Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

0521761921

Download as pdf or txt
Download as pdf or txt
You are on page 1of 262

Parentheticals in Spoken English

Taking both an empirical and a theoretical view of the prosodic ­phrasing


of parentheticals in English, this book reviews the syntactic and ­prosodic
literature on parentheticals along with relevant theoretical work at the
syntax–prosody interface. It offers a detailed prosodic analysis of six
types of parentheticals; full parenthetical clauses, non-restrictive relative
clauses, nominal appositions, comment clauses, reporting verbs and ques-
tion tags, all taken from the spoken part of the British Component of the
International Corpus of English. To date, the common assumption is that,
by default, parentheticals are prosodically phrased separately, an assump-
tion which, as this study shows, is not always in line with the predictions
made by current prosodic theory. The present study provides new empir-
ical evidence for the prosodic phrasing of parentheticals in spontaneous
and semi-spontaneous spoken English, and offers new implications for a
theory of linguistic interfaces.

Nic ol e Dehé is a Professor of General Linguistics at the University of


Konstanz.
STUDI E S I N E NG L I S H L A N G UAG E

General editor
Merja Kytö (Uppsala University)

Editorial Board
Bas Aarts (University College London), John Algeo (University of Georgia), Susan Fitzmaurice
(University of Sheffield), Christian Mair (University of Freiburg), Charles F. Meyer (University of
Massachusetts)

The aim of this series is to provide a framework for original studies of English, both present-day and
past. All books are based securely on empirical research, and represent theoretical and descriptive con-
tributions to our knowledge of national and international varieties of English, both written and spoken.
The series covers a broad range of topics and approaches, including syntax, phonology, grammar,
vocabulary, discourse, pragmatics and sociolinguistics, and is aimed at an international readership.

Already published in this series:

Daniel Schreier, Peter Trudgill, Edgar Schneider and Jeffrey Williams: The Lesser-Known Varieties
of English: An Introduction
Hilde Hasselgård: Adjunct Adverbials in English
Raymond Hickey: Eighteenth-Century English: Ideology and Change
Charles Boberg: The English Language in Canada: Status, History and Comparative Analysis
Thomas Hoffmann: Preposition Placement in English: A Usage-based Approach
Claudia Claridge: Hyperbole in English: A Corpus-based Study of Exaggeration
Päivi Pahta and Andreas H. Jucker (eds.): Communicating Early English Manuscripts
Irma Taavitsainen and Päivi Pahta (eds.): Medical Writing in Early Modern English
Colette Moore: Quoting Speech in Early English
David Denison, Ricardo Bermúdez-Otero, Chris McCully and Emma Moore (eds.): Analysing Older
English
Jim Feist: Premodifiers in English: Their Structure and Significance
Steven Jones, M. Lynne Murphy, Carita Paradis and Caroline Willners: Antonyms in English: Con-
struals, Constructions and Canonicity
Christiane Meierkord: Interactions across Englishes: Linguistic Choices in Local and International Contact
Situations
Haruko Momma: From Philology to English Studies: Language and Culture in the Nineteenth Century
Raymond Hickey (ed.): Standards of English: Codified Varieties Around the World
Benedikt Szmrecsanyi: Grammatical Variation in British English Dialects: A Study in Corpus-Based
Dialectometry
Daniel Schreier and Marianne Hundt (eds.): English as a Contact Language
Bas Aarts, Joanne Close, Geoffrey Leech and Sean Wallis (eds.): The Verb Phrase in English: Investi-
gating Recent Language Change with Corpora
Martin Hilpert: Constructional Change in English: Developments in allomorphy, word formation, and syntax
Jakob R. E. Leimgruber: Singapore English: Structure, Variation and Usage
Christoph Rühlemann: Narrative in English Conversation
Dagmar Deuber: English in the Caribbean: Variation, Style and Standards in Jamaica and Trinidad
Jock Onn Wong: English in Singapore: A Cultural Analysis
Eva Berlage: Noun Phrase Complexity in English
Nicole Dehé: Parentheticals in Spoken English: The Syntax–Prosody Relation

Earlier titles not listed are also available


Parentheticals in Spoken
English
The Syntax–Prosody Relation

NICOLE DEHÉ
University of Konstanz
University Printing House, Cambridge CB2 8BS, United Kingdom

Published in the United States of America by Cambridge University Press, New York


Cambridge University Press is part of the University of Cambridge.
It furthers the University’s mission by disseminating knowledge in the pursuit of education,
learning and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9780521761925
© Nicole Dehé 2014
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2014
Printed in the United Kingdom by Clays, St Ives plc
A catalogue record for this publication is available from the British Library
Library of Congress Cataloguing in Publication data
Dehé, Nicole.
Parentheticals in spoken English : the syntax prosody relation / Nicole Dehé.
  pages  cm. – (Studies in English language)
Includes bibliographical references and index.
ISBN 978-0-521-76192-5 (hardback)
1.  English language–Parenthetical constructions.  2.  Grammar, Comparative and
general–Parenthetical constructions.  3.  English language–Spoken English.  4.  English
language–Syntax.  I.  Title.
P293.35.D34 2014
425–dc23
2014004409
ISBN 978-0-521-76192-5 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
Contents

List of figures page vii


List of tables ix

1  Parentheticals in English: introduction 1


1.1 Parentheticals – a motley crew 2
1.2 The data 13
1.3 This study 17
2 The syntax and prosody of parentheticals 18
2.1 The syntax of parentheticals and the relation between
parenthetical and host 18
2.2 Prosodic features of parentheticals 30
2.3 The syntax and prosody of six selected types of
parentheticals 44
2.3.1 Full parenthetical clauses 44
2.3.2 Non-restrictive (appositive) relative clauses 47
2.3.3 Nominal appositions 54
2.3.4 Comment clauses 64
2.3.5 Reporting verbs 74
2.3.6 Question tags 78
2.4 Concluding remarks 84
3 Parentheticals, intonational phrasing and prosodic theory 87
3.1 Intonational phrasing: the Tone Group and the
Intonational Phrase 88
3.2 Prosodic separation and the correspondence between
the Tone Group and the Intonational Phrase 97
3.3 The syntax–prosody interface: the relation between
syntactic and prosodic constituents 104
4 Data analysis, results and discussion 116
4.1 Data source and data retrieval 117
4.1.1 Full parenthetical clauses 118
4.1.2 Non-restrictive relative clauses 119

v
vi  Contents

4.1.3 Nominal appositions 120


4.1.4 Comment clauses 122
4.1.5 Reporting verbs 123
4.1.6 Question tags 125
4.1.7 General notes on data retrieval 126
4.2 Data treatment 128
4.3 Analysis and results 136
4.3.1 Full parenthetical clauses 137
4.3.2 Non-restrictive relative clauses 145
4.3.3 Nominal appositions 154
4.3.4 Comment clauses 175
4.3.5 Reporting verbs 193
4.3.6 Question tags 205
4.4 Summary 226
5 Final discussion 227

References 234
Index 247
Figures

Figure 2.1 E  xample (2.38): parenthetical insertion with prototypical


parenthetical properties page 37
Figure 2.2 Example (2.43)b: prosodic, but not syntactic, parenthesis 41
Figure 3.1 Non-recursive and recursive pitch structures 95
Figure 3.2 Example (3.12), from Beckman and Pierrehumbert
(1986: 295): two possible realizations of an accented
sentence-final QT 100
Figure 4.1 Example (4.25): example analytic procedure 132
Figure 4.2 Example (4.32): full parenthetical clause, prosodic separation 139
Figure 4.3 Example (4.39): full parenthetical clause, prosodic integration 143
Figure 4.4 Example (4.46): NRRC, prosodic separation 148
Figure 4.5 Example (4.47), phrasing as in (4.51) 150
Figure 4.6 Example (4.59): N-APP, prosodic separation 157
Figure 4.7 Example (4.60): N-APP with pronominal anchor,
prosodic separation 158
Figure 4.8 Example (4.61): N-APP phrased together with anchor 159
Figure 4.9 Example (4.66): N-APP phrased together with anchor
and following host material 162
Figure 4.10 Example (4.83): CC, prosodic separation 178
Figure 4.11 Example (4.84): CC phrased in a CPD 179
Figure 4.12 Example (4.85): CC with nuclear prominence, phrased
together with preceding host material 181
Figure 4.13 Example (4.88): CC unstressed and phrased together
with following host material 184
Figure 4.14 Example (4.89): CC phrased together with preceding and
following host material, unstressed 185
Figure 4.15 Example (4.90): CC phrased together with preceding and
following host material, part of hesitant stretch 187
Figure 4.16 Example (4.94): CC, an alternative analysis to that of
Kaltenböck (2010: 252) 192
Figure 4.17 Example (4.100): simple RV, prosodic separation 196
Figure 4.18 Example (4.104): simple RV, encliticized 199
Figure 4.19 Example (4.105): complex RV, encliticized 201
vii
viii  List of figures

Figure 4.20 Example (4.112): QT, prosodic separation 208


Figure 4.21 Example (4.113): QT, prosodic separation 209
Figure 4.22 Example (4.121): QT unstressed, prosodic integration
with preceding host material 214
Figure 4.23 Example (4.122): QT unstressed, prosodic integration
with preceding and following host material 215
Figure 4.24 Example (4.123): QT unstressed, prosodic integration
with preceding and following host material 217
Tables

Table 1.1  orpus references and text categories (see Nelson, Wallis
C
and Aarts 2002) page 14
Table 2.1 Prosodic phrasing of parentheticals: predictions 85
Table 4.1 Summary of items 118
Table 4.2 Distribution numbers of QTs according to auxiliary verb
in the tag (be, do, have, modal) and type (reversed polarity
tag: pos-neg, neg-pos; constant polarity tag: pos-pos,
neg-neg) 126
Table 4.3 Patterns of prosodic phrasing of nominal appositions
(N-APPs) 156
Table 4.4 Cross-tabulation of N-APPs according to phrasing and type 167
Table 4.5 Cross-tabulation of N-APPs according to phrasing and
syntactic function 167
Table 4.6 Cross-tabulation of N-APPs according to phrasing and
semantic relation between anchor and N-APP 168
Table 4.7 Cross-tabulation of N-APPs of semantic type equivalence
according to phrasing and type of N-APP 168
Table 4.8 Cross-tabulation of N-APPs according to phrasing and
position 168
Table 4.9 Patterns of prosodic phrasing of comment clauses (CCs) 176
Table 4.10 Phrasing/accent patterns of RVs according to position
(final, non-final) 195
Table 4.11 Phrasing/accent patterns of RVs according to complexity
(simple, complex) 195
Table 4.12 Cross-tabulation of QTs according to phrasing and
intonation contour 220
Table 4.13 Cross-tabulation of QTs according to phrasing and polarity 220
Table 4.14 Cross-tabulation of QTs according to phrasing and position 221
Table 4.15 Cross-tabulation of QTs according to phrasing and verb type 221
Table 4.16 Prosodic phrasing: summary of results for six types of
parentheticals 226

ix
1 Parentheticals in English: introduction

A monograph on parentheticals should reasonably begin with a definition of


the phenomenon under investigation. Several definitions of parentheticals
have already been offered in the literature, all of which have in common
that a parenthetical is considered a linguistic entity which is linearly inte-
grated in another linguistic structure but is unrelated to the surrounding
linguistic material in one way or another, i.e. in terms of syntactic struc-
ture, semantic meaning and/or intonation. For example, Burton-Roberts
(2006: 179)  maintains that a parenthetical (P) is “an expression of which
it can be argued that, while in some sense ‘hosted’ by another expression
(H), P makes no contribution to the structure of H”, i.e. it is structurally
unrelated. Bussmann (1996: 349) makes a similar point in defining a paren-
thetical as an “[e]xpression (word, phrase, clause) inserted into a sentence
from which it is structurally independent: Her new boy-friend  – his name
is Jacob – will be coming over tonight.” She thus touches on the diversity in
structural complexity: parentheticals can be anything from a single word to
a full clause. Biber et al. (1999: 1067) add the aspect of meaning, defining a
parenthetical as “an interpolated structure … a digressive structure (often
a clause) which is inserted in the middle of another structure, and which is
unintegrated in the sense that it could be omitted without affecting the rest
of that structure or its meaning”. Taglicht (1998: 195) defines a parenthet-
ical provisionally as a non-initial and non-final “syntactic node for which
the grammar specifies no function in relation to any sister node”. De Vries
(2012a: 153), maintaining that “it is far from obvious how to define paren-
thesis either syntactically or phonologically, even though everyone recog-
nizes it intuitively”, offers the following working definition: “parenthesis is
a grammatical construction type that involves a message that is presented
or perceived as secondary with respect to the host, where message covers
propositions, modal propositions, questions, metalinguistic comments, and
so on”, leaving open what is included in “and so on”. From a semantic point
of view, Potts (2005) argues that parentheticals are perfect illustrations of
conventional implicatures (CIs) as formulated by Grice (1975). According
to Potts (2005, 2007), their content is speaker oriented and discourse-new,
but de-emphasized in the given context; it is outside the regular content of
1
2  Parentheticals in English: introduction

the utterance; it is not contextually determined, but part of the conventional


meaning of the words. Following Potts (2005), Kluck (2011: 229) includes
speaker orientation in her (working) definition of parentheticals and states
that parentheticals are expressions which are “structurally and semantic-
ally independent of [their] host” and express “speaker-oriented content”.
Moreover, it has been observed and often been taken for granted that par-
entheticals are “marked off from their hosts by some form of punctuation
in writing or special intonation contour in speech” (Burton-Roberts 2006:
180). Dashes, commas or parentheses do the job in writing, while these
devices correspond to tonal and temporal prosodic cues in spoken language.
It follows from these definitions that parentheticals are a relevant phenom-
enon in the syntax, phonology and meaning components of the grammar,
as well as at the respective interfaces. It is striking, however, that authors do
not usually commit themselves to a final definition of the phenomenon and
at the same time they fail to find previous definitions convincing enough to
use them in their own work. In this study, I will follow recent accounts by
Potts (2005) and in particular by de Vries (2007, 2012a, b), as outlined below.
To begin, the following section will provide an overview of expressions
considered parentheticals in previous work and introduce ways of classifi-
cation. Throughout the book, parentheticals are in italics. The sources of
the examples are given in parentheses. Following common practice, the sen-
tence/utterance linearly integrating the parenthetical will be referred to as
the ‘host’, ‘host utterance’ or ‘host sentence/clause’ throughout this study.

