with the
antithesis of large-scale formal clarity—a clarity that was initially made
possible by a severe restriction of gesture, but which eventually transcended
that limitation.

1
Eventually, Lewin gave up and I spent the rest of the semester happily writing fugues.
2
For example, Caplin 2004, p. 69.
3
McClary 1991 and Fink 2004.
4
Burnham 1995.
5
One can add transpositions to this template but that does not change the basic Tinctorian situation.
6
Small intervals like the minor second have a small voice exchange but a large voice leading to
their tritone transposition; large intervals have large voice exchanges, but a small voice leading to their
tritone transposition. The minor third is perfectly balanced between these two poles.
7
Examples include the first-movement developments of the Kreutzer (or perhaps “Bridgetower”)
sonata, the Second Symphony, the Waldstein, the Eroica, and the Fifth Symphony (discussed later).
8
Thanks here to Nathan Mitchell, who also pointed out the relevance of Rink’s “Morte” schema
(Figure 4.7.4).
9
Steve Taylor points out that the second half of the opening adagio melody is similar to the third-
movement antiparallel episode discussed in the prelude to chapter 4—thus linking all three
movements thematically. The first-movement introduction can also be interpreted as a disguised
Ludwig variant over a perpetually descending bass; mm. 7–8 are a potential source for Figure 4.6.1
and Tristan more generally.
10
Beethoven exploits the schema’s contrary-motion potential by continuing it through multiple
octaves: in this case, the descending line appears on A4 to A3 and A2, while the ascending line moves
from A1 to A3 and A4.
11
Schindler 1840, p. 197; 1841, II, pp. 83–84. For a generally Beethovenian piece, see Clementi’s
minuet from Op. 10, no. 1; for some Clementi Ludwigs, see Op. 7, no. 2, I, mm. 4–5, Op. 10, no. 1,
III, mm. 1–2, Op. 13, no. 4, mm. 45–49 (initially published as Op. 14, no. 1). For a Beethovenian use
of the diatonic third’s basic voice leading, see Op. 7, no. 2, I, mm. 78–79 (compare Figure 3.2.4).
12
That is, the one consonance that is not invertible at the octave, the fifth, is also the only
consonance that cannot be expanded by a third: C–G is consonant, but C–B is dissonant.
13
Due to registral limitations, continuous contrary-motion Ludwiging is often disguised by octave
displacements.
14
Beethoven sometimes uses this change-of-direction strategy in developmental passages,
including the first-movement developments of his second (mm. 146ff) and third (mm. 186ff)
symphonies.
15
Note that the two kinds of motion exploit two different geometries: the logic of contrary motion
is essentially a dyadic note-against-note logic, while the logic of parallel motion is more robustly
triadic.
16
This is the second theme in Hepokoski and Darcy’s reading.
17
This is broadly speaking a devotion to musical poverty, a kind of asceticism or humility; some
theorists have connected this to Beethoven’s interest in popular or revolutionary music (Biamonte
2006, Ferraguto 2019).
18
This theme resembles one of the second themes of Op. 7, I, mm. 60–68.
19
Kinderman 2009a, p. 357.
 20
Cavett-Dunsby 1988 considers register in the transposed second-theme recapitulations of
Mozart’s “Haydn” quartets. Bergé and D’hoe 2009 note the registral issue here.
21
Beethoven’s description of how a pianist should play the piece (Burnham 2009).
22
Kant (1788) 1997, p. 133.
23
On Beethoven and monumentality, see Rehding 2009 and Dahlhaus 1991.
24
An 1820 conversation book contains the phrase “the starry skies above and the moral law within
—KANT!” (Thayer 1967, p. 747). For the inscription, which Kant considered among the most
sublime ever uttered, see Schindler (1860) 1996, p. 365, and Kant (1790) 1970, p. 160.
25
Bergé and D’hoe 2009 survey approaches to the motivic content.
26
Dahlhaus 1991 suggests that the piece represents a decisive break with earlier conceptions of
form. See Schmalfeldt 1995 and Horton 2014 for discussion.
27
This is the parsing given in Tovey 1931. Caplin 2009a also stresses the continuity between the
Tempest and Beethoven’s style. Vande Moortele 2009 surveys readings of the piece’s form.
28
See Dahlhaus 1991 and especially Schmalfeldt 1995 and 2011. Hamilton 2009 connects the
opening to the practice of improvised preluding.
29
Webster 2010 surveys the equally unusual form of Op. 10, no. 3, I.
30
Schmalfeldt (1995, p. 62) questions the analysis as a period because the consequent is
sequential; however, I believe sequential themes are common in Beethoven. Rothstein 2009, following
Riemann and Uhde, notes that the entire period forms a 2 + 4 structure if one takes the Largo half as
fast as the Allegro. Both the phrase and its constituent halves are examples of Goldenberg’s
“question/answer pairs” (2020).
31
James Hepokoski (2009), John Nathan Martin (2010), and David Damschroder (2016) all read
C♯–E–G–B♭ in m. 11 as a common-tone diminished seventh of G minor. This reading not only breaks
the harmonic sequence but postulates a very rare chord, the minor-key ct°7. Hatten 2009 agrees with
my reading: a dominant C♯–E–G–B♭ moving retrofunctionally to a dominant-of-the-dominant. This is
a common Beethoven progression (§3.8).
32
Op. 18, no. 6, II, m. 18, contains the same ascending-fifth pattern. Damschroder’s Schenkerian
analysis (2016) turns the sequence’s parallel motion into a non-parallel chorale. Burnham 2009 notes
the near omnipresence of first-inversion triads.
33
Hepokoski (2009) suggests that the movement has no second theme since it lacks a medial
caesura; instead, he hears this music as a “dominant lock” (or “standing on the dominant”). For me it
has too much individuality for that, its simple tonic-dominant harmony characteristically
Beethovenian.
34
My reading of the second-theme phrasing is broadly compatible with that of Schmalfeldt and
Caplin. Both of them, however, read the m. 63 cadence as the conclusion of a large thematic group
stretching back to m. 42, asserting a strong formal juncture between the nearly parallel presentations
of the second theme: thus they parse the music as (AB)(B´), with the B´ representing, in Caplin’s
words, “an entirely different thematic region” (Caplin 2009a, p. 111). I find it more natural to imagine
something more like (A)(BB´) with A and B briefly overlapping, the A material forgotten as the music
takes up a new subject. Thus I would consider m. 55 simultaneously weakly cadential with respect to
the preceding music and initiating with respect to what follows.
35
Caplin’s invocation of “invertible counterpoint” is really a metonym for a more general issue,
namely the way registral shifts can extinguish (what he considers) cadences by removing the required
root-position dominant (as in Example 4.7 of Caplin 2009a).
36
I hear the bass articulating Byros’s le–sol–fi–sol schema, but on the wrong scale degrees,
somewhat like the transition in the first movement of Beethoven’s Second Symphony (Byros 2012,
2015). This raises the delicate question, endemic to schematic analysis, of whether to read the passage
as a deliberate deforming of a preexisting norm, or as a new Beethovenian norm of its own.
Schmalfeldt 1995 hears an A major tonic but notes its potential for dominant function.
37
The diminished seventh is in m. 165 and the third-sequence starts in m. 166.
38
Schmalfeldt (1995), Hepokoski (2009), and Rothstein (2009) do not hear closure arriving until
m. 87; Caplin (2009a) instead hears those measures as a “post-cadential codetta.” Both hearings strike
me as reasonable.
39
Tovey 1931.
40
Cohn 2005 reads many of my “stock patterns” as “introverted motives” rather than schemas.
41
The ♮ 6̂–♭6̂–5̂ approach is a common developmental arrival.
42
Kinderman 2009b observes that the D minor lament anticipates the opening of the baritone
recitative in Beethoven’s Ninth Symphony.
43
Since each ascending-step modulation adds five flats, reinterpreting VI of the preceding minor
key as V of the next, the three sequential steps add fifteen total flats and complete a full circle in scale
space.
44
Skipping this tonicization is fairly standard, since the supertonic is diminished in minor.
45
Jander 1996 contextualizes the piece’s storm tropes.
46
Hepokoski 2009 notes this connection.
47
Schenker (1921–1923) 2004. Tovey (1935) 1981 snarkily observes that no piece was ever
derived from just four notes.
48
This same sort of gradual clarification characterizes the second theme of Op. 23, I (Figure 3.1.8),
the variations movement of Op. 1, no. 3 (§4.10), the variations movement of Op. 109 (§9.4), and the
idea of the “disguised model.”
49
If we disregard rotations, then there are four ways of arranging two harmonies in four bars:
AAAA, AAAB, AABB, and ABAB. All four appear on Figure 10.4.3. While I doubt that Beethoven
was consciously working through these possibilities, I do believe he was deliberately searching for
variety in the context of a limited harmonic vocabulary.
50
Lerdahl (2020, p. 89) makes a similar observation about the less-symmetrical opening music.
William Caplin would likely call this a compound sentence with two stages of fragmentation.
51
Schenker (1923–1924) 2005 notes that the dominant form of this arpeggiation gave Beethoven
particular difficulties.
52
Lewin 1986 colorfully describes this moment’s enharmonic play.
53
Riemann 1903, p. 38; Imbrie 1973. My own preference is for the Riemann/Imbrie reading,
though I think both are reasonable. Justin London (2004, p. 92) proposes yet another analysis in which
the second theme begins on a hypermetrical strong beat (m. 63). This disagreement suggests that
hypermeter can be fairly subjective. Temperley 2008 considers hypermetrical transitions in general.
54
In the transformation from horn call to violin melody, some harmonic notes become
nonharmonic; this leads Auerbach to dismiss the motivic connection as “invalid” (2021, p. 91). Like
Reti, I consider it to be significant (Reti 1951).
55
This near sequence often seems to appear in E♭, as in the second paragraph of the Eroica, the
second theme of the first movement of the C minor violin sonata, and the end of Op. 31, no. 3—
though it is also found in the first movement of the D major violin concerto. The shift from the
German augmented sixth to the fully diminished seventh chord, with the slightly odd ♭ 6̂–♮ 6̂–5̂ bass,

appears in a number of other pieces as well (e.g., Op. 10, no. 3, I, m. 199). Here ♭ 6̂ is “sublimated”
(Samarotto 2004).
 56
Beethoven uses almost the same voice leading in the Eroica scherzo’s ascending-step sequence
(m. 143). Schenker (1923–1924) 2005 recognizes a connection between this moment and the closing
theme, but not its schematic identity.
57
B♭ minor continues the “down the ladder” portion of the schema one step too far.
58
The introduction to the second theme, which shifts from horns to bassoons, confronts conductors
with a delicate decision, as it is distinctly less powerful than the exposition; I might consider adding a
single horn to the bassoons, strengthening the texture while still preserving some timbral difference.
59
The last movement of Brahms’s C minor piano quartet Op. 60 also seems to separate the pitch
and rhythm of the Fifth Symphony’s opening notes.
60
Schenker reduces the coda’s upper voice to C–D–E♭, rather than the E♭–F–G of the schema; the
issues here are largely analogous to those surrounding the Op. 22 minuet (§9.4).
61
Sommer 1986 reads the form against the rhythm of repetition. Tovey 1931 is much closer to my
reading.
62
There is a cadence on the dominant of A in m. 59, an F♯ minor theme in m. 63, a weak A major

cadence in m. 70, a return to F ♯ minor in m. 71, and a half cadence on its dominant in m. 77.
Longyear and Covington (1988) compare this to “backing slowly into a parking space.” See also
Hepokoski and Darcy (2006, p. 120) on “tonally migratory” phrases.
63
The upper three voices move in doubly parallel motion, perhaps developing the fauxbourdon
opening; the chordal sevenths resolve up by step in the first and third measures.
64
Indeed, it was this passage that inspired Figure 6.1.3, which is as much Beethoven’s discovery as
it is mine. An analogous passage can be found in the closing theme of the first movement of the A
major sonata Op. 2, no. 2, there with V7 supporting 4̂.
65
It is also possible to hear a non-octave-repeating Ludwig taking D to D, though that reading cuts
against the rhythmic organization.
66
Thanks here to Yuqi Liang.
67
The outer voice round at m. 230 is reminiscent of m. 170 of the first movement of the Second
Symphony, which extends the pattern from triad to seventh chord.
68
See Grant 2018 for a helpful survey of Schubert’s three-part expositions, which generally
articulate similar trajectories. Barry 2014 considers whether the piece may be “Schubert’s most
extreme treatment of sonata form” (32); Webster 1978 suggests it is not in sonata form. Longyear and
Covington (1988) suggest that Beethoven’s Op. 28 is a precursor. Other insightful discussions include
Bruce 1969, Fieldman 2002, Smith 2006, Mak 2008, and Hunt 2014.
69
See Fieldman 2002 for a survey of these criticisms.
70
Webster 1978 notes that these quick transitions are characteristic of Schubert; Bruce 1969, Navia
2016, and Duane 2017 hear an earlier transition starting in m. 13. Smith 2006 considers the various
roles of D♭ major and the Neapolitan in the movement.
71
Here is a place where it is helpful to view a I64 chord as tonic, a functionally tonal analogue to
the harmonic 64 in Figure 5.2.1.
72
In Roman numerals, IV–I–V–ii–[I replacing vi]–V–ii–[I replacing vi]–V–I. The I–ii–V–I and I–
V–ii–I are four-chord double emploi progressions: in the more familiar five-chord double emploi (e.g.,
I–IV–ii–V–I or the pop-music I–V–ii–IV–I), we hear third-related chords in succession; in the four-
chord version, one such chord replaces another.
73
Duane 2017 uses computational analysis to argue that texture can signify thematicity in
Schubert’s expositions.
 74
For example, §2.5 or §7.3. The last movement of Mozart’s E♭ major piano sonata, K.282, mm.
48–56, has a similar sequence, lowering the notes of a dominant seventh so that V7 becomes iv of the
key a step higher.
75
Richard Cohn (1999) has proposed that major-third-related key areas can serve similar
functional roles, implying that the second theme’s A♭ major could be a tonic substitute. I personally

hear A♭ major as a distinct key rather than a flavor of tonic, a hearing that comports with the broader
three-key practice surveyed in Grant 2018.
76
Grant 2018 describes a number of Schubertian three-key movements in which the thematic
material initially has a closing character that is retrospectively reinterpreted as thematic.
77
Many commentators have noted the motivic connection, including Beach 1994 and Barry 2014;
the schematic relationship is less frequently discussed.
78
Again, this bears comparison to the gradual revelation of structure we found in Beethoven.
79
The Neapolitan appears in the next sequential unit, which would continue ♭ V– ♭ II. For a rock
analogue see Led Zeppelin’s “Kashmir,” which similarly extends the 6–5 pattern beyond its standard
length. Peter Smith 2006 associates this moment with the prominent Neapolitan of the opening.
80
The others are the second phrase of the opening section and the G major theme.
81
Compare Figure 10.4.7.
82
Beach 1994 suggests that the B♭ key area is part of a movement-long C–B♭–A♭–G progression. I
do not place much aesthetic weight on these sorts of long-term pitch relations.
83
There is an extensive literature on Schubert’s borrowings, including Cone 1970, Edwards 1997,
and Humphreys 1997. Thanks here to René Rusch.
84
This kind of repositioning of the opening music appears famously appears in the first movement
of Mozart’s D major piano sonata K.311, I. Schubert makes a few other very small changes to the
recapitulation, including using minor chords in the second-theme Waldstein sequence.
85
Yust 2015a also notes the omnipresence of stepwise descending voice leading in Schubert,
though he analyzes it differently.
86
See Liszt 2016, pp. 117–18 (also found in Trippet 2010, p. 19), and Wagner 1907, pp. 232–33.
Kramer 2002 discusses the antisemitism of Wagner’s program and various efforts to resist this reading.
87
Deathridge 2008. The connection to the desert suggests that the antisemitism may have been
intrinsic to the music’s conception, rather than a late-breaking addition: for Wagner, the contemporary
world is an antispiritual Jewish-dominated desert.
88
The anticipations in m. 8 recall the third-based anticipations in the opening of Monteverdi’s
“Ohimé,” while softening the parallel fifths across the phrase break.
89
One could potentially hear this music as juxtaposing full cadences in E with half-cadences in A;
I, however, hear an E center throughout.
90
Kramer 2002 reads this final wedge as the second half of the fourth stanza; to me it sounds new.
91
Wagner’s symbolic use of register bears comparison with the end of Monteverdi’s “Cosi sol d’un
chiara fonte” (Madrigals, Book 8), where a Ludwig schema generates wide registral separation that
mirrors the words “I am so far from my salvation.”
92
Reynolds 2015; Gauldin 2004 explores chromatic wedges in Wagner’s later music.
93
In comparison to other Beethoven pieces, the Ninth Symphony does not strike me as being
heavily saturated with Ludwigs or other classical schemas; in fact, I suspect that its nonschematic
character represents its greatest influence on Wagner.
 94
See Figure 4.1.4. Segall 2010 contains an interesting example from Schoenberg’s Verklärte
Nacht.
 Conclusion
Past and Future
For centuries theorists have dreamed of a Platonic and extra-human music,
with listening the unwitting contemplation of ratio and number.1 There is a
lot wrong with this picture: to take it literally is to imagine that music could
be independent of physics, biology, culture, and history; yet the association
of ratio and consonance depends on the physics of vibrating objects, and our
pleasure in these relationships depends on our bodies and upbringing.
Platonism feeds the temptation to ignore complexity and detail, encouraging
us to think that nonharmonic tones must “stand for” their harmonic partners,
or that the return to the tonic key must create a sense of resolution, or that
beloved pieces must have an Urlinie. The very framework of rigid norms not
actually embodied in practice, of “geniuses” taking liberties with a priori
musical laws, reflects an urge to deny the human elements of music-making.
It is understandable if twenty-first-century musicians want to put Platonism
behind us.
Yet for all that it is not completely misguided. Around the world we can
hear musicians deploying an organizational strategy whose origins date to
the early Renaissance: surface voices moving inside chords that move inside
scales that are themselves moving inside the aggregate. The spiral diagrams
provide an elegant tool for visualizing this structure, a unified set of
geometrical models revealing the interdependence of harmony and
counterpoint: wheels within wheels, each turning independently of the
others. Styles can be characterized, in part, by the settings of this complex
machine: in a consonant, diatonic, and imitative environment, with little
constraint on harmonic progression, it can produce the transcendent stasis of
high-Renaissance polyphony; in a more homophonic and slightly less
diatonic context, the hedonic yawp of AC/DC. Augmented by a complex
harmonic grammar, we can approach the Newtonian clockwork of a Bach
prelude, or the Romantic subjectivity of a Beethoven sonata. With extended-
tertian sonorities and an expanded modal vocabulary, we move toward
impressionism, jazz, and minimalism. No doubt new settings remain to be
discovered, new ways of recombining the same basic ingredients.
 The collectional hierarchy can be the explicit object of attention, as when
musicians move motives along both chord and scale, or when surface voices
ascend against a descending background—as in the Waldstein or the
Revolutionary Etude. More subtle are pieces like “Eight Days a Week,” Heu
me Domine, or “Helpless,” where voices might ascend along a continually
descending scale in order to remain stationary—or those moments where
composers exploit dyadic rather than triadic logic. For here the quadruple
hierarchy can arise not because composers are paying attention to it, but
because musical culture has settled upon idioms that exploit its unique
characteristics. This is most obviously true of those countless embodied
routines that produce efficient voice leading by combining similar operations
at multiple hierarchical levels: big-T little-t transposition, or in the case of
Gesualdo’s trick, big-I little-i inversion (appendix 3). Here the quadruple
hierarchy is like a collective unconscious, supplying hidden and non-obvious
structure to a vast range of practices and techniques. Some parts of this
structure are easily visible, like the tip of an iceberg; others lurk far below
the musical surface.
In its full generality, the collectional hierarchy lies beyond the horizon of
our intuition: there are few if any musicians today who can sit down at their
instrument, choose some arbitrary three-note subset of some arbitrary seven-
note scale, and fluently manipulate the resulting possibilities. Instead, we
tend to internalize the more limited set of musical routines relevant to a
particular style. Over the course of this book, we have identified theoretical
origins for nine of the eleven schemas listed in the appendix of Gjerdingen’s
Music of the Galant Style: dyadic space for the Quiescenza, Fonte, and
Monte (§3.1), triadic space for the Prinner and Romanesca (§3.4), a canonic
fourth-progression for the Fenaroli (Figure 4.5.6), and basic sentential logic
for the Meyer, Do-Re-Mi, and Sol-Fa-Me (§9.1).2 Such patterns are not
isolated idioms specific to the galant period but reflections of very general
voice-leading possibilities known to composers in many genres. This tracing
of schemas to abstract geometrical structure underwrites my claim that
musical knowledge is often implicit, known in ways that may not be directly
accessible to conscious self-reflection. For it is often the case that there are
only a small number of ways to achieve basic musical goals, such as
harmonizing stepwise descending melodies using closely related triads.
Given enough time, musicians will eventually discover these possibilities,
thereby internalizing a kind of geometrical knowledge without realizing it.
Culture will preserve these discoveries, transmitting them as important
idioms or musical figures of speech. Perhaps a particular style will favor
some possibilities to the point where they become markers of its sound-
world. Listeners, having internalized these routines, may develop a sense for
that core vocabulary even without being able to express their knowledge in
words.
Many chapters of this story remain untold. In Bebop and later jazz styles,
the descending voice leading of the diatonic fifth-sequence is chromaticized,
allowing roots to descend by semitone, tritone, or fifth. Here the 2-in-12
diagram of chromatic dyads can model both root progressions and upper-
voice pairs (Figure C1). Charlie Parker’s “Blues for Alice” uses this
chromaticized space to reimagine the blues as a sleek machine, the chord-
roots making three complete turns through the 2-in-12 diagram, moving
radially or by clockwise step, with upper-voice fifths always sliding
downward (Figure C2).3 This is an ancient musical logic, part of a dialogue
beginning before Ockeghem’s “generalized fauxbourdon” and still audible in
Neil Young’s “Helpless” (§2.3, §3.4, §7.3). From this point of view, Bebop is
more Romantic than modernist; for rather than rejecting functional
convention, Bebop musicians complexified it, adding new stories to the
building rather than razing it to the ground.
Figure C1. Bebop harmony extends the circle of dyadic fifths by allowing tritone substitution. This
replaces the 2-in-7 diatonic spiral with the 2-in-12 chromatic spiral. The lower diagram can be read
either as referring to root progressions or to upper-voice pairs connected by efficient voice leading.
Figure C2. The harmonic structure of Charlie Parker’s “Blues for Alice.” The numbers above the staff
refer to motion on the previous example’s 2-in-12 circle. Upper voices include the important tones in
Parker’s melody.

This book has tried to reconceive musical schemas so as to highlight their

deeper structure, taking their most basic forms to be those most closely
connected to abstract musical geometry: a mathematical substrate that is then
embellished, varied, and refined to produce the more specific idioms
characterizing individual styles. Thus we can distinguish the general
Pachelbel progression, the ubiquitous progression that harmonizes parallel
thirds as root-third and third-fifth in alternation, from the particular
“Romanesca” that operates in the galant, a I–V6–vi–I6 opening gambit that is
comparatively infrequent elsewhere. In much the same way, we have
analyzed the “converging cadence” as just one manifestation of the diatonic
third’s basic voice leading, particularly important in the classical period but
belonging to a larger genus. By taking the geometrical substrate to be
primary, we can highlight the ways in which different musical dialects
exploit the same fundamental core. And we can understand why the same
schemas reappear in so many different styles, as composers in different eras
exploit the same set of musical relationships.
The collectional hierarchy represents a point of contact between different
strands of music theory, including schema theory, scale theory, voice-leading
geometry, neo-Riemannian theory, and Schenkerian analysis. The
deemphasizing of voice exchanges is fundamental to Schenkerianism,
emblematic of its ambition to reduce complex surfaces to simpler templates.
Voice exchanges can often be interpreted as the product of motion “on the
surface,” decorating a crossing-free background that is recognizably scalar.
These background patterns can in turn be represented as combining
transposition or inversion at multiple levels, a strategy that encompasses the
basic moves of “neo-Riemannian theory.”4 Conversely, reconceiving chords
as scale-like entities allows us to transcend analytical literalism, permitting
“chords” whose notes may not be literally present, or which may appear in
multiple octaves while belonging to none. The resulting analyses are at once
Schenkerian and scale-theoretical, expressing recognizably Schenkerian
insights in less metaphysically freighted language (e.g., Figures 1.2.2, P5.3,
P5.7, and 9.7.3).
Compositionally, this approach lets us extend the core techniques of
earlier music, including registral inversion, efficient voice leading, contrary-
motion counterpoint, independent voices pursuing their own big-T little-t
trajectories, hierarchically nested collections, musical networks containing
multiple types of transformations, and sequences. This synthesis could
perhaps lead to a new creative practice that detaches compositional
technique from its schematic moorings. If I ever write another book, it will
be about how we might use these ideas to make new music—including new
forms of dissonant and atonal music.
As a preview of that effort, and a conclusion to this one, let us adapt the
contrary-motion technique of the Waldstein opening in accordance with the
Prime Directive. Figure C3 starts with an arbitrary chord progression, the
first measure connecting diatonic fourth chords by efficient descending
voice leading. Every subsequent measure permutes the voices as shown by
the arrows, producing a perpetually descending four-voice canon. This
progression in turn defines a series of abstract scales whose degrees appear
in every octave, along which a melody might move. Figure C4 composes
melodies that move along these scales: here, the top voice sounds offbeat
eighths that ascend by scale degree, while the next line plays the repeating
pattern + 3, –3, +3, +3 at a delay of six quarter notes. The result is simple
and algorithmic but also charming, process music inspired by a classical
sonata. By reconceiving the past, we can help revitalize it.
Figure C3. A four-voice round generating a series of scale degrees. In the top-staff chorale, the
soprano always sings scale degree 65, alto 63, tenor 62, and bass 60. After each three-chord cycle, the
chords are transposed down by step: thus the scale-degree label assigned to a pitch increases by one
with each iteration (cf. Figure 1.2.2).
Figure C4. Melodies moving relative to the previous example’s scale degrees.

This, I think, is just one of many reasons to be optimistic about the

musical future. Computers and abundant musical data are transforming our
sense of what we can know, replacing speculative generalization with solid
empirical claims.5 New electronic instruments are allowing us to combine
algorithm and intuition in unprecedented ways (e.g., Figure P9.1). Pluralism
and multiculturalism are giving rise to a music theory less fixated on
European masterworks and a pedagogy better suited to contemporary music-
making. Theorists have a chance to nudge culture here, reinterpreting basic
musical concepts so that they are more directly relevant to our current needs.
We can begin by painting a picture of earlier music not as a cherished
museum of unapproachable masterpieces, but as a springboard to music that
has yet to be made. In this way we might truly take ownership of the
tradition we have ostensibly inherited.