1.1  Parentheticals – a motley crew


Elements which have been considered parentheticals in previous literature
form a heterogeneous set and there is no general agreement as to the exact
delimitation of a potential class of parentheticals. As Dehé and Kavalova
(2007: 1)  put it, they are “a motley crew”, although, as de Vries (2012a:
153) notes, “everyone recognizes [parenthesis] intuitively”. The examples
in (1.1) through (1.19) provide an overview of this heterogeneous family.
While this list does not claim to be exhaustive, it does illustrate how par-
entheticals vary in length and complexity, syntactic category and projec-
tion level, and in function. According to Espinal (1991: 727), none of the
parenthetical expressions are “parentheticals per se; rather, they are said
to be like parentheses because they are identified as independent syntactic
constituents or, more generally, as independent syntactic structures within
another syntactic structure”. According to de Vries (2012a: 153), the vari-
ous types have in common the particular way in which they are syntactic-
ally related to their host, which he refers to as ‘parenthetical merge’ (see
Chapter 2 below).
The heterogeneous class of parentheticals includes main clauses and con-
tent clauses (e.g. declarative or interrogative, see (1.1)) which may or may
Parentheticals – a motley crew  3

not be introduced by a connector (see (1.1) and (1.2), respectively), ellip-


tical clauses (see (1.3)), adverbial clauses (see (1.4)) and non-finite clauses
(see (1.5)), non-restrictive (appositive) relative clauses (NRRCs; see (1.6))
and nominal appositions (see (1.7)), lexical phrases of categories AP, PP,
NP (see (1.8)), interrogative parentheticals (see (1.9)), question tags (see
(1.10)), statement tags (see (1.11)a) and imperative tags (see (1.11)b), report-
ing verbs (see (1.12)), comment clauses (CCs; also known as parenthetical
verbs; see (1.13)), vocatives (see (1.14)), sentence adverbs (see (1.15)), other
one-word expressions (see (1.16)), and interjections and filled pauses (see
(1.17)). Finally, right-node raising constructions (see (1.18)) and syntactic
amalgamation (see (1.19)) have been analysed in terms of parenthesis.1
(1.1) Clauses
a. When we were on holiday – that reminds me, I must pick up the pho-
tos – we saw so many interesting places. (Wichmann 2001: 178)
b. Newcastle and North you find uhm there’s a marvellous walled
garden I don’t know where it is with hyacinths (Dehé 2009: 579;
ICE-GB: s1a-065 #298)
c. Well esterases are able in organic solvents to carry out a number
of useful can you hear me all right now? organic uhm processes to
produce things like food products (Dehé 2009: 579; ICE-GB:
s2a-034 #13)
(1.2) Clauses introduced by a conjunction
a. Her account – and I must say I’m attracted to it – suggests that
we have to re-think the relationship between meaning and truth.
(Blakemore 2006: 1671)
b. If he checks my story  – and he probably will  – I’ll be sacked.
(Huddleston and Pullum 2002: 1361)
c. Ames, as the FBI eventually discovered, was a spy. (Potts
2002b: 624)
d. The Hawks will win, or at least so I’ve been told, by at least 10
points. (Peterson 1999: 232)
(1.3) Elliptical clause
For those of us who remember nineteen sixty-five one or two of our
listeners may Tory party leadership contests used to be as the cardi-
nals in Rome and leaders would emerge (Dehé 2009: 570; ICE-GB:
s1b-024 #1)
(1.4) Adverbial clauses
a. John smokes, ’cos his place is full of dirty ashtrays. (Haegeman
1991: 232)

1
Not all authors writing on parentheticals include all types listed here in their class of paren-
theticals. For example, Altmann (1981: 64) explicitly excludes vocatives and interjections.
4  Parentheticals in English: introduction

b. I’ve just received the expected letter, if that makes you feel any
better. (Espinal 1991: 726)
c. My idea, if you really want to know, was to treat the phenom-
enon as a conventional implicature. (Blakemore 2006: 1671)
(1.5) Non-finite clauses
a. Having read the report, Max was sure he had nothing to worry
about. (Huddleston and Pullum 2002: 1360)
b. The most fundamental of all parental wishes to educate our chil-
dren in our own morality is indoctrination and a denial of their
free development (ICE-GB: s2b-029 #147)
(1.6) Non-restrictive relative clauses
a. He shouldn’t have pushed that kid, who is so conscientious, out
that door. (Emonds 1979: 226)
b. The singer, who believes she is a rock diva trapped in the body of
a pop star, launched into her new single … (Loock 2010: 83;
corpus example)
c. so the word disability which is this nebulous thing that exists some-
where between the two people has a part on each side (Dehé 2009:
570; ICE-GB: s1a-001 #59)
(1.7) Nominal appositions
a. A university lecturer, Dr Brown, was arrested for the crime.
(Huddleston and Pullum 2002: 1358)
b. A surprise present, a bouquet of roses, was delivered to my door.
(Huddleston and Pullum 2002: 1357)
c. John McClave, my neighbor, is a nice guy. (Heringa 2011: 1)
(1.8) Lexical phrases: AP (a), PP (b), NP (c)
a. The secretary well-mannered as anybody will present an apol-
ogy. (Espinal 1991: 726)
b. Her husband had always been quite irresponsible. Bill on the
contrary appeared to be completely trustworthy. (Espinal 1991:
727)
c. Robert, no genius, is applying for a scholarship to Harvard.
(Huddleston and Pullum 2002: 1358)
(1.9) Interrogative parentheticals
a. Is he going do you know/think (Mittwoch 1979: 402–403)
b. Isn’t that a bit of an imposition don’t you think (Mittwoch
1979: 409)
c. Is it safe, would you say? (Huddleston and Pullum 2002: 895)
(1.10) Question tags
a. He suffered great mental distress didn’t he after the war (Dehé
and Kavalova 2007: 3; ICE-GB: s1b-032 #164)
Parentheticals – a motley crew  5

b. Oh it’s not very valuable is it? (Tottie and Hoffmann 2006: 283)
c. Lucy can play the viola, can she? I didn’t know that. (McCawley
1998: 501)
(1.11) Statement tags (a) and imperative tags (b)
a. John will go to Spain, he will. (Knowles 1980: 380)
b. Turn out the light, won’t you? (Knowles 1980: 380)
(1.12) Reporting verbs
a. The reason for the Prime Minister’s resignation she said was to
enable Cabinet colleagues to enter the ballot (Dehé 2009: 570;
ICE-GB: s2b-020 #13)
b. The Hawks will win, says John, by at least 10 points. (Peterson
1999: 233)
(1.13) Comment clauses2
a. There were no other applicants, I believe, for that job. (Quirk
et al. 1985: 1113)
b. Charles wouldn’t, I imagine, have done such a thing. (Nespor
and Vogel 1986: 190)
c. Only if, I fear, we work like dogs, will we be able to save this
company. (Asher 2000: 39)
d. John and, I think, Mary will play the next round. (Peterson
1999: 238)
(1.14) Vocatives (noun phrases)
a. Today’s topic, ladies and gentlemen, is Nuclear Magnetic
Resonance. (Espinal 1991: 727)
b. If Mary had tutored him, John, Bill would have passed.
(Burton-Roberts 2006: 180)
c. Jean, could you check the mail for me (Hock and Dutta 2010: 2)
(1.15) Sentence adverbs
a. He is, unfortunately, ill. (Urmson 1952: 486)
b. He described himself, engagingly, as an economist on leave.
(Wichmann 2001: 179)
c. Frankly, my dear, I don’t know how to handle that. (Espinal
1991: 726)
d. I don’t agree with you, personally. (Astruc-Aguilera and Nolan
2007a: 242)
(1.16) One-word expressions (other than sentence adverbs)
a. I’ve been dreaming of winning a gold medal for what 20 years
now (Dehé and Kavalova 2006: 289)
2
Comment clauses and reporting verbs are elliptical clauses in some sense. The missing
object of the verb corresponds to the host clause (see also Peterson 1999: 233 and Sections
2.3.4 and 2.3.5 below).
6  Parentheticals in English: introduction

b. I mean it wouldn’t be very proper just to go and – well- “live”


with him. (Nosek 1973: 112)
c. In that case – yes – in that case I think I can speak freely (Nosek
1973: 106)
d. Certainly in an area like that successful known prosecutions
of police officers for racism I think will do more to improve
the confidence of the black community in the complaints
procedure than say a hundred plus programmes uh put on
by the Metropolitan Police (Dehé 2007: 273; ICE-GB: s1b-
033 #125)
(1.17) Interjections and filled pauses
a. He is oh! so smart. (de Vries 2007: 204)
b. My knowledge of this sort of thing, I admit, comes chiefly from
the – um – popular press. (Nosek 1973: 106)
(1.18) Right node raising
a. Amanda is, or at least she used to be, my best friend. (Peterson
1999: 232)
b. Amanda is, and there is no doubt in my mind that she always will
be, my best friend. (Peterson 1999: 232)
(1.19) Syntactic amalgamation
a. John invited you’ll never guess how many people to his party.
(Lakoff 1974: 321)
b. John is going to I think it’s Chicago on Sunday. (Lakoff
1974: 324)
The examples in (1.20) illustrate that it is not uncommon to find combin-
ations of different types of interpolations next to each other, as well as mul-
tiple occurrences of the same type within one host sentence, and we also
find syntactically complex interpolations, which might in turn be the host
for another interpolation. The examples in (1.21) illustrate that strings
which may not be obvious candidates for parentheticals in the syntax may
be marked prosodically and contribute a secondary communication such
that they function as parentheticals. In (1.21)a, a restrictive relative clause
is marked prosodically as a parenthetical, in (1.21)b it is a conjoined noun
together with a comment clause which stands out prosodically (rather than,
for example, the comment clause on its own). See also the discussion of
example (2.43)b/Figure 2.2 in Section 2.2 below.
(1.20) Multiple and complex parenthesis
a. For over three months – since July, Colonel! – we’ve been hav-
ing poisoned darts thrown at our backs. (Nosek 1973: 105)
b. Marcia, who you wanted to meet, didn’t you?, has just arrived.
(Loock 2010: 11)
Parentheticals – a motley crew  7

c. John will finally marry – I should imagine (but that’s his lookout)
he will have a church wedding – next Sunday. (Espinal 1991: 727)
d. Professionally a lawyer, that is to say associated with dignity,
reverse, discipline, with much that is essentially middle-class, he
is compelled by an impossible love to exhibit himself dressed
up, disguised  – that is, paradoxically, revealed  – as a child,
and, worse, as a whore masquerading as a child. (Huddleston
and Pullum 2002: 1356; my italics according to Huddleston
and Pullum’s explanations)
e. When the opportunity comes, and it will, I’ll bet, sooner than you
expect, you’ve got to be ready to grab it. (Bolinger 1989: 186)
f. They both skip around their cities around Los Angeles in Frank’s
case or London in Ron’s case dropping in and out of schools of
architecture dropping in and out of parties dropping in and out of
architects’ gatherings and not least and I think this is very import-
ant dropping in to other people’s studios (ICE-GB: s2a-040 #41)
g. A long-lived scar on the American psyche second only I suspect
to the one marked Vietnam bore the name of Iran (ICE-GB: s2b-
034 #101)
(1.21) Syntactic non-parenthesis marked prosodically as parenthesis
a. The story I told you – that you enjoyed so much, remember? – was
one I made up for our daughter when she was six. (Bolinger
1989: 193)
b. And one must also remember that uh the same Arnold Bax
has written poetry and I think plays under the pseudonym of
Dermot O’Brien (ICE-GB: s1b-032 #114)
The interruption of the main clause/proposition caused by parenthet-
ical insertion is sometimes marked by backtracking, i.e. repetition of some
part of the utterance, leading back to the main proposition; see (1.22), the
repeated material is underlined. Biber et al. (1999: 1067f) describe the use
of backtracking as a marker of disfluency and a direct consequence of online
processing. Kavalova (2007) finds backtracking in 9 per cent of her set of 70
and-parentheticals from the ICE-GB and DCPSE corpora.3 She sees it as
“a device to remind the hearer that what he has just heard was only a quick
diversion, an aside, and is not to be treated as part of the main utterance.
It also assists the hearer to quickly resume the interpretation of the main
utterance, i.e., to pick it up where it was left” (Kavalova 2007: 160).

3
The Diachronic Corpus of Present-day Spoken English (DCPSE), developed by the
Survey of English Usage at University College London, contains more than 400,000 words
from the ICE-GB and 400,000 words from the London-Lund Corpus. See the DCPSE
website for more information: www.ucl.ac.uk/english-usage/projects/dcpse/ (last accessed
on 9 September 2013). The International Corpus of English (ICE-GB) will be introduced
in Section 1.2 below.
8  Parentheticals in English: introduction

(1.22) Backtracking
a. Mr Lehrer are you I want to be clear about this are you telling
us that this is your interpretation of what he was telling you or
what he was actually telling you (ICE-GB: s1b-064 #167)
b. What is unfair and I think what Tony Travis has failed to point
out is that by setting the ceiling and we understand so far there
are seven bands though well they may be dithering into nine by set-
ting a ceiling you’re in fact giving people in high-valued prop-
erty a subsidy and you’re then making those in lower-valued
properties pay more (ICE-GB: s1b-034 #68)
c. But a different role uh because when we get to the time of uh
Ezra as with the more classical Wellhausen uh hypothesis when we
get to the time of Ezra we have the further narrowing of the office
of priest (Dehé and Kavalova 2007: 3; ICE-GB: s1b-001 #9)
d. And I think that the Iraqi people and there are large numbers
of them honourable and decent who have been suppressed and
repressed and tortured and beaten and bombed over all of these
years that they are going to exert their view (Kavalova 2007:
161; ICE-GB: s1b-036 #71)
e. And it would seem to me  – that unless this morning’s exer-
cise – which has been so rewarding and so profitable – unless this
morning’s exercise is to dissipate – into another piece of feel-
ing – and er – pleasurable discussion – then we ought to take
quite seriously – the, the words of Jesus. – Go and do! (Biber
et al. 1999: 1068)
Classifications of clausal parentheticals have been suggested based on the
presence or absence of an ‘anchor’ in the host. For example, Kavalova
(2007: 149–52) distinguishes between anchored and floating parentheti-
cals, Kluck (2011: 234f) between anchored and free parentheticals. Both
dichotomies have in common that anchored parentheticals (or an ana-
phoric element contained in the parenthetical) refer back to an anchor,
often a noun phrase, in the host; see (1.23)a and b (the anchor is under-
lined; the anaphoric element is in small capitals). Kavalova (2007:
150) notes that anchored parentheticals of the clausal kind in (1.23)b cor-
respond to non-restrictive relative clauses (see (1.23)c). Kluck’s (2011)
free parentheticals correspond to Kavalova’s (2007) floating parentheticals
such that parentheticals of these types, or any anaphoric element within
them, do not refer to any constituent in particular in the host but rather
to the main proposition as a whole; see (1.24). While the parentheticals
in (1.24) are not anchored, they are yet semantically related to the host.
Specifically, there is a relation such that the parentheticals provide fur-
ther information about something expressed in the host sentence or com-
ment on the main proposition, which in (1.24) is the reference of this. In
Parentheticals – a motley crew  9

cases like (1.24), it is thus the host clause that functions as the anchor
(Huddleston and Pullum 2002: 1351). Thus anchored and free/floating par-
entheticals have in common that they are semantically related to the host.
In contrast, the parentheticals in (1.25) have no semantic relation with the
host but are related to the host utterance only via the discourse situation
or they contribute to the relation between the interlocutors established
by the situational context. Examples (1.25)a–c are taken from unscripted
speeches or demonstrations such as academic presentations. In (1.25)a, the
speaker interrupts his sentence to make sure that he can be heard all right
and receives the answer ‘Yes’ from the audience before carrying on. In
(1.25)b the speaker interrupts her presentation in order to turn a projector
off and to comment on what she is doing. In (1.25)c, it seems to occur
to the speaker that the audience might not be able to read the slides and
he reacts accordingly. In (1.25)d, the speaker interrupts the utterance in
order to ask the interlocutor to take a seat. I will refer to parentheticals of
this kind as detached parentheticals.
Some of the examples presented so far also show that the illocutionary
force does not have to be identical in parenthetical and host: for example,
in (1.1)c (= (1.25)a) an interrogative parenthetical clause is wedged into
a declarative host sentence; in (1.22)a the host is interrogative, while the
parenthetical is declarative; see also (2.8) in Chapter  2 below for more
examples.
(1.23) Anchored parentheticals
a. I saw that Bob, who just got fired, was booking a flight to Brazil.
(Kluck 2011: 234)
b. Because on this on this theory and it’s very deeply held uh good
educational news is by definition inadmissible as evidence
(Kavalova 2007: 149; ICE-GB: s2a-021 #91)
c. Because on this theory, which is very deeply held, …
(1.24) Free/floating parentheticals
a. Bill – and this is so typical – was dating several women at the
same time. (Kluck 2011: 235)
b. What this graph represents is the fact that for English uhm
and this work owes uh not a little debt to Eileen Whitley uhm we
have syllables which are distinct in being Y or W that is roughly
front spread back rounded being short or long (Kavalova 2007:
151; from ICE-GB: s2a-030 #85)
(1.25) Detached parentheticals
a. Well esterases are able in organic solvents to carry out a number
of useful can you hear me all right now? organic uhm processes
to produce things like food products (= (1.1)c: Dehé 2009: 579;
ICE-GB: s2a-034 #13)
10  Parentheticals in English: introduction

b. So what we can do in fact I’ll just turn it off is to use that sig-
nal to train people’s ability to perceive voicing distinctions in
speech and their ability to actually produce them (ICE-GB:
s2a-056 #87)
c. It’s occurred it occurred to me on the train to sort out this
business of weak and strong learning organisations to try out a
kind of three three level don’t worry if you can’t read it I I’ll read
it for you a sort of three levels or three a s a three step hierarchy
if you like of uh degrees of learning formats in organisations
(ICE-GB: s2a-049 #61)
d. The main point – Why not have a seat? – is outlined in the mid-
dle paragraph. (Burton-Roberts 2006: 180)
While some types of parentheticals, in particular anchored parentheticals
such as nominal appositions, non-restrictive relative clauses and anchored
and-parentheticals (e.g. (1.23)), have been shown to be more restricted in
their distribution than others such that their default position is next to their
anchor (see Section 2.3.3 below for exceptions), the places of other types of
parentheticals are generally variable. They include positions such as between
subject and finite verb (e.g. (1.2)a, c, (1.4)c, (1.5)b, (1.14)a), within the ver-
bal complex (e.g. (1.13)b, (1.27)a, d), between a lexical verb and its com-
plement (e.g. (1.28)a–c), between a preposition or noun and complement
((1.28)d–e), between a nominal head and a postmodifier (e.g. (1.1)b, (1.27)b,
c), interrupting a sequence of premodifiers within a noun phrase (e.g. (1.1)
c), between a possessive pronoun or a determiner and a noun (see (1.29)),
among others; see also the example in (1.26). The position of the parenthet-
ical within its host clause may (but does not have to) affect its semantic scope
and interpretation (e.g. Ifantidou 2001: 138–146; Grenoble 2004: 1966–7).
(1.26) Positional flexibility (from Kavalova 2007: 151; @ indicates pos-
sible positions)
a. I personally take the view and I’ve informed the Soviet
Government of this that that visit of the Ballet would be more
acceptable to all of our people including myself.
b. I personally take the view @ that @ that visit @ of the Ballet @
would be @ more acceptable @ to all of our people @ includ-
ing myself.
(1.27) Parenthetical positioned within verbal complex (a, d) or between
N and postmodifier (b–c)
a. Raids on Baghdad’s forces by the Allies would he said end the
murderous terrorist attacks from Iraq (ICE-GB: s2b-005 #97)
b. but there’d been no response not the slightest sign he said from
Iraq (ICE-GB: s2b-010 #124)
Parentheticals – a motley crew  11

c. Diesel petrol has other pollutants Kate which are just as dan-
gerous (ICE-GB: s1a-085 #103)
d. Do to answer the the the questioner’s main point do you believe
that the British Government and the American Government
are and I’m quoting his quote from Resolution sixty Six Sixty uhm
supporting all efforts in regard to creating intensive negoti-
ations between Iraq and Kuwait (ICE-GB: s1b-035 #68)
(1.28) Parenthetical positioned between verbal (a–c), prepositional (d) or
nominal (e) head and its syntactic complement
a. So the thing really is to identify I think the areas of interest for
you … (ICE-GB: s1a-066 #45)
b. And we’re going to see I think the introduction after just
twelve minutes of Alan MacCarthy into the side (ICE-GB: s2a-
003 #88)
c. The LSE would be doing that principally and you need I argue
an a rule-based knowledge system before you can articulate
what a text grammar should be (ICE-GB: s1a-024 #87)
d. Well they dropped cards in I suppose the doors (ICE-GB: s1a-
020 #177)
e. I mean in in that piece we’ve just heard from The Revenger’s
Tragedy it’s a mixture isn’t it of original instruments and kind
of what sound to me like modern trumpets (ICE-GB: s1b-
023 #140)
(1.29) Parenthetical positioned between possessive pronoun (a) or deter-
miner (b) and head N in NP
a. Your your uh what’s this called thorax is probably vibrating far
too much (ICE-GB: s1a-018 #53)
b. Uh in the uhm I think October issue of Computational uh
Linguistics there’s an attempt to do something of this type
(ICE-GB: s1a-024 #105)
These examples go against claims in the literature that some syntactic posi-
tions are impossible or do not lend themselves easily to being niches for par-
entheticals. These positions include the following: (i) following a preposition
(e.g. Lakoff 1974: 326, referring to Ross); (ii) between a prepositional or
verbal head and its complement (e.g. Jackendoff 1972: 98; Peterson 1999:
239; Potts 2002b: 645f); (iii) between a noun and a following restrictive rela-
tive clause (e.g. McCawley 1998: 460); (iv) between a noun and a preceding
adjectival adjunct (e.g. Taglicht 1998: 196); or (v) between a demonstrative
determiner and a noun (e.g. Taglicht 1998: 196). More generally speak-
ing, parentheticals have been argued not to be allowed between a head and
its complement or between a head and its modifier. In light of data from
actual spoken language as presented here, however, any assumption as to
general positional restrictions must be put more carefully (see, for example,
12  Parentheticals in English: introduction

(1.28)a–c for parentheticals intervening between verb and complement,


(1.28)d for a CC between a preposition and its complement, (1.28)e for a
question tag between a noun and its complement, and (1.27)c for a voca-
tive between a noun and the NRRC modifying it). Peterson (1999: 239–40)
argues that whether a parenthetical is possible or not between verb and
object is a matter of “end-weight preferences”. If the object is long or heav-
ily stressed, an intervening parenthetical is allowed (see his examples in
(1.30); see also Potts 2002b: 645). Some authors further exclude peripheral
positions, including non-initial and non-final positions only (e.g. Taglicht
1998: 196), while other authors explicitly include peripheral positions (see
Huddleston and Pullum 2002: 1355).
(1.30) Parenthetical placement interacting with constituent weight
(examples from Peterson 1999: 239)
a. *John likes, it must be admitted, icecream.
b. John likes, it must be admitted, icecream with peanut butter
topping.
Given the syntactic heterogeneity of parentheticals, some authors have sug-
gested defining and classifying them along functional or pragmatic lines
instead of by their syntactic properties. According to Grenoble (2004), for
example, the morphosyntactic properties of parentheticals are too varied to
allow for a structural definition. Instead, she argues, parentheticals can be
defined functionally since they all serve metacommunicative pragmatic func-
tions “and are united in operating on a distinct discourse plane” (Grenoble
2004: 1954). Grenoble (2004: 1968–72) identifies four broad categories of
shifts in the discourse: (i) subjective evaluation, represented by elements
which encode a subjective stance towards the speaker’s proposition, e.g.
comment clauses and sentence adverbs; (ii) addressee-oriented shifts, which
may be used to elicit a listener’s response, including, for example, 2nd per-
son parentheticals such as English you see, you know, don’t you think and
question tags; (iii) supplementary information such as explicit explanations
or background information, often expressed by clausal parentheticals; and
(iv) metastatements about the text, which serve as signals to the listener as
to how to interpret what is being said. Astruc-Aguilera (2005) comes to a
similar conclusion. Assuming that despite the “lack of syntactic homogen-
eity [parentheticals] function in very similar ways, if not in exactly the same
way”, she argues that parentheticals “are part of the contextual informa-
tion” (Astruc-Aguilera 2005: 13) and should be dealt with in the pragmatic
component. She suggests a functional analysis of parentheticals in terms of
their information structural status in the given context. In particular, she
follows Vallduví’s (1992 and elsewhere) analysis of dislocation in terms of
link, rheme and tail, suggesting that the information structural categor-
ies can be extended to other types of parentheticals and that their prosodic
properties reflect their information structural status.
The data  13

Parentheticals have also been defined prosodically such that they are set
off from the rest of the sentence by phonetic parameters: “the parenthesis
interrupts the prosodic flow of the frame utterance” (Bolinger 1989: 185).
In this book, the main focus is on the prosodic phrasing of parentheti-
cals. However, as Wichmann (2001: 181)  notes, investigating the prosody
of parentheticals demands a non-prosodic definition of parenthesis, other-
wise there is a risk of circularity in the argumentation. Moreover, strings
defined and identified according to prosodic properties need not correspond
to a syntactic constituent or syntactic insertion (see (1.21)b, repeated here as
(1.31); and I think plays has parenthetical prosody; see Wichmann 2001: 188
and Section 2.2 below).
(1.31) And one must also remember that uh the same Arnold Bax has
written poetry and I think plays under the pseudonym of Dermot
O’Brien (ICE-GB: s1b-032 #114)
In investigating the intonational phrasing of parentheticals, I will therefore
follow current syntactic-semantic definitions of parentheticals. In particu-
lar, I follow de Vries 2007, 2012a, b and Potts 2005; see Chapter 2. The gist
of these analyses is that while parentheticals are a motley crew and cover
syntactic construction types as varied as full clauses, appositions, question
tags, etc. (see above), they all have something in common. According to de
Vries, this common property is the way parentheticals are related to their
host; according to Potts this is their conventional implicatures (CI) seman-
tics, marked by a comma feature and resulting in intonational separation;
see Section  2.1 below. It is not the purpose of this book to develop these
syntactic approaches or offer a comprehensive semantic approach to par-
entheticals. Instead, the present study focuses on the prosodic properties
and in particular the prosodic phrasing of the syntactically varied class of
parentheticals and the implications for the syntax–prosody interface. To this
end, a large set of data from actual spoken language has been analysed for
prosodic phrasing. The source and nature of the data is the topic of the fol-
lowing section.

1.2  The data
The empirical part of the present study is entirely based on recorded, spon-
taneous or semi-spontaneous spoken language. The data used here are drawn
from the spoken part of the British Component of the International Corpus
of English (ICE-GB; Releases 1 and 2, ICE-CUP versions 3.0 and 3.1; see
Nelson, Wallis and Aarts 2002 and the ICE-GB website). The spoken part
of this corpus contains 637,682 words from various text types, ranging from
direct casual conversations and private telephone calls, public dialogues and
discussions, to unscripted and scripted monologues and broadcast news, all
recorded in the 1990s. The ICE-GB corpus and accompanying materials
14  Parentheticals in English: introduction

Table 1.1  Corpus references and text categories (see Nelson, Wallis and Aarts 2002)

Corpus reference Domain Text category


s1a-001 – s1a-090 private direct conversations dialogue
s1a-091 – s1a-100 private telephone conversations
s1b-001 – s1b-020 public classroom lessons dialogue
s1b-021 – s1b-040 public broadcast discussions
s1b-041 – s1b-050 public broadcast interviews
s1b-051 – s1b-060 public parliamentary debates
s1b-061 – s1b-070 public legal cross-examinations
s1b-071 – s1b-080 public business transactions
s2a-001 – s2a-020 public spontaneous commentaries unscripted monologue
s2a-021 – s2a-050 public unscripted speeches
s2a-051 – s2a-060 public demonstrations
s2a-061 – s2a-070 public legal presentations
s2b-001 – s2b-020 public news broadcasts mixed
s2b-021 – s2b-040 public broadcast talks (scripted) scripted monologue
s2b-041 – s2b-050 public non-broadcast speeches
(scripted)

(Nelson et al. 2002) provide detailed information on the source of the cor-
pus data. In the spoken part of the corpus, files whose names begin s1a are
from the private domain, comprising 205,627 words of speech material from
private direct conversations (s1a-001–s1a-090) and private telephone con-
versations (s1a-091–s1a-100). They thus represent more casual, informal
speech. Files whose names begin with s1b, s2a or s2b are from the public
domain, representing more formal speech, 432,055 words overall. These
parts of the corpus consist of dialogues (s1b: classroom lessons, broadcast
discussions and interviews, parliamentary debates, legal cross-examinations,
and business transactions), unscripted monologues (s2a: spontaneous com-
mentaries, unscripted speeches, demonstrations, and legal presentations),
scripted monologues (s2b-021–s2b-050: broadcast talks and non-broadcast
speeches), and a category “mixed” from broadcast news (s2b-001–s2b-020).
(See Appendix 2 to Nelson et al. 2002 for more information on the sources
of the ICE-GB texts.) For easier reference, Table 1–1 relates corpus refer-
ences to text categories.
The ICE-GB grammar contains a number of functions and categories,
among them Detached Function (DEFUNC), which includes parentheti-
cal clauses and vocatives (Nelson et al. 2002: 45). The initial corpus search
carried out for the present study was for this category, yielding 4,969 hits in
the spoken part of the corpus. This was then complemented by additional
search processes and subsequent manual sorting, reducing the number of
parentheticals included in the final analysis to 1,160 overall, distributed
among six groups (full parenthetical clauses, non-restrictive relative clauses,
The data  15

nominal appositions, comment clauses, reporting verbs, and question tags;


see Section 4.1).
The spoken part of the ICE-GB corpus provides accompanying sound
files, most of which are accessible and were used in this study for pro-
sodic analysis. The corpus is fully parsed syntactically, but not prosodically,
except that pauses (i.e. silent stretches of all kinds) are indicated in the cor-
pus transcriptions. In these transcriptions, silent stretches are indicated by
one or more commas between angled brackets (<,>, <,,>, etc.); the num-
ber of commas indicates the length of a pause as perceived (but not mea-
sured) by the corpus annotators. For the purpose of the prosodic analyses
reported on here, these pause indicators have been removed and they do
not occur in any of the corpus examples given throughout the book, even if
present in the ICE-GB corpus transcriptions. This was done because silent
stretches do not automatically correspond to boundaries of phrasal prosodic
constituents (see Section 3.1 below) and the intention was to avoid drawing
any conclusions from the transcriptions as to the status of silent stretches as
structure-related breath pauses or mere hesitation as well as any premature
conclusions regarding their function as cues to prosodic boundaries. After
removal of the pause indicators, all ICE-GB data appear without punctua-
tion in the present study since the transcriptions of the spoken material in
the corpus do not contain any punctuation marks.
The sound files were analysed auditorily and instrumentally. All prosodic
annotations were done by the author, using Praat (Boersma and Weenink
2012). However, given the well-known difficulties in the analysis of naturally
occurring data and given the general listener variability in the perception of
prosody (see Cole, Mo and Baek 2010 for recent discussion), all problematic
cases were referred to one or two experts trained in intonational analysis
and cases for which the annotators did not agree were removed from the
analysis. (I would like to thank Anne Wichmann and Jill House for their
intonational analysis of numerous sound files.)
The nature of the data comes with the obvious drawback that it does
not allow for experimental control and manipulation of the materials (see,
for example, Xu 2010 for a discussion of the disadvantages of using spo-
ken corpus data). Some of the questions remaining open due to the nature
of the data used in the present study are probably best addressed in an
experimental design (see Chapters 4 and 5), thus corpus data and experi-
mentally elicited data should ideally complement each other. However, the
use of spoken – spontaneous and semi-spontaneous – corpus data is a valu-
able undertaking for the following reasons. From an empirical perspec-
tive, many previous studies on parentheticals have been based on native
speaker intuition, on a relatively small number of made-up examples,
or on carefully manipulated, often contrived, experimental data elicited
in reading tasks (e.g. Kutik, Cooper and Boyce 1983; Frota 2000; Fagyal
2002; Astruc-Aguilera 2005; Astruc-Aguilera and Nolan 2007b; Kawahara
16  Parentheticals in English: introduction