1
Both Leibniz and Thelonious Monk described music as a kind of unconscious mathematics
(Rohrmeier 2020).
2
See appendix 3 for another perspective on both the Fenaroli and the “horn fifths” schemas.
3
My diagram factors the upper voices into fifth-related pairs such as third-seventh, fifth-ninth, and
so on. When these pairs are related by chromatic transposition (e.g., two tritones or two fifths) they
can be modeled by the diagram in Figure C1. In other cases, we have to postulate that upper voices
move within a diatonic field whose modulations are directed by the roots.
4
See §1.1, appendix 3, and Tymoczko 2023.
5
Particularly important here is the development of musicxml, a cross-platform format for
representing musical scores, and music21, a software package for analyzing and creating scores
(Cuthbert and Ariza 2010).
 APPENDIX 1

Fundamentals

Geometry begins in the commitment to take distance seriously. In this book

the relevant notion of distance is usually voice-leading distance,
corresponding to the physical distance a finger moves on an instrument such
as the piano. Our analytical atoms are “paths in pitch class space,”
combining an initial pitch class (such as “D” or “E quarter tone sharp”) with
a directed magnitude (such as “fourteen semitones up” or “two and a half
semitones down”). These objects are conceptually unproblematic and easy to
hear; we can notate them using a number to indicate how a pitch class
moves: E ♭ for the descending four-semitone path, E ♭ for the
ascending eight-semitone path, and so on.1 Alternatively we can use pitch
labels, writing G4→E♭4 and G4→E♭5 to identify the corresponding pitch-
class paths. For legibility I will eliminate the numbers when a path connects
its pitch classes by the shortest possible route, using G→E ♭ for the
descending four-semitone path.
Distance is always measured in scale steps. This means that the size of an
interval is measured along some contextually relevant scale: E–G has
chromatic size 3, diatonic size 2, pentatonic size 1, and triadic size 1 (e.g.,
along the C major triad). Notions such as “transposition,” “inversion,” “path
in pitch-class space,” and “chord type” are also scale-dependent: in the
diatonic scale (C, E, G) is transpositionally related to (D, F, A), and hence
the same type of chord, but this is not true in the chromatic scale. I generally
use Tx, Tx, and tx for chromatic, scalar, and chordal transposition by x steps;
sometimes I use τ x for transposition along a chordal subset. Pitches are
labeled using a generalized “midi note numbering” in which scale steps are
size 1 and the number 60 is as close to middle C as possible. For the
chromatic scale, this means B3 = 59, C4 = 60, C♯4 = 61, D4 = 62, and so
on; for the white-note scale, B3 = 59, C4 = 60, D4 = 61, etc.; and for the C
major triad, G3 = 59, C4 = 60, E4 = 61.
This relativity of musical distances underwrites a hierarchical set theory
in which transposition and inversion can apply not just chromatically or
diatonically, but at the chordal level as well. This is represented pictorially in
Figure 1.2.6, embedding the C major triad inside a C diatonic scale that is
itself embedded in the twelve-tone equal tempered chromatic scale.
Transposition is represented by addition on input scale degrees: to transpose
(C4, E4, G4) by ascending triadic step, we add 1 to its triadic representation
(60, 61, 62), producing (61, 62, 63) or (E4, G4, C5) (i.e., t1); to transpose it
along the diatonic scale, we add 1 to its diatonic representation (60, 62, 64),
obtaining (61, 63, 65) or (D4, F4, A4) (T1); to transpose chromatically, add 1
to get (61, 65, 68) or (D ♭ 4, F4, A ♭ 4) (T1). Inversion is represented by
subtraction from a constant value, sending x to c – x for some constant c.
The pitch c/2 is a “fixed point” held constant by this inversion; when C is
odd, this fixed point lies halfway between scale degrees (e.g., a quarter tone
in twelve-tone equal temperament). To invert (C4, E4, G4) around C4 within
the intrinsic scale C–E–G, we subtract each number in its triadic
representation (60, 61, 62) from 120, which is twice the fixed point; this
produces (60, 59, 58) or (C4, G3, E2). To invert diatonically around C4, we
subtract each number in its diatonic representation (60, 62, 64) from 120,
producing (60, 58, 56) or (C4, A3, F3). To invert chromatically we do the
same with (60, 64, 67), obtaining (60, 56, 53) or (C4, A♭3, F3).
Transformations at one level often counteract their analogues at another.
Figure 2.1.4 shows how chromatic transposition by four ascending
semitones can nearly counteract transposition downward by one major-triad
step, producing efficient voice leading as its residue. Figures 3.1.2 and 3.3.1
show how this works for diatonic dyads, Figure 3.4.1 for diatonic triads, and
Figure P8.2 for the diatonic scale considered as a seven-note chord in
twelve-note chromatic space. Figure A1.1 shows that inversion can produce
the same sort of hierarchical cancellation: we begin by labeling the notes of
the C major triad in two different ways, once using triadic scale degrees and
once using chromatic scale degrees; choosing two notes x and y to remain
fixed, we invert twice, sending each note n to (x + y) – n; the figure shows
this calculation for each of the three pairs of notes within the triad. Once
again the combination of triadic and chromatic transformations nearly cancel
out, here keeping two notes fixed and moving the third “parsimoniously,” by
one or two semitones. The result is a voice leading between two similarly
voiced, inversionally related pitch-class sets, known to theorists as “neo-
Riemannian transformations.” These transformations are generalized as
“Gesualdo’s trick” in §1.1.2 There, two voices move in parallel, combining a
generalized neo-Riemannian transformation with a transposition. Such voice
leadings always preserve the distance between at least two voices, and are
discussed further in appendix 3. Interested readers can also download
computer code allowing for the arbitrary embedding of arbitrary collections,
and providing a variety of set-theoretical transformations at every
hierarchical level.3

Figure A1.1. Inversion occurring at both the triadic and chromatic levels. The two inversions preserve
the same pair of notes and combine to produce an efficient voice leading.
 Formally, a voice leading can be defined as a collection of paths in pitch-
class space, measured along some scale. Voice leadings are an important
component of implicit musical knowledge, representing possible routes from

one chord to another. We can write (C, E, G) (C, F, A), meaning “C,
whatever octave it is in, stays fixed, while E, whatever octave it is in, moves
up by semitone, and G, whatever octave it is in, moves up by two
semitones.” This voice leading is measured in chromatic space; measured

diatonically it would be (C, E, G) (C, F, A), since the distance from G

to A is one diatonic step. The written ordering of the paths is immaterial,

with (C, E, G) (C, F, A) and (E, G, C) (F, A, C) equally good

representations of one and the same voice leading. As with paths in pitch-
class space, the absolute register of the voices is unspecified, but the distance
moved by each voice is; given a voice leading and the pitches in the first
chord, we can calculate the pitches in the second. We can eliminate the
numbers when notes move by the shortest possible distance, writing (C, E,
G) → (C, F, A) for the voice leading described above. Alternatively, we can
use pitch-space voice leadings to represent pitch-class voice leadings,
writing (C4, E4, G4) → (C4, F4, A4) to identify general pitch-class paths
whose voices could be in any octave.
There are two different reasons why we might use voice leadings as just
defined. The first is to group together related but distinct pitch-space
transformations, as when we consider the voice leadings (C4, E4, G4) →
(C4, F4, A4) and (E3, C4, G4) → (F3, C4, A4) to instantiate the same basic
pattern. In this case, the notion of voice leading involves a certain amount of
abstraction from the musical surface. The second is to model scalar objects
whose “notes” are melodic slots available in any octave. Here, the musical
object is inherently abstract, and the pitch-class perspective a faithful
reflection of its underlying ontology. What is interesting is that the same
formalism covers both cases: we can think of the voice leading (C, E, G) →
(C, F A) either as a general description for a class of pitch-space voice
leadings, or as a voice leading between two scales whose notes exist in every
octave.
There are several important categories of voice leading. Bijective (or note-
to-note) voice leadings send each note in the first chord to exactly one note
in the second: thus (C, E, G) → (C, F, A) is a bijective voice leading from C
major to F major, while (C, C, E, G) → (A, C, F, A) is a nonbijective voice
leading between these same chords.4 I usually focus on bijective voice
leadings. A voice exchange is a bijective voice leading from a chord to itself
whose paths sum to zero, like (C, E, G) → (E, C, G). Voice exchanges can
be decomposed into a series of pairwise voice exchanges that move two
voices in contrary motion by opposite amounts, exchanging their notes.5
Voice exchanges, thus defined, involve contrary motion (i.e., paths that sum
to zero): the voice leading (C4, E4) → (E4, C5) is not a voice exchange even
though the two voices exchange pitch classes; instead it is a transposition
along the chord.
Any voice leading, between any chords whatsoever, can be factored into a
voice exchange and a remainder that is spacing-preserving or strongly
crossing free. Spacing-preserving voice leadings preserve chordal-step
distance and therefore cannot be arranged in register so that their voices
cross: if plotted on the pitch-class circle, each voice can glide smoothly from
its starting point to its destination without ever sounding the same pitch class
as any other voice, except perhaps at its endpoints. If a voice leading is not
spacing-preserving, then its voices will cross when plotted on the pitch-class
circle; this means they can be arranged in register to produce literal crossings
(Figure A1.2). A spacing-preserving voice leading between transpositionally
related chords can always be represented as a combination of transposition
along the scale (T) and transposition along the chord (t). Similarly, a
spacing-preserving voice leading between inversionally related chords
combines inversion along a scale (I) with inversion along the chord (i). For
unrelated chords, a spacing-preserving voice leading is called an interscalar
transposition; all the interscalar transpositions between any two chords can
be derived by combining any one interscalar transposition with
transpositions along the chord.6 Voice crossings never make a voice leading
smaller: hence between any two chords, there is always a minimal voice
leading that is crossing-free. This means that hierarchical self-similarity can
arise as the byproduct of the search for efficient voice leading, in both the
melodic and harmonic domains.
Figure A1.2. The voices in the leftmost voice leading cross each other; those in the center do not, even
though they contain the same paths. The paths on the right will never cross no matter what octave they
are in.

The spiral diagrams represent the spacing-preserving voice leadings

between the transpositions of one or more chords—in other words, the
possibilities that remain when we ignore voice exchanges. Appendix 2
derives the diagrams in two different ways. These representations are
abstract and topological rather than concrete and geometrical, more like
subway maps than topographical maps. As a general matter, they show us
how hierarchically nested transpositions combine: we can see from the 2-in-
7 diagram that two applications of t1T–4 produces T–1 and that seven
applications produces t–1, for any two-note chord in any seven-note scale;
similarly, it is clear from the structure of 3-in-12 space that there is always a
voice leading that, when repeatedly applied, cycles through the chord’s
major-third transpositions before returning every voice to its starting note;
by contrast, it is clear from the 7-in-12 diagram that there can be no such
cycle, which is why enharmonic respelling will always be necessary
whenever any modulatory schema is repeated until it returns to its starting
key. Perhaps most importantly, the diagrams show us which combinations of
transposition-along-the-chord and transposition-along-the-scale come closest
to counteracting one another, producing purely radial or nearly radial
motion. Of course, the degree to which they actually do so depends on a
chord’s intervallic structure.
When a chord divides the octave nearly evenly, the two kinds of
transposition are similar and the radial voice leadings are efficient or small
(Figures A1.3–4). As I emphasized in §2.3, this sort of contrapuntal
proximity does not imply any perceptible harmonic relationship. Sometimes,
however, the spiral diagrams do represent harmonic similarity, or shared
note content. This occurs when we have a maximally even chord whose size
is relatively prime to the size of our scale: such a chord will be a “near
interval cycle” or stack of intervals k scale-steps large, with one additional
interval of size k ± 1; the basic voice leading moves one of these notes by
just one step, changing the unusual interval to size k and producing another
interval cycle that is the k-step transposition of the first. (This is the starting
point for the “generalized set theory” of §8.7.) It follows that chords adjacent
on the spiral diagram share all but one of their notes, with the mismatched
note differing by only one step; hence they are similar both contrapuntally
(in the sense of being connected by efficient voice leading) and harmonically
(in the sense of sharing all but one of their notes). Such chords can often
play similar musical roles: the most familiar examples are the diatonic triad,
where adjacent (third-related) chords have similar functional roles (§3.4,
§7.1), and the seven-note diatonic scale, where adjacent collections give rise
to similar modes (e.g., mixolydian and ionian, §8.6) and constitute nearby
tonal regions (§8.3).7
Figure A1.3. A symbolic diagram showing how three near symmetries affect the shape of the graph.
For a nearly T-symmetrical chord, the distance between the rings of the spiral is small; for a nearly I-
symmetrical chord, the distance between a chord and its inversion is small when we superimpose their
spiral diagrams; for a nearly P-symmetrical chord, the curved arrow representing one or more pairwise
voice exchanges is small. In this book I do not try to represent these distances faithfully.
Figure A1.4. Chords sharing the same angular position need not be contrapuntally close in an absolute
sense.

It is in this context that we can best understand the problem of doublings

and incomplete chords. For while it is mathematically possible to construct
circular spaces for chords with a fixed number of doublings, these inevitably
spread sonically similar objects throughout the spiral (e.g., CCEG, CEGG,
and EGCE on Figure A1.5). In other words, contrapuntal proximity diverges
from harmonic similarity in a particularly egregious fashion. There is no
easy fix to this problem, since harmonically similar states such as (C, C, E,
G) and (C, E, E, G) are genuinely different contrapuntally: the former is two
diatonic steps away from an F major triad whereas the latter is not. This
means we have to choose whether to model voice-leading distance,
harmonic similarity, or some other music-theoretical notion. We cannot do
everything at once.
Figure A1.5. Circular diagrams for diatonic triads with one doubling. (top) Superimposing three
copies of 4-in-7 spiral diagram; here r, t, and f indicate whether the root, third, or fifth is doubled.
(bottom) A more abstract version of the graph.

What we can do is model music hierarchically, treating chords as abstract

scale-like objects whose notes may not all be present, or may appear in
multiple octaves simultaneously. Recall the Waldstein analysis in §3.7,
where the four voices pass through a range of different harmonic
configurations, all represented by a background voice leading with three
logical voices. Here, doubling and voice exchanges are a byproduct of the
surface-level motion of voices rather than of the deeper counterpoint. The
two-tiered picture thus suggests a strategy for implementing what is
sometimes called cardinality equivalence—the perspective on which CGG
and CCG are “the same,” and hence represented by a single location in
geometrical space.8 We can model chords like CGG and CCG as different
configurations of surface voices within a single underlying two-note “scale”
CG, just as we represented the second and third chords in Figure 3.7.2
(GGBD and GGBB) as configurations of surface voices inside a scale-like
triad. Similarly, we can postulate background voice leadings containing
notes not found on the surface, so that two surface voices (e.g., the third CE)
could represent a complete triad at the background level.9 In this way, the
hierarchical perspective can bridge the gap between mathematics and
musical intuition. It also suggests the interesting analytical project of
providing hierarchical analyses for passages like Figures P5.3 and P5.7.
One possibility, not discussed in this book, is to allow surface voices to
move from chord to chord by a process of quantization, a useful musical
transformation that snaps “out-of-grid” notes into a scale. Consider the
Palestrina passage in the upper-left corner of Figure 4.9.6: there the bass was
interpreted as alternating between two different moves along the triad (t2 and
t1). An alternative conceives the bass as always moving to the next chordal
tone by the smallest possible nondescending interval: when the fifth is in the
bass, it stays fixed to become the root of the next chord (0 steps); when the
root is in the bass, it ascends by step to become the fifth of the next chord
(+1 step). Instead of a frog hopping along a gently shifting circular
arrangement of lily pads (§1.2), we can imagine one set of lily pads sinking
under the water while a new set rises, so that the frog has no option to “move
with the lily pads”; it must choose to hop clockwise or counterclockwise (or
hop vertically when that is possible). The advantage of this approach is that
it allows us to deal with situations where different groups of voices move in
different ways. Indeed, it allows us to avoid postulating any specific voice
leading at the background: lily pads rise and fall, while surface voices move
in various ways.
Sometimes, however, we do want to postulate specific voice leadings at
the scalar level, and sometimes the number of background voices actually
changes. This can happen in two ways, depending on whether the change
occurs at the arrival of a new chord or while a chord is sounding. In the first
and simpler case, a voice simply enters or exits the texture at the start of a
new harmony, with the existing voices articulating a bijective voice leading
between the smaller chord and a subset of the larger (Figure A1.6).10 More
complex are those situations where the number of voices changes during a
chord’s lifetime: in the first four right-hand chords of Figure 9.3.1, for
example, the alto doubles the bass at the octave, the four right-hand voices
implying a three-voice triad at the background level; from the fifth to the
sixth chords, however, the two surface B voices move by contrary motion to
D7, (G, B1, B2, D) → (F♯, A, C, D). Here we might want to say that a single
voice “splits” into multiple voices.11 In other words, the two right-hand Bs
begin as surface-level doublings of a single abstract voice, and end as two
different abstract voices—the “splitting” promoting them from surface to the
background; the converse process of “merging” moves them from
background to surface.
Figure A1.6. Two ways in which the number of background voices can change. In the first, a new
voice enters at the moment of a chord change; in the second, the number of voices changes while the
chord is sounding. On the right, the top line features a three-note scale, whose degrees are numbered
above the example; the lower lines feature a four-note scale, whose degrees are numbered below.

That chords can function in an abstract and scale-like way is an important

Schenkerian insight, one that significantly increases the power of our
geometrical models. This perspective tends to deemphasize voice crossings
by assigning them to the surface. A complementary approach, more
Tinctorian or set-theoretical, deemphasizes transposition in order to
represent voice crossings more perspicuously. Figure A1.7 shows an abstract
graph representing the cross section of three-note voice-leading space, an
intrinsically two-dimensional space that is compressed into a line-segment
on the circular diagrams. Here each chord appears three times,
corresponding to its three modes, while the mirror boundaries represent the
three pairwise voice exchanges swapping the chord’s adjacent pitch classes.
As in Figure 3.5.4, transposition comes out of the page. We can use this
figure to represent classes of voice leadings equivalent under the Tinctoris
Transform, including those with voice crossings. Figure A1.8 simplifies and
analyzes the opening of Beethoven’s Op. 109 variations theme. Where a
traditional set-theoretic perspective would simply observe the repeated
presence of the same set class (here, the diatonic triad 024), our more
contrapuntal set theory labels the specific mappings between the notes,
assigning different voice leadings to different geometrical paths. These
transformations generalize the concept of voice leading from specific chords
to generalized chord types, endowing set theory with rich contrapuntal
content. We can construct analogous graphs for chords of any size, using
two-dimensional polygons to represent complicated higher-dimensional
geometries—and providing subway-style maps for otherwise-
incomprehensible spaces.12

Figure A1.7. The cross section of three-note chord space. Each chord appears in all three of its modes;
“mode 1” is an ordering with the smallest interval between the first two notes. Each boundary acts
geometrically like a mirror, exchanging voices’ pitches and labels: if voice 1 has C and voice 2 has E,
then exchanging them sends voice 1 to E and relabels it voice 2, while sending voice 2 to C and
relabeling it voice 1.
Figure A1.8. A simplified version of Beethoven’s Op. 109 variations theme, with its voice leadings
analyzed into three components: voice crossings (cxy, with x and y identifying the crossed voices),
transposition along the diatonic scale (Tx), and transposition along the triad (tx). Underneath the
example, I give a (diatonic) set-theoretical analysis, consisting solely in the observation that every
chord is a diatonic triad (024). Beneath that, I identify the passage’s set-class voice leadings, obtained
by omitting the T component. The second staff of music shows an alternate realization of these same
set-class voice leadings, keeping the E major harmony fixed. This music is related by the Tinctoris
Transform to Beethoven’s original. For clarity, I separate out the two components of the “c23t1” voice
leading. When voices cross, they exchange numbers, so that c21 exchanges the notes in the first and
second voices while also renumbering them accordingly. Below, I graph the music in set class space.

1
These objects differ from the pitch-class intervals of earlier music theory in distinguishing
multiple routes between the same points: thus “start at G and move down by four semitones” is
different from “start at G and move up by eight semitones,” even though both connect G to E♭. See
Tymoczko 2011a and 2023.
2
The opening of Schoenberg’s Op. 36 violin concerto is a nice example, moving from F♯2, B2,

D♯3, E3, A3, B♭3 to the inversionally related A♭2, D3, F3, G3, C4, D♭4, with both hexachords spaced
(3, 1, 1, 2, 1) in chordal steps. The registral distance between the three-note chromatic cluster is
preserved.
3
See http://www.madmusicalscience.com/quadruple.html.
4
A subtlety: if we think of chords as multisets with specific doublings, then (C, C, E, G) → (A, C,
F, A) could be understood as a bijective voice leading from C-with-doubled-root to F-with-doubled-
third (Callender, Quinn, and Tymoczko 2008, Tymoczko 2011a).
5
This conception of “voice exchange” is minimal and technical, referring to a specific type of
voice leading. Schenkerian theory uses a more robust notion (Cutler 2009).
6
See Tymoczko 2008 and 2011a. In my earlier work I used scalar and interscalar interval matrices
to represent these voice leadings; the spiral diagrams provide a geometrical representation of these
algebraic objects.
7
If we take 7-in-12 space to represent diatonic modes, then angular neighbors are modally similar:
for instance, a one-step clockwise move changes C ionian to C mixolydian.
8
I sometimes treat doubling as part of chordal identity (so that, for example, CGG and CCG lie on
different rings of Figure 3.5.3), representing changes of spacing by curved arrows.
9
Headlam 2012 criticized voice-leading geometry on the grounds that it could not accommodate
cardinality equivalence or incomplete chords. I think that the hierarchical picture allows us to model
many of these phenomena.
10
For example, Figure P5.7 when the bass sounds D. The upper voices continue to move along the
three-note subset.
11
Callender 1998.
12
Tymoczko 2020b.
 APPENDIX 2

Deriving the Spiral Diagrams

The spiral diagrams represent all possible combinations of big-T

transposition along the scale and little-t transposition along the chord, for
any choice of chord and scale. As we have seen, these transformations are
fundamental to a wide range of music, capturing some substantial
component of intuitive knowledge. The diagrams can be derived either as the
union of helices, or as subspaces of the higher-dimensional space of n-note
chords.
(a) The union of helices. The easiest way to arrive at the spiral diagrams is
to build them out of helices.1 Start with the two-note spiral shown on the
upper right in Figure P2.9; we will work in chromatic space for simplicity.
Choose some point to represent the chord (C4, G4). Now imagine a third
dimension, extending outward from the paper and representing chords’
“center of gravity” or pitch sum. In this three-dimensional space, position in
the xy plane will represent pitch-class content, so that chords with the same
pitch-class content will always be vertically aligned.2 As we transpose the
(C4, G4) dyad upward, we spiral outward from the paper to form a helix or
corkscrew, returning to our original position when each voice has moved up
by octave. (One can imagine a glissando in continuous space, ascending
smoothly between equal-tempered perfect fifths; this gives meaning to every
point on the helix.) The helix contains all and only those ordered chromatic
dyads whose first note is a perfect fifth below its second; it contains no
dyads whose first note is a perfect fourth below its second. Starting at (C4,
G4), or (60, 67) in chromatic pitch numbers, we do not return to (C, G) until
we reach (C5, G5) or (72, 79). The sum of this second dyad is 151, which is
24 more than the sum of the first, 127.
Now consider the effect of transposition along the chord (t1). This turns
every perfect fifth into a perfect fourth, increasing its sum by 12: (C4, G4) or
(60, 67) becomes (G4, C5) or (67, 72), which sums to 139; (D4, A4) or (62,
69) becomes (A4, D4) or (69, 74). Geometrically, this creates a second helix,
interleaved with the first, whose points represent fourths rather than fifths
(Figures A2.1–2). Every point on this second helix shares the same pitch-
class content with the dyad immediately below it vertically, but sums to 12
more than its lower neighbor. Thus as we ascend vertically from any point,
we alternate between helices, moving from fourth to fifth and back,
repeatedly applying t1 to increase the pitch-sum by 12.3 Looking down on
the double helix from above gives us the two-note spiral diagram shown in
Figure A2.1. Each helix is continuous, with each of its points representing a
specific dyad; points not on either helix have no musical meaning.

Figure A2.1. (a) Three-dimensional helices each containing the transpositions of an individual mode.
(b) When viewed from above these produce the spiral diagrams.
Figure A2.2. The modes of (C, G). Perfect fifths sum to 7 + 24i where i is an integer, while perfect
fourths sum to 19 + 24i. As the sum increases by 12, we move from one mode to the other, hopping
between the helices in the previous example.

In this three-dimensional space, every path between helical points

represents the voice leading taking the starting pitches to the destination
pitches, first note to first note and second to second; the two helices do not
intersect themselves or one another. To use the 2D version of the diagram,
we need to ensure that different voice leadings can be distinguished when
seen from above; this means disallowing purely vertical paths that would
disappear from that vantage. We can accomplish this by requiring that paths
either move parallel to a helix or horizontally between them. This guarantees
that a transposition along the chord will make a “loop” when seen from
above; it also creates the misleading impression of a self-intersecting curve.
The same reasoning applies for chords with three or more notes; for an n-
note chord in an o-note scale, transposition by octave adds on to the sum of
the chord’s scale degrees and creates a helix; transposition along the chord
adds o to its sum and creates a new helix interleaved with the first. We can
do this n times before we return to the original helix, creating n interleaved
helices, one for each mode, and generating the spiral diagram when viewed
from above.
(b) As subspaces of the higher-dimensional space of all chords.
Alternatively, we can derive the spirals as lines passing through the higher-
dimensional geometrical space containing all possible n-note chords. These
are twisted, mirrored donuts, technically known as orbifolds, whose structure
can be quite hard to intuit. Recipe 1 in Figure A2.3 gives general instructions
for forming voice-leading spaces whose points represent objects such as
chords or set classes and whose paths represent voice leadings between their
endpoints. This recipe allows us to associate any voice-leading-in-a-score
with a path-in-the-space; to go from path to voice leading, we need Recipe 2
in Figure A2.4.

Figure A2.3. Recipe 1, for constructing a range of musical geometries.

Figure A2.4. Recipe 2, for interpreting paths in a geometry.

Both recipes exploit the fact that musical space is structured like
wallpaper, composed of many equivalent regions containing different
versions of the same basic objects (e.g., chords or set classes, illustrated
metaphorically by Figure A2.5). We can represent music using any one of
these regions, each of which contains just one point for each object (except
at their boundaries, which contain multiple points representing the same
object). As we move off a region’s boundary, we either “bounce off” a
mirrorlike point, or “teleport” between boundary points representing the
same musical object. In either case, we have to apply some musical
transformation to all the coordinates in our region. This is because we have
entered a new region of pitch-space that is equivalent to the first but contains
different pitches (e.g., reordered and transposed by octave, relative to the
previous region). It follows that geometrical boundaries can be identified
with specific musical transformations, and hence that geometrical modeling
can illuminate musically significant operations such as transposition along
the chord.

Figure A2.5. Musical space is like a piece of wallpaper, constructed out of many similar regions. Any
path in this larger space can be represented in a single region, though paths may appear to “bounce
off” certain boundary points (β), or “teleport” between others (α). These boundaries can be associated
with specific musical transformations.

The spiral diagrams are lines or circles in the higher-dimensional spaces

containing all the transpositions of some chord. The simplest is pitch-class
space, which can be derived from the line representing every conceivable
pitch (Figure A2.6). Recipe 1 tells us to choose a region containing one point
for every pitch class, with duplications only on the boundary; here, I choose
the octave from C4 to C5, 60 to 72 in numbers. This is chromatic pitch-class
space. Applying T12 turns the left boundary into the right, and applying T–12
turns the right boundary into the left. Within the region, a path’s length and
direction give us the relevant voice leading: one-unit leftward motion lowers
a note by semitone, two-unit rightward motion raises it by two semitones,
and so on. (For pitches in the region, we can calculate these paths by simply
subtracting the starting coordinate from the destination coordinate.) By
imagining smooth glissandos, we find that an ascending octave moves off
the rightward boundary, reappearing on the left to return to its starting point,
while descending-octave motion does the same in the opposite direction. We
can check this using Recipe 2, which tells us that when we move off the
right boundary, we need to apply T12 to every point’s coordinate, since that
is the transformation that turns the destination point (left boundary) into the
departure point (right boundary); conversely, we subtract 12 from every
point when moving off the left boundary. The figure shows the path
representing ascending-third motion from B♭ to D.

Figure A2.6. Representing music using a one-octave region of chromatic pitch space. Within the
region, paths can be determined by subtracting start-coordinate from end-coordinate; when moving off
the boundary, one first has to apply a transformation to all coordinates.

For two-note chords, we begin with the Cartesian plane representing all
ordered pairs of chromatic pitches. The square at the center of Figure A2.7 is
defined by the equations 12 ≤ x + y ≤ 24, and 0 ≤ y – x ≤ 12. Every ordered
pair of pitches is equivalent to either a single point in the square’s interior or
a pair of points on its boundary.4 Figure A2.8 rotates the square and labels
the points representing unisons, major thirds, and their inversions. Points on
the left edge (summing to 12) are related by t1 to points on the right
(summing to 24): for example, the left-edge sixth (2, 10) is duplicated as the
right-edge third (10, 14), and the left-edge third (4, 8) is duplicated as the
right-edge sixth (8, 16).5 According to Recipe 2, moving off the right
boundary requires us to apply t1 to every point in the space, while moving
off the left boundary requires us to apply t–1. (The top and bottom
boundaries meanwhile reorder each dyad while perhaps transposing notes by
octave.6) Figure A2.8 uses curved lines to represent the process of
“teleportation,” but this is merely a graphical contrivance that does not
substantively alter the figure. The result is recognizable, modulo a little
stretching and twisting, as the spiral diagram for chromatic major thirds and
minor sixths.
Figure A2.7. A derivation of two-dimensional chord space. The square is bounded by the lines y = x
(unisons), y = x + 12 (octaves), x + y = 12, and x + y = 24. The two lines y = x + 4 and y = x + 8
contain major thirds and minor sixths respectively; these are the two strands of the spiral. The dashed
arrows represent transposition up by one step along the chord, sending root up to third and third up to
root; these link chords on the x + y = 12 face with their equivalents on the x + y = 24 face.
Figure A2.8. The region shown in the previous figure, rotated by 45 degrees. For ease of reading, I
have shown only the unisons, octaves, major thirds, and minor sixths. Transposition moves
horizontally, while chords lying on a vertical line sum to the same value. Chords on the left boundary
relate by t1 to chords on the right boundary; when we move off the right to reappear on the left, we
need to apply t1 to all the points in the region.

Exactly the same reasoning applies in higher dimensions. Figure A2.9

gives the general equations identifying the region of higher-dimensional
Euclidean space containing one point for every chord (except at the
boundaries, which have multiple points representing the same chord).
Figures A1.10–11 show the line segments containing major triads in three-
note (three-dimensional) chord space and four-note seventh chords in four-
note (four-dimensional) chord space. Since the extra dimensions are not
relevant here, these diagrams simply draw those segments in two
dimensions. Once again, we see that points on the left boundaries relate by t1
to those on the right.7 As we move off the right boundary, we reappear on the
left, applying t1 to all our coordinates; when moving off the left, we reappear
on the right and apply t–1. Thus horizontal “loops” have the effect of
transposition-along-the-chord. The spiral diagrams connect these horizontal
line segments to form a continuous curve, replacing “teleportation” with
motion along a spiral. This continuous curve appears to self-intersect, but
that is an artifact of representing higher-dimensional structures on a 2D sheet
of paper. As we have seen, these graphs can be extended by adding multiple
chord types, or introducing curved arrows to represent pairwise voice
exchanges, incorporating additional features of the higher-dimensional
geometry.

Figure A2.9. Two equations that determine the region representing chord space for chords of any size:
the first determines a “generalized triangle” containing all set classes whose notes are in order while
spanning no more than an octave; the second determines the circular dimension whose coordinate is
given by the sum of chords’ pitch classes. Here c is an arbitrary constant. In three voices with an
octave of size 12, and choosing c as 0, the equations are x1 ≤ x2 ≤ x3 ≤ x1 + 12 and 0 ≤ x1 + x2 + x3 ≤
12.
Figure A2.10. Chromatic major triads lie on three line segments in the region determined by the
equations in Figure A2.9. This region is three-dimensional, though here I draw it in two dimensions.
Chords on the same vertical line sum to the same value. Chords on the left boundary are related by t1
to chords on the right. Recipe 2 tells us that as we move off the right boundary we need to apply t1 to
every chord in the region.
Figure A2.11. Dominant seventh chords lie on four line segments in the region determined by the
equations in Figure A2.9. This region is intrinsically four-dimensional. Each line contains a different
mode of the dominant seventh. Chords on the same vertical line sum to the same value. Chords on the
left boundary are related by t1 to chords on the right. Recipe 2 tells us that as we move off the right
boundary, reappearing on the left, we need to apply t1 to every chord in the region.