2012). Despite yielding important insights into the prosodic properties


of parentheticals, there are also a number of drawbacks with this kind
of experimental work. First, as Speer, Warren and Schafer (2011: 38–9)
explain, there are a number of relevant differences between read speech
and spontaneously produced speech, among them the pragmatic goals of
readers vs. speakers and the actual prosodic structures. For example, Price
et  al. (1991) assume that in non-professional, spontaneous speech the
prosodic structure will be affected by hesitation phenomena and speech
errors and prosodic cues may be marked less clearly than in professional
speech. Howell and Kadi-Hanifi (1991) report that read speech and spon-
taneous speech differ in the location of prosodic boundaries, the position
of stresses, and the presence of (silent) pauses within prosodic constitu-
ents. Blaauw (1994) shows that read speech and spontaneous speech differ
in the distribution and realization of prosodic boundaries and that these
differences result in differences in perception. Hirschberg (2000) reports
considerable differences in speech rate (such that read speech is faster)
and intonation contours between read and spontaneously produced utter-
ances. Savino (2012) argues for an effect of style (read speech as obtained
by reading tasks used in previous research vs. spontaneous language as
obtained in a map task in her own work) on intonational contours associ-
ated with polar questions in Italian.
Second, experimental work studying parentheticals using contrived data
and read speech has been found to be flawed. For example, Kutik et  al.
(1983) performed a reading study designed to test fundamental frequency
(F0) effects in and around a parenthetical insertion of increasing length.
Wichmann (2000: 97) comments on their methodology, limited data set and
on the unnaturalness of the data, which she considers “one of the hazards
using contrived data”. Fagyal’s (2002) study was designed to yield evidence
for or against strong prosodic boundaries at the edges of utterance-medial
parentheticals in French. In her material, which was read by three speakers,
the parentheticals were separated from the surrounding sentences by com-
mas. It is perhaps not surprising that the participants produced strong pros-
odic boundaries in these positions. Kawahara’s (2012) study was designed
to provide evidence in favour of the level of the Intonational Phrase in the
Japanese prosodic hierarchy; the parentheticals in his experimental mate-
rials were marked by em dashes on either side, perhaps unsurprisingly
resulting in strong prosodic breaks. According to Kawahara (2012: 310), em
dashes are used specifically for parentheticals in Japanese orthography, thus
they unambiguously indicate the insertion and are therefore likely to result
in a speaker’s bias towards a prosodic boundary. Experimental paradigms
avoiding reading tasks and punctuation and eliciting more spontaneous data
might produce different results. It is one aim of the present study to inves-
tigate whether the assumptions about the behaviour of parentheticals and
related aspects of intonational phrasing hold to be true for spontaneous and
This study  17

semi-spontaneous spoken language, as archived in corpus material. From


a theoretical perspective, data of this kind has not yet been systematically
exploited in prosodic theory. This study aims to contribute to the debate
in this area, in particular to the discussion of intonational phrasing and its
relation to syntactic constituency.

1.3  This study
It is one major aim of this study to describe the intonational phrasing of
strings which are parentheticals in the syntax and to compare the results
with previous assumptions in the literature on parentheticals. Another is
to investigate the relation between syntactic parenthesis on the one hand
and intonational phrasing on the other to test and develop current pros-
odic theory on the basis of the obtained results. To this end, the empirical
sections will focus on the prosodic phrasing of six selected types of paren-
theticals: (i) full parenthetical clauses, (ii) non-restrictive relative clauses,
(iii) nominal appositions, (iv) comment clauses, (v) reporting verbs and (vi)
question tags.
Areas this study does not intend to make a major novel contribution to
include the syntax of parentheticals (in particular the syntactic relation
to the host), a semantic-pragmatic analysis of parentheticals and the rela-
tion between intonation and meaning (unless relevant to the syntax–pros-
ody relation), and the phonetics of intonational phrasing. I hope that the
results presented here will still be useful to researchers working on paren-
theticals in theoretical syntax and/or semantics. In particular, the relation
between prosody and meaning deserves much more attention than it can be
given here.
The remainder of the study is organized as follows. Chapter 2 will intro-
duce the reader to the syntactic and prosodic properties of parentheticals
as discussed in the continuously growing descriptive and theoretical lit-
erature on the topic. Following a general overview of syntactic accounts of
parentheticals and their prosodic properties, Chapter 2 focuses on previous
work on the syntax and prosody of the six types of parentheticals analysed
in Chapter 4. Based on this, predictions as to their prosodic phrasing will
be formulated. Since previous work on parentheticals in English has been
couched in different frameworks (in particular the British system of inton-
ation analysis and the Autosegmental-Metrical approach), Chapter  3 will
introduce the relevant prosodic categories. The notion of ‘prosodic separ-
ation’ will be defined for the purpose of this study. Moreover, the current
theoretical approaches to the interface between syntax and prosody will be
reviewed, leading to hypotheses as to the prosodic phrasing of parentheti-
cals. Chapter 4 represents the core of the empirical analysis. It reports the
data analysis and compares the results obtained here against previous find-
ings. Chapter 5 will discuss the results in light of prosodic theory.
2 The syntax and prosody
of parentheticals

This chapter introduces the reader to the syntactic relation between paren-
thetical and host (Section 2.1) and to the prosodic properties of parentheti-
cals (Section 2.2) as described in previous literature. Section 2.3 will focus
on the syntax and prosody of the six types of parentheticals that will be the
focus of the empirical study further investigated in Chapter 4.

2.1 The syntax of parentheticals and the relation between


parenthetical and host
It is a common assumption in the relevant literature that, unlike syntac-
tic arguments or adjuncts, parentheticals are ‘outside’ of, ‘external’ to or
‘invisible’ to the syntactic structure of the host sentence in which they
are linearly integrated, i.e. they are in some loose and specific syntactic
relation to their host (e.g. Haegeman 1991; Peterson 1999, 2004; Espinal
1991; Burton-Roberts 1999; Safir 1986; Fabb 1990; de Vries 2007, 2012a,
2012b). Taglicht (1998: 195) defines a parenthetical as corresponding to
a “syntactic node for which the grammar specifies no function in rela-
tion to any sister node” and as being “a licenced intruder in its phrase”.
There is ample evidence for this relative independence, which is outlined
at length in the relevant literature and will be only briefly summarized
here. For example, unlike syntactic arguments and adjuncts, parentheti-
cals cannot be the focus of a cleft construction (e.g. Haegeman 1991: 233;
Espinal 1991: 729; see (2.1) through to (2.3)), and they cannot be the
focus of a question (e.g. Haegeman 1991: 233f; Espinal 1991: 729; see
(2.4)). Parentheticals may be deleted without affecting the grammatical-
ity of the overall structure (e.g. Espinal 1991: 730; see (2.5): the examples
correspond to the a-examples in (1.1) through to (1.18) in Chapter  1:
the parentheticals have been omitted without rendering the sentences
ungrammatical or infelicitous). In verb-second (V2) languages such as
German and Dutch, parentheticals do not count as first constituents (e.g.
Espinal 1991: 730f; Ackema and Neeleman 2004: 97; see (2.6)). Movement
from the parenthetical into the host is impossible (e.g. de Vries 2007: 209f;
see (2.7)). The illocutionary force of the parenthetical is not dependent
18
The syntax of parentheticals  19

on that of its host (e.g. de Vries 2007: 217; see (2.8)). Moreover, bind-
ing of an element in the parenthetical from outside is impossible. For
example, a pronoun in the parenthetical cannot be bound by a quanti-
fier in the host (Q-binding; e.g. Haider 2005: 281; d’Avis 2005: 263f; de
Vries 2007: 212; see (2.9)); a reflexive or an anaphor in the parenthet-
ical cannot be bound from outside the parenthetical (A-binding; de Vries
2007: 213; see (2.10)); and a negative polarity item (NPI) in the paren-
thetical is outside the scope of a negative element in the host clause (e.g.
Peterson 1999: 234f; Burton-Roberts 2006: 181; de Vries 2007: 215f; see
(2.11)). Moreover, Condition C effects do not apply in parentheticals (de
Vries 2007, 2012a; Burton-Roberts 2006; see (2.12)). As de Vries (2007:
207) puts it, parentheticals are “invisible”, i.e. they do not “interact with
the host in terms of c-command-based relations”.
(2.1) It-cleft (adverbial clause; Haegeman 1991: 232–3)
a. John studied mathematics in Cambridge, while his son is studying
physics in Oxford.
b. *It is while his son is studying in Oxford, that John studied math-
ematics in Cambridge.
(2.2) It-cleft (sentence adverb; Espinal 1991: 729)
a. John, confidentially, committed suicide.
b. *It is confidentially that John committed suicide.
Compare:
c. John deliberately committed suicide.
d. It is deliberately that John committed suicide.
(2.3) It-cleft (nominal apposition; Heringa 2011: 110)
a. Joe asked Bill, a famous trumpet player, to teach him.
b. *It is a famous trumpet player that Joe asked Bill to teach him.
(2.4) Question formation (Espinal 1991: 729)
a. Beth is, honestly, my worst neighbour.
b. How is that?
*Honestly.
(2.5) Omission of parentheticals (cf. the a-examples in (1.1) through to
(1.18) in Chapter 1)
a. When we were on holiday we saw so many interesting places.
b. Her account suggests that we have to re-think the relationship
between meaning and truth.
c. For those of us who remember nineteen sixty-five Tory party
leadership contests used to be as the cardinals in Rome and lead-
ers would emerge
d. John smokes.
e. Max was sure he had nothing to worry about.
20  The syntax and prosody of parentheticals

f. He shouldn’t have pushed that kid out that door.


g. A university lecturer was arrested for the crime.
h. The secretary will present an apology.
i. Is he going?
j. He suffered great mental distress after the war.
k. John will go to Spain.
l. The reason for the Prime Minister’s resignation was to enable
Cabinet colleagues to enter the ballot.
m. There were no other applicants for that job.
n. Today’s topic is Nuclear Magnetic Resonance.
o. He is ill.
p. I’ve been dreaming of winning a gold medal for 20 years now.
q. He is so smart.
r. Amanda is my best friend.
(2.6) V2 (Dutch; from Ackema and Neeleman 2004: 97)
a. Jan, zo heb ik  gehood, bezit een verdacht boek
John so have  I heard possesses  a   suspect book
‘John, so I have heard, possesses a suspect book.’
b. *Zo heb ik gehoord  bezit Jan  een verdacht boek
so have I heard possesses  John  a suspect book
(2.7) Movement from parenthetical into host (adapted from de Vries
2007: 209)
a. The police, the commissioner suspected Hank stole the money,
searched his house.
b. *Whati did the police, the commissioner suspected Hank stole
what, search his house?
(2.8) Illocutionary force not identical in parenthetical and host
a. Well esterases are able in organic solvents to carry out a
number of useful can you hear me all right now? organic uhm
processes to produce things like food products (ICE-GB: s2a-
034 #13)
b. Jake said – why am I not surprised? – that he hates bicycles. (de
Vries 2007: 217)
c. Did Jake, John pondered, own a car? (de Vries 2007: 217)
(2.9) Q-binding
a. *[Every guest]i – hei just arrived – was talking about Hank. (de
Vries 2007: 212)
b. *Nobodyi was, hei claimed, the dumbest guy in the room. (de
Vries 2007: 212)
c. *Every woman, a talkative person, participated in the discussion.
(Heringa 2012: 564)
d. *[No climber]i talked about the K2, which hei conquered last
month. (de Vries 2012a: 155)
The syntax of parentheticals  21

(2.10)  A-binding (Dutch; de Vries 2007: 213)


*Joopi heeft – wie zal het zichselfi kwalijk nemen? – een
Joop has who will it se-self evil take a
nieuw huis gekocht.
new house bought
‘Joopi has – who will blame himi/*himselfi for it? – bought a new
house.’
(2.11) Negative polarity items
a. I didn’t predict, *(any of) you bastards, that we would win.
(Peterson 1999: 234)
b. John doesn’t work, while (*any of) his childeren are very (/* at
all) busy. (Burton-Roberts 2006: 181)
c. None of the authors, who had (*any) imagination, remained.
(Burton-Roberts 2006: 181)
(2.12) Condition C effects
a. Hei said – this is typical for Joopi – that he(i) didn’t like veggie
burgers. (de Vries 2012a: 155)
b. Johni studies mathematics, while hisi/Johni’s wife studies physics.
(Burton-Roberts 2006: 180)
c. Johni gets on best with private firms, who employ himi/Johni
often. (Burton-Roberts 2006: 181)
The fact that parentheticals may be omitted as a criterion for parenthetical
status seems to disqualify syntactic amalgamation as a type of parentheticals.
Syntactic amalgamations are constructions which lack a syntactic constituent,
and instead have an “interrupting clause” (IC; Kluck 2011) in the position of
the missing constituent (see (1.19), repeated as (2.13)a and (2.14)a below; e.g.
Lakoff 1974; Kluck 2011, and references given in the latter). That the IC cannot
be omitted easily is illustrated in (2.13)b and (2.14)b. To render the sentences in
(2.13)b and (2.14)b grammatical, the IC, instead of being omitted, would have
to be replaced by the syntactic constituent which the amalgam lacks.
Most recently, Kluck (2011) argues for an analysis of amalgams as anchored
parentheticals (see also de Vries 2012a). Kluck’s conclusion is based on prop-
erties of the relation between the host and the IC, in particular the syntactic
opacity and structural independence of the IC (e.g. elements of the matrix/
host clause do not c-command into the IC), the truth-conditional independ-
ence of the IC, and the inherent speaker orientation of the IC involving, for
instance, epistemic modality. Sentential amalgams will not be considered
further in this study; however, given their degree of syntactic integration,
their prosodic behaviour will be an interesting topic for future research.
(2.13) Syntactic amalgamation: omission of interrupting clause
a. John invited you’ll never guess how many people to his party.
(Lakoff 1974: 321)
b. *John invited to his party.
22  The syntax and prosody of parentheticals

(2.14) Syntactic amalgamation: omission of interrupting clause


a. John is going to I think it’s Chicago on Sunday. (Lakoff
1974: 324)
b. *John is going to on Sunday.
Approaches to the syntax of parentheticals differ in the ways they handle the
apparent structural independence from their host. For example, Haegeman
(1991) argues that parentheticals are orphan constituents which are syn-
tactically unattached at all levels of representation and whose interpret-
ation follows from general principles of utterance interpretation. Similarly,
Peterson (1999) maintains that there is no syntactic relationship between
parenthetical and host, but that there is a semantic ‘bond’ between the two,
which accounts for their association, including pronoun-antecedent and
gap-antecedent relationships. Espinal (1991) agrees with Haegeman (1991)
on (i) “the existence of syntactic independence between host and disjunct(s)
at every syntactic level of representation” and (ii) “the existence of concep-
tual relationships between the logical representations” involved (Espinal
1991: 741), but she also argues that linearization is nevertheless syntactic.
In her alternative account, complex syntactic structures have multiple root
nodes and involve separate planes in a three-dimensional space. Kaltenböck,
Heine and Kuteva (2011) argue that host and parenthetical are formed
in two separate domains of the grammar: sentence grammar and thetical
grammar, respectively, but how linearization works in their account is left
unexplained. Haider (2005) accounts for the linear order along the lines of
serialization in the phonetic component. Safir (1986) suggests the additional
grammatical module of LF-bar (LF′) to account for the specific behaviour
of NRRCs and other types of parentheticals, and McCawley (1982) allows
for crossing branches in the syntactic representation.
As noted in previous work (e.g. Burton-Roberts 2006), the apparent
syntactic independence of parentheticals is problematic in a strictly syn-
tactic account because in syntactic theory, at least since Kayne (1994), lin-
ear order is usually assumed to be determined by hierarchical relations
(but cf. de Vries 2003). Nevertheless, the assumption that parentheticals
are orphans is quite common in the literature. However, several authors
have argued that certain syntactic relations and scope effects do indeed
exist between parenthetical and host. Based on examples such as (2.15)
for instance, Hoffmann (1998: 302) argues that anaphors in a parenthetical
can sometimes be bound by antecedents in the host clause. In (2.15)a the
reflexive sich in the parenthetical is bound by the matrix subject, and in
(2.15)b it is bound by the matrix object. In syntactic theory, binding serves
as evidence for a syntactic c-command relation between the bindee (e.g.
an anaphor) and its binder (i.e. the antecedent). In (2.15) it should thus
be possible for the binder (Hanna in (2.15)a, ihn in (2.15)b) in the host
to c-command the reflexive in the parenthetical. Ackema and Neeleman
The syntax of parentheticals  23

(2004: 98)  show that parentheticals can be secondary predicates taking


a DP in the host structure as subject (see (2.16)a). They emphasize that
“[t]his is not just a matter of interpretation, since the usual c-command
restriction on predication holds” (Ackema and Neeleman 2004: 98; see
example (2.16)b). Moreover, they argue that “parentheticals can con-
tain parasitic gaps that are licensed by A′-movement in the host clause”
(Ackema and Neeleman 2004: 98; see example (2.17)a), and that “this can-
not be a matter of interpretation only, since the usual anti-c-command
restriction holds: the parasitic gap cannot be c-commanded by the trace of
the movement that licenses it” (see (2.17)b).
(2.15)  Binding into the parenthetical (from Hoffmann 1998: 302)
a.   Hannai  hat,  sichi  nicht  schonend,  die  Arbeit  zu Ende  gebracht.
Hanna  has  herself not   sparing   the  work  to end  brought
‘Without sparing herself, Hanna has finished the work.’
b. Hanna  hat  ihni,  sichi    nicht  schonend,  bei  der  Arbeit  gesehen.
Hanna  has  him  himself not  sparing   with the  work  seen
‘Hanna saw him working not sparing himself.’
(2.16) Parentheticals as secondary predicates (from Ackema and Neeleman
2004: 98)
a.  Jan,     naakt  en   dronken,  stond  weer  op  de  deur  te  bonken.
John,  naked  and  drunk,  stood  again on  the door  to  bang
‘John, naked and drunk, was banging on the door again.’
b. *Met Jan, naakt en dronken, valt niet te praten.
with John, naked and  drunk, falls not to speak
‘When he is naked and drunk it is impossible to talk to John.’
(2.17)  Parasitic gaps (from Ackema and Neeleman 2004: 98)
a.  [Welk boek]i zei Jan [dat je, althans zonder ei te
Which book said John that you, at least without to
lezen, ti niet moet veroordelen].
read, not must condemn
‘Which book did John say that you should not condemn without
reading it.’
b.   *[Welk boek]i maakt ti [dat je, althans zonder ei te
which book makes that you, at least without to
lezen, de bibliotheek niet wilt verlaten].
read, the library not want leave
‘Which book is such that you do not want to leave the library
without reading it.’
However, discussing Hoffmann’s example given in (2.15), de Vries (2007:
213, fn. 7) notes that “since Chomsky (1981) it is standard to assume that
present participle constructions, like infinitival to clauses, have a PRO sub-
ject” which can then serve as a local binder for the reflexive. According to de
24  The syntax and prosody of parentheticals

Vries, binding from outside the parenthetical is thus unnecessary (and in fact
impossible), which clearly weakens Hoffmann’s (1998) argument. A similar
analysis would also go against Ackema and Neeleman’s (2004) example in
(2.16), if PRO is assumed as subject to the secondary predicates.
In an attempt to nevertheless account for apparent relations between host
and parenthetical and for linearization, parentheticals have been analysed as
syntactic adjuncts in a series of publications in the 1970s and early 1980s.
For example, Ross (1973), Emonds (1973, 1976, 1979) and McCawley (1982)
all assume the same underlying structure in which the parenthetical is sister
and daughter of the root node S. The accounts differ in the way they derive
the surface word order in which a parenthetical appears in a non-peripheral
position. While under Ross’s (1973) Slifting analysis the parenthetical starts
out as a main clause whose complement moves across it to the left resulting
in the adjunction structure, Emonds (1973, 1976, 1979) suggests postposing
of a constituent of the host structure, and McCawley (1982) proposes cross-
ing branches. McCawley (1982) treats parenthetical placement on a par with
other “order-changing transformations”, among them scrambling, Heavy NP
Shift, Relative Clause Extraposition and Right Node Raising. More recently,
Corver and Thiersch (2002) argue that speaker-oriented parentheticals (cf.
also Reinhart 1983) are structurally identical to adverbs. D’Avis (2005) sug-
gests adjunction of the parenthetical to the closest phrasal projection (the
projection in whose vicinity it is interpreted) along with intonation-guided
interpretation of the interpolated constituent in order to account for the
ambivalent properties of parentheticals.
Put together, there are two main approaches to be found in the literature
on parentheticals, syntactic independence (orphan analysis) on the one hand
and syntactic integration (adjunction) on the other. These two approaches
have been referred to as the ‘unintegrated’ approach and the ‘integrated’
approach, respectively, by Dehé and Kavalova (2007), a terminology I will
stick to here. Within the unintegrated (or orphanage) approach, Heringa
(2011: 123)  distinguishes between radical orphanage analyses on the one
hand and non-radical analyses on the other. Radical approaches assume no
syntactic relation between parenthetical and host whatsoever (Fabb 1990;
Haegeman 1991; Peterson 1999; among others), while non-radical analyses
assume some relation, but at a late point in the derivation (e.g. Emonds
1979; McCawley 1982; Safir 1986; Espinal 1991).
Both accounts, unintegrated and integrated, have been found to have
quite obvious drawbacks. The unintegrated approach fails to account for
any apparent structural relationships between parenthetical and host (see
also Heringa 2011: 123–6, de Vries 2012a: 154–5). For example, appositions
and NRRCs are usually adjacent to their anchors and do not show as much
positional flexibility as some other types of parentheticals. De Vries (2012a:
154) reports that in Dutch these kinds of parentheticals seem to form a con-
stituent with their host. Evidence comes from topicalization: NRRCs and
appositions can be topicalized together with their anchor, but they cannot
The syntax of parentheticals  25

be stranded. Moreover, appositions have the same case as their anchor in


many languages (de Vries 2012a: 155; Heringa 2011). Furthermore, the
unintegrated approach is ruled out by prosodic evidence (Dehé 2009): if
prosodic structure has as its input the output of the syntactic component,
and if parenthetical and host are separate syntactic structures and lineariza-
tion is a matter of one of the interface modules, in particular if this is the
interpretational module, prosodic theory does not have access to both the
host and the parenthetical in the same mapping process. Any prosodic effect
of the parenthetical onto the host structure will thus be hard to explain
under a syntactic analysis lacking a structural relation between parenthetical
and host (see also de Vries 2007: 220, 2012a: 154). For example, Nespor and
Vogel (1986: 189f) assume that the strings preceding and following the par-
enthetical form intonational domains on their own, even if they would not
do so otherwise, i.e. the intonational phrasing of the host is affected by the
presence of the parenthetical, suggesting that host and parenthetical are part
of the same process relating syntactic and prosodic structure. Other pros-
odic relations between parenthetical and host have been observed in previ-
ous research on the prosody of parentheticals (e.g. Crystal 1969; Wichmann
2000, 2001; Watson and Gibson 2004; Peters 2006; Dehé 2007, 2009). Any
kind of effect of the parenthetical on the prosodic properties of the host,
such as prosodic phrasing, accent position and the shape of the intonational
contour, is hard to explain in standard theories of the syntax–prosody inter-
face under an orphan approach to parentheticals.
The idea behind the integrated approach to the syntax of parentheticals
is to account for linearization and for certain apparent syntactic relations
that exist between the parenthetical and its host in the syntactic structure,
which suggest an analysis along the lines of adjunction. However, integrated
accounts are incompatible with the observation that parentheticals and
adverbials (adjuncts) behave differently from other kinds of adjuncts in many
respects. For example, as is well known, true adjuncts can be clefted, they can
be the focus of a question, they are allowed as first constituents in V2 con-
texts, and so on, all properties which parentheticals do not have; see above.
Based on this dilemma and on the discussion of their examples given in
(2.16) and (2.17), Ackema and Neeleman (2004: 99) conclude that “a paren-
thetical cannot affect the syntax of the host clause, but grammatical require-
ments imposed by material in the parenthetical can be satisfied by elements
in the host clause”. They account for this generalization in a theory of fea-
ture matching called Insertion, which crucially builds on the Inclusiveness
condition (Chomsky 1995: 228; cf. also Neeleman and Van de Koot 2002 and
Ackema and Neeleman 2004: 99ff). In this theory, the syntactic representa-
tion of the parenthetical is connected to the representation of the host via
insertion of the former in a non-terminal node of the latter. The desired
results of this process are that first, inserted elements (parentheticals) are
invisible to operations that apply in the host structure, and second, it is
impossible for parentheticals to introduce licensing functions.
26  The syntax and prosody of parentheticals

Two recent prominent approaches – those of Christopher Potts and Mark


de Vries – suggest that the relation between parenthetical and host is indeed
one of adjunction, but that the connection between the two is nevertheless “of
a particular nature” (de Vries 2012a: 158). Potts (2002a, b, 2005) approaches
parentheticals from the semantic perspective (see Blakemore 2007 for a crit-
ical discussion of Potts’ approach). He argues that parentheticals (or sup-
plements in his terminology) are syntactically adjuncts; they are adjoined
to the linguistic material from which they obtain their meaning. In (2.20),
for example, the apposition a cyclist relates to the DP Lance. Any differences
between other adjoined elements and parentheticals, Potts argues, are best
explained in the semantics. Specifically, according to Potts, parentheticals (e.g.
as-parentheticals; and-parentheticals – his niched conjunctions, Potts 2005: 14;
non-restrictive relative clauses; nominal appositions; speaker-oriented, topic-
oriented and utterance-modifying adverbs) and expressives such as expres-
sive attributive adjectives (‘Shut that blasted window!’) and epithets (‘Sami,
this idiot, forgot the appointment’) are expressions of conventional implica-
tures (CIs; Grice 1975). Their meaning is assessed independently from that of
the host. They contribute discourse-new, speaker-oriented entailments, thus
match the definition of CIs; “their primary discourse function is to introduce
new, but deemphasized material” (Potts 2005: 33). Potts (2005: 11) extracts
four abstract properties of CIs from Grice (1975), all given in (2.18).
According to (2.18)a, CIs, unlike conversational implicatures, are inherently
linguistic, i.e. their content can never be contextually determined or variable
(Potts 2005: 26–30). Potts’ “at-issue entailment” (see (2.18)d) refers to what
Grice (1975) identifies as ‘what is said’ and what has elsewhere been referred
to as ‘descriptive meaning’, ‘truth-conditional content’, ‘regular content’ or
‘assertion’ (Potts 2005, 2007). Parentheticals are independent of the at-issue
entailments of the host (see (2.18)d) and are instead speaker-oriented entail-
ments or “speaker-oriented comments on a semantic core (at-issue entail-
ments)” (Potts 2005: 11; see (2.18)c). CIs are “secondary entailments that
cooperative speakers rarely use to express controversial propositions or carry
the main themes of a discourse [but which] are used to guide the discourse
in a particular direction or to help the hearer to better understand why the
at-issue content is important at that stage” (Potts 2005: 7). The example in
(2.19) illustrates the independence of CIs and at-issue entailment. The CI
‘Lance Armstrong is a former British road racing cyclist’, expressed by the
parenthetical insertion, is false: Armstrong is American. However, the at-issue
meaning, i.e. that Lance Armstrong had won the Tour de France seven times
in a row between 1999 and 2005, is true. Moreover, we can easily imagine this
sentence being uttered in a context where the relative clause or apposition is
added in order to provide additional information, which is secondary in the
given context but which may help the hearer interpret the sentence and to
The syntax of parentheticals  27

further clarify and contextualize the at-issue content of the host (compare
Potts’ 2005 antibackgrounding requirement).
(2.18) Abstract properties of CIs (Potts 2005: 11; extracted from
Grice 1975)
a. CIs are part of the conventional meaning of words.
b. CIs are commitments, and thus give rise to entailments.
c. These commitments are made by the speaker of the utterance
‘by virtue of the meaning of ’ the words he chooses.
d. CIs are logically and compositionally independent of what is
‘said (in the favored sense)’, i.e. independent of the at-issue
entailments.
(2.19) Apposition (after Potts 2005: 32)
Lance Armstrong, (who was) a former British road racing cyclist,
had won the Tour de France seven times in a row between 1999
and 2005 (before being disqualified from each of these races and
banned for life for doping offenses in 2012).
The semantic independence between parenthetical and host is central in the
present context. Unlike other authors, Potts does not think it necessary for
this independence to be reflected in the syntactic structure, i.e. the syntactic
connection between parenthetical and host. Instead, parenthetical and host
are connected via regular adjunction. However, unlike regular adjuncts, the
adjoined parenthetical is marked by a comma feature (see (2.20)).
(2.20) The syntax of supplements (from Potts 2005: 97)
a. Lance, a cyclist, is training.

DP VP

DP NP is training

COMMA
Lance

D° NP

a cyclist

The comma feature essentially serves two functions: semantically, “it per-
forms a type shift: it takes at-issue content to CI content”; prosodically,
it signals isolation, thus “accounting for the commas in print and the
28  The syntax and prosody of parentheticals

intonational boundary marks in speech” (Potts 2005: 98). Given this fea-
ture, he argues, no extra syntactic effort is needed; supplements appear as
regular adjuncts and the comma feature serves to distinguish parentheticals
from other kinds of modifiers/adjuncts. The effects of the comma feature
are illustrated along the lines of (2.21). Syntactically, both the as-parenthet-
ical in (2.21)a and its manner-adverbial counterpart in (2.21)b are adjoined
to the VP, to which they relate. They are distinguished from one another by
punctuation and intonation. The as-parenthetical in (2.21)a has a comma
feature, which signals intonational separation, i.e. comma intonation.
Accordingly, (2.21)a, but not (2.21)b, has CI semantics, its content being
independent of the main proposition. For example, the as-clause is outside
the negation in (2.22)a, but not in (2.22)b. (In writing, the comma feature is
expressed by the comma separating the as-parenthetical but not the manner
adverbial from the preceding material.)
(2.21) As-parenthetical (a) vs. manner adverbial (b) (Potts 2005: 135)
a. Alonzo built the canoe, as the foreman said he would (though not
in the manner they expected him to build it).
b. Alonzo built the canoe as the foreman said he would (with an
ax and elbow grease).
(2.22) As-parenthetical (a) vs. manner adverbial (b)
a. Alonzo built the canoe, as the foreman said he would (though not
in the manner they expected him to build it).
a1. No, he didn’t. = Alonzo did not build the canoe.
a2. No, he didn’t. = The foreman did not say he would.
b. Alonzo built the canoe as the foreman said he would (with an
ax and elbow grease).
b1. No, he didn’t. = Alonzo did not build the canoe in the
way the foreman said he would.
b2. No, he didn’t. ≠ The foreman did not say he would.
In a similar spirit, but anchoring the differences between regular adjuncts
and parentheticals in the syntax rather than the semantics, de Vries also
argues that parenthetical phrases are adjoined to their host (see de Vries
2005, 2007, 2012a and elsewhere; see also Griffiths and de Vries 2013 for
yet more recent evidence from ellipsis in fragment answers). However,
according to de Vries, the internal structure of parenthetical adjuncts is
built up by a non-canonical, “second type of Merge” (de Vries 2012a:
153). According to de Vries, given the syntactic properties of parentheti-
cals, they cannot be connected to their hosts along the lines of regular
merge, because regular merge builds up a hierarchical structure such that
the merge-mates are dominated by (or contained in) the merged object,
with the result that c-command relationships hold. However, despite being
syntactically included in the host, parentheticals do not interact with it
The syntax of parentheticals  29

in terms of c-command-based relations (see e.g. (2.9) and (2.11) above;


this is de Vries’ 2007 Invisibility). Building on the traditional fundamen-
tal distinction between syntactic subordination and non-subordination,
de Vries suggests “two basic types of inclusion; regular inclusion (which
corresponds to traditional dominance) and parenthetical inclusion” (de
Vries 2012a: 156) as two primitives of the grammar. Accordingly, there are
two types of merge (see (2.23)): regular merge resulting in regular inclu-
sion/dominance, and parenthetical (par-)merge, resulting in a paratactic
hierarchy and rendering the par-merged constituent invisible for c-com-
mand-based relations such as binding and other dependency relations.1
Two par-merged constituents are built up internally by regular merge
(unless another parenthetical is involved), but the connection between
parenthetical and host is one of par-merge. A structure involving both
regular merge and par-merge is represented in (2.24). While C and D are
c-commanded by F, A and B are not; this is because they are par-merged,
thus par-included in D, which blocks c-command. (In (2.24) solid lines
between syntactic objects, e.g. between F and E, represent the merge-
mate relationship, or sisterhood; the dotted arrow represents the inclusion
relationship; the stars next to the dotted arrow indicate par-merge; see de
Vries 2012a: 149, 157).
(2.23) Two types of merge (de Vries 2012a: 156)
a. (Regular) merge (A, B) yields C such that
i. C directly (regularly) includes A,
ii. C directly (regularly) includes B, and
iii. A is the merge-mate of B.
b. Par(enthetical) merge (A, B) yields C such that
i. C directly par-includes A,
ii. C directly par-includes B, and
iii. A is the merge-mate of B.
(2.24) Two types of merge (de Vries 2012a: 156-7)