The spiral diagrams come in three flavors (Figure A2.12). When the size
of the chord n divides the size of the scale o, then there are n distinct chord-
forms sharing the same angular position, linked by the transposition o/n. In
this case there is a diagonal voice leading (To/nt–1) that moves chords
radially.8 By repeating it, we can generate all the spacing-preserving
leadings between chords at the same angular position, passing through each
of the chord’s inversions or modes and eventually returning each note to its
starting pitch. When the size of the chord is instead relatively prime to the
size of the scale, there is exactly one chord at each and every radial position;
here a basic voice leading moves chords by one angular position, generating
by repeated application all the voice leadings on the spiral.9 The third type of
graph occurs when the size of the chord shares a common factor with the
size of the scale; in this case, there are between 1 and n chords sharing the
same angular position. We can think of these spaces as having both a
diagonal voice leading (which moves radially, passing through some subset
of the modes before returning each voice to its starting position) and a basic
voice leading (which connects angular neighbors).

Figure A2.12. The three kinds of spiral diagram. (left) When the size of the chord (n) evenly divides
the size of the scale (o) we have n copies of each chord sharing the same angular position; the
diagonal action D moves inward or outward by one radial unit and generates all the voice leadings
connecting chords at the same angular position; these combine with chromatic transpositions (T) to
generate the remaining voice leadings. (middle) When the size of the chord is relatively prime to the
size of the scale, no two transpositions share the same angular position; here the basic voice leading B
moves chords by one angular unit; repeated applications of this basic voice leading generate all the
voice leadings on the figure. (right) When the size of the chord shares a common factor with the size
of the scale, there is both a diagonal action and a basic voice leading.

1
Thanks to Nori Jacoby, who helped me work out the multiple-helix derivation.
2
If we want distance in the xy plane to represent voice-leading distance, we need the n-fold spiral
arrangement used in the book. This is because To/nt–1 is small for nearly even n-note chords.
3
This double helix is the two-note analogue to Roger Shepard’s helical representation of pitches
(Shepard 1964).
4
Imagine a pair of pitch classes x and y: you can always arrange them in pitch so that one is less
than an octave above the other, satisfying 0 ≤ y – x ≤ 12. Since transposition along the chord increases
or decreases their sum by 12, repeated transpositions will eventually satisfy 12 ≤ x + y ≤ 24, producing
either a unique point in the interior or a pair of points on the boundaries. The two quantities x + y and y
– x are the two Tinctorian coordinates from the prelude to chapter 3.
5
In each case, the right-edge duplicate has the left edge’s second pitch as its first pitch, and the
octave-transposition of the left edge’s first pitch as its second pitch.
6
The bottom boundary swaps a dyad’s notes; the top boundary swaps them but also transposes the
first note up by octave and the second note down by octave. These octave shifts change as we move
from region to region.
 7
For details see Tymoczko 2020b.
8
The adjective “diagonal” comes from mathematics, reflecting the fact that a single group acts
simultaneously as a transposition along both chord and scale.
9
Hook 2008 and 2011 was the first to explore the basic voice leading, beginning with the case of
the diatonic scale and eventually extending this idea to other seven-note collections. He also
introduced the big-T, little-t notation used in this book. Tymoczko 2008 and 2011a explored the
general combination of transposition along both chord and scale, but had no way to represent these
relationships geometrically.
 APPENDIX 3

Sequence and Transformation

The topic of sequences belongs to a larger transformational theory that

extends the apparatus of chapter 4 in two ways: allowing transformations
other than transposition and giving arrows an explicit parameter controlling
temporal delay.1 Explicitly notated temporal delays can be used to model
nonrepeating canons. In Figure A3.1, for example, the middle voice imitates
the top voice down a fourth at a distance of one measure while the lower
voice imitates the middle voice down a fifth at a distance of two measures.2
They are also useful when a sequence contains imitation at a temporal
interval smaller than its simple period. For example, the sequential unit in
Figure A3.2 is three beats long, but the top voice imitates the middle at the
distance of a single beat. The model in chapter 4 is unable to capture the
one-beat imitative relationship; an accurate analysis would have the top part
imitating the middle part up a sixth and at a one-beat delay, and the bottom
parts imitating themselves up a third and at a four-beat delay. These more-
complicated arrows move music through pitch and time like the transposing
delay lines of modern audio production.
Figure A3.1. The opening of “Non nobis, Domine,” an anonymous sixteenth-century canon featuring
multiple transpositions and temporal delays. “Dx” represents delay by x beats.
Figure A3.2. The sequence in mm. 37ff of the second F major fugue in The Well-Tempered Clavier.

When the temporal delay is zero, we have a vertical process in which

input and output occur simultaneously. In Figure A3.3 this produces parallel
motion along the diatonic scale, the chromatic scale, and both chord and
scale together.3 (As discussed in §3.6, the lower-left passage is characteristic
of chorale textures in which the upper voices move in doubly parallel motion
along a rapidly changing triad.) Sometimes, however, parallel motion cannot
be modeled with vertical arrows: in Figure A3.4, Kabalevsky moves two
voices along collections of different sizes, creating constantly changing
vertical intervals; here it is easier to use coupled melodic motion, horizontal
arrows whose subscripts indicate that their melodic intervals are always the
same.4 Figure A3.5 shows a vertical process called quantization, a
transformation that sends notes downward until they lie within a chosen
collection.5

Figure A3.3. Vertical sequences involving transposition.

Figure A3.4. Vertical sequences represented with coupled melodic intervals. (top left) Kabalevsky’s
“Etude” (Op. 27, no. 24) features parallel motion along two different collections, changing the vertical
interval between the parts. (top right) Stravinsky’s Rite of Spring (R46 + 2) features contrary motion
in which the upper part takes three steps for every step in the lower part, the two always moving in
opposite directions. (bottom) Bartók’s “Subject and Reflection” (Mikrokosmos 141) moves its parts in
exact chromatic contrary motion. This last sequence can also be analyzed with vertical inversional
arrows.
Figure A3.5. Pärt’s Für Alina transposes the top voice down by eight diatonic steps and then quantizes
to the B minor triad.

We can think of these transformational arrows as algorithms, little chunks

of computer code that generate one part of a score from another. Traditional
sequences are a special class of algorithm that has the property of being self-
replicating: since all parts have the same temporal delay, equal to the length
of the sequential unit, the resulting structure continues indefinitely,
generating an infinite amount of music from a finite source. Twentieth-
century composers often use sequences as a tool for accessing unusual
harmonic states: in Schoenberg’s “Farben,” a short melody moves
canonically throughout the ensemble, producing a variety of different set
classes (Figure A3.6).6 In this context, harmonically unstable sequences are
preferable to harmonically stable sequences as they generate a wider range
of harmonies.
Figure A3.6. The harmonies at the start of Schoenberg’s Five Pieces for Orchestra, Op. 16, III
(“Farben”). The middle five chords are a repeating contrapuntal pattern; since it is not harmonically
stable it generates a variety of set classes.

Also noteworthy are those harmonically stable repeating contrapuntal

patterns that exploit contrary motion. One common idiom uses subsets of the
whole-tone scale, conceived as a harmony (typically an extended dominant,
V9♯5♯11). Two superimposed whole-tone subsets can be transposed by either
an even or an odd number of semitones to produce a whole-tone resultant, as
in Figure A3.7.7 In The Rite of Spring, Stravinsky applies an analogous
technique to diminished seventh chords, descending major seconds in the
first violins against rising major thirds in the lower voices (Figure A3.8). He
then uses the Principle of Musical Approximation, replacing the completely
symmetrical diminished seventh with the nearly symmetrical dominant
seventh, producing slightly irregular melodies.8 This suggests a robust
though perhaps intuitive understanding of the Prime Directive, applying one
and the same technique to the whole-tone scale, the diminished seventh
chord, and the nonsymmetrical dominant seventh.9
Figure A3.7. The Firebird (1919 Suite), R7. (a) The upper two staves move in exact contrary motion,
producing whole-tone subsets; the bottom staff completes the whole-tone scale. (b) The strings move
within the contrary-motion augmented triads.

Figure A3.8. Contrary motion in the string parts at R70 of The Rite of Spring. (a) Descending seconds
against ascending major thirds, forming a series of diminished seventh chords. (b) The same technique
applied to nearly symmetrical dominant seventh chords.

Postwar jazz synthesizes these contrary-motion sequences with tonally

functional harmony (Figure A3.9).10 In Figure A3.10 the left hand plays
“Viennese fourth chords,” a tritone plus a perfect fourth; if the right hand
plays any octatonic subset, then the two hands can move in contrary motion
to produce another octatonic subset, ascending whole tones and descending
semitones combining to express descending-fifth harmony. A related strategy
superimposes stacks of fifths moving independently: because the diatonic
scale is a large stack of fifths, the result tends to be diatonic (Figure
A3.11).11 Pioneered by McCoy Tyner, this technique is so distinctive that
one can evoke modern jazz simply by combining random fourth chords with
equally random melodies composed of quartal trichords.12

Figure A3.9. Jazz sequences in which the upper voices descend semitonally while the bass descends
by fifth. As long as the initial scale contains the tritone transposition of the initial bass note, then the
resulting harmonies will always belong to one of its transpositions.
Figure A3.10. Here the left-hand voicings ascend by two semitones while the right-hand voicings
descend by one semitone; because the voicing forms an octatonic subset, the resultant harmonies are
octatonic as well. The entire passage expresses a sequence of descending-fifth harmonies.
Figure A3.11. At 2’57” in “Passion Dance,” McCoy Tyner plays four-note pentatonic subsets in his
right hand, moving in doubly parallel motion along both chromatic and pentatonic scales; the left hand
plays chromatic fourth chords. Though his two hands move independently and in contrary motion, the
music articulates diatonic subsets. This need not require deliberate planning as it tends to happen
naturally. Compare Figure 1.1.8.

It is also possible to use inversional arrows to form sequences that turn

each successive unit upside down. In Figure A3.12 the lines echo one
another in both a one-bar canon-by-inversion and a two-bar canon-by-
transposition.13 A closer look shows why these sequences are not very
common: since IxIx = T0 for any inversion Ix, every other unit contains
exactly the same pitch classes. It follows that harmonically stable inversional
sequences tend to oscillate between unchanging sonorities. For this reason,
inversion is usually used to create near rather than exact sequences: Figure
A3.13, for example, obtains harmonic variety by transposing each successive
unit; this comes at the cost of destroying the one-bar canon-by-inversion, but
that was not very salient to begin with. Figure A3.14 shows an interesting
canon that combines transpositional and inversional arrows.14
Figure A3.12. A sequence using inversion rather than transposition. Notes are numbered diatonically
with C4 = 60, D4 = 61, and so on; Ix(y) sends x to y – x. The grand period is six bars long.

Figure A3.13. We can obtain greater harmonic variety by transposing each successive unit.
Figure A3.14. Measures 15–18 of Bach’s first two-part invention present a near sequence combining
transposition and inversion (the music is in the lower system of Figure 7.7.5). The transposition moves
the upper voice downward by 10 and 11 diatonic steps in alternation; the inversion moves the top note
of the original sixteenth-note figure up by octave to become the bottom note of the inverted figure.

When we turn to repeating contrapuntal patterns, we find ourselves

unexpectedly close to the concerns of neo-Riemannian theory. The passages
in Figure A3.15 analyze the basic neo-Riemannian voice leadings as one-
note sequences-by- inversion, repeatedly applying the same operation so that
harmonies oscillate between major and minor. As in the transpositional case,
we can eliminate the crossed arrows by reconceiving the sequences
hierarchically: here, we combine “little i” inversion-within-the-chord and
“big I” inversion-within-the-chromatic-scale (compare Figure 4.2.4). The
two inversions nearly cancel out, producing an efficient voice leading as
residue (cf. Figure A1.1).15
Figure A3.15. The neo-Riemannian L, P, and R voice leadings as repeating contrapuntal patterns. See
Figure A1.1 for the calculation of index numbers.

As before, we can obtain greater harmonic variety by transposing

successive units. Figure A3.16 generates the “Moro Lasso” progression from
the “relative” transformation, adding –1 to each chromatic inversion; this
causes the sequence to slide down by chromatic step. One and the same set
of arrows can be used to produce a number of the three-voice sequences
found in §1.1, formalizing the observation they are all “the same” (Figure
A3.17).16 Figure A3.18, meanwhile, analyzes the transpositional and
inversional relationships at play in Figure 1.1.7. The two analyses capture
two different features of the music—first, that the lines move in doubly
parallel motion along three separate diatonic scales, and second, that the
harmonies are inversionally related. It is unusual to find a single passage
possessing both properties.
Figure A3.16. (top) The relative transform as pair of hierarchical inversions. (bottom) The “Moro
Lasso” sequence applies the Tinctoris Transform by subtracting 1 from its chromatic subscripts.
Figure A3.17. Two sequences from §1.1, described with the same contextual inversions used in the
previous example. In the top sequence, the letters a, b, c refer to chromatic scale degrees; in the
bottom, they are diatonic.

Figure A3.18. Two analyses of the sequence in Figure 1.1.7, the first using transposition and the
second using inversion.

At this point, what began as a theory of sequences has evolved into a

general framework for analyzing a very wide range of transformations.
Figure A3.19 illustrates its scope by analyzing “Angst und Hoffen,” the
seventh song in Schoenberg’s Book of the Hanging Gardens, Op. 15.17 The
music begins with a voice leading from an augmented triad (perhaps
“Angst”) to a Viennese fourth chord (perhaps “Hoffen”); I describe this as
schema α, modeling it with a diagram that includes a parameter c
determining the second chord’s transposition.18 This schema appears right-
side-up with c = 0, and then upside-down with c = –3, –1 (recreating the
ascending-step melody of the opening), and –9. The final chord is distorted,
its middle voice a semitone too low—a flexible treatment of intervals
characteristic of Schoenberg’s early atonal music.19 My analysis imagines
the large melodic leap as resulting from an elided transposition-along-the-
chord that turns B4–E♭5–G5 into E♭4–G4–B4 (t–2), thus relating this voice
leading to its predecessor.20 The rest of the example identifies a series of
sequences, the first using Gesualdo’s trick, the second a contrary-motion
wedge, and the third a two-chord transpositional sequence extending I(α)
into a chromatic melody. It ends two chordal steps higher than we expect
(t2), overlapping with the final appearance of α. Though this analysis is
technical and specific, I feel it captures something of Schoenberg’s intuitive
play-with-shapes: the basic musical objects here are not sets, abstract
collections of disembodied pitch classes, but more specific “hand shapes” or
Tinctorian vertical configurations, linked by familiar sequences and voice-
leading patterns. These sequences link the music to the past even while
generating futuristic harmonies.
Figure A3.19. Some salient progressions in Schoenberg’s “Angst und Hoffen.” I focus on the piano
part and omit mm. 5–6 and 11–12.

Approached in this way, the theory of sequences can be generalized into a

transformational theory that is more responsive to information about voice,
register, and time.21 A final generalization permits the transposition and
inversion of any collection along any other. Figure A3.20 does this with
fractional scale degrees.22 An alternative and somewhat more intuitive
approach allows for intervallic networks containing unstated notes (§4.8).
Figure A3.21 moves a two-note motive along an unstated C major triad, the
tones of resolution being merely conceptual and implied—observable by
their transformational effects, much as a planet might be discovered through
its influence on another planet’s orbit.
Figure A3.20. Using fractional scale degrees to transpose an arbitrary collection along the C diatonic
scale.

Figure A3.21. Using unstated chord tones to transpose an arbitrary collection along the C major triad.
Here B and D are interpreted relative to tacit chordal degrees.
 1
Taneyev (1909) 1962 discusses displacements in time as well as pitch.
2
This canon repeats after nine measures, but we can choose not to observe the repeat. The example
comes from Noam Elkies, who notes that the use of multiple temporal delays is unusual; most canons
involve just a single temporal delay. See also Gauldin 1996.
3
The figure’s Petrushka analysis originates in a collaboration with Rachel Hall and Jason Yust,
never published because I felt dissatisfied with its formalism; in retrospect, I was looking for
something like the generalized transformational theory described here (Hall, Tymoczko, and Yust
2010).
4
Horizontal arrows are generally more flexible than vertical arrows. In Figure A3.4, we cannot
analyze the Stravinsky passage using arrows as the upper intervals are three times as large as the lower
intervals; in the Bartók passage, we can use a vertical inversional arrow, as the two parts move in
exact contrary motion. Vertical inversion arrows are characteristic of “K-nets” (Klumpenhouwer 1991
and Lewin 1990).
5
This “quantization” operation is essentially the mathematical floor function: it leaves notes in the
B minor triad untouched, sending any other note downward until it reaches a note in the B minor triad.
Avoiding octaves thus requires diatonic transposition by eight rather than seven steps: F♯5 quantizes

to F♯5, but Pärt harmonizes melodic F♯6 with D5 rather than F♯5.
6
Ligeti borrowed this technique fifty years later in “Monument.”
7
Russell 2018. Early Stravinsky often moves instrumental lines relative to surface voice leading,
as in the opening bassoon parts of The Firebird or R43 of The Rite of Spring. This orchestrational
habit may have led him to consider how the technique could be used in more obviously contrapuntal
ways. In Figure A3.7, Stravinsky moves the string parts within the augmented triads, forming diatonic
melodies. This is not how the passage appears in the original ballet or 1911 suite; it may have been
recomposed after The Rite of Spring.
8
Alternatively one could apply this sort of thinking within the chord, for instance by playing a two-
step interval as measured within a four-note chord (i.e. notes 1–3 or 2–4). The two voices can then be
moved by either an even or odd number of tetrachordal steps to produce subsets of just two harmonies
(notes 1–3 or 2–4). Versions of this technique are used in the Fenaroli and “horn fifths” schemas, as
well as a Prinner variant (cf. Figures 4.5.6, 4.6.1–2, and 4.8.6).
9
The Rite of Spring is full of interesting contrary-motion patterns, for instance at R44 + 4 (seventh
chords in contrary motion forming large octatonic subsets), R60 (triads moving in contrary motion in
winds and trumpets), throughout the “Sacrificial Dance” (e.g., R192–R194, upper voices against
bassoons, or R196, trombones moving down octatonically while winds ascend along the B♭ minor and
C major triads forming a nearly octatonic composite). See Russell 2018.
10
Stuckenschmidt 1965. Here again contrary motion functions to produce harmonic expansion.
11
Nine of the twelve superimpositions of three-note fourth chords produce a diatonic subset.
Different patterns of left-and-right hand motion are more or less likely to preserve diatonicism.
Suppose, for example, left and right hand are each playing three-note fourth chords. If they are close
together on the circle of fifths, they will belong to the same diatonic or pentatonic collection. If one
hand moves up by three semitones while the other moves down by two, their relative position on the
circle of fifths will change by only one; by contrast, if one hand stays fixed and the other moves by
semitone, their configuration changes by 5. The latter change is more likely to turn a diatonic subset
nondiatonic.
12
See https://www.madmusicalscience.com/robots.html.
 13
Since the index numbers are related by octave, this collection of arrows is harmonically stable,
preserving the pitch-class intervals between lines.
14
Compare m. 61 of the first A minor fugue in The Well-Tempered Clavier.
15
Crossed Tx and Ix arrows describe sequences using symmetry operations that can apply to
extended passages of music; dual transformations like Txty and Ixiy describe sequences using voice
leadings that apply to just a single chord. The latter description has the advantage of being insensitive
to a chord’s registral position and internal intervallic structure. Thus it is often useful to go back and
forth between the two. We could, for example, convert a neo-Riemannian voice leading into a large-
unit sequence by writing down an instance of that voice leading, describing it with crossed Ix arrows,
and using these arrows to form a larger sequence. See Tymoczko 2020b on the difference between the
symmetry group and the homotopy group.
16
These patterns are nondualist since both major and minor chords descend; a dualist sequence
would move them in opposite directions (Tymoczko 2020b).
17
This piece is the subject of two extended analyses by David Lewin (1981, 1987, pp. 128ff) and a
third by Jack Boss (2019). Lewin’s second analysis prefigures his later interest in K-nets by analyzing
contrary motion using vertical inversional arrows. I prefer horizontal transpositional arrows instead.
18
I do not provide any notation for the pattern’s absolute transpositional level. Voice leadings
conforming to the schema are related by the Tinctoris Transform.
19
Ethan Haimo suggests that the basic objects of this music are more elastic than set classes,
neighborhoods containing set classes related by small “perturbations” (i.e., efficient voice leadings);
see Haimo 1996 and Tymoczko 2020b.
20
In other words, the augmented triad (B, E♭, G) allows for an ascending-step melody with both

Viennese chords that follow it, the (A, D, A♭) in m. 7 and its near-transposition down by minor sixth,

the distorted (D♭, F = G♭, C) in m. 8. In both cases, the ascending registral semitone is audible in the

music (G5–A♭5 and B4–C5). The idea of elided inversion is useful at the end of the piece, connecting
the final chord of sequence δ to the return of α.
21
Traditional transformational theory often focuses on disembodied pitch classes, assigned to
octave and instrument according to unformalized compositional whim (Lewin 1987).
22
Tymoczko 2011a, chapter 4.
 APPENDIX 4

Corpus Analysis, Statistics, and Grammar

Though it has one foot in the creative arts and the other in the sciences,
music theory has been somewhat slow to embrace empirical methods.
Figures such as David Huron and Fred Lerdahl have tried to push the field
in that direction, the former influenced by empirical psychology, the latter
by the introspective methods of Chomskian linguistics. But I think it is fair
to say that many theorists do not consider their profession to be even
broadly continuous with science. Those who do tend to follow Huron in
adopting the language and style of psychological discourse: formalized
theories, papers written as experiments, and null-hypothesis significance
testing, often in a context of skepticism toward traditional music-theoretical
verities. In this book I have instead chosen to adopt a more traditional style,
for three reasons.
The first, I admit, is constitutional: I believe I will have more fun and
reach more readers if I speak somewhat informally and leave out the
significance tests. After all, corpus analysis can be a powerful tool for
traditional scholarship—as it was for both Knud Jeppesen and Allen Irvine
McHose.1 Having programmed a computer to identify non-harmonic tones,
I discovered the Renaissance sevenths when they caused my algorithm to
fail, at which point I was able to theorize about them in a relatively
traditional way. Similarly, having programmed a computer to identify
cadences in Renaissance music (operationalized as two voices converging
stepwise on an octave or unison with one decorated by a suspension), I was
able to search hundreds of pieces for patterns like that at the bottom of
Figure 6.3.5. Here the computer pointed me toward examples whose
analytical significance was intuitively clear. In such cases, machines act as
virtual assistants, increasing our access to musical examples. Humans can
take advantage of these tools without adopting the language of statistical
science.
 A second reason is that large effects can often be seen without
sophisticated statistics. My Bach-chorale analyses contain 45 instances of
I–X–I6, where X is a single chord over bass scale-degree 2̂; 34 of these are
I–vii°6–I6 progressions, 10 are I–V43–I6, and only 1 is I–ii–I6. Meanwhile
the Mozart piano sonatas contain about 115 progressions of this form
without a single indisputable I–ii–I6. Such regularities speak for themselves.
Statistics are necessary when we are dealing with subtle effects not obvious
to the naked eye—a new drug that slightly decreases the length of an
illness, an early-childhood intervention that slightly improves test scores a
decade later, and so on.2 The musical behaviors I care about tend to be
considerably more obvious than this. And even when we are dealing with
subtle topics, such as the gradual increase in root-position fifth-progressions
over the course of the sixteenth century, we can often observe the relevant
phenomena analytically, once we have been alerted to them by corpus study
(§6.5).
The final issue is that there are significant methodological questions that
need to be resolved before we can profit from rigorous statistical
methodology. Consider David Huron’s treatment of the ii–vii°6 idiom,
discussed in connection with Figure P7.1. Huron considers two possible
analyses of the passage in Figure A4.1, neither of which I endorse; the best
analysis, in my view, recognizes two eighth-note chords rather than one
quarter-note harmony.3 Huron’s analysis of this idiom underwrites his claim
that more than 20% of Bach’s supertonic chords progress to tonics—a
number that exceeds my own by a factor of at least 10.4 Or consider the
Tempest’s m. 11 G–B ♭ –C ♯ –E chord, which Hepokoski and Martin call a
“ct°7/iv” and which Hatten and I consider a vii°7/V (m. 11, shown in Figure
10.3.4). These cases show us that simple Roman-numeral labeling,
sometimes disparaged as “trivial transliteration of note content,” requires a
nontrivial sense of a composer’s vocabulary. The paradox of analysis is that
we have to know what we’re looking for in order to find it.
Figure A4.1. Three analyses of the sixth measure of Bach’s “Liebster Jesu, wir sind hier” (BWV 373,
Riemenschneider 328). The first two are considered by David Huron; the third is my preferred
analysis.

For these reasons, I have decided to follow the philosophy of open data—
writing in a relatively informal mode but releasing all my code and data so
that readers can reproduce and criticize my results.5 Those who worry about
my subjective choices can reanalyze my data, for instance automatically
labeling all the “nonharmonic consonances” in the Bach chorales, or
removing any modulations I have introduced into Renaissance pieces.
Readers will no doubt discover errors in both my analyses and my code.
But a policy of openness allows others not only to correct those mistakes
but also to use my data for work I have not anticipated. My hope is that this
combination of humanistic writing and data transparency will satisfy both
nonspecialists who are primarily interested in high-level conclusions and
quantitative researchers who are more interested in the fine details.
This data serves as the input to musical “grammars,” or theories of what
does or does not typically happen in a particular musical genre. In this
book, I have simultaneously tried to bolster traditional theories while
identifying their limits. The first effort involves, for example, the
suggestion that traditional theories of nonharmonicity work reasonably well
in Renaissance, baroque, and classical music, allowing us to identify
consonant harmonies decorated by familiar nonharmonic formulae (§5.1–
§5.4); the proposal that the descending-third arrangement of triads, from I
down to V, captures the most common progressions in functional music
(§7.1–§7.3); and the claim that sequential progressions are harmonically
significant (§7.6). Such proposals are meant to counter skeptics who want
to replace traditional theory with something radically different. The second
part of the project acknowledges the schema-theoretical point that music is
fundamentally idiomatic, reconceiving traditional precepts not as
exceptionless laws but as largely true generalizations. We cannot expect that
every passage in every piece will conform to those generalizations, nor
even that there will be any explanation for anomalous occurrences.
Sometimes exceptions happen.
It is a truism of linguistics that grammaticality is not the same as
frequency: that there can be perfectly grammatical sentences that have never
been uttered, and are unlikely ever to be: Noam Chomsky famously gave
the example of the rare-but-grammatical “colorless green ideas sleep
furiously.” Is it wrong to link “grammaticality” with frequency in the
musical case?
My first answer is that Chomsky’s example was meant to highlight the
rarity of particular sequences of words rather than sequences of abstract
grammatical categories. The extremely rare sentence “colorless green ideas
sleep furiously” corresponds to the very common sequence <ADJECTIVE>
<ADJECTIVE> <NOUN> <VERB> <ADVERB>. If we translate this
point to the musical domain, then the analogues to word-sequences are
something like actual segments of musical notation, with Roman numeral
sequences like V–IV being analogous to abstract sequences like <NOUN>
<ADJECTIVE>. And at the level of grammatical categories, the question
about frequency and grammaticality is by no means resolved. There are
some sequences, like <NOUN> <NOUN> <NOUN> <VERB> <VERB>
<VERB>, that are almost unknown in English (e.g., “people cats dogs like
hate sigh,” a description of the sadness expressed by people who are hated
by those cats that are liked by dogs). Some linguists do take the rarity of
this abstract sequence to indicate that it is indeed ungrammatical, while
others consider the form to be grammatical in an idealized sense.
Second, the linguistic divergence between frequency and grammaticality
is underwritten by semantics: though we might initially be baffled by
“people cats dogs like hate sigh,” we can upon reflection extract its
meaning. (Even with “colorless green ideas sleep furiously” we can answer
questions like: “were the ideas awake?” or “how did they sleep?”6) By
contrast, it is much harder to extract any meaning from “like hate sigh
people cats dogs.” Lacking semantics, it is unclear whether music supports
any analogous distinction. If asked “is root-position I–V–IV–I a
grammatical chord progression?” I can say only that it is common in rock
and rare in classical music; in that sense, it is “grammatical in,” or
“characteristic of,” the one style but not the other. At the same time, were I
to hear a I–V–IV–I progression in classical music, I would register it as
unusual without experiencing any further difficulties.7 In particular, I have
no experience of being able to “comprehend” I–V–IV–I in rock but not
classical music; instead, I have the expectation that the progression occurs
in some styles but not others. Violating these expectations leads to mild
surprise rather than genuine perplexity.
It might be possible to separate frequency and acceptability were
classical music a living tradition. For suppose two Viennese classical-era
composers disagreed about whether the I–ii–I6 progression was
grammatical in their shared language: if one were to compose successful
pieces making use of the progression, and if these pieces were accepted by
the community as exemplars of the classical style (rather than archaic
evocations of earlier music), then this would provide evidence that the
progressions were grammatical despite their infrequency. It is not clear that
composition can still function in this way, as contemporary ears are no
longer capable of delivering eighteenth-century judgments. For suppose that
a charismatic music theorist constructed a theory implying that I–ii–I6 was
acceptable (say, because the supertonic was “merely passing”); and suppose
that theorist’s followers began composing mock-classical music featuring
extensive I–ii–I6 progressions.8 It seems to me that this would tell us less
about the classical style than about the contemporary theoretical
community. The question is not what we think about classical harmony but
what its native speakers thought.9
For all these reasons, I think it is best to understand musical grammars as
interpretive summaries of statistical tendencies. That is, they aspire to
provide a useful description of musical behavior as being “standard” or
“unusual” to varying degrees. (While the simplest grammars might use just
two categories, we have seen that it is often useful to adopt a more fine-
grained picture in which standardness comes in degrees.10) This sorting
serves a variety of practical functions. For the analyst, it directs attention
toward those features of a piece that are genuinely exceptional, preventing
us from wasting time “explaining” common occurrences. For the student,
trying to compose in the style, it can provide suggestions about what might
come next. For the theorist, it can point toward hypotheses about why
musical practice might have developed as it did (e.g., that mature
functionality fleshes out the simpler I–IV–V template with a series of
descending-third progressions [§7.2]). For the cognitive scientist, it can hint
at the mental representations that constitute competence in a musical style.
My word “interpretive” is meant to serve a double function. First, it
registers the reality that we can construct multiple theories accounting for
the same set of observable facts. For instance, I have proposed that the
progression I–ii–I6 is ungrammatical or unusual in mature functional music;
I explain most of the exceptions as instances of the pseudochord idiom
(§7.5).11 Others may assert that I–ii–I6 is grammatical, proposing auxiliary
explanations that restrict its appearance to the “pseudochord” idiom. This
kind of pluralism, or “underdetermination,” is endemic to science, where
equivalent theories often account for the same observations.
The word “interpretive” also points to the holistic nature of the
grammatical enterprise. Consider once again the interdependence of
harmonic labeling, nonharmonic-tone identification, and harmonic
theorizing discussed in the prelude to chapter 7. This holism gives us yet
another reason to be cautious about the prospects for a simple approach to
corpus study—and in particular about the possibility of theory-neutral
datasets of analyzed music. For suppose it is true that the identification of
nonharmonic tones needs to be carried out on the basis of expectations
about what is common in a particular repertoire; if so, then one of the basic
canons of scientific method, the separation of data collection and hypothesis
formation, may be inappropriate. If analysis is theory-dependent, then it is
circular—though not necessarily viciously so.
Music theory is a difficult discipline, requiring equal comfort with the
two cultures of art and science, and on occasion the third culture of
philosophical reasoning as well. It is also an activity of diminishing cultural
prestige, once fit to inhabit the Quadrivium, now demoted to the study of a
relatively marginal aspect of human behavior—extremely marginal if we
concern ourselves with notated composition. If the new era of
computational corpus study exacerbates the first problem, requiring
theorists to master yet another complex skill, it ameliorates the second,
putting music theory in dialogue with computer science, information theory,
linguistics, 3D visualization, and cognitive science. To me this is yet
another reason to be encouraged about its future.

1
Working before computers, both scholars had to gather data separately for each music-
theoretical question—an earlier study of parallel fifths would not facilitate a later study of root
motions, for example. One of the main virtues of computers is that they allow us to ask multiple
questions of the same data.
2
There is a “replication crisis” currently unfolding across the social and biomedical sciences,
sometimes described as an epidemic of untrustworthy and implausible results. See Simmons, Nelson,
and Simonsohn 2011. These difficulties provide another reason to focus on larger effects.
3
Huron 2007b.
4
Huron 2007a, p. 266. The other issue is Huron’s decision to consider I64 a standard tonic rather
than an anomalous cadential sonority, hierarchically dependent on the following dominant (§9.5). He
and I once had a chance to examine his data, determining that a large proportion of his ii–I
progressions belonged to these two categories.
5
Readers can go to https://www.madmusicalscience.com/taom to download my code, data, and a
python program allowing them to reproduce my quantitative examples, as well as the other programs
described in the text.
6
See Camp 2004. The issue is complicated, as there are semantically defective sentences that are
judged grammatical (e.g., “more people have been to Russia than I have,” Wellwood et al. 2018).
7
For example, there are retrofunctional moments in Brahms (chapter 7, footnote 29) and Bach:
“Nun freut euch, lieben Christen g’mein” (BWV 388, Riemenschneider 183), mm. 3–4; “Komm,
Gott Schöpfer, Heiliger Geist” (BWV 370, Riemenschneider 187), m. 1; and “So gibst du nun, mein
Jesu, gute Nacht” (BWV 412, Riemenschneider 206), mm. 26–27.
8
Indeed, one periodically finds this progression in theory texts, despite its extreme rarity in actual
music.
9
My suggestion here is vaguely reminiscent of Wittgenstein’s “Private Language Argument”
(Wittgenstein 1953), which argues that linguistic meaning crucially requires the presence of a
community. My argument is that judgments of grammaticality may require the presence of an
appropriate community, at least insofar as they depart from judgments about frequency.
10
Thus the V2–I progression in Figure 1.5.1 can be simultaneously unusual relative to a coarse-
grained norm, while still occurring frequently enough to be characteristic of Bach’s style. For a very
different understanding of “musical grammar” see Zbikowski 2017.
11
For me, the almost-invariable absence of the ii chord’s fifth is an important datum, suggesting a
linear configuration rather than a true harmony.
 Terms, Symbols, and Abbreviations

(C, E, G) → (C, F, A). A voice leading from (C, E, G) to (C, F, A) in which

the first note in the first chord moves to the first note in the second chord,
the second note in the first chord moves to the second note in the second
chord, and so on. Here, C stays fixed, E moves up to F, and G moves up to
A. These intervals are paths in pitch-class space that can occur in any
octave (appendix 1).

C4, C ♯ 5, and so on. A method of referring to pitches, combining a pitch

class (C) with an octave number. Octave numbers run from written C to the
B above, with octave 4 starting at middle C. Thus B3 is the note below
middle C, while B4 is a major seventh above.

60, 61, 62 . . . scale degrees are typically numbered so that 60 is as close to

middle C as possible. The exact numbering is unimportant; what matters is
the difference between scale degrees.

A3D2 (etc.). A sequence in which roots alternately ascend by diatonic third

(A3) and descend by diatonic second (D2). Octaves typically do not matter,
with ascending fourth (A4) equal to descending fifth (D5).

Abstract voice. A scale degree or “melodic slot” available in every octave

and moving in specific ways (§1.2, appendix 1).

Alphabet (musical). A chordal, scalar, or other collection along which

surface voices can move.

Basic voice leading. A voice-leading schema that, when repeated, generates

all the voice leadings on a spiral diagram; it is found only when the size of
the chord is relatively prime to the size of the scale (§3.1, §3.4, appendices
1–2).

Bijective voice leading. A voice leading in which each note of one chord is
connected to exactly one note of the other (appendix 1).

Chase schema. A three-voice canonic schema dating back to the

Renaissance (§4.4).

Chord. A collection of either pitches or pitch classes, depending on context.

Chord-loop family. A collection of repeating progressions that can be

derived from a single repeating four-chord progression by eliminating
chords; the underlying four-chord progression typically moves clockwise
through the space of chromatic major triads, selecting one chord at each
angular position (§2.3).

Chord type. A collection of chords related by transposition.

Collectional hierarchy. A more general name for the quadruple hierarchy,

useful for situations in which there are different numbers of layers.

Concrete voice. See surface voice.

Disguised model. A sequence whose initial unit is altered so as to conceal

the sequential structure.

Extrinsic scale. A scale in the standard sense, containing smaller chords;

contrasts with intrinsic scale.

Fauxbourdon ROTO (or Fauxbourdon Rule of the Octave). A schema for

harmonizing descending scalar bass lines (§7.5).

Functional analysis. A style of Roman-numeral analysis that tries to record

the meaning, origin, or function of chords, as opposed to the more neutral
scale-degree analysis (prelude to chapter 6).

Functional tonality. A generic term for what is sometimes called “tonal” as

opposed to “modal” harmony, or the harmony of the “common-practice
period,” making frequent use of progressions such as I–IV–I, I–IV–V–I, I–
V–I. The term is a very general label that does not favor the views of any
particular theorist (§7.1–§7.3, chapters 6–7).

Garden-path modulation. A modulation that juxtaposes two passages, each

syntactic in its own key, so as to create the impression of a nonfunctional
progression across the key boundary (prelude to chapter 7).

Gesualdo’s trick. A sequence of chords sharing the same intrinsic spacing in

which two voices move in parallel, while the remainder move so that each
chord is the pitch-class inversion of its predecessor (§1.1, prelude to chapter
2, appendix 3).

Harmonic skeleton. What remains when nonharmonic notes are removed.

Helicopter drop. A modulatory schema in which a sudden modulation to a

distant key is followed by a long sequence returning to the tonic (§8.4).

Ix, Ix, ix, ιx. These symbols refer to x-step inversion (pitch-space reflection)
along various collections: the chromatic scale (Ix), a scale such as the
diatonic (Ix), a chord (ix), and a chordal subset (ι x) (appendix 1).

Index number. A tool used in calculating the effects of inversion (pitch-

space reflection). If we number the notes in a collection, inversion along
that collection sends every note x to note c – x for some fixed c known as
the index number (appendix 1).

Intrinsic scale. The octave-repeating scale formed by a chord’s own notes

(prelude to chapter 2, appendix 1).

Intrinsic spacing. See spacing in chordal steps.

Inversion. An ambiguous term referring to two unrelated phenomena:

registral inversion, the rearranging of notes into different octaves (prelude
to chapter 2); and pitch-space reflection, which turns a collection of
intervals upside down (appendix 1).
Inversional equivalence. The view that chordal identity is determined by
pitch-class content. Here, C3–G3–E4 is the same chord as E3–G3–C4 as
they both contain the pitch classes C, E, and G. This view is often
associated with Rameau, though it was anticipated by earlier theorists such
as Lippius.

Irreducible seventh. See Renaissance seventh.

Loop family. See chord-loop family.

The Ludwig. An outer-voice contrary-motion schema that exploits the

intrinsic geometry of diatonic space (§10.1–§10.2, §4.7, and §3.2).

Melodic slot. A scale degree or abstract voice.

Modality. Largely diatonic music in which each diatonic collection can

have a wide range of tonal centers; typically chords progress freely.

Morte. A contrary-motion schema moving from tonic to dominant (Figure

4.7.4).

Neo-Riemannian progression. A spacing-preserving voice leading between

inversionally related chords; it can be decomposed into a pair of inversions
along intrinsic and extrinsic scales (prelude to chapter 2, appendix 3). Any
such progression will preserve the distance between at least two voices, and
can be used in Gesualdo’s trick.

Nonharmonic. A tone that is not part of the underlying harmony, according

to standard contrapuntal theory. It is a misleading term insofar as
nonharmonic tones (in the technical sense) can produce harmonic effects (in
the nontechnical sense). Yet it is accurate insofar as it captures the thinking
of many theorists and composers.

OUCH theory. A configuration-based strategy for writing four-voice

counterpoint (§3.6).

Pachelbel progression. A descending-fourth, ascending-step sequence of

root progressions (D4A2) allowing two voices to descend in parallel. It can
be analyzed as an instance of Gesualdo’s trick (§1.1).

Paired-sequence design. A pair of sequences of comparable size, one of

which undoes the other (§8.4).

Pairwise voice exchange. A voice exchange in which two voices move

along equal and opposite paths to exchange pitch classes, as in (C4, E5) →
(E4, C5); here, the first voice moves up by major third from C4 to E4 and
the second moves down by major third from E5 to C5. More complicated
voice exchanges can be expressed as the product of pairwise voice
exchanges.

Paratonal century. The period from 1550–1650 when modal and

functionally tonal techniques mix freely. Similar to our own time in some
respects.

Path in pitch-class space. A combination of an octave-free pitch class with a

distance and direction in pitch space, for example, “C, in any octave, moves
down by eight semitones” (appendix 1).

Pitch class. An octave-free note type such as C, C♯, or D. Spelling does not
matter and the note can occur in any octave.

Prime Directive. “Whenever you find an interesting musical technique, try

to generalize it to every possible chordal and scalar environment” (§1.1).

Principle of Musical Approximation. The technique of viewing

nonsymmetrical chords as perturbations of, or approximations to,
symmetrical chords (§2.1, Figure A3.8).

Prinner. A common schema in which 6̂–5̂–4̂–3̂ and 4̂–3̂–2̂–1̂, in two

different voices, are harmonized by IV–I–vii°–I. This schema plays
different roles in different styles, serving as a cadential progression in the
Renaissance and a “riposte” in the Galant. I use the term to refer to the
general progression rather than its style-specific manifestation.
Protofunctionality. A set of ionian-mode harmonic routines first appearing
around 1500 and emphasizing root-position I, IV, and V chords (§6.1).

Quadruple hierarchy. A hierarchical arrangement of nested collections, or

alphabets, in which surface voices can move along collections of various
types (§1.2).

Renaissance seventh. A sounding seventh chord in which the seventh acts

like a suspension but which cannot be reduced to a triad according to
standard contrapuntal theory (§5.1).

Retrofunctional. Harmonic techniques that reverse the descending-fifth

norm of functional tonality, often motivated by voice leading (chapter 2,
§3.8).

Rogue note. A nonharmonic tone that is leapt-to and leapt-away-from; rare

from 1500 to 1900.

Roman-numeral analysis. A very general style of musical analysis that

labels tertian harmonies relative to a tonal center (prelude to chapter 6).

Scale. An abstract musical object whose notes can appear in any octave; it
can be modeled as a circular ordering of pitch classes.

Scale degree. A way of identifying the abstract voices in a scale. A scale’s

pitch classes are typically labeled relative to the tonic, which is scale degree
1 (e.g., Figure 8.1.3). Its specific pitches are typically numbered so that note
60 is as close to middle C as possible.

Scale-degree analysis. A style of Roman-numeral analysis that aspires to a

relatively theory-neutral labeling of chords’ scale-degree content, as
opposed to functional analysis (prelude to chapter 6).

Sequential reduction. A style of analysis that traces a quasi-regular surface

back to a regular sequential background (§4.10).

Set class. A collection of chords related by transposition or inversion. On its

own, set class usually refers to collections of pitch classes while pitch-set
set class refers to collections of pitches. The term transpositional set class
refers to collections related by transposition but not inversion.

Shepard-tone passacaglia. A repeating passage articulating a nontrivial

voice leading across the repeat, producing the effect of continual descent or
ascent (§2.5). I believe the phenomenon was first described by David
Feurzeig.

SNAP system. A system of nonharmonic tones comprising Suspensions,

Neighbor tones, Anticipations, and Passing tones (§5.1).

Spacing in chordal steps. The intervals between a collection of pitches as

measured along the intrinsic scale; a transpositional pitch-set set-class
measured intrinsically (prelude to chapter 3).

Spiral diagram. The main theoretical models used in this book, representing
the geometry of a pair of hierarchically nested transpositions (prelude to
chapter 2, prelude to chapter 3, §2.1, appendices 1–2).

Structured arpeggiation. A melodic sequence in which a voice moves

regularly along a shifting chordal background (§4.8, prelude to chapter 5).

Surface voice. A surface melodic stream containing notes that are felt to
connect.

Tx, Tx, tx, τ x. These symbols refer to x-step transposition along various
collections: the chromatic scale (Tx), another scale such as the diatonic (Tx),
a chord (tx), and a chordal subset (τ x) (prelude to chapter 2).

Thinking within the chord. Treating chords as scale-like collections along

which surface voices move (§1.2, prelude to chapter 2, §3.7).

Tinctoris Transform. The independent transposition of each chord in a voice

leading (prelude to chapter 3). Two progressions are related by the Tinctoris
Transform if they exhibit the same sequence of vertical configurations.

Tonal. An ambiguous term that can refer to music that is functionally tonal
(see above), or non-atonal. In addition the adjective sometimes refers to
keys and modes (“tonal center,” “tonal region,” “tonal plan”). I typically
use the term as a synonym for “non-atonal,” encompassing techniques
common to a range of modal and functionally tonal styles.

Up-and-down-the-ladder. A paired-sequence design in which a tension-

increasing series of modulations up the circle of fifths is undone by a
tension-releasing series of modulations down the circle of fifths (§8.5).

Voice. An ambiguous term that can either refer to abstract or concrete

voices.

Voice exchange. A voice leading from a chord to itself in which the sum of
all the intervals moved by all voices equals zero.

Voicing. A way of arranging a chord in register, which can be formalized as

a pattern of spacing in chordal steps.

Voice leading. A collection of paths in pitch class space that indicate how
the notes of one chord move to those of another; see (C, E, G) → (C, F, A)
(appendix 1).

Voice-leading schema. A collection of voice leadings related by

transposition, such as (C, E, G) → (C, F, A) and (G, B, D) → (G, C, E). It
can be modeled as a set of instructions for moving chordal elements: “hold
the root fixed, move the third up by step, and move the fifth up by step.”
 Bibliography