F E

C D

* *
A B

1
Par(enthetical) merge corresponds to b(ehindance) merge in de Vries (2005, 2007). If A and
B are b-merged into C, then C b-includes A and B, and A and B are paratactically construed
with respect to C. Crucially, b-inclusion blocks c-command relations.
30  The syntax and prosody of parentheticals

De Vries (e.g. 2007: 229f, 2012a: 158) further argues that the parenthetical
is embedded in an abstract parenthetical phrase ParP (see (2.25)). This is
necessary to avoid invisibility of a potential (par-) merge-mate (see de Vries
2012a: 157–8). ParP is adjoined to the host, resulting in the linear order of
parenthetical and host. The lack of c-command relations between paren-
thetical and host follows from the invisibility of the internal structure of
ParP, which is due to the nature of par-merge. De Vries (2012a: 158) thinks
of the syntactic head Par as a “specialized discourse connector, a ‘paren-
thetical specifying coordinator’” (see also de Vries 2009), which may in
fact be spelled out, for example as a conjunction in and-parentheticals
(see (2.26)a) or appositions (see (2.26)b). He compares the semantic effect
of Par to the comma operator introduced by Potts (2005) but argues that
the semantic effects must be evoked by syntactic elements. Syntactically,
“par-inclusion starts a new c-command domain” (de Vries 2012a: 158).
Semantically, this may start a new lambda term and account for the par-
enthetical as being perceived as secondary with respect to the host. In the
prosody, de Vries (2012a: 159) suggests, the syntactic parenthetical marker
may be interpreted as the beginning of a new intonational domain.
(2.25) Parenthetical structure

… YP

ParP YP
* * …
Par XPpar

(2.26) Spell-out of Par (see de Vries 2012a)


a. I still owe Anna – and Anna disappeared last night – 250 dollars.
b. Elvis Presley, or The King, remains unforgotten.
Despite the controversial syntactic debate, there seems to be wide agree-
ment that the syntactic relation between parenthetical and host is special. I
will follow de Vries’ syntactic account, with reference to Potts for the seman-
tics. In particular, par-merge makes it possible to account for surface rela-
tions between parenthetical and host, including prosodic effects, while at
the same time accounting for the syntactic invisibility and semantic separ-
ateness of parentheticals. The next section will survey the prosodic features
of parentheticals as observed in previous research.

2.2  Prosodic features of parentheticals


The prosodic properties of parentheticals have been approached from vari-
ous perspectives and have already been reviewed elsewhere (e.g. Dehé and
Prosodic features of parentheticals  31

Kavalova 2007: 12–15). Generally speaking, there is wide agreement that


intonation is a defining feature of parentheticals, and that parentheticals are
prosodically separate and independent from their host. Among linguists
discussing the syntax and semantics of parentheticals, intonational separate-
ness, i.e. the phrasing of parentheticals in intonation domains of their own,
typically the Intonational Phrase (IP) or the Tone Group (TG) in the British
framework (see Section 3.1 below for these categories) is commonly taken
for granted, often without supporting empirical evidence (e.g. Safir 1986;
Haegeman 1991; Potts 2002b, 2005; D’Avis 2005: 262; Burton-Roberts
2006: 180, among many others). For example, Huddleston and Pullum
(2002: 1350), in The Cambridge Grammar of the English Language, claim that
parentheticals “are intonationally separate from the rest of the sentence”;
Haegeman (1991: 250) maintains that “[c]omma intonation is a syntactically
determined PF property of parentheticals”; Potts (2002b: 650) claims that
examples such as the clausal as-parenthetical in (2.27) “are ungrammatical
if read without an intonation break at the onset of the parenthetical and
another such pause at its coda”.
(2.27) Joan as Chuck can attest owns a unicycle.
The idea that parentheticals are obligatorily phrased separately into their
own IP is mostly based on the assumption that IPs are syntactically con-
strained and that parentheticals are syntactically independent. As out-
lined in the previous section, parentheticals have long been assumed to be
either external to the syntactic structure of the host or only very loosely
related to it. The idea is, therefore, that constituents which are syntactically
non-integrated or loosely related cannot be prosodically integrated at the
same time. The closest syntactic relation suggested between host and par-
enthetical is that of adjunction. For adjuncts it has been argued that they
are in a syntactic position not relevant to the syntax–prosody mapping (e.g.
Truckenbrodt 1995, 1999), predicting prosodic separateness for adjuncts.
In Truckenbrodt’s account, this idea is based on Chomsky’s proposal that
“interface constraints generally make reference to syntactic categories, not
syntactic segments” (Truckenbrodt 1999: 235; he refers to class lectures of
Chomsky in 1994). If parentheticals are adjuncts, the same prediction holds
for them; see also Section 3.3 below.
In the syntactic-semantic literature, intonational breaks around paren-
theticals are usually assumed to coincide with pauses. Among the empir-
ically attested cues to these breaks are pauses indeed (e.g. Schwyzer 1939;
Altmann 1981; Bolinger 1989; Taglicht 1998; Payà 2003; Astruc-Aguilera
2005; Astruc-Aguilera and Nolan 2007b), but also a falling-rising pitch at
the end of the immediately preceding domain (Local 1992), and the block-
ing of sandhi rules (e.g. Frota 2000 for fricative voicing, syllable degemina-
tion and vowel adjacency resolution in European Portuguese).
32  The syntax and prosody of parentheticals

In the prosodic literature, the properties of parentheticals have been


studied in terms of prosodic phrasing as well as intonation (see e.g. Astruc-
Aguilera 2005 for an overview). The idea of intonational separateness, as
put forward in the syntactic literature, is also prominent in the prosodic lit-
erature. For example, Nespor and Vogel (1986: 188–90) assume that paren-
theticals are obligatorily phrased in a separate IP (see also Selkirk 1981: 131,
1984: 295, 1995: 567; Truckenbrodt 2005: 275; the assumption goes back
at least to Downing 1970, 1973). Moreover, Nespor and Vogel (1986: 189–
90) assume that the strings preceding and following a parenthetical form
IPs on their own, even if they would not do so otherwise. This yields the
phrasing patterns in (2.28). The examples in (2.29) (from Nespor and Vogel
1986: 189, 190) illustrate (2.28)a. Selkirk (1981: 131) assumes that the pres-
ence of a sentence-medial parenthetical leads to “at least three intonational
phrases (and intonation contours) [since the] medial intonation phrase will
be bounded by one on either side”.
(2.28) Prosodic separation of parentheticals
a. … IP[…]IP IP[parenthetical]IP IP[…]IP…
b.  IP[parenthetical]IP IP[…]IP…
c. … IP[…]IP IP[parenthetical]IP
(2.29) Prosodic separation of parentheticals (Nespor and Vogel 1986:
189–90)
a. IP[Isabelle]IP IP[as you know]IP IP[is an artist]IP
b. IP[Isabelle is]IP IP[as you know]IP IP[an artist]IP
c. IP[Charles wouldn’t]IP IP[I imagine]IP IP[have done such a
thing]IP
Selkirk (1984: 296)  assumes for NRRCs specifically that they “are always
separate [Intonational Phrases], not included in the same [Intonational
Phrase] with the constituent they modify”. Experimental work using read
speech has confirmed the idea about separate intonational domains (e.g.
Frota 2000 for European Portuguese; Fagyal 2002 for French; Astruc-
Aguilera 2005, Astruc-Aguilera and Nolan 2007b for Catalan and English;
Kawahara 2012 for Japanese).
Elsewhere in the literature, it has been argued that the presence of the
parenthetical does not force preceding and following material into sep-
arate IPs; i.e. a non-recursive structure such as (2.28)a is not mandatory.
Specifically, experimental research by Cooper and Sorensen (1981) using
data of the kind given in (2.30)a shows that the presence of a parenthet-
ical does not have to affect the overall declination, i.e. the gradual lower-
ing of pitch, of the host utterance. Rather, the downward trend begun
on material preceding the parenthetical may be continued after the end
of the parenthetical insertion (see also Section  3.1 below). Ladd (1986,
Prosodic features of parentheticals  33

1996) argues that under the assumption that declination applies across
an intonational domain of type IP, Cooper and Sorensen’s finding can
be accounted for in terms of limited recursivity in prosodic structure.
Specifically, he suggests an analysis in terms of a compound prosodic
domain (CPD; see (2.31)), a structure which is analogous to a compound
in morphosyntax. The phrasing in (2.30)b illustrates the non-recursive
(default) structure as suggested by Nespor and Vogel (1986); (2.30)c gives
the CPD structure as suggested by Ladd. In a CPD structure, the matrix
clause forms a single domain (i.e. the outer IP) across which declination
applies, regardless of whether or not it is interrupted by another domain
(here the parenthetical; see (2.32)).
(2.30) Parentheticals used in experimental research (Cooper and
Sorensen 1981)
a. The book on the table, it seems to me, was a gift from my
mother.
b. [The book on the table] [it seems to me] [was a gift from my
mother]
c. [The book on the table [it seems to me] was a gift from my
mother]
(2.31) Compound Prosodic Domain (Ladd 1996: 244, 2008: 297)
A CPD is a prosodic domain of a given type X whose immediate
constituents are themselves of type X.
(2.32) Parenthetical phrased in a CPD
IP1[… IP2[parenthetical]IP2 …]IP1

Further empirical support for the CPD has been provided by Frota (2000:
60–77), also using data involving parenthetical insertions (see (2.33) for an
example). She finds that the IP which embeds the parenthetical, i.e. the
outer IP, has one intonational contour, while sandhi processes bound by the
IP apply at the boundary separating the IP spanning the parenthetical from
the matrix IP. Dehé and Kavalova (2006) argue for an analysis of the pros-
odic phrasing of parenthetical what along the lines of the CPD. Specifically,
they suggest that parenthetical what introduces a prosodic boundary where
there would be no boundary otherwise (for example, between a syntactic
head such as verb or preposition and its complement), that what always joins
the preceding domain, but that the resulting structure is recursive and is of
type CPD (see (2.34)).
(2.33) CPD (from Frota 2000: 70; European Portuguese; U = prosodic
utterance)
a. As alunas, até onde sabemos,  obtiveram boas  avaliações.
The students  as far as we know got good marks
34  The syntax and prosody of parentheticals

b. [ [ [As alunas]IP1 IP2[até onde sabemos]IP2]IP3 IP4[obtiveram


U IP3 IP1
boas avaliações]IP4]U
(2.34) Parenthetical what phrased in a CPD (from Dehé and Kavalova
2006: 313)
a. There is no ‘unless’ coz no one has said ‘whoopsy daisies’ for
what 50 years and even then it was uhm it was just little girls
with blonde ringlets (Julia Roberts in Notting Hill)
b. (no one)(has said whoopsy daisies)((for what)(fifty years))
The CPD analysis is compatible with Wichmann’s (2001: 181) assumption
that if parentheticals are asides, they should not interrupt the overall down-
trend of the utterance, and editing out the parenthetical in an instrumental
analysis should result in a coherent contour. This has been tested infor-
mally, e.g. by Wichmann (2000) and for parenthetical what by Dehé and
Kavalova (2006). These studies find essentially that editing parentheticals
out of the host utterance results in coherent contours. Wichmann takes this
as evidence for the assumption that speakers plan parentheticals as separate
utterances, interpolated in another utterance, and that they have a memory
for where they left off before a parenthetical. However, it would be too hasty
to generalize from these results, because prosodic patterns of parentheti-
cals, in particular of short parentheticals, are variant and depend on many
factors (Bolinger 1989; Wichmann 2001, among many others), and editing
them out may not result in coherent contours across the board. Moreover,
as has been observed elsewhere, the “strategy of compression is the canon-
ical prosody of parenthesis” (Wichmann 2001: 186), i.e. pitch drop as well
as pitch compression for the duration of the parenthetical. It is also pos-
sible, however, to use pitch expansion rather than compression as a means
of marking the parenthetical (Bolinger 1989; Wichmann 2001; see example
(2.43)b/Figure 2.2). These strategies (see Wichmann 2001: 186–8 for fur-
ther examples), commonly found in spontaneous spoken language, may well
interfere with the overall downtrend of the utterance. For example, a down-
ward trend begun before the parenthetical may be picked up at a higher level
after a parenthetical with expanded pitch, or at a lower level following a par-
enthetical with compressed pitch; this would have to be tested. Naturally,
neither compression nor expansion are necessary strategies to prosodically
mark syntactic parenthesis – any of the typical features may be suspended
(Bolinger 1989).
Also relevant in the context of the prosodic phrasing of parentheticals
is Gussenhoven’s (2004: 290–2) analysis of certain “extra-sentential elem-
ents”, including reporting verbs, comment clauses and vocatives, among
others (Bing’s 1985 ‘Class 0 expressions’). According to Gussenhoven
(2004: 291), these come in two types, both of which are unaccented: (i) they
may be included in the preceding IP (‘incorporation’), or (ii) they may be
encliticized. By contrast, accented extra-sentential elements are phrased
Prosodic features of parentheticals  35