Aarden, Bret, and Paul T. von Hippel. 2004. “Rules for Chord Doubling (and Spacing): Which Ones
Do We Need?” Music Theory Online 10(2).
Agmon, Eytan. 1989. “A Mathematical Model of the Diatonic System.” Journal of Music Theory
33(1): 1–25.
Agmon, Eytan. 1995. “Functional Harmony Revisited: A Prototype-Theoretic Approach.” Music
Theory Spectrum 17(2): 196–214.
Agmon, Eytan. 2013. The Languages of Western Tonality. New York: Springer.
Agmon, Eytan. 2020. “The Webern in Mozart: Systems of Chromatic Harmony and Their Twelve-
Tone Content.” Music Theory Spectrum 42(2): 173–92.
Aldwell, Edward, and Carl Schachter. (1969) 2002. Harmony and Voice Leading. 3rd ed. Belmont:
Wadsworth.
Allsop, Peter. 1992. The Italian “Trio” Sonata: From Its Origins until Corelli. New York: Oxford
University Press.
Arndt, Matthew. 2011. “Schenker and Schoenberg on the Will of the Tone.” Journal of Music Theory
55(1): 89–146.
Arndt, Matthew. 2013. “Schenker and Schoenberg on the Eye of the Genius.” Theoria 20: 39–120.
Arndt, Matthew. 2018. The Musical Thought and Spiritual Lives of Heinrich Schenker and Arnold
Schoenberg. New York: Routledge.
Arndt, Matthew. 2019. “Schoenberg—Schenker—Bach: A Harmonic, Contrapuntal, Formal Braid.”
Zeitschrift der Gesellschaft für Musiktheorie 16(1): 67–97.
Arthur, Claire. 2021. “Vicentino versus Palestrina: A Computational Investigation of Voice Leading
across Changing Vocal Densities.” Journal of New Music Research 50(1): 74–101.
Auerbach, Brent. 2021. Musical Motives: A Theory and Method for Analyzing Shape in Music. New
York: Oxford University Press.
Babbitt, Milton. 1987. Words about Music. Edited by Stephen Dembski and Joseph N. Straus.
Madison: University of Wisconsin Press.
Bach, Carl Philipp Emanuel. (1762) 1949. Essay on the True Art of Playing Keyboard Instruments.
Edited and translated by William Mitchell. New York: W. W. Norton.
BaileyShea, Matthew. 2004. “Beyond the Beethoven Model: Sentence Types and Limits.” Current
Musicology 77: 5–33.
Balzano, Gerald. 1980. “The Group-Theoretic Description of 12-Fold and Microtonal Systems.”
Computer Music Journal 4(4): 66–84.
Barry, Barbara. 2014. “Schubert’s ‘Quartettsatz’: A Case Study in Confrontation.” Musical Times
155(1928): 31–49.
Bass, Richard. 1996. “From Gretchen to Tristan: The Changing Role of Harmonic Sequences in the
Nineteenth Century.” 19th-Century Music 19: 263–68.
Beach, David. 1967. “The Functions of the Six-Four Chord in Tonal Music.” Journal of Music
Theory 11(1): 2–31.
Beach, David. 1974. “The Origins of Harmonic Analysis.” Journal of Music Theory 18(2): 274–306.
Beach, David. 1983. “A Recurring Pattern in Mozart’s Music.” Journal of Music Theory 27: 1–29.
Beach, David. 1994. “Harmony and Linear Progression in Schubert’s Music.” Journal of Music
Theory 38(1): 13–17.
Bent, Margaret. 1998. “The Grammar of Early Music: Preconditions for Analysis.” In Tonal
Structures in Early Music, edited by Cristle Collins Judd, 15–59. New York: Garland.
Bergé, Pieter, and Jeroen D’hoe. 2009. “To Play or Not to Play: Motivic Connections in the First
Movement of Beethoven’s Tempest Sonata.” In Beethoven’s Tempest Sonata: Perspectives of
Analysis and Performance, edited by Peter Berge, Jeroen D’hoe, and William Caplin, 9–38.
Walpole: Peters.
Bernhard, Christopher. (~1640) 1973. “The Treatises of Christoph Bernhard.” Translated by Walter
Hilse. The Music Forum 3: 1–197.
Bernstein, Leonard. 1956. The Joy of Music. New York: Simon and Schuster.
Bernstein, Leonard. 1976. The Unanswered Question. Cambridge, MA: Harvard University Press.
Besseler, Heinrich. 1950. Bourdon und Fauxbourdon. Leipzig: Breitkopf und Härtel.
Biamonte, Nicole. 2006. “Modality in Beethoven’s Folk-Song Settings.” Beethoven Forum 13(1):
28–63.
Biamonte, Nicole. 2010. “Triadic Modal and Pentatonic Patterns in Rock Music.” Music Theory
Spectrum 32(2): 95–110.
Biamonte, Nicole. 2012. “Variations on a Scheme: Bach’s ‘Crucifixus’ and Chopin’s and Scriabin’s
E-minor Preludes.” Intégral 26: 47–89.
Bobbitt, Richard. 1955. “Harmonic Tendencies in the Missa Papae Marcelli.” The Music Review 14:
273–88.
Boone, Graeme. 2000. “Marking Mensural Time.” Music Theory Spectrum 22(1): 1–43.
Boretz, Benjamin. 1970. “Sketch of a Musical System (Meta-Variations, Part II).” Perspectives of
New Music 8(2): 49–111.
Boretz, Benjamin. 1972. “Meta-Variations, Part IV: Analytic Fallout (1).” Perspectives of New Music
11(1): 146–223.
Boss, Jack. 2019. Schoenberg’s Atonal Music: Musical Idea, Basic Image, and Specters of Tonal
Function. Cambridge: Cambridge University Press.
Brewer, William F. 2015. “Perception Is Theory Laden: The Naturalized and Philosophical
Implications.” Journal for General Philosophy of Science 46: 121–38.
Brinkman, Alexander R. 1986. “A Binomial Representation of Pitch for Computer Processing of
Musical Data.” Music Theory Spectrum 8: 44–57.
Brody, Christopher. 2016. “Parametric Interaction in Tonal Repertoires.” Journal of Music Theory
60(2): 97–148.
Brody, Christopher. 2021. “Second-Reprise Opening Schemas in Bach’s Binary Movements.” Music
Theory Spectrum 43(2): 257–79.
Brown, Matthew. 2005. Explaining Tonality. Rochester, NY: University of Rochester Press.
Bruce, Robert. 1969. “The Lyrical Element in Schubert’s Instrumental Forms.” Music Review 30(2):
131–37.
Buelow, George. 1992. Thorough-Bass Accompaniment According to Johann David Heinichen.
Lincoln: University of Nebraska Press.
Bukofzer, Manfred. 1947. Music in the Baroque Era: From Monteverdi to Bach. New York: W. W.
Norton.
Burke, Edmund. (1757) 1999. A Philosophical Enquiry into the Origins of the Sublime and Beautiful
and Other Pre-Revolutionary Writings. London: Penguin.
Burnham, Scott. 1995. Beethoven Hero. Princeton: Princeton University Press.
Burnham, Scott. 2009. “Singularities and Extremes: Dramatic Impulse in the First Movement of
Beethoven’s Tempest Sonata.” In Beethoven’s Tempest Sonata: Perspectives of Analysis and
Performance, edited by Peter Berge, Jeroen D’hoe, and William Caplin, 39–60. Walpole: Peters.
Burns, Lori. 1993. “J. S. Bach’s Mixolydian Chorale Harmonizations.” Music Theory Spectrum
15(2): 144–72.
Burns, Lori. 1995. Bach’s Modal Chorales. Hillsdale: Pendragon Press.
Burstein, L. Poundie. 2015. “The Half Cadence and Related Analytic Fictions.” In What Is a
Cadence? Theoretical and Analytical Perspectives on Cadences in the Classical Repertoire, edited
by Markus Neuwirth and Pieter Bergé, 85–116. Leuven: Leuven University Press.
Burstein, L. Poundie. 2018. “The Practice of Music Theory, and Music Theory versus Practice.” In
Teaching Music Theory, edited by Rachel Lumsden and Jeffrey Swinkin, 1–12. New York: W. W.
Norton.
Burstein, L. Poundie. 2020. “Those Boring, Arcane Part-Writing Exercises.” Gamut: Online Journal
of the Music Theory Society of the Mid-Atlantic 9(1).
Burstein, L. Poundie. forthcoming. “Analysis nach Schenker.” In Handbuch Musikanalyse. Methode
und Pluralität, edited by Ariane Jeßulat, Oliver Schwab-Felisch, Jan Philipp Sprick, and Christian
Thorau, translated by Oliver Schwab-Felisch. Kassel: Bärenreiter.
Byros, Vasili. 2012. “Meyer’s Anvil: Revisiting the Schema Concept.” Music Analysis 31(3): 273–
346.
Byros, Vasili. 2015. “Hauptruhepuncte des geistes: Punctuation Schemas and the Late-Eighteenth-
Century Sonata.” In What Is a Cadence? Theoretical and Analytical Perspectives on Cadences in
the Classical Repertoire, edited by Markus Neuwirth and Pieter Bergé, 215–51. Leuven: Leuven
University Press.
Byros, Vasili. 2017. “Mozart’s Vintage Corelli: The Microstory of a Fonte-Romanesca.” Intégral 31:
63–89.
Callender, Clifton. 1998. “Voice-Leading Parsimony in the Music of Alexander Scriabin.” Journal of
Music Theory 42(2): 219–33.
Callender, Clifton, Ian Quinn, and Dmitri Tymoczko. 2008. “Generalized Voice Leading Spaces.”
Science 320: 346–48.
Camp, Elisabeth. 2004. “The Generality Constraint, Nonsense, and Categorial Restrictions.”
Philosophical Quarterly 54(215): 209–31.
Camp, Elisabeth. 2017. “Perspectives in Imaginative Engagement with Fiction.” Philosophical
Perspectives 31: 74–102.
Camp, Elisabeth. 2018. “Why Maps Are Not Propositional.” In Non-Propositional Intentionality,
edited by A. Grzankowski and M. Montague, 19–45. New York: Oxford University Press.
Caplin, William E. 1998. Classical Form: A Theory of Formal Functions for the Instrumental Music
of Haydn, Mozart, and Beethoven. New York: Oxford University Press.
Caplin, William E. 2004. “The Classical Cadence: Conceptions and Misconceptions.” Journal of the
American Musicological Society 57(1): 51–118.
Caplin, William E. 2009a. “Beethoven’s Tempest Exposition: A Springboard for Form-Functional
Considerations.” In Beethoven’s Tempest Sonata: Perspectives of Analysis and Performance,
edited by Peter Berge, Jeroen D’hoe, and William Caplin, 87–125. Walpole: Peters.
Caplin, William E. 2009b. “What Are Formal Functions?” In Musical Form, Forms & Formenlehre:
Three Methodological Reflections, edited by Pieter Bergé, 71–89. Leuven: Leuven University
Press.
Caplin, William E. 2013. “Teaching Classical Form: Strict Categories versus Flexible Analyses.”
Tijdschrift voor Muziektheorie 18(3): 119–35.
Caplin, William E. 2015. “Harmony and Cadence in Gjerdingen’s ‘Prinner.’” In What Is a Cadence?
Theoretical and Analytical Perspectives on Cadences in the Classical Repertoire, edited by
Markus Neuwirth and Pieter Bergé, 17–58. Leuven: Leuven University Press.
Cavett-Dunsby, Esther. 1988. “Mozart’s ‘Haydn’ Quartets: Composing Up and Down without Rules.”
Journal of the Royal Music Association 113(1): 57–80.
Ceulemans, Anne-Emmanuelle. 2017. “Lasso, Meier, Powers. The Reality of the Modes under
Scrutiny.” Paper presented to the 9th European Music Analysis Conference, Strasbourg.
Chafe, Eric. 1992. Monteverdi’s Tonal Language. New York: Schirmer.
Christensen, Thomas. 1992. “The ‘Règle de l’Octave’ in Thorough-Bass Theory and Practice.” Acta
Musicologica 64(2): 91–117.
Christensen, Thomas. 1993a. Rameau and Musical Thought in the Enlightenment. Cambridge:
Cambridge University Press.
Christensen, Thomas. 1993b. Review of Carl Dahlhaus’s Studies on the Origins of Harmonic
Tonality and Joel Lester’s Between Mode and Keys. Music Theory Spectrum 15(1): 94–111.
Christensen, Thomas. 2019. Stories of Tonality in the Age of François-Joseph Fétis. Chicago:
University of Chicago Press.
Chua, Daniel K. L., and Alexander Rehding. 2021. Alien Listening: Voyager's Golden Record and
Music from Earth. New York: Zone Books.
Clendinning, Jane Piper, and Elizabeth West Marvin. 2016. The Musician’s Guide to Theory and
Analysis. 3rd ed. New York: W. W. Norton.
Clough, John. 1979a. “Aspects of Diatonic Sets.” Journal of Music Theory 23: 45–61.
Clough, John. 1979b. “Diatonic Interval Sets and Transformational Structures.” Perspectives of New
Music 18(1–2): 461–82.
Clough, John. 1994. “Diatonic Interval Cycles and Hierarchical Structure.” Perspectives of New
Music 32(1): 228–53.
Clough, John. 2000. “Flip-Flop Circles and Their Groups.” Paper presented to the national meeting
of the Society for Music Theory, Toronto.
Cohn, Richard. 1992. “The Autonomy of Motives in Schenkerian Accounts of Tonal Music.” Music
Theory Spectrum 14(2): 150–70.
Cohn, Richard. 1996. “Maximally Smooth Cycles, Hexatonic Systems, and the Analysis of Late-
Romantic Triadic Progressions.” Music Analysis 15(1): 9–40.
Cohn, Richard. 1997. “Neo-Riemannian Operations, Parsimonious Trichords, and Their ‘Tonnetz’
Representations.” Journal of Music Theory 41(1): 1–66.
Cohn, Richard. 1999. “As Wonderful as Star Clusters: Instruments for Gazing at Tonality in
Schubert.” 19th-Century Music 22(3): 213–32.
Cohn, Richard. 2005. “‘This Music Crept by Me upon the Waters’: Introverted Motives in
Beethoven’s ‘Tempest’ Sonata.” In Engaging Music: Essays in Analysis, edited by Deborah Stein,
226–35. New York: Oxford University Press.
Cohn, Richard. 2012. Audacious Euphony. New York: Oxford University Press.
Collins, Denis. 2015. “Taneyev’s Theories of Moveable Counterpoint in the Music of J. S. Bach.”
Bach 46(2): 22–45.
Collins, Denis. 2018. “Approaching Renaissance Music Using Taneyev’s Theories of Movable
Counterpoint.” Acta Musicologica 90(2): 178–201.
Collins Judd, Cristle. 1992. “Modal Types and ‘Ut, Re, Mi’ Tonalities: Tonal Coherence in Sacred
Vocal Polyphony from about 1500.” Journal of the American Musicological Society 45(3): 428–
67.
Coluzzi, Seth J. 2015. “Speaking in (and out of) Mode: Structure and Rhetoric in Marenzio’s O
Mirtillo, Mirtillo anima mia (1595).” Music Theory Spectrum 37(2): 253–74.
Cone, Edward T. 1970. “Schubert’s Beethoven.” The Musical Quarterly 56(4): 779–93.
Cook, Nicholas. 1987. “The Perception of Large-Scale Tonal Closure.” Music Perception 5(2): 197–
205.
Cook, Nicholas. 1989. “Music Theory and ‘Good Comparison’: A Viennese Perspective.” Journal of
Music Theory 33(1): 117–41.
Crane, Hewitt D., and Thomas P. Piantanida. 1983. “On Seeing Reddish Green and Yellowish Blue.”
Science 221: 1078–80.
Cuthbert, Michael Scott, and Christopher Ariza. 2010. “Music21: A Toolkit for Computer-Aided
Musicology and Symbolic Music Data.” In Proceedings of the 11th International Society for Music
Information Retrieval (ISMIR 2010), August 9–13, 2010, Utrecht, Netherlands, edited by J.
Stephen Downie and Remco C. Veltkamp, 637–42.
Cutler, Timothy. 2009. “On Voice Exchanges.” Journal of Music Theory 53(2): 191–226.
Czerny, Carl. (1830) 1848. School of Practical Composition. Translated by John Bishop. London:
Robert Cocks & Co.
Dahlhaus, Carl. 1967. “Zur chromatischen Technik Carlo Gesualdos.” Analecta Musicologica 4: 77–
96.
Dahlhaus, Carl. 1990. Studies in the Origins of Harmonic Tonality. Translated by Robert Gjerdingen.
Princeton: Princeton University Press.
Dahlhaus, Carl. 1991. Ludwig van Beethoven: Approaches to His Music. Translated by Mary
Whittall. New York: Oxford University Press.
Damschroder, David. 2016. Harmony in Beethoven. Cambridge: Cambridge University Press.
Damschroder, David. 2018. Tonal Analysis: A Schenkerian Perspective. New York: W. W. Norton.
Danto, Arthur. 1981. The Transfiguration of the Commonplace: A Philosophy of Art. Cambridge,
MA: Harvard University Press.
Davies, Stephen. 1983. “Attributing Significance to Unobvious Musical Relationships.” Journal of
Music Theory 27(2): 203–13.
de Clerq, Trevor, David Temperley, David Lustig, and Ethan Tan. 2011. “A Corpus Study of Rock
Music.” http://rockcorpus.midside.com. Accessed September 26, 2020.
Deathridge, John. 2008. Wagner Beyond Good and Evil. Berkeley: University of California Press.
DeBellis, Mark. 1999. “The Paradox of Musical Analysis.” Journal of Music Theory 43(1): 83–99.
Deutsch, Diana, and John Feroe. 1981. “The Internal Representation of Pitch Sequences in Tonal
Music.” Psychological Review 88(6): 503–22.
Doll, Christopher. 2017. Hearing Harmony: Toward a Tonal Theory for the Rock Era. Ann Arbor:
University of Michigan Press.
Douthett, Jack, and Peter Steinbach. 1998. “Parsimonious Graphs: A Study in Parsimony, Contextual
Transformations, and Modes of Limited Transposition.” Journal of Music Theory 42(2): 241–63.
Duane, Ben. 2017. “Thematic and Non-Thematic Textures in Schubert’s Three-Key Expositions.”
Music Theory Spectrum 39(1): 36–65.
Edwards, George. 1997. “A Palimpsest of Mozart in Schubert’s Symphony No. 5.” Current
Musicology 62: 18–39.
Einstein, Alfred. 1948. The Italian Madrigal. Princeton: Princeton University Press.
Everett, Walter. 2004. “Making Sense of Rock’s Tonal Systems.” Music Theory Online 10(4).
Everett, Walter, and Tim Riley. 2019. What Goes On: The Beatles, Their Music, and Their Time. New
York: Oxford University Press.
Ewell, Philip A. 2020. “Music Theory and the White Racial Frame.” Music Theory Online 26(2).
Ferraguto, Mark. 2019. Beethoven 1806. New York: Oxford University Press.
Fétis, François-Joseph. (1840) 1994. Esquisse de l’histoire de l’harmony. Translated by Mary Arlin.
Hillsdale: Pendragon Press.
Feurzeig, David. 2010. “On Shifting Grounds: Meandering, Modulating, and Möbius Passacaglias.”
Paper presented to the national meeting of the Society for Music Theory, Indianapolis.
Fieldman, Hali. 2002. “Schubert’s Quartettsatz and Sonata Form’s New Way.” Journal of
Musicological Research 21: 99–146.
Fink, Robert. 2004. “Beethoven Antihero: Sex, Violence, and the Aesthetics of Failure, or Listening
to the Ninth Symphony as Postmodern Sublime.” In Postmodern Modes of Hearing, edited by
Andrew Dell’Antonio, 109–53. Berkeley: University of California Press.
Flew, Antony. 1975. Thinking about Thinking: Do I Sincerely Want to Be Right? London: Collins
Fontana.
Forte, Allen. 1959. “Schenker’s Conception of Musical Structure.” Journal of Music Theory 3(1): 1–
30.
Forte, Alan, and Stephen Gilbert. 1982. Introduction to Schenkerian Analysis. New York: W. W.
Norton.
Froebe, Folker. 2015. “On Synergies of Schema Theory and Theory of Levels: A Perspective from
Riepel’s Fonte and Monte.” Zeitschrift der Gesellschaft für Musiktheorie 12(1): 9–25.
Fux, Johann. (1725) 1971. The Study of Counterpoint. Edited and translated by Alfred Mann. New
York: W. W. Norton.
Gauldin, Robert. 1995a. A Practical Approach to Sixteenth-Century Counterpoint. Long Grove:
Waveland.
Gauldin, Robert. 1995b. A Practical Approach to Eighteenth-Century Counterpoint. Long Grove:
Waveland.
Gauldin, Robert. 1996. “The Composition of Late Renaissance Stretto Canons.” Theory and Practice
21: 29–54.
Gauldin, Robert. 1997. Harmonic Practice in Tonal Music. New York: W. W. Norton.
Gauldin, Robert. 2004. “The Theory and Practice of Chromatic Wedge Progressions in Romantic
Music.” Music Theory Spectrum 26(1): 1–22.
Gelman, Andrew, John Carlin, Hal Stern, David Dunson, Aki Vehtari, and Donald Rubin. 2013.
Bayesian Data Analysis. New York: Chapman and Hall.
Georgiades, Thrasybulos. 1937. Englische Diskantetheorie aus der ersten hälfte 15. Jahrhunderts.
Munich: Musikwissenschaftliches Seminar der Universität München.
Gjerdingen, Robert. 1988. A Classic Turn of Phrase: Music and the Psychology of Perception.
Philadelphia: University of Pennsylvania Press.
Gjerdingen, Robert. 1996. “Courtly Behaviors.” Music Perception 13(3): 365–82.
Gjerdingen, Robert. 1999. “An Experimental Music Theory?” In Rethinking Music, edited by
Nicholas Cook and Mark Everist, 161–70. New York: Oxford University Press.
Gjerdingen, Robert. 2007. Music in the Galant Style. New York: Oxford University Press.
Gjerdingen, Robert. 2015. “Musical Grammar.” In The Oxford Handbook of Critical Concepts in
Music Theory, edited by Alexander Rehding and Steven Rings, 651–72. New York: Oxford
University Press.
Gjerdingen, Robert. 2020. Child Composers in the Old Conservatories: How Orphans Became Elite
Musicians. New York: Oxford University Press.
Gjerdingen, Robert, and Janet Bourne. 2015. “Schema Theory as a Construction Grammar.” Music
Theory Online 21(2).
Goldenberg, Yosef. 2020. “Continuous Question-Answer Pairs.” Music Theory Online 26(3).
Goldenberg, Yosef. 2021. “Intriguing Interpretation of Dyads in Common-Practice Tonal Music.”
Gamut: Online Journal of the Music Theory Society of the Mid-Atlantic 10(1).
Goldstein, Robin, Johan Almenberg, Anna Dreber, John Emerson, Alexis Herschkowitsch, and Jacob
Katz. 2008. “Do More Expensive Wines Taste Better? Evidence from a Large Sample of Blind
Tastings.” Journal of Wine Economics 3(1): 1–9.
Gosman, Alan. 2009. “Canonic Patterns: Reframing the Puzzle.” Theory and Practice 34: 1–29.
Gosman, Alan. 2012. “Canonic Threads and Large-Scale Structure in Canons.” Gamut: Online
Journal of the Music Theory Society of the Mid-Atlantic 5(1).
Grant, Aaron. 2018. “Schubert’s Three-Key Expositions.” PhD diss., Eastman School of Music.
Greenberg, Yoel. 2017. “Of Beginnings and Ends: A Corpus-Based Inquiry into the Rise of the
Recapitulation.” Journal of Music Theory 61(2): 171–200.
Guillotel-Nothmann, Christophe. 2018. “Conditional Asymmetry and Spontaneous Asymmetry of
Harmonic Progressions: The Changing Status of the Directional Tendency of Root Progressions in
Madrigal Cycles from Verdelot to Monteverdi (c. 1530–1638).” Music Theory Online 24(4).
Hacking, Ian. 1975. The Emergence of Probability: A Philosophical Study of Early Ideas about
Probability, Induction and Statistical Inference. Cambridge: Cambridge University Press.
Hadjeres, Gaëtan, François Pachet, and Frank Nielsen. 2017. “DeepBach: A Steerable Model for
Bach Chorales Generation.” Proceedings of Machine Learning Research 70: 1362–71.
Haimo, Ethan. 1996. “Atonality, Analysis, and the Intentional Fallacy.” Music Theory Spectrum
18(2): 167–99.
Hall, Rachel, Dmitri Tymoczko, and Jason Yust. 2010. “Upright Petrouchka, Sideways Neapolitans,
and Proper Scales.” Paper delivered to the national meeting of the Society for Music Theory,
Indianapolis.
Hamilton, Kenneth. 2009. “Beethoven’s Tempest Sonata in Performance.” In Beethoven’s Tempest
Sonata: Perspectives of Analysis and Performance, edited by Peter Berge, Jeroen D’hoe, and
William Caplin, 127–62. Walpole: Peters.
Hansberry, Benjamin. 2017. “What Are Scale-Degree Qualia?” Music Theory Spectrum 39(2): 182–
99.
Harrison, Daniel. 1988. “Some Group Properties of Triple Counterpoint and Their Influence on
Compositions by J. S. Bach.” Journal of Music Theory 32(1): 23–49.
Harrison, Daniel. 1994. Harmonic Function in Tonal Music: A Renewed Dualist Theory and an
Account of Its Precedents. Chicago: University of Chicago Press.
Harrison, Daniel. 2003. “Rosalia, Aloysius, and Arcangelo: A Genealogy of the Sequence.” Journal
of Music Theory 47: 225–72.
Harrison, Daniel. 2016. Pieces of Tradition: An Analysis of Contemporary Tonality. New York:
Oxford University Press.
Hasty, Christopher. 1981. “Segmentation and Process in Post-Tonal Music.” Music Theory Spectrum
3: 54–73.
Hatten, Robert S. 2009. “Interpreting Beethoven’s Tempest Sonata through Topics, Gestures, and
Agency.” In Beethoven’s Tempest Sonata: Perspectives of Analysis and Performance, edited by
Peter Berge, Jeroen D’hoe, and William Caplin, 163–80. Walpole: Peters.
Hatten, Robert S. 2014. “Performing Expressive Closure in Structurally Open Contexts: Chopin’s
Prelude in A minor and the Last Two Dances of Schumann’s Davidsbündlertänze.” Music Theory
Online 20(4).
Headlam, David. 2012. Review of A Geometry of Music. Music Theory Spectrum 34(1): 123–43.
Hepokoski, James. 2009. “Approaching the First Movement of Beethoven’s Tempest Sonata through
Sonata Theory.” In Beethoven’s Tempest Sonata: Perspectives of Analysis and Performance, edited
by Peter Berge, Jeroen D’hoe, and William Caplin, 181–212. Walpole: Peters.
Hepokoski, James. 2010. “Sonata Theory and Dialogic Form.” In Musical Form, Forms &
Formenlehre: Three Methodological Reflections, edited by Pieter Bergé, 71–89. Leuven: Leuven
University Press.
Hepokoski, James. 2021. A Sonata Theory Handbook. New York: Oxford University Press.
Hepokoski, James, and Warren Darcy. 2006. Elements of Sonata Theory: Norms, Types, and
Deformations in the Late-Eighteenth-Century Sonata. New York: Oxford University Press.
Hinton, Stephen. 2010. “The Emancipation of Dissonance: Schoenberg’s Two Practices of
Composition.” Music & Letters 91(4): 568–79.
Hodes, Jeffrey. 2012. “A New Roman-Numeral Analyzer for Classical Music.” Undergraduate thesis,
Princeton University.
Hoffmann, E. T. A. 1989. E. T. A. Hoffmann’s Musical Writings. Edited by David Charlton.
Translated by Martyn Clarke. Cambridge: Cambridge University Press.
Holtmeier, Ludwig. 2007. “Heinichen, Rameau, and the Italian Thoroughbass Tradition: Concepts of
Tonality and Chord in the Rule of the Octave.” Journal of Music Theory 51(1): 5–49.
Holtmeier, Ludwig. 2011. Review of Robert Gjerdingen’s Music in the Galant Style. Eighteenth-
Century Music 8(2): 307–48.
Hook, Julian. 2002. “Uniform Triadic Transformations.” Journal of Music Theory 46(1/2): 57–126.
Hook, Julian. 2003. “Signature Transformations.” Paper presented to the national meeting of the
Society for Music Theory, Madison.
Hook, Julian. 2008. “Signature Transformations.” In Music Theory and Mathematics: Chords,
Collections, and Transformations, edited by Jack Douthett, Martha M. Hyde, and Charles J. Smith,
137–60. Rochester, NY: University of Rochester Press.
Hook, Julian. 2011. “Spelled Heptachords.” In Mathematics and Computation in Music, Proceedings
of the Third International Conference of the Society for Mathematics and Computation in Music,
edited by Carlos Agon, Emmanuel Amiot, Moreno Andreatta, Gérard Assayag, Jean Bresson, and
John Mandereau, 84–97. New York: Springer.
Hook, Julian. 2020. “Generic Sequences and the Generic Tonnetz.” Journal of Music Theory 64(1):
63–103.
Horn, Katelyn, and David Huron. 2015. “On the Changing Use of the Major and Minor Modes 1750–
1900.” Music Theory Online 21(1).
Horton, Julian. 2014. “Dialectics and Musical Analysis.” In Aesthetics of Music: Musicological
Perspectives, edited by Stephen Downes. New York: Routledge.
Huddleston, Rodney, and Geoffrey Pullum. 2002. The Cambridge Grammar of the English
Language. Cambridge: Cambridge University Press.
Humphreys, David. 1997. “Something Borrowed.” Musical Times 138(1853): 19–24.
Hunt, Graham G. 2014. “When Structure and Design Collide: The Three-Key Exposition Revisited.”
Music Theory Spectrum 36(2): 247–69.
Huovinen, Erkki. 2002. Pitch-Class Constellations: Studies in the Perception of Tonal Centricity.
Turku: Acta musicologica Fennica.
Huron, David. 1991. “The Avoidance of Part-Crossing in Polyphonic Music: Perceptual Evidence
and Musical Practice.” Music Perception 9(1): 93–104.
Huron, David. 2007a. Sweet Anticipation: Music and the Psychology of Expectation. Cambridge,
MA: MIT Press.
Huron, David. 2007b. “On the Role of Embellishment Tones in the Perceptual Segregation of
Concurrent Musical Parts.” Empirical Musicology Review 2(4): 123–39.
Huron, David. 2016. Voice Leading: The Science behind a Musical Art. Cambridge, MA: MIT Press.
Imbrie, Andrew. 1973. “‘Extra’ Measures and Metrical Ambiguity in Beethoven.” In Beethoven
Studies, edited by Alan Tyson, 45–66. New York: W. W. Norton.
Jacoby, Nori, Eduardo A. Undurraga, Malinda J. McPherson, Joaquín Valdés, Tomás Ossandón, and
Josh H. McDermott. 2019. “Universal and Non-Universal Features of Musical Pitch Perception
Revealed by Singing.” Current Biology 29: 3229–43.
Jander, Owen. 1996. “Genius in the Arena of Charlatanry: The First Movement of Beethoven’s
Tempest Sonata in Cultural Context.” In Musica Franca: Essays in Honor of Frank A. D’Accone,
edited by Irene Aim, Alyson McLamore, and Colleen Reardon, 585–630. Hillsdale: Pendragon
Press.
Jeppesen, Knud. (1931) 1939. Counterpoint: The Polyphonic Vocal Style of the Sixteenth Century.
Translated by Glen Haydon. Englewood Cliffs, NJ: Prentice Hall.
Jeppesen, Knud. (1944–1945) 1975. “Problems of the Pope Marcellus Mass.” In Palestrina: Pope
Marcellus Mass, edited and translated by Lewis Lockwood, 99–130. New York: W. W. Norton.
Jeppesen, Knud. (1946) 1970. The Style of Palestrina and the Dissonance. New York: Dover.
Jost, John, Laurie A. Rudman, Irene V. Blair, Dana R. Carney, Nilanjana Dasgupta, Jack Glaser, and
Curtis D. Harding. 2009. “The Existence of Implicit Bias Is beyond Reasonable Doubt: A
Refutation of Ideological and Methodological Objections and Executive Summary of Ten Studies
That No Manager Should Ignore.” Research in Organizational Behavior 29: 39–69.
Junod, Julien, Pierre Audétat, Carlos Agon, and Moreno Andreatta. 2009. “A Generalization of
Diatonicism and the Discrete Fourier Transform as a Mean [sic.] for Classifying and
 Characterising Musical Scales.” In Mathematics and Computation in Music, edited by Elaine
Chew, Adrian Childs, and Ching-Hua Chuan, 166–79. New York: Springer.
Kant, Immanuel. (1788) 1997. Critique of Practical Reason. Edited and translated by Mary Gregor.
Cambridge: Cambridge University Press.
Kant, Immanuel. (1790) 1970. Critique of Judgment. Translated by John Henry Bernard. New York:
Free Press.
Katz, J., and David Pesetsky. 2011. “The Identity Thesis for Language and Music.” Unpublished
manuscript available at https://ling.auf.net/lingbuzz/000959.
Kerman, Joseph. 1981. The Masses and Motets of William Byrd. Berkeley: University of California
Press.
Kinderman, William. 2009a. Beethoven. 2nd ed. New York: Oxford.
Kinderman, William. 2009b. “The First Movement of Beethoven’s Tempest Sonata: Genesis, Form,
and Dramatic Meaning.” In Beethoven’s Tempest Sonata: Perspectives of Analysis and
Performance, edited by Peter Berge, Jeroen D’hoe, and William Caplin, 213–34. Walpole: Peters.
Kinderman, William, and Harold Krebs, eds. 1996. The Second Practice of Nineteenth-Century
Tonality. Lincoln: University of Nebraska Press.
Kirlin, Phillip B. 2014. “A Probabilistic Model of Hierarchical Music Analysis.” PhD diss.,
University of Massachusetts.
Kloss, Jürgen. 2012. “‘. . . Tell Her to Make Me a Cambric Shirt’: From the ‘Elfin Knight’ to
‘Scarborough Fair.’” https://www.justanothertune.com/html/cambricshirt.html.
Klumpenhouwer, Henry. 1991. “A Generalized Model of Voice-Leading for Atonal Music.” PhD
diss., Harvard University.
Klumpenhouwer, Henry. 2011. “History and Structure in Cohn’s Audacious Euphony.” Intégral 25:
159–75.
Korte, Werner Fritz. 1929. Die Harmonik des frühen XV. Jahrhunderts in ihrem Zusammenhang mit
der Formtechnik. Münster: Suhrbier & Bröcker.
Kostka, Stefan, and Dorothy Payne. 2003. Tonal Harmony. 4th ed. New York: Alfred A. Knopf.
Kramer, Lawrence. 1990. Music as Cultural Practice: 1800–1900. Berkeley: University of California
Press.
Kramer, Lawrence. 2002. “Contesting Wagner: The Lohengrin Prelude and Anti-Anti-Semitism.”
19th-Century Music 25(2–3): 190–211.
Labov, William. 1975. “Empirical Foundations of Linguistic Theory.” In The Scope of American
Linguistics, edited by Robert Austerlitz, 77–134. Lisse: Peter de Ridder Press.
Laitz, Steven. 2008. The Complete Musician. New York: Oxford University Press.
Lalite, Philippe, Emmanuel Bigand, Joanna Kantor-Martynuska, and Charles Delbé. 2009. “On
Listening to Atonal Variants of Two Piano Sonatas by Beethoven.” Music Perception 26(3): 223–
34.
Laukens, Dirk. 2019. Jazz Guitar Chord Dictionary. Antwerp: Jazz Guitar Online. Free ebook
available from http://www.jazzguitar.be.
Lawrence, John Y. 2020. “Toward a Predictive Theory of Theme Types.” Journal of Music Theory
64(1): 1–36.
Lendvai, Erno. 1971. Bela Bartók: An Analysis of His Music. London: Kahn and Averill.
Lerdahl, Fred. 1988. “Cognitive Constraints on Compositional Systems.” In Generative Processes in
Music: The Psychology of Performance, Improvisation, and Composition, edited by John Sloboda,
231–59. New York: Oxford University Press.
Lerdahl, Fred. 2001. Tonal Pitch Space. New York: Oxford University Press.
Lerdahl, Fred. 2020. Composition and Cognition. Berkeley: University of California Press.
Lerdahl, Fred, and Ray Jackendoff. 1983. A Generative Theory of Tonal Music. Cambridge, MA:
MIT Press.
Lester, Joel. 1998. “J. S. Bach Teaches Us How to Compose: Four Pattern Preludes of the Well-
Tempered Clavier.” College Music Symposium 38: 33–46.
Levine, Mark. 1989. The Jazz Piano Book. Petaluma: Sher Music.
Levinson, Jerrold. 1997. Music in the Moment. Ithaca, NY: Cornell University Press.
Lewin, David. 1981. “A Way into Schoenberg’s Opus 15, Number 7.” In Theory Only 6(1): 3–24.
Lewin, David. 1984. “Amfortas’s Prayer to Titurel and the Role of D in ‘Parsifal’: The Tonal Spaces
of the Drama and the Enharmonic C♭/B.” 19th-Century Music 7(3): 336–49.
Lewin, David. 1986. “Music Theory, Phenomenology, and Modes of Perception.” Music Perception
3(4): 327–92.
Lewin, David. 1987. Generalized Musical Intervals and Transformations. New Haven, CT: Yale
University Press.
Lewin, David. 1990. “Klumpenhouwer Networks and Some Isographies That Involve Them.” Music
Theory Spectrum 12: 83–120.
Lewis, Christopher Orlo. 1983. “Tonality and Structure in the Ninth Symphony of Gustav Mahler.”
PhD diss., University of Rochester.
Lewis, Clive Staples. 1964. The Discarded Image. Cambridge: Cambridge University Press.
Liszt, Franz. 2016. The Collected Writings of Franz Liszt, vol. 3, part 2. Edited and translated by
Janita R. Hall-Swadley. Lanham, MD: Rowman & Littlefield.
London, Justin. 2004. Hearing in Time. New York: Oxford University Press.
Long, Megan Kaes. 2018. “Cadential Syntax and Tonal Expectation in Late Sixteenth-Century
Homophony.” Music Theory Spectrum 40(1): 52–83.
Long, Megan Kaes. 2020. Hearing Homophony. New York: Oxford University Press.
Longyear, Rey M., and Kate R. Covington. 1988. “Sources of the Three-Key Exposition.” Journal of
Musicology 6(4): 448–70.
Lowinsky, Edward. 1961. Tonality and Atonality in Sixteenth-Century Music. Berkeley: University of
California Press.
Lund, Jeb. 2014. “Is a Hot Dog a Sandwich? An Extended Meditation on the Nature of America.”
Guardian, July 3. Available at https://www.theguardian.com/commentisfree/2014/jul/03/is-a-hot-
dog-a-sandwich-nature-america.
Mak, Su Yin. 2008. “Et in Arcadia Ego: The Elegiac Structure of Schubert’s Quartettsatz in C Minor
(D. 703).” In The Unknown Schubert, edited by Barbara M. Reul and Lorraine Byrne Bodley, 145–
53. New York: Routledge.
Margolis, Eric, and Stephen Laurence. 2003. “Concepts.” In The Blackwell Guide to the Philosophy
of Mind, edited by Stephen P. Stich and Ted A. Warfield, 190–213. Oxford: Blackwell.
Margulis, Elizabeth. 2013. On Repeat. New York: Oxford University Press.
Martin, Nathan John. 2010. Review of Beethoven’s ‘Tempest’ Sonata: Perspectives of Analysis and
Performance. Theory and Practice 35: 169–89.
Martin, Nathan John. 2018. “Schenker and/or Rameau.” Paper presented to the national meeting of
the Society for Music Theory, San Antonio.
Marvin, Elizabeth West, and Alexander Brinkman. 1999. “The Effect of Modulation and Formal
Manipulation on Perception of Tonic Closure by Expert Listeners.” Music Perception 16(4): 389–
408.
Marvin, William. 2012–2013. “ ‘Und so weiter’: Schenker, Sonata Theory, and the Problem of the
Recapitulation.” Theory and Practice 37–38: 221–40.
Marx, Adolf Bernhard. (1859) 1997. Musical Form in the Age of Beethoven: Selected Writings on
Theory and Method. Translated by Scott Burnham. Cambridge: Cambridge University Press.
McClary, Susan. 1991. Feminine Endings: Music, Gender, and Sexuality. Minneapolis: University of
Minnesota Press.
McClary, Susan. 2004. Modal Subjectivities. Berkeley: University of California Press.
McCreless, Patrick. 1996. “An Evolutionary Perspective on Nineteenth-Century Semitonal
Relations.” In The Second Practice of Nineteenth-Century Tonality, edited by William Kinderman
and Harald Krebs, 87–113. Lincoln: University of Nebraska Press.
McCreless, Patrick. 2011. “Ownership, in Music and Music Theory.” Music Theory Online 17(1).
McDermott, Josh, and Marc Hauser. 2005. “The Origins of Music: Innateness, Uniqueness, and
Evolution.” Music Perception 23(1): 29–59.
McEnery, Tony, and Andrew Wilson. 1996. Corpus Linguistics. Edinburgh: Edinburgh University
Press.
McHose, Allen. 1947. The Contrapuntal Harmonic Technique of the 18th Century. New York: Crofts.
Meeùs, Nicolas. 2000. “Toward a Post-Schoenbergian Grammar of Tonal and Pre-Tonal Harmonic
Progressions.” Music Theory Online 6(1).
Meeùs, Nicolas. 2018. “Harmonic Vectors and the Constraints of Tonality.” Music Theory Online
24(4).
Meier, Bernhard. 1988. The Modes of Classical Vocal Polyphony: Described According to the
Sources, with Revisions by the Author. Translated by Ellen S. Beebe. London: Broude Brothers.
Meyer, Leonard. 1956. Emotion and Meaning in Music. Chicago: University of Chicago Press.
Meyer, Leonard. 1982. “Process and Morphology in the Music of Mozart.” Journal of Musicology
1(1): 67–94.
Meyer, Leonard. 1996. Style and Music. Chicago: University of Chicago Press.
Milsom, John. 2005. “Imitatio, Intertextuality and Early Music.” In Citation and Authority in
Medieval and Renaissance Musical Culture: Learning from the Learned, edited by Suzannah Clark
and Elizabeth Eva Leach, 141–51. Woodbridge: Boydell.
Mirka, Danuta. 2015. “The Mystery of the Cadential Six-Four.” In What Is a Cadence? Theoretical
and Analytical Perspectives on Cadences in the Classical Repertoire, edited by Markus Neuwirth
and Pieter Bergé, 157–84. Leuven: Leuven University Press.
Moore, Alan. 2012. Song Means: Analyzing and Interpreting Popular Song. Burlington: Ashgate.
Moreno, Jairo. 1996. “Theoretical Reception of the Sequence and Its Conceptual Implications.” PhD
diss., Yale University.
Morgan, Robert. 1978. “Schenker and the Theoretical Tradition: The Concept of Musical Reduction.”
College Music Symposium 18(1): 72–96.
Morgan, Robert. 2014. Becoming Heinrich Schenker. Cambridge: Cambridge University Press.
Morris, Robert D. 1995. “Equivalence and Similarity in Pitch and Their Interaction with PCSet.”
Journal of Music Theory 39(2): 207–43.
Muns, Lodewijk. 2008. “Why I Am Not a Schenkerian.” Unpublished manuscript available at
https://lodewijkmuns.nl/WhyIamNotaSchenkerian.pdf.
Nápoles López, Néstor, Claire Arthur, and Ichiro Fujinaga. 2019. “Key-Finding Based on a Hidden
Markov Model and Key Profiles.” In 6th International Conference on Digital Libraries for
Musicology, 33–37. New York: Association for Computing Machinery.
Narmour, Eugene. 1988. “Melodic Structuring of Harmonic Dissonance: A Method for Analysing
Chopin’s Contribution to the Development of Harmony.” In Chopin Studies, edited by Jim
Samson, 77–114. Cambridge: Cambridge University Press.
National Research Council. 2014. Identifying the Culprit: Assessing Eyewitness Identification.
Washington, DC: National Academies Press.
Navia, Gabriel Henrique Bianco. 2016. “The Medial Caesura in Schubert’s Sonata Forms: Formal
and Rhetorical Complications.” PhD diss., University of Arizona.
Neff, Severine. 2012. “Schoenberg as Theorist.” In Schoenberg and His World, edited by Walter
Frisch, 55–84. Princeton: Princeton University Press.
Neumeyer, David. 1987. “The Ascending Urlinie.” Journal of Music Theory 31(2): 275–303.
Newcomb, Anthony. 2007. “Marenzio’s Riffs.” In Luca Marenzio e il madrigale romano, edited by
Franco Piperno, 115–43. Rome: Accademia Nazionale di Santa Cecilia.
Nicolas, Patrice. 2019. “Challenging Some Misconceptions about the Règle de l’octave.” Music
Theory Online 25(4).
Ninov, Dimitar. 2016. “Functional Nature of the Cadential Six-Four.” Musicological Annual 52(1):
73–96.
Nobile, Drew. 2015. “Counterpoint in Rock Music: Unpacking the ‘Melodic-Harmonic Divorce.’”
Music Theory Spectrum 37(2): 189–203.
Nobile, Drew. 2016. “Harmonic Function in Rock Music: A Syntactic Approach.” Journal of Music
Theory 60(2): 149–80.
Nobile, Drew. 2020. “Double-Tonic Complexes in Rock Music.” Music Theory Spectrum 42(2): 207–
26.
Nolan, Catherine. 2003. “Combinatorial Space in Nineteenth- and Early Twentieth-Century Music
Theory.” Music Theory Spectrum 25(2): 205–41.
Okazaki, Miles. 2015. Fundamentals of Guitar. Fenton, MO: Mel Bay.Osborn, Brad. 2017. “Rock
Harmony Reconsidered: Tonal, Modal and Contrapuntal Voice-Leading Systems in Radiohead.”
Music Analysis 36(1): 59–93.
Palisca, Claude. 1956. “Vincenzo Galilei’s Counterpoint Treatise: A Code for the Seconda Pratica.”
Journal of the American Musicological Society 9(2): 81–96.
Palisca, Claude. 1994. Studies in the History of Italian Music and Music Theory. New York: Oxford
University Press.
Parncutt, Richard. 1988. “Revision of Terhardt’s Psychoacoustical Model of the Root(s) of a Musical
Chord.” Music Perception 6: 65–94.
Perry, Jeffrey, and Paul Lansky. 1996. “The Inner Voices of Simple Things: A Conversation with Paul
Lansky.” Perspectives of New Music 34(2): 40–60.
Pinter, D. 2019. “The Magnificent Flat-Seventh.” Soundscapes 22. Available at
https://www.icce.rug.nl/~soundscapes/VOLUME22/Magnificent_flat-seventh.shtml.
Piston, Walter. 1941. Harmony. New York: W. W. Norton.
Powers, Harold. 1958. “Mode and Raga.” The Musical Quarterly 44(4): 448–60.
Powers, Harold. 1981. “Tonal Types and Modal Categories in Renaissance Polyphony.” Journal of
the American Musicological Society 34(3): 428–70.
Powers, Harold. 1992a. “Is Mode Real? Pietro Aron, the Octenary System, and Polyphony.” Basler
Jahrbuch für Historische Musikpraxis 16: 9–52.
Powers, Harold. 1992b. “Modality as a European Cultural Construct.” In Secondo convegno europeo
di analisi musicale: Atti, edited by Rosanna Dalmonte and Mario Baroni, 207–20. Trento:
Università degli studi di Trento, Dipartimento di storia della civiltà europea.
Proctor, Gregory. 2004. “A Schenkerian Look at Lully.” Journal of Seventeenth-Century Music 10(1).
Putnam, Hilary. 1988. Realism and Reason. Cambridge: Cambridge University Press.
Quine, Willard van Orman. 1953. From a Logical Point of View. Cambridge, MA: Harvard
University Press.
Quinn, Ian. 2005. “Harmonic Function without Primary Triads.” Paper presented to the national
meeting of the Society for Music Theory, Boston.
Quinn, Ian. 2018. “Tonal Harmony.” In The Oxford Handbook of Critical Concepts in Music Theory,
edited by Alexander Rehding and Steven Rings, 467–97. New York: Oxford University Press.
Rabinovitch, Gilad. 2018. “Gjerdingen’s Schemata Reexamined.” Journal of Music Theory 62(1):
41–84.
Rabinovitch, Gilad. 2019. “Implicit Counterpoint in Gjerdingen’s Schemata.” Music Theory &
Analysis 6(1): 1–49.
Rabinovitch, Gilad. 2020. “Hidden Polyphony, Linear Hierarchy, and Scale-Degree Associations in
Galant Schemata.” Indiana Theory Review 36(1–2): 114–66.
Rameau, Jean Paul. (1722) 1971. Treatise on Harmony. Translated by Philip Gossett. New York:
Dover.
Reef, John. 2020. “A ‘Proto-Theme’ in Some of J. S. Bach’s Fugal Works.” Music Theory Online
26(4).
Rehding, Alexander. 2009. Music and Monumentality: Commemoration and Wonderment in
Nineteenth-Century Germany. New York: Oxford University Press.
Reti, Rudolf. 1951. The Thematic Process in Music. New York: Macmillan.
Reynolds, Christopher Alan. 2015. Wagner, Schumann, and the Lessons of Beethoven’s Ninth.
Berkeley: University of California Press.
Ricci, Adam. 2004. “A Theory of the Harmonic Sequence.” PhD diss., University of Rochester.
Rice, John. 2015. “The Morte: A Galant Voice-Leading Schema as Emblem of Lament and
Compositional Building-Block.” Eighteenth-Century Music 12(2): 157–81.
Richards, Mark. 2010. “Closure in Classical Themes: The Role of Melody and Texture in Cadences,
Closural Function, and the Separated Cadence.” Intersections: Canadian Journal of Music 31(1):
25–45.
Richards, Mark. 2011. “Viennese Classicism and the Sentential Idea: Broadening the Sentence
Paradigm.” Theory and Practice 36: 179–224.
Richards, Mark. 2013a. “Beethoven and the Obscured Medial Caesura: A Study in the
Transformation of Style.” Music Theory Spectrum 35(2): 166–93.
Richards, Mark. 2013b. “Sonata Form and the Problem of Second-Theme Beginnings.” Music
Analysis 32(1): 3–45.
Richards, Mark. 2017. “Tonal Ambiguity in Popular Music’s Axis Progressions.” Music Theory
Online 23(3).
Richter, Ernst Friedrich. 1852. Die Grundzige der musikalischen Formen und ihre Analyse. Leipzig:
Georg Wigand.
Riemann, Hugo. 1896. Harmony Simplified. London: Augener.
Riemann, Hugo. 1903. System der musikalischen Rhythmik und Metrik. Leipzig: Breitkopf and
Härtel.
Rings, Steven. 2011. Tonality and Transformation. New York: Oxford University Press.
Rings, Steven. 2013. “A Foreign Sound to Your Ear: Bob Dylan Performs ‘It’s Alright, Ma (I’m Only
Bleeding),’ 1964–2009.” Music Theory Online 19(4).
Rohrmeier, Martin. 2011. “Towards a Generative Syntax of Tonal Harmony.” Journal of Mathematics
and Music 5(1): 35–53.
Rohrmeier, Martin. 2020. “The Syntax of Jazz Harmony: Diatonic Tonality, Phrase Structure, and
Form.” Music Theory & Analysis 7(1): 1–62.
Roig-Francolí, Miguel A. 2003. Harmony in Context. New York: McGraw-Hill.
Roig-Francolí, Miguel A. 2018. “From Renaissance to Baroque: Tonal Structures in Tomás Luis de
Victoria’s Masses.” Music Theory Spectrum 40(1): 27–51.
Rosch, Eleanor. 1978. “Principles of Categorization.” In Cognition and Categorization, edited by
Eleanor Rosch and Barbara Bloom Lloyd, 28–49. Hillsdale, NJ: Erlbaum.
Rosen, Charles. 1971. The Classical Style. New York: W. W. Norton.
Rosen, Charles. 1988. Sonata Forms. New York: W. W. Norton.
Rosen, Charles. 1998. The Romantic Generation. Cambridge, MA: Harvard University Press.
Rosen, Charles. 2001. Critical Entertainments. Cambridge, MA: Harvard University Press.
Rotem, Elam, and Peter Schubert. 2021. “Stretto fuga—How to Make a Masterpiece out of Two
Intervals.” Video, 17:50. https://youtu.be/zGZjzBvXCzg.
Rothenberg, David. 1978. “A Model for Pattern Perception with Musical Applications. Part I: Pitch
Structures as Order-Preserving Maps.” Mathematical Systems Theory 11: 199–234.
Rothgeb, John. 1975. “Strict Counterpoint and Tonal Theory.” Journal of Music Theory 19(2): 260–
84.
Rothgeb, John. 1996. “Re: Eytan Agmon on Functional Theory.” Music Theory Online 2(1).
Rothstein, William. 1989. Phrase Rhythm in Tonal Music. New York: Schirmer.
Rothstein, William. 2009. “Riding the Storm Clouds: Tempo, Rhythm, and Meter in Beethoven’s
Tempest Sonata.” In Beethoven’s Tempest Sonata: Perspectives of Analysis and Performance,
edited by Peter Berge, Jeroen D’hoe, and William Caplin, 235–72. Walpole: Peters.
Rowe, Dorothy. 1983. Depression: The Way Out of Your Prison. New York: Routledge and Kegan
Paul.
Russell, Jonathan. 2018. “Harmony and Voice-Leading in The Rite of Spring.” PhD diss., Princeton
University.
Sailor, Malcolm, and Andie Sigler. 2017. “The Insufficiently Stimulated Ear: A Corpus Study of
Dissonance Treatment from DuFay to Victoria.” Paper presented to the national meeting of the
Society for Music Theory, Arlington.
Salzer, Felix. (1952) 1962. Structural Hearing. New York: Dover.
Salzer, Felix. 1983. “Heinrich Schenker and Historical Research: Monteverdi’s Madrigal Oimè, se
tanto amate.” In Aspects of Schenkerian Theory, edited by David Beach, 135–52. New Haven, CT:
Yale University Press.
Samarotto, Frank. 2004. “Sublimating Sharp 4: An Exercise in Schenkerian Energetics.” Music
Theory Online 10(3).
Sanguinetti, Giorgio. 2012. The Art of Partimento. New York: Oxford University Press.
Saslaw, Janna. 1992. “Gottfried Weber and the Concept of Mehrduetigkeit.” PhD diss., Columbia
University.
Savage, Patrick E., Steven Brown, Emi Sakai, and Thomas E. Currie. 2015. “Statistical Universals
Reveal the Structures and Functions of Human Music.” Proceedings of the National Academy of
Sciences 112(29): 8987–92.
Savage, Patrick, Adam Tierney, and Annirudh Patel. 2017. “Global Music Recordings Support the
Motor Constraint Hypothesis for Human and Avian Song Contour.” Music Perception 34(3): 327–
34.
Schachter, Carl. 1990. “Either/Or.” In Schenker Studies, edited by Heidi Siegel, 165–80. Cambridge:
Cambridge University Press.
Schachter, Carl. 2016. The Art of Tonal Analysis. Edited by Joseph Straus. New York: Oxford
University Press.
Schenker, Heinrich. (1910) 1987. Counterpoint, vol. 1. Edited by John Rothgeb. Translated by John
Rothgeb and Jürgen Thym. New York: Schirmer.
Schenker, Heinrich. (1921–1923) 2004. Der Tonwille: Pamphlets in Witness of the Immutable Laws
of Music, vol. 1. Edited by William Drabkin. Translated by Ian Bent, William Drabkin, Joseph
Dubiel, Timothy Jackson, Joseph Lubben, and Robert Snarrenberg. New York: Oxford University
Press.
Schenker, Heinrich. (1923–1924) 2005. Der Tonwille: Pamphlets in Witness of the Immutable Laws
of Music, vol. 2. Edited by William Drabkin. Translated by Ian Bent, William Drabkin, Joseph
Dubiel, Joseph Lubben, William Renwick, and Robert Snarrenberg. New York: Oxford University
Press.
Schenker, Heinrich. (1926) 1996. The Masterwork in Music, vol. 2. Edited by William Drabkin.
Translated by Ian Bent, William Drabkin, John Rothgeb, and Heidi Siegel. Cambridge: Cambridge
University Press.
Schenker, Heinrich. (1930) 1997. The Masterwork in Music, vol. 3. Edited by William Drabkin.
Translated by Alfred Clayton and Derrick Puffett. Cambridge: Cambridge University Press.
Schenker, Heinrich. (1933) 1969. Five Graphical Music Analyses. Translated by Felix Salzer. New
York: Dover.
Schenker, Heinrich. (1935) 1979. Free Composition. Edited and translated by Ernst Oster. New York:
Longman.
Scherer, Josh. 2015. “A Bro and a Philosopher Debate the True Meaning of a Sandwich.”
https://firstwefeast.com/eat/2015/04/philosophy-of-meat-bread.
Schindler, Anton. 1840. Biographie von Ludwig van Beethoven. Münster: Achendorff.
Schindler, Anton. 1841. The Life of Beethoven. Translated by Ignaz Moscheles. London: Henry
Colburn.
Schindler, Anton. (1860) 1996. Beethoven as I Knew Him. Edited by Donald W. MacArdle.
Translated by Constance S. Jolly. Mineola: Dover.
Schmalfeldt, Janet. 1991. “Towards a Reconciliation of Schenkerian Concepts with Traditional and
Recent Theories of Form.” Music Analysis 10(3): 233–87.
Schmalfeldt, Janet. 1995. “Form as the Process of Becoming: The Beethoven-Hegelian Tradition and
the Tempest Sonata.” Beethoven Forum 4: 37–71.
Schmalfeldt, Janet. 2011. In the Process of Becoming. New York: Oxford University Press.
Schoenberg, Arnold. (1911) 1983. Theory of Harmony. Berkeley: University of California Press.
Schoenberg, Arnold. 1984. Style and Idea. Edited by Leonard Stein. Translated by Leo Black.
Berkeley: University of California Press.
Schubert, Peter. 2008. Modal Counterpoint, Renaissance Style. New York: Oxford University Press.
Schubert, Peter. 2018. “Thomas Campion’s ‘Chordal Counterpoint’ and Tallis’s Famous Forty-Part
Motet.” Music Theory Online 24(1).
Schuijer, Michiel. 2008. Analyzing Atonal Music: Pitch-Class Set Theory and Its Contents.
Rochester, NY: University of Rochester Press.
Segall, Christopher. 2010. “K-Nets, Inversion, and Gravitational Balance.” Theory and Practice 35:
119–45.
Segall, Christopher. 2014. “Sergei Taneev’s Vertical-Shifting Counterpoint: An Introduction.” Music
Theory Online 20(3).
Sewell, Ian. 2021. “When All You Have Is a Hammer: Surface/Depth as Good Comparison.” Music
Theory Spectrum 43(2): 197–220.
Shepard, Roger. 1964. “Circularity in Judgements of Relative Pitch.” Journal of the Acoustical
Society of America 36(12): 2346–53.
Simmons, Joseph P., Leif D. Nelson, and Uri Simonsohn. 2011. “False-Positive Psychology:
Undisclosed Flexibility in Data Collection and Analysis Allows Presenting Anything as
Significant.” Psychological Science 22(11): 1359–66.
Simon, Herbert, and Kenneth Kotovsky. 1963. “Human Acquisition of Concepts for Sequential
Patterns.” Psychological Review 70(6): 534–46.
Simon, Herbert, and Richard Sumner. 1968. “Pattern in Music.” In Formal Representation of Human
Judgment, edited by Benjamin Kleinmuntz, 219–50. New York: Wiley.
Slottow, Steven. 2018. “To Be or Not to Be: Schenker’s Versus Schenkerian Attitudes towards
Sequences.” Gamut: Online Journal of the Music Theory Society of the Mid-Atlantic 8(1).
Smith, Charles. 1986. “The Functional Extravagance of Chromatic Chords.” Music Theory Spectrum
8: 94–139.
Smith, Charles. 1996. “Musical Form and Fundamental Structure: An Investigation of Schenker’s
Formenlehre.” Music Analysis 15: 191–92.
Smith, Peter. 1994. “Brahms and Schenker: A Mutual Response to Sonata Form.” Music Theory
Spectrum 16: 77–71.
Smith, Peter. 2006. “Harmonic Cross-Reference and the Dialectic of Articulation and Continuity in
Sonata Expositions of Schubert and Brahms.” Journal of Music Theory 50(2): 143–79.
Sommer, Heinz-Dieter. 1986. “Beethovens kleine Pastorale Zum 1. Satz der Klaviersonate op. 28.”
Archiv für Musikwissenschaft 43(2): 109–27.
Sprick, Jan Philipp. 2018. “Sequences: Between Affirmation and Destruction of Tonality.” Music
Theory Online 24(4).
Stein, Beverly. 2002. “Carissimi’s Tonal System and the Function of Transposition in the Expansion
of Tonality.” Journal of Musicology 19(2): 264–305.
Stephenson, Kenneth. 2002. What to Listen for in Rock: A Stylistic Analysis. New Haven, CT: Yale
University Press.
Strunk, Oliver. 1950. Source Readings in Music History. New York: W. W. Norton.
Stuckenschmidt, Hans Heinz. 1965. “Debussy or Berg? The Mystery of a Chord Progression.” The
Musical Quarterly 51(3): 453–59.
Subotnik, Rose Rosengard. 1991. Developing Variations: Style and Ideology in Western Music.
Minneapolis: University of Minnesota Press.
Sutcliffe, W. Dean. 2003. The Keyboard Sonatas of Domenico Scarlatti and Eighteenth-Century
Musical Style. Cambridge: Cambridge University Press.
Swinden, Kevin. 2005. “When Functions Collide: Aspects of Plural Function in Chromatic Music.”
Music Theory Spectrum 27(2): 249–82.
Taneyev, Sergei. (1909) 1962. Convertible Counterpoint in the Strict Style. Translated by G. Ackley
Brower. Boston: Bruce Humphries.
Taruskin, Richard. 2005. The Oxford History of Western Music. New York: Oxford University Press.
Taylor, Benedict. 2019. Towards a Harmonic Grammar of Grieg’s Late Piano Music. New York:
Routledge.
Telesco, Paula J. 1998. “Enharmonicism and the Omnibus Progression in Classical-Era Music.”
Music Theory Spectrum 20(2): 242–79.
Temperley, David. 2006. Review of Music in the Galant Style. Journal of Music Theory 50(2): 277–
90.
Temperley, David. 2007. “The Melodic-Harmonic ‘Divorce’ in Rock.” Popular Music 26(2): 323–42.
Temperley, David. 2008. “Hypermetrical Transitions.” Music Theory Spectrum 30(2): 305–25.
Temperley, David. 2011a. “Composition, Perception, and Schenkerian Theory.” Music Theory
Spectrum 33(2):146–68.
Temperley, David. 2011b. “The Cadential IV in Rock.” Music Theory Online 17(1).
Temperley, David. 2018. The Musical Language of Rock. New York: Oxford University Press.
Thayer, Alexander Wheelock. 1967. Thayer’s Life of Beethoven. Edited by Elliot Forbes. Princeton:
Princeton University Press.
Theune, Michael. 2007. Structure & Surprise: Engaging Poetic Turns. New York: Teachers &
Writers Collaborative.
Thomas, Keith. 1971. Religion and the Decline of Magic. New York: Scribner.
Tierney, Adam, Frank Russo, and Annirudh Patel. 2008. “Empirical Comparisons of Pitch Patterns in
Music, Speech, and Birdsong.” The Journal of the Acoustical Society of America 123: 3721.
Tierney, Adam, Frank Russo, and Annirudh Patel. 2011. “The Motor Origins of Human and Avian
Song Structure.” Proceedings of the National Academy of Sciences of the United States of America
108(37): 15510–15.
Tomlinson, Gary. 1990. Monteverdi and the End of the Renaissance. Berkeley: University of
California Press.
Tovey, Donald. (1935) 1981. Symphonies and Other Orchestral Works: Essays in Musical Analysis.
New York: Oxford University Press.
Tovey, Donald. 1931. A Companion to Beethoven’s Pianoforte Sonatas. London: Associated Board of
the Royal Schools of Music.
Trippet, David. 2010. “Liszt on Lohengrin (or: Wagner in absentia).” The Wagner Journal 4(1): 4–21.
Tymoczko, Dmitri. 2003. “Progressions Fondamentales, Functions, Degrés, une Grammaire de
l’Harmonie Tonale Élémentaire.” Musurgia 10(3–4): 35–64.
Tymoczko, Dmitri. 2004. “Scale Networks in Debussy.” Journal of Music Theory 48(2): 215–92.
Tymoczko, Dmitri. 2008. “Scale Theory, Serial Theory, and Voice Leading.” Music Analysis 27(1):
1–49.
Tymoczko, Dmitri. 2011a. A Geometry of Music. New York: Oxford University Press.
Tymoczko, Dmitri. 2011b. “Dualism and the Beholder’s Eye.” In The Oxford Handbook of
Riemannian and Neo-Riemannian Music Theories, edited by Ed Gollin and Alex Rehding, 246–67.
New York: Oxford University Press.
Tymoczko, Dmitri. 2020a. “Review-Essay on Fred Lerdahl’s Composition and Cognition.” Music
Theory Online 26(1).
Tymoczko, Dmitri. 2020b. “Why Topology?” Journal of Mathematics and Music 14(2): 1–56.
Tymoczko, Dmitri. 2023. “Hierarchical Set Theory.” Journal of Mathematics and Music 17(2): 282–
90.
Tymoczko, Dmitri, and Jason Yust. 2019. “Fourier Phase and Pitch-Class Sum.” In Proceedings of
the 7th International Conference of Mathematics and Computation in Music, edited by Mariana
Montiel, Francisco Gómez-Martín, and Octavio A. Agustín-Aquino, 46–58. New York: Springer.
Vande Moortele, Steven. 2009. “The First Movement of Beethoven’s Tempest Sonata and the
Tradition of Twentieth-Century Formenlehre.” In Beethoven’s Tempest Sonata: Perspectives of
Analysis and Performance, edited by Peter Berge, Jeroen D’hoe, and William Caplin, 293–314.
Walpole: Peters.
Varricchio, David J., and Frankie D. Jackson. 2016. “Reproduction in Mesozoic Birds and Evolution
of the Modern Avian Reproductive Mode.” The Auk 133(4): 654–84.
Vicentino, Nicola. (1555) 1996. Ancient Music Adapted to Modern Practice. Edited by Claude V.
Palisca. Translated by Maria Rika Maniates. New Haven, CT: Yale University Press.
Von Hippel, P., and David Huron. 2000. “Why Do Skips Precede Reversals? The Effect of Tessitura
on Melodic Structure.” Music Perception 18: 59–85.
Wagner, Napthali. 2003. “ ‘Domestication’ of Blue Notes in the Beatles’ Songs.” Music Theory
Spectrum 25(2): 353–65.
Wagner, Richard. 1907. Richard Wagner’s Prose Works, vol. 3. Translated by William Ashton Ellis.
London: Kegan, Paul.
Waltham-Smith, Naomi. 2018. “Sequence.” In The Oxford Handbook of Critical Concepts in Music
Theory, edited by Alexander Rehding and Steven Rings, 577–601. New York: Oxford University
Press.
Warfield, Patrick. 2011. “The March as Musical Drama and the Spectacle of John Philip Sousa.”
Journal of the American Musicological Society 64(2): 289–318.
Watkins, Holly. 2017. “Toward a Post-Humanist Organicism.” Nineteenth-Century Music Review 14:
93–114.
Weber, Gottfried. (1817–1821) 1846. Theory of Musical Composition. Translated by James F.
Warner. Boston: Wilkins, Carter, and Company, and O. C. B. Carter.
Webster, James. 1978. “Schubert’s Sonata Form and Brahms’s First Maturity.” 19th-Century Music
2(1): 18–35.
Webster, James. 2010. “Formenlehre in Theory and Practice.” In Musical Form, Forms &
Formenlehre: Three Methodological Reflections, edited by Pieter Bergé, 123–39. Leuven: Leuven
University Press.
Wellwood, Alexis, Roumyana Pancheva, Valentine Hacquard, and Colin Phillips. 2018. “The
Anatomy of a Comparative Illusion.” Journal of Semantics 35(3): 543–83.
Westergaard, Peter. 1975. An Introduction to Tonal Theory. New York: W. W. Norton.
Wiering, Franz. 2001. The Language of the Modes: Studies in the History of Polyphonic Modality.
New York: Routledge.
Wittgenstein, Ludwig. 1953. Philosophical Investigations. Edited by G. E. M. Anscombe and R.
Rhees. Translated by G. E. M. Anscombe. Oxford: Blackwell.
Wolff, Christoph. 2001. Johann Sebastian Bach: The Learned Musician. New York: W. W. Norton.
Wootton, David. 2015. The Invention of Science. New York: HarperCollins.
Yellin, Victor. 1998. The Omnibus Idea. Warren, MI: Harmonie Park Press.
Yust, Jason. 2009. “The Step-Class Automorphism Group in Tonal Analysis.” In Mathematics and
Computation in Music: MCM 2007, edited by T. Klouche and T. Noll, 512–520. New York:
Springer.
Yust, Jason. 2015a. “Distorted Continuity: Chromatic Harmony, Uniform Sequences, and Quantized
Voice Leadings.” Music Theory Spectrum 37(1): 120–43.
Yust, Jason. 2015b. “Voice-Leading Transformation and Generative Theories of Tonal Structure.”
Music Theory Online 21(4).
Yust, Jason. 2018. Organized Time: Rhythm, Tonality, and Form. New York: Oxford University
Press.
Zarlino, Gioseffo. (1558) 1968. The Art of Counterpoint. Translated by Guy A. Marco and Claude V.
Palisca. New Haven, CT: Yale University Press.
Zarlino, Gioseffo. (1558) 1983. On the Modes. Edited by Claude V. Palisca. Translated by Vered
Cohen. New Haven, CT: Yale University Press.
Zbikowski, Lawrence M. 2017. Foundations of Musical Grammar. New York: Oxford University
Press.
 Index