separately. The respective phrasings are given in (2.35), with examples in


(2.36). According to Gussenhoven (2004: 291), reporting verbs “are either
incorporating or cliticizing” (see (2.36)a, b), while question tags may be
prominent and separate (see (2.36)c).
(2.35) Adapted from Gussenhoven (2004: 291; ES = extra-sentential
element)
a. IP[… ES]IP incorporation
b. IP[IP[…]IP ES]IP encliticization
c. IP[…]IP IP[ES]IP separation
(2.36) Adapted from Gussenhoven (2004: 292; capitals indicate
prominence)
a. IP[Is it TRUE she asked]IP incorporation
b. IP[IP[Is it TRUE]IP she asked]IP encliticization
c. IP[It’s TRUE]IP IP[ISn’t it]IP separation
Unlike incorporated material, encliticized material is set off from preced-
ing material by a boundary tone (Gussenhoven 1990). It receives a copy of
the tones after the last stressed syllable. For example, if the nuclear pattern
is H*LL% (i.e. a nuclear peak with subsequent falling pitch towards a low
domain boundary; see Chapter 3), then LL is copied onto the encliticized
material. If the nuclear pattern is H*LH% (i.e. a nuclear peak with subse-
quent falling then rising pitch towards a high domain boundary), then LH
is copied onto the encliticized material (see (2.37)). This has been referred
to as “tone copy” (Gussenhoven 1990, 2004).
(2.37) Tone copy (Gussenhoven 2004: 291)
Shouldn’t we dis  cuss  this the two of us
%L H*L H%  L H%
Along with prosodic phrasing, other prosodic properties have been
observed for parentheticals. “By definition”, Bolinger (1989: 185)  states,
“the parenthesis interrupts the prosodic flow of the frame utterance”. He
maintains that the “intonation is more than a reflection of the fact that a
given segment of discourse is a parenthesis; it is often the main cue dif-
ferentiating it as such”. In other words, while in writing there are ortho-
graphical means to mark a parenthetical (e.g. round brackets around the
interpolation, hyphens, commas, dashes and the like), there are intonational
features which clearly signal the beginning and end of a parenthetical inser-
tion in spoken language. Parentheticals are prosodically marked by varia-
tions in pitch, tempo and loudness (Schwyzer 1939; Crystal 1969; Bolinger
1989; Wichmann 2000, 2001; Astruc-Aguilera 2005, among many others).
According to Bolinger (1989: 186), the “typical parenthesis has three pros-
odic characteristics: it is lower in pitch than the matrix sentence, it is set
off by pause(s), and it has a rising terminal”. This is complemented by
36  The syntax and prosody of parentheticals

diminished loudness in Crystal (1969: 174). As for pitch, while lower pitch
has been assumed to be typical (Crystal 1969: 174; Bolinger 1989: 186),
it is also possible that the parenthetical is marked by higher pitch (pitch
expansion) rather than pitch compression (Bolinger 1989: 188; Wichmann
2001: 188).
While a parenthetical string typically differs from its immediate environ-
ment in at least one of these prosodic parameters in order to be recogniz-
able, any one of the features can be suspended, depending on the function,
length and position of the parenthetical (Bolinger 1989; Wichmann 2000).
Therefore, none of the defining prosodic features qualifies as a necessary
condition for parenthesis.
The example of an elliptical parenthetical clause in (2.38)/Figure  2.1
serves to illustrate the ‘prototypical’ prosodic features of parentheticals as
assumed in previous literature (i.e. prosodic separation in phrasing, accom-
panied by intonational features). In this example, there is every reason to
assume prosodic separation of the parenthetical, i.e. phrasing into its own
intonational domain. The prosodic annotation in Figure  2.1 follows the
ToBI conventions (Silverman et  al. 1992; Beckman and Elam 1993; see
Chapter 3 for details).
(2.38) Parenthetical insertion with ‘prototypical’ prosodic properties
For those of us who remember nineteen sixty-five one or two of our
listeners may Tory party uh leadership contests used to be uh as
the cardinals in Rome and leaders would emerge (ICE-GB: s1b-
024 #001)
The parenthetical string one or two of our listeners may shows the follow-
ing prosodic features which have been argued to be characteristic of paren-
theticals. First, it is lower in pitch than the matrix sentence. This concerns
the maximum pitch value (194 Hz in the domain preceding the parenthet-
ical, 165.5 Hz inside the parenthetical, 172 Hz in the domain following the
parenthetical), the pitch range (104 Hz before the parenthetical, 75.8 Hz
in the parenthetical, 77 Hz after the parenthetical), and average pitch (124
Hz before the parenthetical, 109.5 Hz in the parenthetical, 122 Hz after the
parenthetical). Second, the parenthetical is set off by pauses (0.23 seconds
before, 0.911 seconds after the parenthetical). Third, it has a rising terminal
(see the high boundary tone H%). It is also faster in speech rate than the
host utterance (parenthetical: 8.4 syll/sec; preceding unit for those of us who
remember nineteen sixty-five: 6.6 syll/sec; following unit Tory party uh lead-
ership contests used to be uh as the cardinals in Rome: 4.5 syll/sec), another
prosodic feature often associated with parentheticals (see Crystal 1969;
Schwyzer 1939: 32, among others). The parenthetical is phrased in its own
separate IP, which can be identified on the basis, for example, of the follow-
ing criteria (see also Section 3.1 below): (i) the high boundary tones ter-
minating the preceding domain and the domain of the parenthetical; (ii) the
250

200
Pitch (Hz)

150

100

50
L+H* L+!H* L+!H* L*+H H-H% H* L*+H H-H% H*

parenthetical

for those of us who remember nineteen sixty five <SIL> one or two of our listeners may <SIL> Tory Party <SIL>

0 4.97

Time (s)
Figure 2.1  Example (2.38): parenthetical insertion with prototypical parenthetical properties
38  The syntax and prosody of parentheticals

structure-related pauses immediately to the left and right of the parenthet-


ical; and (iii) the fact that the relevant IP is the domain of a complete tonal
contour (CTC), i.e. it has a main prominence (pitch accent L*+H) followed
by the H-H% sequence of edge tones, or, in the British tradition, a rising
nuclear tone (see Chapter 3 below for an introduction to the relevant frame-
works). The nuclear syllable is may, which is associated with a local pitch
rise terminating in H%. The criteria for the identification of IPs and their
boundaries will be further explained in Chapter 3 below.
The specific prosodic properties of parentheticals serve semantic, prag-
matic and communicative functions. For example, prosodic separation may
mark the parenthetical as material contributing additional information,
outside the main proposition, but relevant to the discourse in one way or
another. First, according to Potts (2005), intonational separation goes along
with CI semantics, i.e. it marks the beginning and end of the inserted string,
which provides new, relevant information outside the at-issue entailment.
The parenthetical one or two of our listeners may in (2.38), for example, is
outside the main proposition. It contributes CI meaning along the lines that
it is possible that some of the listeners do remember 1965, thus strengthen-
ing the overall relevance of the utterance. The main proposition is directly
addressed to those listeners that do remember 1965, thus it may be neces-
sary, for example, to remind other listeners that these people may be among
the audience.
Moreover, the prosodic phrasing of utterances containing parentheticals
as well as their intonational properties may have scope effects and serve to
disambiguate between different meanings which arise from structural ambi-
guities. For example, the string you know in (2.39)a (example from Price
et  al. 1991: 2967; Hirschberg 2004: 523; see also Crystal 1969: 264), rep-
resented there without commas, is ambiguous between a parenthetical (or
discourse marker), roughly paraphrased as As you are aware, Mary knows
many languages (see the context in (2.39)b, from Price et al. 1991: 2967), and
an abbreviated restrictive relative clause (languages that you also know, see
the context in (2.39)c, from Price et al. 1991: 2967). Under the parenthet-
ical interpretation, the assumption is that you know would be preceded by a
phrasal prosodic boundary and would typically be associated with a falling-
rising pitch pattern. An abbreviated relative clause would not be predicted to
be preceded by a prosodic boundary of similar strength. Similarly, in (2.40),
they all knew is ambiguous between a simple complement involving that-
omission (We only suspected THAT they all knew that …) and a parenthetical
interpolation meaning They all knew that we only suspected that a burglary
had been committed (Hirschberg 2004: 524). A parenthetical would be pre-
dicted to be separated from the rest of the utterance by strong intonational
breaks, i.e. typically pauses and rising pitch associated with suspected and
again with knew. A third interpretation, which would also involve phrasal
prosodic boundaries before and after they all knew, can be paraphrased as
Prosodic features of parentheticals  39

We only suspected but they all knew that a burglary had been committed. This
interpretation would require contrastive prominence on the pronouns we
and they, and on the verbs suspected and knew. In (2.41), the ambiguity is
between an appositional ((2.41)a) and a non-appositional (object) use ((2.41)
b) of the NP the lady in red.
Sentence pairs like the ones in (2.39) and (2.41), among other ambigu-
ous sentence pairs, were tested in a perception experiment based on speech
material recorded by professional public radio newscasters with stronger
prosodic boundaries separating parentheticals than their alternatives (Price
et al. 1991). Phonetic cues to boundary strength were segment duration and
intonational cues (boundary tones, pitch-range changes and pitch accents).
The initial hypothesis was that because parentheticals and appositions (e.g.
(2.39)b and (2.41)a) have stronger prosodic boundaries than their alterna-
tives (e.g. subordinate clauses such as you know in (2.39)c and object NPs
in (2.41)b), parentheticals and appositions would be more reliably identified
by the listeners. While this hypothesis was not borne out across the board
(parentheticals such as you know in (2.39)b were less reliably identified than
their alternatives such as the subordinate clause you know in (2.39)c), paren-
theticals and appositions were identified at levels greater than chance, sug-
gesting that listeners used the available prosodic evidence to disambiguate
between parenthetical/appositional use on the one hand and non-paren-
thetical/non-appositional use on the other.
(2.39) Structural ambiguity; here: parenthetical vs. subordinate clause
(Price et al. 1991; Hirschberg 2004)
a. Mary knows many languages you know.
b. Mary leaves on Tuesday. She will have no problem in Europe.
Mary knows many languages, you know.
c. Mary and you have similar backgrounds and have both learned
many languages. Mary knows many languages you know.
(2.40) Structural ambiguity (Hirschberg 2004)
We only suspected they all knew that a burglary had been
committed.
(2.41) Structural ambiguity; here: nominal apposition vs. attached NP
(Price et al. 1991)
a. Most of the women had forgotten the strange event by the next
week. Only one remembered, the lady in red.
b. Most of the people forgot about the strange visitor. Only one
remembered the lady in red.
The example in (2.42) is a case of semantic scope ambiguity. The comment
clause (CC) I think may have scope over what precedes, meaning I think it
was my friend who got it twelve years ago, or over what follows, meaning I
think it was about twelve years ago that my friend got it. The attested phrasing
40  The syntax and prosody of parentheticals

in the actual corpus example is such that the CC is not phrased separately
but forms a prosodic constituent with the following material, suggesting the
latter interpretation and avoiding ambiguity.
(2.42) Scope ambiguity (Dehé 2009: 604)
But my friend got it I think about twelve years ago (ICE-GB: s1a-
071 #90)
Looking at the relation between syntactic and prosodic parentheses, it is
important to realize that the one-to-one correspondence between the two
is a strong tendency at most, in particular if formulated in an absolute
sense for all syntactic types of parentheticals. First, the different functions
served by parentheticals may demand differences in prosodic realization: for
example, while supplementary information, often encoded in clausal syn-
tactic constituents, is likely to be prosodically separated, scope disambigu-
ation may demand prosodic integration, i.e. phrasing of a parenthetical in
one domain together with host material. Second, ‘prosodic parenthesis’ is
not identical to ‘syntactic parenthesis marked in the prosody’. If parenthesis
“expresses a secondary communication” (Nosek 1973: 100), and if, “[b]y
definition, the parenthesis interrupts the prosodic flow of the frame utter-
ance” (Bolinger 1989: 185), then any string of words suitable for express-
ing a secondary communication and marked prosodically may function as a
parenthetical, whether the string is a syntactic constituent or not and even
if it does not belong to one of the types listed in Chapter 1. For example,
the that-clause in (2.43)a, from Bolinger (1989: 193), is a restrictive rela-
tive clause in form and meaning, realized with parenthetical prosody, indi-
cated here by dashes (see also Arnold 2007: 282). In (2.43)b/Figure 2.2, the
conjoined phrase and plays, not a parenthetical in the syntax, is interrupted
by the CC I think and the whole string and I think plays has parenthetical
prosody and is one prosodic constituent (e.g. pauses of 335 ms before and
and 280 ms after plays, and pitch expansion; see also Wichmann 2001: 188).
Based on these prosodic properties, the whole string can be interpreted as
one parenthetical insertion. Similarly, if a token of one of the types listed in
Chapter 1 is realized without parenthetical prosody (i.e. without being set
off from the host by prosodic features), it does not, strictly speaking, qualify
as prosodic parenthesis (see also Dehé 2007: 266–7).
(2.43)  a. The story I told you – that you enjoyed so much, remember? – was
one I made up for our daughter when she was six.
b. And one must also remember that uh the same Arnold Bax
has written poetry and I think plays under the pseudonym of
Dermot O’Brien (ICE-GB: s1b-032 #114)
That is, if any string which is a parenthetical according to its syntactic
properties fails to “interrupt the prosodic flow” of the utterance (Bolinger
1989: 185), syntactic and prosodic parenthesis do not coincide. The lack of
300

250
Pitch (Hz)

200

150

100

50
L+H* L+!H* L-L % L+H* L-L % L+H* !H* L-L % L+!H* L+!H* L+H* L-L %

CC

the Same Arnold Bax has written poetry <SIL> and I think plays <SIL> under the pseudonym of Dermot O’ Brian

(yeah)

0 6.242
Time (s)
Figure 2.2  Example (2.43)b: prosodic, but not syntactic, parenthesis
42  The syntax and prosody of parentheticals

correspondence may be viewed from two perspectives. First, we have to be


clear about whether we speak about ‘prosodic realization of syntactic par-
enthesis’, or whether ‘prosodic parenthesis’ is at issue. Prosodic parenthesis
may be seen as a different theoretical notion to begin with, relying exclu-
sively on prosodic criteria. As Wichmann (2001) rightly points out, there
is always the issue of circularity in methodology and argumentation. “To
investigate the possibility that there are different ways of realising parenthet-
ical elements, we need to identify … parenthetical structures on the basis of
independent, non-prosodic criteria [and the] prosodic features of these will
give us a clear idea of what speakers do with parentheticals” (Wichmann
2001: 181). If other elements are realized with the same features, then these
elements can be argued to be treated by the speaker as parenthetical, regard-
less of whether or not they are parentheticals syntactically. Non-prosodic
criteria include syntactic ones, but again one has to be careful, because for
types such as NRRCs and apposition it may be the prosody which decides
between a non-restrictive (loose, parenthetical) meaning and a restrictive
(close, non-parenthetical) one (Bolinger 1989: 191; Wichmann 2001: 182;
Heringa 2011: 2, among many others), and again there is the issue of cir-
cularity in the argumentation. According to Wichmann, another helpful
criterion, if present, is backtracking (see Section 1.1). The present study
focuses on the prosodic realization of syntactic parenthesis.
Second, a potential mismatch between syntactic parenthesis and pros-
odic constituency may be due to the fact that syntax–prosody interface
constraints may be overridden by other constraints. It has long been under-
stood on the basis of theoretical and experimental research in prosody that
prosodic phrasing in general and IP boundary placement in particular are
not determined by syntax alone, even if the syntactic boundaries are strong
(i.e. if they are clause boundaries). Other relevant factors include syntac-
tic length/complexity (e.g. Cooper and Paccia-Cooper 1980; Gee and
Grosjean 1983; Nespor and Vogel 1986; Ferreira 1991; Watson and Gibson
2004), prosodic length in terms of number of syllables or number of pros-
odic words in one phrase (e.g. Frota 2000; Sandalo and Truckenbrodt 2002;
Jun 2003; Prieto et al. 2005; Elordieta, Frota and Vigário 2005; D’Imperio
et al. 2005; Hellmuth 2008), prosodic weight in terms of prosodic branching
or number of accents (e.g. Selkirk 2000; Sandalo and Truckenbrodt 2002;
Watson and Gibson 2004; Elordieta et al. 2005; D’Imperio et al. 2005), bal-
anced prosodic constituent size (e.g. Gee and Grosjean 1983; Ghini 1993;
Frota 2000), eurhythmy (e.g. Truckenbrodt 2007; Selkirk 2011 for reviews),
performance factors such as speech rate (e.g. Nespor and Vogel 1986; Ghini
1993; Frota 2000; Jun 2003; Hellmuth 2008) and style of speech (e.g. Nespor
and Vogel 1986), focus and contrastive prominence (e.g. Ferreira 1993;
Selkirk 2000, 2005; Jun 2003), pragmatic goals (e.g. Schafer et al. 2000) and
semantic coherence (e.g. Selkirk 1984; Frazier, Clifton and Carlson 2004).
Particularly relevant in the present context is the strong empirical evidence
Prosodic features of parentheticals  43