For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion,
appear on only one of those pages.

Notes are indicated by “n” following the page number. Figures are indicated by f following the page
number.

Aarden, Bret 130–31, 131n.41

AC/DC
“Back in Black” 61, 527
Acuff, Roy and his Smokey Mountain Boys 83–84
aggregate (chromatic) 15, 42, 45–46, 366–67, 401, 527
Agmon, Eytan 75n.34, 117n.19, 362n.2, 374n.10
Alberti bass 205, 206n.5, 408–9
Aldwell, Edward 237n.49
Allsop, Peter 334n.36
alphabet, musical 15n.15, 25f, 206–7, 250–51, 411
Anderson, Laurie
“O Superman” 384
Andriessen, Louis
De Staat 179–81, 180f
Armstrong, Louis 37, 189, 250–51
“West End Blues” 38f, 189f
Arndt, Matthew 21n.36, 170n.20, 248n.67, 407n.1, 431n.33
Arthur, Claire 211n.6
Artusi, Giovanni 226–27, 234, 235, 249
Ashley, Clarence “Tom”
“Rising Sun Blues” 83–84, 83f, 248–49
Auerbach, Brent 486n.54
“Autumn Leaves” (Kosma) 138–40

Babbitt, Milton 3, 136n.46, 222–23

Bach, C. P. E. 113n.14, 120n.21, 154n.4, 281n.45, 344, 354–55, 409
Bach, J. S. 3, 16n.20, 27–28, 34, 141, 154n.4, 161–62, 168f, 173–75, 173n.22, 175f, 181–83, 182f,
188, 190, 203, 209n.9, 210–11, 215n.15, 218–20, 223–24, 224f, 237–39, 257, 258–59, 259f, 291,
Ch. 7 (prelude), §7.3, 334f, 335n.38, 336–40 §7.7, 371n.7, 388n.27, 391, 392f, 412–13, 451, 453,
527, 573n.10
Brandenburg Concerto No. 5 331–32, 332f, 429n.28, 510–11
Cello Suite No. 1 in G major 188–89, 189f, 355–56, 356f, 379n.13
chorales 28f, 118n.20 §3.6, 144n.53, 208f, 217f, 218f, 225f, 238f, 254f, 279–82, 281f, 307f, 311–
12, 312f, 314f, 329f, 387n.24, 413f, 415f, 434f, 436, 437f, 569–71, 570f, 572n.7
 “Ach, lieben Christen, seid getrost” §7.7c
“Es woll uns Gott genädig sein” 77f
“Ich dank’ dich, lieber Herr” 128, 129f
“Jesu, du mein liebstes Leben” 304–5, 305f
“Was mein Gott will, das g’scheh allzeit” 181–83, 183f
computational imitation of 412–13, 414f
dualism of § 7.7
Partita in B♭ 204f
two-part inventions
C major §7.7a, 562f
D minor 204f, 305–6, 306f
F major 348n.55, 379n.13
The Well-Tempered Clavier 23–24, 24f, 144n.53, 173n.22, 176n.23, 177–78, 190n.31, 371n.7,
379n.13, 409
Book I:
Fugue in C minor 166f, 338f
Fugue in E minor 162–63, 163f
Fugue in F major 186–87, 187f
Fugue in A♭ major 327–28, 328f

Fugue in G♯ minor 157f

Fugue in A minor 418f, 563n.14
Fugue in B♭ major 237f

Fugue in B♭ minor 27f, 191f

Prelude in C major 13, 355–56, 355f
Prelude in C♯ major §7.7b, 379n.13
Prelude in D major §7.3, 379n.13, 426n.25
Prelude in G major C5.0F3 f, 336n.41
Prelude in A major 317f
Prelude in B♭ major 434–35n.39
Book 2:
Fugue in C major 169f, 379n.13
Fugue in C♯ minor 171–72, 173f
Fugue in E major 141n.49
Fugue in F major 141n.49, 177f, 556f
Fugue in F♯ minor 169f, 305n.11, 350n.60

Fugue in B♭ major 237f

Fugue in B minor §7.7d
Prelude in C minor 144n.52
Prelude in E minor 158f, 167f
BaileyShea, Matthew 444n.55
“The Ballad of the Green Berets” (Sadler and Moore) 82–83
Balzano, Gerald 394n.29
Barry, Barbara 504–5n.68, 511n.77
Bass, Richard 158n.4, 178n.25
Beach, Amy
“Fire-Flies” (Op. 15, no. 4) 189–90, 190f
Beach, David 221n.20, 253n.3, 380n.14, 511n.77, 514n.82
The Beatles 76, 147–48, 148f, 339n.44
“A Day in the Life” 71–72
“Eight Days a Week” 10–11, 11n.11, 58–59, 147–48
“Here Comes the Sun” 71–72
“I’m a Loser” 63–64
“She Loves You” 148n.58
“With a Little Help from My Friends” 78–79, 78f, 147–48
“You Never Give Me Your Money” 339n.44
“You Won’t See Me” 66, 66f
Beethoven, Ludwig van 3, 17, 20n.31, 23–24, 33, 34, Ch. 4 (prelude), 159–60, 172–73, 188, 197f,
200–1, 206n.5, 241, 258–59, 267–68, 291, 295, 311, 314–15, 345, 380 §9.4, 448n.65, Ch. 10
(prelude), Ch. 10, 527
Piano Sonata, Op. 2, no. 1 107–9, 108f, 110–11, 144–46, 145f, 378, 379f, 444n.55, 460, 462f, 466–
67, 470f
Piano Sonata Op. 2, no. 2 442n.51, 466–67, 467f, 497n.64
Piano Sonata Op. 2, no. 3 208f, 383f, 469f
Piano Sonata Op.7, no. 1 376n.12
Piano Sonata Op.10, no. 1 144n.53, 166n.13, 454f
Piano Sonata Op. 10, no. 2 471f
Piano Sonata Op. 10, no. 3 177–79, 178f, 179f, 461–63, 464f, 486n.55
Piano Sonata Op. 13 (Pathétique) Ch. 4 (prelude), 160f, 167–68, 167n.14, 198–200, 201–2, 376–
77, 377f, 460, 463f, 483
Piano Sonata Op. 22 320f §9.4
Piano Sonata Op. 26 105–6, 106f
Piano Sonata Op. 27, no. 1 100f
Piano Sonata Op. 28 (Pastorale) 436n.42, 458 §10.5, 504–5n.68, 507–8, 509–10, 522
Piano Sonata Op. 31, no. 1 63–64, 461–63, 464f, 465f
Piano Sonata Op. 31, no. 2 (Tempest) 169–70, 376n.12, 451, 452f, 454n.8, 458 §10.3, 482, 486–88,
489–90, 504
Piano Sonata Op. 31, no. 3 C8P23 n.12, 439f, 440–42, 442f, 486n.55
Piano Sonata Op. 49, no. 2 382f
Piano Sonata Op. 53 (Waldstein) 63–64, 132–36, 408–9, 451, 452f, 453, 453f, 460n.7, 483–84,
491, 516, 520, 521–22, 522f, 526
Piano Sonata Op. 54 10f, 204f
Piano Sonata Op. 57 (Appassionata) 461f
Piano Sonata Op. 79 439f
Piano Sonata Op. 81a (Lebewohl) 224–26, 240
Piano Sonata Op. 90 469f
Piano Sonata, Op. 106 (Hammerklavier) 186–87, 187f, 320n.13, 361n.1, 438–39, 441f, 471–72,
471f, 476–77
Piano Sonata Op. 109 §9.4, 437–38, 466–67, 467f, 542–43, 543f
String Quartet Op. 18, no. 3 461f
String Quartet Op. 59, no. 1 144f, 170n.19
String Quartet Op. 59, no. 2 194, 376–77, 377f
String Quartet Op. 59, no. 3 142f
 String Quartet Op. 127 20n.30
String Quartet Op. 130 470f
String Quartets Op. 18 331n.29, 380–82
Symphony No. 1 194–95n.35, 429, 430f
Symphony No. 2 142n.51, 173n.22, 460n.7, 467n.14, 478, 478n.36, 503n.67
Symphony No. 3 (Eroica) 175f, 240, 460n.7, 467n.14, 469, 470f, 489n.56
Symphony No. 5 194, 454n.9, 458, 460n.7 §10.4, 504, 524
“Knock of Fate” 454n.9, 482, 491, 492
Symphony No. 7 5f, 207f
Symphony No. 9 240, 241f, Ch. 8 (prelude), 374–75, 396–97, 481n.42, 524, 524n.93
Trio, Op. 1, no. 3 §4.10
Violin Concerto, Op. 61 486n.55
Violin Sonata, Op. 23 101–2, 101f
Violin Sonata, Op. 30, no. 2 486n.55
Violin Sonata, Op. 47 (Kreutzer) 460n.8
Bent, Margaret 258n.5, 269n.23, 271–72, 274n.30, 279–82
Bernhard, Christoph 204n.2, 221n.20, 227n.31, 235n.46, 236n.47
Bernstein, Leonard 435n.40, 486–88
The Joy of Music 451
Besseler, Heinrich 257n.1
Biamonte, Nicole 66n.23, 335n.38, 471n.17
von Bingen, Hildegard
“Laus Trinitati” 388, 389f
The Blues Brothers 449–50
Bobbitt, Richard 259n.7
Bologne, Joseph, Chevalier de Saint-Georges 334f
Boone, Graeme 262n.14
Boretz, Benjamin 223n.23
Boyce, Tommy and Hart, Bobby
“Steppin’ Stone” 72–74, 73f
Brahms, Johannes 23, 227, 291, 448n.65, 453, 526, 572n.7
Fantasien Op. 116, no. 2 176n.23
Klavierstücke Op. 76, no. 4 335n.39
Piano Quartet, Op. 60 491n.59
Symphony No. 4 320n.13
Vier ernste Gesänge Op. 121, no. 3 320n.13
Violin Sonata in G major, Op. 78 332–33, 333f
Brewer, William 402n.1
Brinkman, Alexander 16n.17, 291n.51, 362n.2, 404n.6, 447n.60
Brody, Christopher 439n.48, 447n.59
Brothomstates (Lassi Nikko) 404
Claro, 405f
Brown, Matthew 20n.32, 35n.57, 436n.41
Bruce, Robert 504–5n.68, 507n.70
Buelow, George 113n.14, 120n.21, 281n.45
Bukofzer, Manfred 257n.1
Bunting, Edward 83n.47
Burke, Edmund 452–53n.3, 504
Burnham, Scott 448n.65, 455n.10, 457–58, 472n.21, 476n.32
Burns, Lori 387n.24
Burstein, L. Poundie 20n.30, 21n.34, 35n.56, 205n.4, 350n.58
Byrd, William
Mass in Three Voices 121f, 124, 211–12, 390n.28
Byros, Vasili 16, 29n.52, 74, 77n.37, 356, 449n.70, 478n.36

cadence 19–21, 28, 30, 61–63, 99–101, 211n.6, 218, 219f, 233, 254f, 258n.4, 261–62, 265, 266, 267f,
270–74, 276n.33, 283f, 287f, 290, 296–97, 302, 348–49, 349n.57, 377–78, 380, 386f, 422n.20,
437–39, 438f, 442, 443, 449n.71, 455–56, 473–75, 476–77, 477n.35, 477f
as harmonic cycle 413–14
authentic 388–89, 521–22
converging 102–3, 151–52, 201–2, 377f, 519–20, 529–30
deceptive 253, 265, 270–74, 290, 321n.15
final 412, 414
half 78, 265, 290, 444f
mixolydian 306–8
perfect authentic 510–11
Phrygian 301–2, 385–86, 388–89, 520, 521–22
Renaissance 256n.9, 279n.42, 569
Schenkerian interpretation 416–18, 418f
caesura 377–78
filled 453–54, 454f, 455f
medial 17, 20, 380–82, 449, 451, 453–54, 454f, 455f, 455–56, 473–75, 476, 483–84, 496, 504–5,
508–9, 516f, 516–17
Callahan, Homer
“Rounder’s Luck” 83–84
Callender, Clifton 155n.2, 536n.4, 542n.11
Camp, Elizabeth 52n.6, 234n.43
Campion, Thomas 91, 91f
Caplin, William 20, 21–22, 116n.17, 151n.1, 333n.34, 334n.37, 340n.47, 414n.9, 415n.11, 439–42,
443n.52, 444n.55, 448, 455–56, 457n.2, 475n.27, 477–78, 477n.34, 478n.38, 483n.50
The Carter Family 269
Carthy, Martin 85
Cavett-Dunsby, Esther 472n.20
centricity 68, 148–49, 247–48, 275, 288–90, 299, 322n.16, 384, 390–91, 416–17, 434
Ceulmans, Anne-Emmanuelle 385n.18
Chafe, Eric 229–11nn.35–6, 235n.45
Chopin, Frederic 3, 9–10, 17, 34, 116n.18, 200–1, 224–26, 247–48, 249, 311, 334–35, 379n.13 §9.3,
516
Etude, Op. 10, no. 1 420–21, 420f
Etude, Op. 10, no. 12 (Revolutionary) 136 §9.3, 432, 528
Prelude, Op. 28, no. 2 (A minor) 188–89 §9.3
Prelude, Op. 28, no. 4 (E minor) 198, 198f, 201–2
Prelude, Op. 28, no. 5 (D major) 374–75, 375f
Mazurka, Op. 6, no. 2 227n.31
Mazurka, Op. 6, no. 3 213f, 225f
Mazurka, Op. 7, no. 2 422, 423f
Mazurka, Op. 17, no. 3 433f
Mazurka, Op. 17, no. 4 334–35 §9.3
Mazurka, Op. 24, no. 4 422n.20
Mazurka, Op. 30, no. 2 207f
 Mazurka, Op. 33, no. 1 236f
Mazurka, Op. 41, no. 1 225f
Mazurka, Op. 41, no. 2 225f
Mazurka, Op. 50, no. 2 144f
Mazurka, Op. 50, no. 3 434–35n.39
Nocturne, Op. 27, no. 2 250f
chorale(s) 13, 35, 82, 96, 134–36, 149–50, 149f, 188–89, 203, 204–5, 205f, 206–7, 412–13, 424–25,
431, 432, 476n.32, 531f
chord
added-sixth 236f, 241–42
as objects of compositional focus 264–65
as scale §1.2, Ch. 2 (prelude), 53 §3.10, Ch. 5 (prelude), 540
augmented 52–53, 53f, 69, 70–71, 71f, 275, 559f, 565–66
augmented sixth 141, 186f, 377–78, 380, 480, 486n.55, 514
common-tone diminished seventh 96–97 §3.11, 375, 476n.31, 496–98
common-tone dominant seventh §3.11, 242–44
common-tone (other chords) 496–98
diatonic distance between§2.2
dyads, geometry of §§3.2–4
Farben 95f, 558f
fourth 9–10, 227n.31, 228f, 560, 561f
grammar 29–30, 119, 239 §6.6, Ch. 7, Appendix 4
pivot 305–6, 307f
power 387
seventh 12, 19–20, 27, 28 §3.11, 257–58 §7.3, 329–30
seventh chords, geometry of §3.11
seven-note 137, 149–50, 149f, 240, 241f, 366–67, 396–97, 534–35
six-four 214–16, 222, 226, 257–58, 286, 333–34, 508n.71
So What 95f
symmetrical 52–53, 394–96, 538f, 558–60
thinking within 31–33 §3.10, 179–81, 231, 495–96, 497f, 559n.8
triads, geometry of §§3.6–10
Tristan 224–26, 225f
unprepared seventh 173–75, 218n.18, 228–29, 228f
voicing Ch. 3 (prelude)
Christensen, Thomas 202n.38, 218n.18, 221n.20, 257n.2, 259n.7, 260n.11, 313n.4, 332n.30, 386n.21
chromatic 4–5, Ch. 2 (prelude), 59–60, 63–64n.20, 87, 89f §3.12, 190n.32, 198, 242–44, 246–48,
249, 250f, 251f, 254f, Ch. 8 (prelude), §§8.1–3, 384 §8.7, §9.3, 511–12, 512f, 514, 519–20,
524n.92, 525f, 565–66
cluster 68–69
figures, transposing diatonically 11–12, 566–67
neighbor 40f, 204–5, 204f
sequence §1.1, 53f, 63–64, 64f, 69–71, 147–48, 163–65, 164f, 260–61, 334–35, 353, 481, 488,
489–90, 510f, 515–16, 517f, 520, 528–29, 563–65
set theory 14–15, 94–95, 95f
sevenths 140, 140f, 420f
space 10, 68–69, 134, 137, 344
substitution 114, 114f, 118–19
transposition Ch. 2 (prelude), §2.1, 107, 135f
triad 33, 45–46, 49f §2.1, 69–70, 92 see also modulation
chromaticism 75, 78–79, 96, 132–34, 148–49, 227, 242–44, 358–60, 374, 419, 422, 520
heavily diatonic 75, 146
Chua, Daniel 310n.18
circle of fifth(s) 54–55, 99–100, 198, 199f, 321–22, 371–72, 380–82, 394–96, 396f, 398f, 399f,
560n.11
generalized 397–99
clausula vera 99–100, 271–72, 555
Clementi, Muzio 463–65
Clendinning, Jane
The Musician’s Guide to Theory and Analysis 142n.50, 414–16
Clough, John 6n.6, 15n.15, 316n.8
Cobain, Kurt 59
Cohen, Leonard
“Hallelujah” 330, 330f
Cohn, Richard 3, 4n.5, 21–22, 60n.17, 75n.34, 78–79, 147n.57, 148–49, 149n.63, 358n.3, 369–71,
374, 421n.19, 425n.23, 480n.40, 509n.75
Collins, Denis 156n.3
Collins, Shirley 85
Collins Judd, Cristle 385n.18
Coltrane, John
Giant Steps 290–91
Coluzzi, Seth 299n.62
computational analysis 212, 214, 238–39, 240f, 255 §6.4, 295, 308, 309–10, 336n.42, 385f, 386f,
447–48, 508n.73, 532, 534–35, 569, 573–74
computational harmonization 412–13, 414f
Cone, Edward 514n.83
contrary motion 3, 31, 31f, 32f, 53, 69, 87, 90–91, 93–94, 104, 105–6, 105f, 110–11, 125–26, 134,
137, 151–52, 153–54, 153f, 155, 158–60, 158f, 165, 167–68, 171, 175–76, 178–81, 178f, 180f,
181f, 182f §4.7, 260–62, 263, 264f, 331n.29, 341–42, 350–51, 360, 409–11, 427–28, 428f, 429,
430–31, 432, 438–39, 451, 457, §§10.1–3, 478–80, 489–91, 489f, 495–98, 501–3, 510–11, 518,
522–24, 525f, 526, 530–31, 536, Appendix 3
Cook, Nicholas 16n.17, 209n.8, 291n.51, 404n.6, 447n.60
Cooke, Sam
“Cupid” 440f
Corelli, Archangelo 75–76, 257, 280–81n.44, 311, 334–35 §7.6, 457
Op. 1 336
counterpoint 2–3, 33, 60, 74, 87, 89f, 90–91, 96–98, 104, 113, 119–20, 122f, 125–26, 128–29 §3.10,
§3.12, Ch. 4 (prelude), 161–62, 186, 203, 227, 231, 234–35, 238–39, 249, 275, 276–77, 286, 296–
97, 351–53, 420–21, 432, 459–60 §10.2, 492, 527, 530–31, 540
Baroque 11–12
chromatic 516
geometry of elementary functional 261–62
instruction 2, 23, 24, 205–6, 212n.8
invertible Ch. 4 (prelude), 158–59, 158f, 167n.17, 169f, 328n.22, 466f, 477–78, 477n.35
of J.S. Bach 3 §7.7
language of 8–9
of Mahler 246
modal 203–4, 222–23, 445–46
Morte schema 186, 186f
negative and positive 90–91, 93–94, 154
 parallel 6–8, 265f, 387, 468–69, 495, 498–500
parallel and contrary 110–11, 126f, 467–68
Renaissance 222–23
species 47, 209, 214n.14, 218, 432
of Stravinsky 6–8 see also chord (thinking within), contrary motion, doubly parallel motion,
OUCH theory
Crane, Hewitt 404n.8
Cuthbert, Michael Scott 3, 532n.5
Cutler, Timothy 536n.5
Czerny, Carl 420n.17

Dahlhaus, Carl 26n.46, 149n.63, 223n.23, 256n.9, 257, 257n.1, 259n.7, 265n.16, 265n.17, 269,
269n.23, 270, 271–72, 274–75, 274n.30, 277n.36, 279–82, 431n.32, 453n.5, 453n.7, 473n.23,
475n.26, 475n.28
Dalza, Joan Ambrosio 259–60, 260f, 265–66, 265n.17, 265f, 266f, 267, 268n.22, 277–78, 295
Pavana alla Venetiana 259–60, 260f, 265f
Tastar de Corde 269f
Damschroder, David 202n.39, 254n.4, 476n.31, 476n.32
Danto, Arthur 234n.43
Darcy, Warren 20, 21–22, 380–82, 448, 453–54, 454f, 455–56, 455n.11, 469n.16
Davies, Stephen 404n.7
Davis, Miles 66n.22
“So What” 361, 361n.1
De Clerq, Trevor 56n.11
Deathridge, John 518n.87
DeBellis, Mark 403n.3
Debussy, Claude 4, 79–81, 84–85, 88–89, 192, 203, 226–27, 247–48, 448n.65
Prelude to the Afternoon of a Faun 5f
Deutsch, Diana 15n.15
disguised model 198, 476, 483n.48, 486–88
Doll, Christopher 68n.26, 72n.32, 247n.66
doubly parallel motion 41–42, 53, 54f, 60, 125–26, 132, 195–96, 496n.63, 555, 561f, 563–65
Douthett, Jack 70n.28
Duane, Ben 507n.70, 508n.73
Dufay, Guillaume 23, 214–16, 215–16n.16, 291
Duran Duran
“Rio” 63–64, 67f
Dylan, Bob
“The House of the Rising Sun” 84–85, 84f, 84n.48
“Knockin’ on Heaven’s Door” 59–60, 60f
“Tangled Up in Blue” 61

Edwards, George 514n.83

Einstein, Alfred 296n.59, 387n.25
Ellington, Duke
“Satin Doll” 194, 196f
del Encina, Juan 295
enharmonicism 34 §8.2
epistemic circularity 303–4
Evans, Bill
“So What” 95f
Everett, Walter 78n.39, 148n.58
Ewell, Philip 82n.46, 419n.16

Faithfull, Marianne 85
Farrenc, Louise
Souvenir des Huguenots, Op. 19 193, 194f
fauxbourdon 34, 107–9, 108f, 165–66, 177n.24, 178f, 191–92, 233–34, 241–42, 297, 311, 312–13
§7.5, 336–39, 340n.46, 350–53, 429n.28, 433, 476, 476f, 488, 489–90, 495–500, 496f, 496n.63,
498f, 499f, 501
generalized 114–16, 521–22, 528–29
rule of the octave (ROTO) 76, 77n.36, 116, 148n.60, 331–32, 383f, 409–11, 498f, 501, 506f, 507f,
507–8, 510–11, 514–16, 517f, 519–20, 526
Fekaris, Dino
“I Will Survive” 138–40, 140f
Fenaroli, Fedele 221n.20
Feroe, John 15n.15
Ferraguto, Mark 471n.17
Fétis, François-Joseph 21–22, 218–20, 227n.32, 257, 295n.58
Feurzeig, David 65–66, 158n.4
Fieldman, Hail 504–460nn.68–9
Fink, Robert 457n.3
Flew, Antony 20n.33
flow 17, 407 §10.2
folk music 21n.35, 23, 82–84, 85–86, 227n.31
Formenlehre 21, 407, 449, 454–55
Forte, Allen 35n.58, 428–29, 429f, 431n.33
Foster, Gwen
“Rising Sun Blues” 83–84, 83f
Franklin, Aretha 63–64, 74–75
Froebe, Folker 426n.24
frottola 262–66, 265n.16, 269–71, 270f
Fux, Johann 221n.20

Gauldin, Robert 120n.21, 215–16n.16, 237n.49, 257n.1, 524n.92, 555n.2

Gaynor, Gloria
“I Will Survive” 138–40, 140f
Gelman, Andrew 308n.17
Geneva Psalter 269, 292
Georgiades, Thrasybulos 257n.1
genius 2, 21n.36, 35–36, 173–75, 311, 431n.33, 451, 452–53n.3, 486–88, 490–91, 504, 527
Gesualdo, Carlo §1.1, 41, 69–70, 96
“Moro Lasso” 4–5, 147–48
trick of §1.1, 15, 63–64n.20, 202, 341, 528, 534–35, 537
Gibbons, Orlando 172–73
In Nomine a 4 174f
Gilbert, Stephen 428–29, 429f
Gjerdingen, Robert 3, 6, 16–17, 21–22, 25n.45, 67n.24 §1.5, 74, 151–52, 209n.9, 210n.3, 255n.6,
261–62, 266, 331–32, 333n.34, 356, 409–11, 412n.6, 431n.34, 445n.56, 447n.60, 449, 460, 462f,
500
Music in the Galant Style 102–3, 409, 528 see also schema
Glass, Philip
 Einstein on the Beach 66n.22
Goffin, Gerry
“Natural Woman” 63–64, 63f, 67–68
Goldenberg, Yosef 304n.9, 475–76n.30
Goldstein, Robin 402n.1
Gottschalk, Louis
Le Mancenillier, Op. 11 368, 369f
Goudimel, Claude 269, 292, 292f
grammar 1–2, 3, 17n.21, 22, 29–30, 47, 64–65, 76, 81f, 205–6, 220–21, 226–27, 235, 239, 250–51,
253, 258, 267, 295, 296f, 311, Ch. 7, 376, 407, 408, 412–13, 412n.6, 425, 433, 436n.41, 443, 467,
527, Appendix 4
construction 26–27
Green, Al
“Take Me to the River” 71–72
Grieg, Edvard 76, 79–81
Op. 12, no. 3 (“Watchman’s Song”) 75–76, 76f
Op. 43, no. 3 (“In My Homeland”) 75f–76, 79–81, 81f
Op. 47, no. 7 (“Elegie”) 84n.48, 393–96, 394f
Op. 57, no. 2 (“Gade”) 306–8, 307f
Op. 65, no. 4 (“Salon”) 146n.55
Op. 68, no. 3 143f
Op. 68, no. 4 143f
Guarini, Giovanni Battista 227
Guns-N-Roses
“Sweet Child O’ Mine” 63–64

Hamilton, Kenneth 475n.28

Hansberry, Benjamin 404n.7
harmonic cycle 34, 191–92, 198, 199f, 260–61, 265, 267–68, 286, 290, 294, 297, 311, §§7.1–3, 327,
331–32, 331n.28, 339–42, 354, 355–56, 386–87, 391, 412–16, 413f, 416f, 418–19, 434, 438–39,
443–44, 451, 495
harmonic skeleton 128, 203, 210, 212, 217–18, 217f, 220–21, 228–29, 241, 246, 249, 416–17
Harrison, Daniel 68n.26, 156n.3, 158n.4, 186n.28, 223n.25, 334n.36
Hasty, Christopher 306n.14
Hatten, Robert 421n.19, 476n.31, 570
Haydn, Franz Joseph 75–76, 141–44, 211, 295, 314–15, 334f, 354–55, 380, 409, 453, 455–56, 468,
472n.20, 493–95
Divertimento in C major, Hob.XVI: 1 381f
Divertimento in G major, Hob.XVI:8 439f
Quartet in C major, Op. 50, no. 2 Hob. III:45 178n.25
Sonata in E minor, Hob. XVI:34 181, 182f
Sonata in G minor, Hob. XVI: 44 416f
Sonata in E♭ major, Hob. XVI: 49 339f

Sonata in E♭ major, Hob. XVI: 52 435f

Symphony No. 62, Hob. XVI:62 377–78, 378f
Symphony No. 100, Hob.I:100 142n.51, 207f
Headlam, David 148n.61, 330n.25, 540n.9
hearing as Ch. 9 (prelude), 417–18
Heath, Gordon 85, 85f
“Heavenly Spark” 109–10, 109f see also Sacred Harp
Heinichen, Johann David 362–63, 364f, 373n.8
Hendrix, Jimi 71–72, 74–75
Hensel, Fanny Mendelssohn 4, 31–33, 458–59
“April” (Das Jahr, H. 385) 459f
“January” (Das Jahr, H. 385) 32f
“October” (Das Jahr, H. 385) 192f
Sechs Lieder, Op. 1, no. 2 331f
Hepokoski, James 17n.22, 20, 21–22, 380–82, 448, 449n.71, 453–54, 454f, 455–56, 469n.16,
476n.31, 476n.33, 478n.38, 482n.46, 496n.62, 570
hierarchy
collectional 10, 14–15, 360, 396–97, 409–11, 528, 530
quadruple §1.2, 401, 409–11, 528
Hinton, Stephen 248n.67
Hodes, Jeffrey 239n.54
Hoffmann, E.T.A. 453n.4
holism 308, 309–10
Holtmeier, Ludwig 27n.48, 29n.52, 30n.54, 103n.7, 124n.30, 210n.3, 255n.6, 281n.45, 312n.1
Hook, Julian 6n.6, 42n.5, 156n.3, 337n.43, 553n.9
Horn, Katelyn 315n.7
Horton, Julian 475n.26
“The House of the Rising Sun” 72–74, 83–85, 83f, 84f
Huddleston, Rodney 18n.23
Humphreys, David 514n.83
Hunt, Graham 504–5n.68
Huron, David 3, 23–24, 24n.44, 56n.9, 124–97nn.31–2, 131n.40, 303n.1, 315n.7, 569, 570, 570f

Imbrie, Andrew 485–86

impressionism 12, 527
inversional equivalence 19, 21, 124, 161–62, 223n.24, 256, 257–58, 266, 268, 274, 284–86, 313,
413–14

Jackendoff, Ray 135n.45, 417f, 435n.40

Jacoby, Nori 310n.18, 545n.1
Jacquet de La Guerre, Élisabeth
Sonata no. 1 213f
Suite in D minor 466–67, 468f
Jagger, Mick 61
Jander, Owen 482n.45
jazz 2–3, 4, 12, 21n.35, 31–33, 94–95, 107–10, 226–27, 244n.61, 246–47, 294n.55, 361, 376, 394–
96, 527, 528–29
Jeppesen, Knud 214n.14, 221n.20, 222n.22, 276–77, 283n.47, 295, 385n.20, 569
Joplin, Scott 1, 449
Pine Apple Rag 145f
Jost, John 402n.1
Josquin 23–24, 93f, 122, 159–60, 172–73, 211–12, 212n.8, 214n.12, 218–20, 259, 274, 275, 275n.32,
276–78, 276f, 276n.33, 278f, 282n.46, 291–92, 292f, 295
Benedicite, omnia opera (motet) 174f
Liber generationis (motet) 212f
Mass De beata virgine 272f
Mass L’ami baudichon 160f
 Mass Sub tuum presidium 272f
Junod, Julien 398n.30

Kamakawiwoʻole, Israel
“Over the Rainbow” 111f, 114–16
Kant, Immanuel 473, 473n.24
Katz, Jacob 435n.40
Kellner 362–63, 364f, 373n.8
Kerman, Joseph 385n.20
Kern, Jerome
“All the Things You Are” 138–40, 139f
Kinderman, William 421n.19, 472n.19, 481n.42
King, Carole 4, 63–64
“Natural Woman” 63–64, 63f, 67–68
Kirlin, Phillip 209n.11, 448n.62
Kloss, Jürgen 85n.52
Klumpenhouwer, Henry 156n.3, 369n.6, 555n.4
knowledge 96, 237, 528
background 309–10
conscious 41
geometrical 528
implicit 47, 257, 471–72
intuitive 90–91
musical 1–4, 45–46, 332n.30, 528
public 188
self 402
stylistic 308
Korte, Werner Fritz 257n.1
Kostka, Stephen 128n.36, 237n.49, 312n.2
Kramer, Lawrence 421n.19, 518n.86, 522n.90

Labov, William 22n.38, 211n.6

Laitz, Steven
The Complete Musician 414–16, 416f
Lalite, Philippe 453n.6
Lansky, Paul 65n.21
Laukens, Dirk 94n.10
Lawrence, John 20n.29
Led Zeppelin
“Kashmir” 511n.79
“Stairway to Heaven” 61, 85n.50
Lendvai, Erno 60n.17, 140n.48
Lerdahl, Fred 15n.15, 15n.16, 16, 135n.45, 149n.63, 209n.11, 220n.19, 249n.71, 256n.7, 331n.28,
417f, 435n.40, 438n.46, 444n.54, 483n.50, 569
Levine, Mark 94n.10, 244n.61
Levinson, Jerrold 446n.57
Lewin, David 11n.11, 361n.1, 369–71, 374, 457, 484n.52, 555n.4, 565n.17, 566n.21
Lewis, Christopher 244n.61, 245n.65
Lewis, C.S. (Clive Staples) 87–88
Ligeti, György 12, 17, 66n.22
Passacaglia Ungherese 162–63, 163f
Liszt, Franz 518–19
“Locomotive Breath” (Jethro Tull) 61
London, Justin 486n.53
loops
chord-loop families §§2.3–5, §2.7
and enharmonicism §8.2
and functional harmony 444–45
in spiral diagrams Ch. 2 (prelude), 92, 98–99
Lowinsky, Edward 257, 265n.17, 290n.49
Lund, Jeb 21n.34
Lusitano, Vicente
Heu me Domine 12–13, 13f, 528

MacColl, Ewan 85
Maggio, Antonio
“I Got the Blues” 145f
Mahanthappa, Rudresh 4, 6
“The Decider” 7f
Mahler, Gustav 96, 210–11 §5.5, 247–48
Symphony no. 2 240n.57
Symphony no. 9 34, 142f §5.5
Marenzio, Luca 34, 227 §6.7
“Ahi dispietata morte” §6.7
“Vorria parlare e dire” 387, 388f
Margulis, Elizabeth 342n.50, 448n.66
Marshmello
“Happier” 440f
Martin, Nathan 323n.17, 344n.52, 476n.31, 570
Martines, Marianna
Sonata in A major 208f
“Vo solcando un mar crudele” 157f
Marvin, Elizabeth West 16n.17, 142n.50, 291n.51, 404n.6
The Musician’s Guide to Theory and Analysis 414–16
Marvin, William 447n.59, 448n.65
Marx, Adolf Bernhard 455n.10
mathematics 2, 45–46, 96, 320, 367n.4, 457–58, 540, 553n.8
musical 154, 183–84, 316, 471–72, 527n.1
McClary, Susan 227–10nn.32–3, 234n.43, 267n.20, 267n.21, 457n.3
McCreless, Patrick 4n.3, 361n.1
McDermott, Josh 310n.18
McEnery, Tony 23n.40
McHose, Allen 128n.36, 312n.2, 569
Meeùs, Nicolas 48n.2, 259n.7
Meier, Bernhard 79n.42, 277n.36, 385n.18
melodic slot 12, 13–14, 14f, 134–36, 357, 359f, 363–64, 367–68, 432, 535–36
Meyer, Leonard 3, 16, 24n.44, 26n.47, 74, 202n.38, 240n.59, 356, 407n.1, 421n.19
Michaelis
“Passando per una rezzola” 262–63, 263f
Milsom, John 162n.9
minimalism 201–2, 394–96, 527
Mirka, Danuta 211n.7
modernism 1–2, 85–86, 184, 249n.70
modes 2, 56f, 98, 257–58, 268, 275n.32, 290, 291f, 320, 361, §§8.6–7, 537–38
modulation 2–3, 15, 34–35, 45–46, 68, 151–52, 269, 320, 347–48, 348f, 355–56, Ch. 8 (prelude), Ch.
8, 403–4, 422–25, 447, 481n.43, 483–84, 522, 524–26, 529n.3, 570–71
choral and scalar 375
chord-first 375
direct 306, 307f
garden-path 306, 307f
pivot-chord 306, 307f
Monk, Meredith 4, 31–33, 104
Anthem 32f
Monk, Thelonious
“Pannonica” 109–10, 109f, 527n.1
Monteverdi, Claudio 19–20, 23–24, 104, 210–11, 218n.18 §5.3, 237–38, 248–49, 257, 291, 292,
292f, 468
controversy with Artusi 226–27, 234
“Cosi sol d’un chiara fonte” (Madrigal 8.4) 194–95n.35, 522n.91
“Cruda Amarilli” (Madrigal 5.1) 296–97
“Longe da te, cor mio” (Madrigal 4.19) 335f
“Ohimè, se tanto amate” (Madrigal 4.11) 34 §5.3, 241–42, 519n.88
“Tu se’ morta” (Orfeo) 224–26
Vespro della Beata Vergine 263n.15
Monteverdi, Giulio Cesare 212n.8, 234
Moore, Alan 63n.19
Morgan, Robert 169n.18, 418n.15, 428n.27
Morley, Thomas 79–81
“April Is in My Mistress’ Face” 79, 80f, 81f, 173n.22, 227, 273n.29, 511–12
“Leave, alas, this tormenting” 391–93
Morris, Robert 94n.11
Mozart, W.A. 6, 69–70, 76–77, 141–44, 161–62, 166n.13, 188, 191–92, 206n.5, 206f, 226–27, 237,
291, 295, 311–12, 312f, 314–15, 314n.6, 329f, 333n.34, 334f, 336, 337f, 338f, 339f, 354–55, 380,
380n.15, 409–11, 444n.55, 453, 455–56, 463–65, 468–69, 472n.20, 515n.84, 569–70
Clarinet Concerto, K.622 144n.53
Piano Fantasy, K.475 178n.25
Piano Concerto, K.488 207f
Piano Sonata, K.279 193, 195f, 239, 240f, 451, 452f
Piano Sonata, K. 282 445f, 445–46, 508n.74
Piano Sonata, K. 283 14f
Piano Sonata, K. 284 409–11, 411f
Piano Sonata, K. 309 6f, 143f
Piano Sonata, K. 310 20n.30, 191f, 433f
Piano Sonata, K. 311 438, 441f
Piano Sonata, K. 331 417f
Piano Sonata, K. 332 381f, 382f, 514–15, 515f
Piano Sonata, K. 333 77f, 208f
Piano Sonata, K. 533 193f, 380n.14
Piano Sonata, K. 545 166f
String Quintet, K. 516 116n.18
Symphony no. 36, K.425 250f
 Symphony no. 40, K.550 176n.23, 408–09, 414
Violin Sonata, K. 304 439f
Muns, Lodewijk 434n.38
musica ficta 218, 260n.9, 290

Nápoles López, Néstor 308n.17

Narmour, Eugene 24n.44, 210n.2, 224n.27
Navia, Gabriel 507n.70
Neff, Severine 346n.54
neo-Riemannian theory 41, 42n.6, 50–51, 358–60, 358f, 530, 534–35, 563, 563f
Neumeyer, David 426n.24
Newcomb, Anthony 300n.64
Newman, Mark 308n.17
Newman, Randy
“When She Loved Me” 147–48, 147f
Newton, Isaac 26–27
Nicolas, Patrice 100n.4, 344n.53
Ninov, Dimitar 253n.3
Nobile, Drew 68n.26, 247n.66, 256n.7, 443–44, 444f
Nolan, Catherine 91n.7
nonharmonic consonance 212, 254–55, Ch. 7 (prelude), 570–71

Ockeghem, Johannes 93f, 114–16, 212n.8, 275–78, 276f, 277f, 295, 526, 528–29
Mass De plus en plus 116f
oenophile(s) 402
orbifold 547
ornament 303
Osborn, Brad 78n.39
OUCH theory 96–97 §3.6, 132, 136–37, 271
O-Zone
“Dragostea Din Tei” 68

Pachelbel, Johann 6 see also sequence

Palestrina, Giovanni Pierluigi da 1, 4, 22, 130–31, 161–62, 197f, 203, 210–11 §5.1, 222–23, 226,
227, 237–38, 259, 275–77, 276f, 278–79, 279f, 291–92, 296n.59, 385–86, 385f, 386f, 540–41
masses of 23–24, 24f, 90–91, 93f, 122, 126, 127f, 128, 130–31, 131f, 223f, 385f
Mass Aeterna Christi munera 220f
Mass Assumpta est Maria 217f, 220f
Mass Ave regina coelorum 99–100, 100f, 162f
Mass De beata Marie virginis 213f, 214f, 255f
Mass Descendit angelus Domini 116f, 222f, 222
Mass Dilexi quoniam 215f
Mass Fratres ego enim accepi 173n.22
Mass In duplicibus minoribus 214f
Mass In minoribus duplicibus 255f, 289f
Mass In semiduplicibus majoribus 220f
Mass Io mi son giovinetta 215f
Mass Papae Marcelli 34, 216f
Mass Quando lieta sperai 433f
Mass Spem in alium 114, 115f, 128–29, 130f, 392f
Mass Veni Sancta Spiritus 213f
 Stabat Mater 114n.15, 518–19
Palisca, Claude 71n.31, 73n.33, 221n.20, 235n.45
parallelism (thematic) 434, 444–45, 482, 516
Parncutt, Richard 272n.25
“Paranoid” (Black Sabbath) 61
Paul Revere and the Raiders
“Steppin’ Stone” 72–74
Payant, Lee
“Scarborough Fair” 85, 85f
Perren, Freddie
“I Will Survive” 138–40, 140f
Perry, Jeffrey 65n.21
Petrucci, Ottaviano 259–60, 262–63
Petzold, Christian
Minuet in G 354–55, 355f
philosophy §1.3, 33, 61, 126, 203, 218–20, 253–55, 303n.4, 403, 417–18, 432, 447, 449, 452–53,
455–56, 473, 570–71, 573–74
analytic 3
Picardy third 350, 353–54
Picasso
Guernica 449–50
Pinter, D. 56n.11, 267n.21
Piston, Walter 312n.2
Plato 2, 210–11, 220–21, 419, 527
polyphony 74–75, 96, 205–6, 291n.52
liturgical 282
in the Renaissance 2–3, 25–26 §6.4, 295, 384n.16, 520, 527
in Sacred Harp 23–24, 86n.53, 269
postminimalism 394–96
Powers, Harold 21n.35, 273n.26, 289n.48, 295n.58, 384n.16, 388n.26
pre-predominant 294
Principle of Musical Approximation 52–53, 176, 558–60
Prime Directive 5, 6, 9–10, 15, 96–97, 149–50, 172–73, 251–52, 401 §9.3, 516, 531–32, 558–60
protofunctionality 34, 257, §§6.1–3, 282, 286–88, 290, 294, 302, 313, 317–19
Prout, Ebeneezer 17–19, 18f, 221n.20
pseudochord 335–36, 336f, 347–48, 573
Pullum, Geoffrey 18n.23
Putnam, Hilary 19n.27

quadruple grouping structure 437–38

Quine, Willard van Orman 303n.4
Quinn, Ian 25n.45, 117n.19, 155n.2, 204n.1, 210n.2, 223n.24, 303–4, 536n.4

Rabinovitch, Gilad 29n.52, 103n.7

Rameau, Jean-Philippe 26–27, 33, 34, 201–2, 221n.20, 253, 259n.7, 260n.11, 313 §7.3, 326, 331n.28,
340, 345–46, 407–8
double emploi 68, 471–72
reduction 13, 35, Ch. 5 (prelude), §9.1, 416–18, 420n.17 §9.4, 480
nonharmonic 214, 215–16n.16, 218–20, 227, 228–29, 237, 238–39, 241, 245–46, 248–49, 250
scalar 330
schematic 175f, 480, 490–91
 Schenkerian 446–47
sequential §4.10, 202n.38, 490–91
Reef, John 209n.9
Rehding, Alexander 310n.18, 473n.23
Reich, Steve 12
Renaissance 2–3, 26, 33–34, 79, 83, 85, 86, 88–89, 114–16, 114n.15, 124–25, 124f, 125n.34, 128–29,
146–47, 154, 155, 162n.9, 172–73, 178–79, 181, 218, 222–24, 227, 231–32, 255, 263n.15, 271–
75, 288–90, 299, 344n.52, 350, 376, 407, 432, 520, 521–22, 526, 527, 570–71
cadence 30, 256n.9, 266, 271–72, 272f, 279–82
counterpoint 222–23, 230n.36
harmony 230n.37, 233, 277n.36, 279–82
inversional equivalence in 124, 161–62, 256, 266, 268, 274, 284–86
modes §8.6
sequences 79–81, 336–37, 339–40, 507–8 see also polyphony, schema
Renaissance (band) 82–83
repeating contrapuntal pattern 31–35, 155, §§4.2–3, 172f, 181–83, 184, 197f, 274, 321–22, 342, 357,
358f, 370f, 371, 525f
Reti, Rudolf 446n.58, 486n.54
retrofunctional 56–57, 61, 68, 74–75, 76–77, 77f §2.9, 84–85, 116, 128–29, 138–40, 144–48,
148n.62, 244, 262–63, 288–90, 321, 332–33, 333f, 419, 433, 443, 476n.31, 489–90, 507–8
Reynolds, Christopher 524
Ricci, Adam 158n.4
Rice, John 188n.29
Richards, Mark 20–3nn.30–1, 72n.32, 438n.44, 441n.50, 454n.8
Richter, Ernst 380n.14
Riemann, Hugo 34, 210n.3, 221n.20, 259n.7, 312n.2, 313, 325–26, 340n.47, 345–46, 475–76n.30,
485–86 see also neo-Riemannian theory
Rings, Steven 84n.48, 352n.62, 362n.2, 420n.18
The Rocky Horror Picture Show
“Time Warp” 71–72
rogue note 204–5, 228–29, 239n.54
Rohrmeier, Martin 135n.45, 435n.40, 436n.41, 527n.1
Roig-Francolí, Miguel 277n.37
Roman-numeral analysis 19–20, 21, 23, 25–26, 33, 34–35, 56–57, 57f, 101–2, 114n.15, 119, 223–24,
241, Ch. 6 (prelude), 291, Ch. 7 (prelude), 312–13, 343–44, 345–46, 350, 354, 355–56, 374–75,
461–63, 488–89, 508n.72, Appendix 4
Rosch, Eleanor 20n.29
Rosen, Charles 16n.17, 154, 320n.13, 380n.15, 403n.4, 448n.65, 448n.67
Rotem, Elam 263n.15
Rothenberg, David 394n.29
Rothgeb, John 249n.71, 250n.72, 253n.1, 254f, 254n.4, 255n.5, 256n.10, 333n.31
Rothstein, William 437n.43, 438n.45, 448n.67, 475–76n.30, 478n.38
Rowe, Dorothy 59
Russell, Jonathan 193n.34, 559n.7, 560n.9
Rydell, Bobby
“Forget Him” 147–48

Sacred Harp 23–24, 25f, 269

“Heavenly Spark” 109–10, 109f
Sailor, Malcolm 212n.9
Salzer, Felix 229n.35, 230n.37, 233n.41, 434–35n.39
Samarotto, Frank 59n.14, 486n.55
Sanguinetti, Giorgio 221n.20, 332n.30
Saslaw, Janna 303n.2
Satie, Erik
Gnossienne no. 3 384, 385f, 394–96
Savage, Patrick 47n.1, 310n.18
“Scarborough Fair” 82–83, 85, 85f
Scarlatti, Domenico 37, 188, 210–11, 224–26
Sonata in A minor, K. 3 39f
Schachter, Carl 208f, 237n.49, 254n.4, 354n.63, 355f
Scheidt, Samuel
“Gelobet siest du, Jesu Christ” 386–87, 387f
Schein, Johann
“Aus tiefer Noth” 388–89, 389f
schema 3, 5f, 8–9, 15, 16, 17 §1.5, 34–35, 74, 77n.37, 96, 102–3, 103f, 104–6, 107–9, 116, 151–52,
154, 157f, 159, 163–66, 171, 205–6, 207–9, 286–88, 290, 297–99, 311, 312–13, 331–32, 356, 371,
409–11, 416f, 432, 437–38, 449, 457, 467, 478–80, 478n.36, 482, 483, 486–88, 492n.60, 498–500,
501, 504–5, 510–11, 514–15, 516, 519–20, 522–24, 526, 528, 529–31, 565–66, 571
ascending-fifth 56–57, 67–68, 74, 76–77, 96–97, 114–16, 161–62, 222, 229–30, 336–39 §7.7,
§8.5, 429, 438n.47, §§10.1–2, 476f, 507–8, 517f
ascending subdominant 117–18, 223–24, 241–42, 284–86 §7.7, 430–31, 522
basic voice leading, dyadic §§3.1–3, 118–19, 124, 152–53 §4.2, 154n.4, 165–66, 174f, 187f, 198–
201, 228–29, 233–34, 286, 297, 300, 316–17, 318–19, §7.7, 379, 411n.5 §9.4, §§10.1–2, 529–30
cadential 349–50
Chase 172–75, 174f, 175f, 181–83, 201–2, 299f, 299, 349–50, 355–56, 390–91
contrary-motion 32f, 106 §4.7, §6.1, 331–32, 457, §§10.1–2, 478, 496–98, 501, 526, 530–31, 560
deep 67–68
descending-third 522–23
development 379–80
Do-Re-Mi 261–62, 410f, 528
Fenaroli 178–79, 221n.20, 261–62, 290, 341–42, 528, 559n.8
Fonte 102–3, 528
formal 342, 449, 453–54, 516–17
helicopter drop 376–78, 480
hidden 183f, 209
horn fifths 192n.33, 192f, 495–96, 500, 501–3, 528n.2, 559n.8
Indugio 207–9, 333n.34
Ludwig 33, 110–11, 186, 327–28, 428n.27, 429, 431, 432, 438–39 §10.1, 465–66, 489n.56, 490–
91, 522n.91
Ludwiging §10.2
Meyer 410f, 445n.56, 528
modulatory 361, 362, §§8.4–5, 537
Monte 102–3, 528
Morte 186, 186f, 207–9, 460n.8, 461–63
nonharmonic 220–21
omnibus 10, 10f, 132–34, 181, 181f
one more time 449
paired-sequence design 377–78
parallel motion 113, 327f, 335–36, 467, 468f, 555
 pedal tone 237, 242–44, 279–82, 335–36
Prinner 30, 116, 118–19, 178–79, 207–9, 261–62, 263, 266, 267f, 282–83, 286–88, 290, 297–99,
314n.6, 331–32, 341–42, 356, 430–31, 496, 528, 559n.8
protofunctional 260–62, 262f
Quiescenza 96–97, 102–3, 409–11, 460, 462f, 500–1, 528
Renaissance seventh 19–20, 214–17, 216f, 217f, 218–20, 222, 226, 261–62, 280–81n.44, 296–97,
304n.7, 334–35, 351–53, 436–37
Romanesca 6, 29–30, 29f, 528, 529–30
Schenkerian 426n.24
Sol-fa-mi, 410f, 528
Standing on the dominant 178–79, 377–78, 380–82, 434, 449, 454–55, 476n.33, 480, 488, 489–90,
491, 526
Standing on the predominant 333–34, 496
two-tiered 371
up and down the ladder 101–2, 152n.2, 323, 347–48, 355–56, 375, 377–78 §8.5, 390–91, 488–91,
501–3, 514–16
voice-leading 30, 369–71
Waldstein 63–64, 76, 511–12, 515n.84, 516, 520, 526, 531–32 see also cadence (converging),
clausula vera, fauxbourdon ROTO, Gjerdingen (Robert), pseudochord, repeating contrapuntal
pattern
Schenker, Heinrich 19, 21–22, 25n.45, 33, 34–35, 135n.45, 171, 221n.20, 249–50, 311, 331n.28,
340n.47, 343–44, 355–56, 355f, 407–9, 411–12, 416n.12, 418–19, 422, 424–25, 424f, 429n.28,
431, 431f, 431n.33, 432, 432n.35, 446–47, 482–83, 484n.51, 485–86, 489n.56, 490–91, 490f,
492n.60
Schenkerian 13, 20n.32, 21, 22n.37, 33, 34–35, 48–49, 103–4, 106, 119, 134–36, 148–49, 171, 202,
202f, 205n.3, 209, 249n.70, 261–62, 330n.25, 331n.28, 343–44, 354–56, 380n.14, 407, 409–11,
414–19, 425–26, 428–29, 431, 432, 444n.54, 446–48, 448n.65, 449, 476n.32, 480, 530, 536n.5,
542–43
Scherer, Josh 21n.34
Schindler, Anton 465n.11, 473n.24
Schmalfeldt, Jane 425n.23, 453n.7, 475n.26, 475n.28, 475–76n.30, 477n.34, 478n.36, 478n.38, 480,
480f
Schoenberg, Arnold 19, 21–22, 25n.45, 95f, 131n.41, 210n.2, 213n.11 §5.2, 227, 246, 248n.67,
346n.54, 414n.10, 444n.55, 565–66
“Angst und Hoffen” (Op. 15, no. 7) 158f, 565–66, 565f
The Book of the Hanging Gardens, Op. 15 524–26
“Farben” (Op. 16, no. 3) 95f, 558, 558f
Harmonielehre 221
Verklarte Nacht, Op. 4 524n.94
Violin concerto, Op. 36 535n.2
Schubert, Franz 23, 34, 63–64, 75, 88–89
C major string quintet, D. 956 227n.31
“Morgengruss,” D. 795.8 5f, 11n.11
Quartettsatz, D. 703 136, 458 §10.6, 519–20
Schubert, Peter 91n.8, 123n.28, 137n.47, 155n.1, 263n.15
Schuijer, Michiel 91n.7
Schumann, Clara Wieck
“Liebesfrühling” (Op. 12, no. 4) 118–19, 119f
Soirées Musicales, Op. 6 101–2, 102f
Schumann, Robert
 “Am Kamin” (Op. 15, no. 8) 254f
Dichterliebe, Op. 48 341, 379n.13
“Estrella” (Op. 9, no. 14) 144n.52
“Ich grolle nicht” (Op. 48, no. 7) 341, 342f
“Liebesfrühling” (Op. 37, no. 4) 118–19, 119f
Schütz, Heinrich 4, 75–76
St. Matthew Passion, SWV 479 390f, 390–91, 391f, 392–93
“Wie lieblich sind deine Wohnungen,” SWV29 74–75, 75f
Seals and Crofts
“Summer Breeze” 61
Seeger, Peggy 85
Segall, Christopher 156n.3, 524n.94
set theory 14–15, 37–38, 41, 91, 94–95, 96, 121–22 §8.7, 402, 533–34, 537–38, 542–43
sequence
Baroque 109–10, 334–35, 351–53
canonic 92, 92f, Ch. 4 (prelude), Ch. 4, 222, 227, 229–30, 230f, 263, 290, 422
chromatic 4–5, 53f, 69–70, 481, 515–16, 517f, 520
contrary-motion 3, 31, 31f, 32f, 105–6, 105f §4.7, 260–61, 426–27, 432, 451, §§10.1–2, 478, 489f,
496–98, 526
endless 12, 57–58, 64–66, 70–72, 284–86
melodic 134–36 §4.8, 205f, 208f
modulatory 366, 367–68, 367f, 374–75, 376–77
near 61, §§4.8–9, 262n.13, 324–25, 351, 377–78, 409–11, 426–27, 430–31, 432, 486–88, 512–13
Pachelbel 6, 7f, 29–30, 116, 118–19, 194, 231–32, 338–40, 346, 388–89, 453, 529–30
Renaissance 29–30, 30f, 79–81, 222, 274 see also schema
Sewell, Ian 205n.3, 209n.8, 414n.10
Shelley, Percy Bysshe
Ozymandias 452–53
Shepard, Roger 545n.3
Shepard-tone passacaglias 12 §2.5, 72–74, 171–72, 284–86, 286f, 324–25, 508–9
Shostakovich, Dmitri 4, 66n.22, 70–71, 71f, 76, 190n.32, 247–48
Ninth String Quartet 246–47, 247f
Simmons, Joseph 570n.2
Simon and Garfunkel 85
Simon, Herbert 15n.15
sliding (on a spiral diagram) 44, 50–51, 50n.3, 68–69, 90, 92, 97–98, 137, 357–58, 359f
Slottow, Steven 171n.21, 340n.47
Smith, Charles 149n.63, 447n.59
Smith, Elliot
“Rose Parade” 59
Smith, Patti 300–1
Smith, Peter 447n.59, 504–5n.68, 507n.70, 511n.79
SNAP system §5.1, 239n.54
Sommer, Heinz-Dieter 495n.61
Sousa, John Philip 448–49
sonata (form) 2, 15–16, 17, 20, 35–36, 82, 361–62, 376–77, 380, 402, 403–4, 443, 446–47, 448–50,
453–56, 495, 504–5, 508–9, 516–17, 519–20, 526 see also caesura
spacing (in chordal steps) Ch. 2 (prelude), 60, 70n.29, 94, 95f, 103–4, 106, 107f, 113, 124–25, 128,
149–50, 162–63, 165n.11, 195–96, 197f, 495–96, 536–37, 540n.8, 553
Sprick, Jan Philipp 158n.4, 340n.47
“Stand by Your Man” (Sherrill and Wynette) 82–83
Starr, Ringo 78
Status Quo
“Pictures of Matchstick Men” 61
Stein, Beverly 290n.50
Steinbach, Peter 70n.28
Stephenson, Kenneth 17n.21, 48n.2, 57n.13
Stravinsky, Igor 4, 6–8, 88–89, 159–60, 247–48, 555n.4
The Firebird 160f, 384, 525f
The Rite of Spring 8f, 240, 241f, 250–51, 557f, 558–60
Symphony of Psalms 388
Strozzi, Barbara
“L’amante modesto” (Op. 1, no. 13) 171–72, 173f
“Il Contrasto dei Cinque Sensi” (Op. 1, no. 14) 170f
“Gli amanti falliti” (Op. 1, no. 16) 172f
structured arpeggiation 15, 188–89, 189f, 192, 204–5, 420f
Strunk, Oliver 212n.8
Stuckenschmidt, Hans 560n.10
subjectivity 23, 74, 245–46, 256, 308, 354–55, 362n.2, Ch. 9 (prelude), 439–40, 442, 446–47, 570–
71
Subotnik, Rose 421n.19
Sutcliffe, W. Dean 210n.1
Swinden, Kevin 256n.8
“Sympathy for the Devil” (Rolling Stones) 61–63, 62f, 78
synthesizer 68, 404–5, 405f
Szymanowska, Maria
Polonaise no. 1, from 18 Dances of Different Genres 169–70

Tallis, Thomas
Spem in alium 91n.8, 136–37, 138f
Taneyev, Sergei 155–57, 159n.5, 555n.1
Taruskin, Richard 212n.8, 256n.9, 257n.2, 258n.4, 272n.24, 282n.46, 448n.65
Taylor, Benedict 75n.35
Taylor, Steve 151n.1, 240n.57, 460n.9
Telesco, Paula 10n.9
Temperley, David 29n.52, 56n.11, 61n.18, 63–64n.20, 67n.24, 78n.39, 209n.8, 247n.66, 303n.3,
320n.12, 333n.31, 418n.15, 426n.24, 439n.48, 486n.53
Thayer, Alexander 473n.24
Theune, Michael 233n.40
thinking within the chord 31–33 §3.7, 179–81, 231, 495–96, 497f, 559n.8
Thomas, Keith 402n.1
“Three Blinde Mice” (traditional) 157–58, 157f
Tierney, Adam 47n.1
Tinctoris, Johannes Ch. 3 (prelude), 220–21
Liber de arte contrapuncti 87–88
Tinctoris Transform Ch. 3 (prelude), 110–11, 119–20, 141, 172f, 178f, 184–86, 187–88, 459n.5, 467,
534–35, 542–43, 543f, 549n.4, 564f, 565–66
Tomlinson, Gary 235n.44
tonicization 132–34, 151–53, 193, 336–37, 347–48, 349–50, 380, 477–78, 481, 500, 520
Tovey, Donald Francis 475n.27, 478n.39, 482–83, 485–86, 495n.61
transcription 83n.47, 85, 404
Trevathan, Charles
“Bully Song” 145f
triad see chord
Trippet, David 518n.86
Tromboncino, Bartolomeo 271
“Ah partiale e cruda morte!” 270, 270f, 271f
“Se ben hor non scopro el foco” 263, 263f
Turner, Georgia 83–84
Tymoczko, Dmitri 16n.18, 41n.4, 42n.6, 44n.8, 50n.4, 70n.29, 155n.2, 367n.4, 398n.30, 457–58,
530n.4, 533n.1, 536n.4, 537n.6, 543n.12, 552n.7, 553n.9, 555n.3, 563n.15, 563n.16, 566n.19
A Geometry of Music 13–14, 44n.8, 48–49, 50n.4, 55n.8, 75n.34, 91, 97n.2, 110n.12, 148n.59,
148n.61, 155n.2, 198, 213n.11, 280n.43, 312–13, 334n.37, 335n.39, 374n.9, 375, 398n.30,
416n.13, 421n.19, 422n.21, 430n.29, 533n.1, 536n.4, 537n.6, 553n.9, 566n.22
The Thousand Faces of Form 149–50
Wheels within Wheels 396–97, 397f
Tyner, McCoy 4, 9–10, 9f, 114–16, 560
“Pursuance” (A Love Supreme) 248f
“Passion Dance” (The Real McCoy) 9f, 561f

Vande Moortele, Steven 475n.27

VanderWaal, Grace
“So Much More Than This” 440f
Van Ronk, Dave 72–74, 84–85
Varricchio, David 257n.2
Vial, F.G. 371–72, 372f
Vicentino, Nicola 211n.6
Victoria, Tomas Luis da 122, 277–78, 278f
Victorian 227
voice
abstract 13–14, 25f, 132–34, 204n.2, 330, 362, 363–64, 363f, 365, 366–68, 372f, 445–46, 541–42
surface 10, 13–14, 132–36, 137, 188–89, 204–5, 207n.6, 330, 360, 365, 367–68, 407, 409, 411,
424–25, 426–27, 445–46, 527, 528, 540–41, 559n.7 see also melodic slot
voice exchange 34–35, 45–46, 79, 81f §3.3, §3.7, 134–36, 151–52, 152f, 164f §4.3, 341f, 426–27,
427f, 429, 432, 458–59, 505–7, 530, 536–37, 538f, 540, 542–43, 550–53
pairwise 125f, 164f, 536, 538f, 542–43, 550–53
voicing Ch. 2 (prelude), Ch. 3 (prelude), 120, 128, 561f
von Hippel, Paul 23–24, 24n.44, 130–31, 131n.41

Wagner, Napthali 84n.49

Wagner, Richard 34, 35–36, 88–89, 516
Lohengrin 136, 458, 504 §10.7
Parsifal 147n.57
Tristan und Isolde 79n.43, 136f, 137f, 180f, 460n.9, 524–26
Waltham-Smith, Naomi 158n.4, 369n.6
Warfield, Patrick 449n.68
Watkins, Holly 407n.1
Watson, Doc 83–84
Weber, Gottfried 18n.23, 22n.39, 131, 207n.7, 211n.4, 253n.3, 303–4, 304n.5, 304n.6, 331n.28,
340n.47
Webster, James 407n.1, 475n.29, 504–5n.68, 507n.70
Wellwood, Alexis 572n.6
Westergaard, Peter 13n.14, 223n.23
White Stripes
“The Air Near My Fingers” 61, 62f, 67–68
The Who
“I Can See for Miles” 61
“I Can’t Explain” 61–63, 78
Wiering, Franz 275n.31
Wittgenstein, Ludwig 19–20, 403n.3, 572n.9
Wolff, Christoph 154n.4
Woolf, Virginia
To the Lighthouse 449–50
Wootton, David 22n.37, 235n.44, 273n.27

Yellin, Victor 10n.9, 188n.29

Young, Neil 76
“Helpless” 47f, 57–58, 64, 70–71, 330, 332–33, 528–29
“Keep on Rockin’ in the Free World” 72–74, 73f
Yust, Jason 53n.7, 148n.61, 202n.38, 209n.10, 260n.11, 316n.8, 330n.25, 367n.4, 429n.28, 447n.61,
516n.85, 555n.3

Zarlino, Gioseffo 123, 334n.35, 385n.20, 386, 388

Zbikowski, Lawrence 573n.10