that intonational phrases and syntactic phrases are not always isomorphic. It
seems that while it has been claimed that certain syntactic structures such as
parentheticals, NRRCs, preposed adverbials and certain other moved elem-
ents, tag questions, expletives, and vocatives are obligatorily parsed into sep-
arate IPs (e.g. Nespor and Vogel 1986: 188, 190; Selkirk 1984: 295, 1995:
567; Bing 1985: Chapter  2), various factors have been identified as being
involved in the placement of IP boundaries in more general terms.
Moreover, even though the assumption about the intonational separate-
ness of parentheticals is generally accepted, it has also been argued that the
intonational features of parentheticals specifically depend on various factors,
among them their length/relative weight (e.g. Schubiger 1958; Bolinger
1989), syntactic make-up (Bolinger 1989) and position (e.g. Bolinger 1989),
and that any one of the defining prosodic features can thus be suspended
(e.g. Bolinger 1989; Wichmann 2000). It has in fact been shown in previ-
ous work that certain types of relatively short parentheticals such as com-
ment clauses, reporting verbs, question tags, vocatives and one-word
parentheticals such as parenthetical what may come without a preceding
pause and may be prosodically integrated into either the preceding or fol-
lowing intonation domain (e.g. Armstrong and Ward 1926; Schubiger 1958;
Crystal 1969; Taglicht 1998; Wichmann 2001; Gussenhoven 2004; Peters
2006; Wells 2006; Dehé and Kavalova 2006; Dehé 2007, 2009; Dehé and
Wichmann 2010a). Peters’ (2006) data reveal a link between prosodic inte-
gration and the size of the interpolation in terms of number of syllables,
such that shorter parentheticals are more likely to be prosodically integrated
than longer ones. For NRRCs, it is the result of psycholinguistic research
that the placement of an intonational boundary before the NRRC is far from
obligatory; instead, it depends on position in the sentence along with dis-
course status (Watson and Gibson 2004). Based on the acoustic analysis
of 157 NRRCs from spoken corpora of English, Auran and Loock (2011)
point out the “descriptive inadequacy of a global characterisation of par-
enthetical structures, based on such traditionally advocated characteristics
as lower pitch … diminished loudness … and onset depression” (Auran
and Loock 2011: 193) and underline the role of the surrounding discourse.
Astruc-Aguilera (2005: 5)  emphasizes that parentheticals do not “form a
homogenous group from a prosodic point of view” and that they are more
homogeneous with regard to prosodic phrasing than with regard to their
intonational properties. However, according to Astruc-Aguilera (2005), it is
compulsory only for some types of parentheticals to form prosodic constitu-
ents of their own.
Finally, one has to be careful as to which prosodic framework the work is
based in. A lot of work on parentheticals (e.g. Crystal 1969; Bolinger 1989;
Wichmann 2001) is couched in the British tradition of intonational analysis
(e.g. Palmer 1922; Crystal 1969; Kingdon 1958; O’Connor and Arnold 1971),
while other work is based in the Autosegmental-Metrical (AM) framework
44  The syntax and prosody of parentheticals

(Pierrehumbert 1980; Beckman and Pierrehumbert 1986; Pierrehumbert


and Hirschberg 1990) and a syntax–prosody relationship along the lines
of the edge-based and relation-based accounts of Selkirk (1986, 1995) and
Nespor and Vogel (1986), respectively. The two frameworks (British vs. AM)
differ, among other ways, in the nature of the prosodic hierarchy. While in
the British framework linguists usually assume only one major phrasal pros-
odic level known as the tone group or tone unit, in the AM system there are
at least two phrasal prosodic constituency levels above the prosodic word:
the prosodic phrase (or accentual phrase, phonological phrase, intermediate
phrase) and the intonational phrase (see Chapter 3 below). The difference is
important in the present context since a parenthetical analysed as prosodi-
cally separate in the British framework may at the same time be analysed as
prosodically integrated (i.e. not phrased in its own intonational phrase) in
the AM framework. The two frameworks and their relevance in the present
context will be further outlined in Chapter 3 below, and the notion of pros-
odic separation will be defined for the purpose of this study.
The following section will review from previous research the syntactic
and prosodic properties of the six types of parentheticals selected for fur-
ther prosodic analysis in this study. They will be addressed in turn.

2.3  The syntax and prosody of six selected types of parentheticals


The empirical study in Chapter 4 below will focus on six types of parentheti-
cals in particular. The current section will therefore concentrate on approaches
to the syntax of these types and on their prosodic properties, as suggested
in previous work. The six types are: full parenthetical clauses (FPCs), non-
restrictive relative clauses (NRRCs), nominal appositions (N-APPs), com-
ment clauses (CCs), reporting verbs (RVs) and question tags (QTs).

2.3.1  Full parenthetical clauses


In this study, a full parenthetical clause (sentential parenthetical in Dehé
2009) is a complete, non-elliptical clause which belongs to one of the follow-
ing types (see also (1.1) and (1.2) in Chapter 1 above):
(i) interpolations which are syntactically complete and could stand alone
as independent sentences (see (2.44)a–c and f for declaratives, (2.44)d
and e for interrogatives);
(ii) clausal as-parentheticals (e.g. Potts 2002b; see (2.45));
(iii) and-parentheticals (e.g. Blakemore 2005; Kavalova 2007; see (2.46)).

(2.44) Complete, independent sentences


a. It’s been a mixture of extreme pleasure I’ve had hundreds of
letters from all sorts of people who have enjoyed the book and
Syntax and prosody of six selected types of parentheticals  45

c­ onsiderable irritation because of being constantly interviewed


(ICE-GB: s1b-046 #2)
b. Newcastle and North you find uhm there’s a marvellous walled
garden I don’t know where it is with hyacinths (ICE-GB: s1a-
065 #298)
c. So what we can do in fact I’ll just turn it off is to use that sig-
nal to train people’s ability to perceive voicing distinctions in
speech and their ability to actually produce them (ICE-GB:
s2a-056 #87)
d. Well esterases are able in organic solvents to carry out a number
of useful can you hear me all right now organic uhm processes to
produce things like food products (ICE-GB: s2a-034 #13)
e. That’s a little bit of uh how shall I put it uh uhm uh arrogance
that has still got to be eliminated uh in my life (ICE-GB: s1b-
041 #13)
f. Joop  is  gisteren – het  verbaast  me  trouwens  niets – gezakt
Joop  is  yesterday  it  surprises  me  by.the.way nothing  failed
voor  het  tentamen.
for  the  exam
‘Joop has – it doesn’t surprise me at all – failed the examination
yesterday.’ (de Vries 2012b: 185)
(2.45) Clausal as-parentheticals
a. Ames, as the FBI eventually discovered, was a spy. (Potts
2002b: 624)
b. Heliopolis today is effectively a suburb a north-eastern suburb
of Cairo and is as you can see perhaps from this slide remarkably
industrial wasteland apart from the archaeological sites the
green area and the cultivation which represents the main tem-
ple of the ancient city (ICE-GB: s2a-026 #54)
c. Secondly as most reviewers say, at the end of the book we still
don’t fully understand where Bush is coming from. (Lee-
Goldman 2012: 588)
(2.46) and-parentheticals
a.  In fact I was very candidly told and I repeat my acknowledge-
ment of the candour that it was placed before him in January last
(ICE-GB: s2a-063 #83)
b.  Will the Minister confirm that come the single uh Common
Market that three hundred million EEC nationals could and I
emphasise could seek employment in this country without the
need to obtain a work permit (ICE-GB: s1b-059 #40)
c.  And during the last five years and I take the figures quite arbi-
trarily the gross national product has gone up by forty per
46  The syntax and prosody of parentheticals

cent in that period eighty-five to eighty-nine (ICE-GB: s2b-


036 #101)
d. I had to take the whole class – and I’m talking about a hundred
students – all around the campus until I found an empty lecture
theatre. (Blakemore 2005: 1166)
Complete independent sentences qualify perhaps most straightforwardly
as parentheticals. They may be anchored (see (2.44)b), floating (see (2.44)e)
or detached (see (2.44)c, d); they contribute additional information to the
discourse, their propositional meaning being independent of that of the
host. As shown most recently by de Vries (2012b), they behave as main
clauses: typical structural and semantic/pragmatic characteristics of main
clauses apply, yet they are linearly integrated in the syntactic host. For
example, as illustrated in (2.44)f, parenthetical main clauses in Dutch are
verb second (V2) while subordinate clauses are verb final, they may contain
high adverbs such as by the way and may be speaker-oriented. Other paren-
thetical (or semantic main clause) properties include strong islandhood for
movement and scope, and binding phenomena such as Condition C effects,
as illustrated for main clauses in (2.12)a above (see de Vries 2012b: 184–6).
Semantically, their meaning is interpreted independently from that of the
host. They contribute discourse-new meaning, adding information which
may help the hearer to interpret the main proposition. For example, the par-
enthetical in (2.44)a provides a reason for the extreme pleasure felt by the
speaker: many readers enjoyed her book. The interrogative parenthetical in
(2.44)e conveys to the listener that the speaker is uncertain about the word-
ing and that the host proposition may have to be interpreted with caution.
As root clauses, full parenthetical clauses are commonly assumed to be pro-
sodically independent.
According to Potts (2002b), as-parentheticals are as-clauses with a miss-
ing constituent, involving movement of a null operator from the comple-
ment position of the verb to the specifier of the CP which is the complement
of the preposition as. Evidence for the movement analysis comes from syn-
tactic island facts and (pseudo-) parasitic gaps. Potts argues for adjunction
of the as-parenthetical to the host. The parenthetical nature, or, in Potts’
(2002b: 649–50) words, “the intuition that parentheticals contrast with
other adverbials in being somehow separate from the rest of the sentence”,
is argued to be a semantic effect. A feature which makes “parentheticals in
general, and As-clauses in particular” structurally unique is, according to
Potts (2002b: 650), their “distinctive comma intonation”, without which, he
argues, they are ungrammatical. Potts (2005) marks as-parentheticals, like
other types of parentheticals, with the comma feature, signalling CI seman-
tics and intonational separation.
According to Blakemore (2005: 1166), “and-conjuncts are parentheti-
cals in the sense that they are discontinuous constituents … licensed by
Syntax and prosody of six selected types of parentheticals  47

grammar even though they have no syntactically specified function in the


structure that contains them (they are not heads, complements, specifiers or
adjuncts)”. Blakemore (2005: 1167) maintains that and-parentheticals “are
the result of a deliberate stylistic choice” rather than an example of dis-
fluency. Couched in a relevance theoretic framework (Sperber and Wilson
1986/1995), Blakemore (2005) argues that and-parentheticals are used in
the pursuit of optimal relevance (see also Kavalova 2007). Based on 70
corpus examples, Kavalova (2007: 147–9) shows that the surface positions
of and-parentheticals are varied, including the position between a lexical
head and its phrasal or clausal complement (e.g. (2.46)a), between copula
and predicate, between an adverbial and the main clause it adjoins to (e.g.
(2.46)c), between subject and verb (e.g. (1.2)a in Chapter  1), and within
the verbal complex (e.g. (2.46)b), with a tendency to occur at syntactic XP
edges rather than within an XP. And-parentheticals behave differently from
standard sentential coordination with regard to their contribution to utter-
ance interpretation (Blakemore 2005) and in terms of their syntactic prop-
erties, e.g. the way they license ellipsis and the readiness with which they
lend themselves to omission (Kavalova 2007). In de Vries’ (2009, 2012a:
159)  account, and in and-parentheticals is a spell-out of the Par head.
Consequently, it bears information about the clause it introduces such that
this clause is marked in the syntax as parenthetical, which may be used by
the phonological component to begin a new intonational domain, thus sug-
gesting prosodic separation for and-parentheticals. Potts (2005) assigns a
comma feature to and-parentheticals, signalling comma intonation and CI
interpretation.
Based on their syntactic and semantic properties as discussed in previous
literature, the prediction at this stage is that all three subtypes of full paren-
thetical clauses are phrased in their own separate intonational domain. This
is formulated in (2.47) and (2.48) below.
(2.47) Prosodic phrasing of full parenthetical clauses: hypotheses
The default prosodic phrasing of full parenthetical clauses is sep-
aration. Parenthetical clauses are phrased in their own intonational
domain, preceded and followed by the respective boundaries.
(2.48) Full parenthetical clauses (FPC): predicted prosodic phrasing
prosodic separation: (…) IP[…]IP IP[FPC]IP (IP[…]IP) (…)

2.3.2  Non-restrictive (appositive) relative clauses


Non-restrictive (or appositive) relative clauses (NRRCs) have received a
lot of attention in the literature (e.g. Emonds 1979; Safir 1986; Fabb 1990;
Burton-Roberts 1999; Del Gobbo 2007; Arnold 2007; Loock 2010; de Vries
2012b; Lee-Goldman 2012). Examples of NRRCs are given in (2.49) and
(2.50).
48  The syntax and prosody of parentheticals

(2.49) NRRCs
a. My brother, who is an engineer, lives in America. (Quirk et al.
1985: 1240)
b.  I bought the cheapest book, which was not a paperback. (Arnold
2007: 272)
c.  Too much sun made these tomatoes, which we paid a lot for, rot
on the vine. (Emonds 1979: 211)
d. Kim won the race, which was a relief. (Arnold 2007: 272)
e.  Kim put it on his back, which was the right place. (Arnold
2007: 274)
f.  We persuaded Bill, who is a good person, most people agree, to
leave immediately. (Loock 2010: 10)
(2.50) NRRCs
a. Walker gets a hand on it but the ball in fact runs out to long-
on where Andy Roberts who is by the way the smallest first-class
cricketer we checked it up earlier on does the fielding and that’s
the end of Worcestershire’s innings (ICE-GB: s2a-013 #137)
b. But Sheedy who normally appears as the girlfriend in movies like
Wargames The Breakfast Club and Short Circuit clearly relishes
this unaccustomed leading role and hurls herself into it with
engaging enthusiasm as if she really believed all the nonsense
(ICE-GB: s2b-033 #105)
c. Diesel petrol has other pollutants Kate which are just as danger-
ous (ICE-GB: s1a-096 #157)
d. But what has amazed me about a lot of Elgar’s music is that
since then I don’t think uh a great deal of his music’s played
in Germany uh which I think is a great pity (ICE-GB: s1b-
032 #50)
e. and people now do things like art therapy and dance therapy
uh which is great (ICE-GB: s1a-004 #94)
f. ‘Well, on the up side,’ said Ron finally, who was sitting watch-
ing the skin on his hands regrow, ‘we got the Horcrux …’ (J. K.
Rowling, Harry Potter and the Deathly Hallows, p. 441)
NRRCs are modifiers that add information to an entity in the host clause
which is not essential for identification (Quirk et al. 1985: 1239), i.e. they
contribute supplementary information in Grenoble’s (2004) sense. Unlike
restrictive relative clauses, NRRCs do not delimit the interpretation of
their referent. The NRRC in (2.49)b, for example, “simply [adds] infor-
mation” about the cheapest book, but its “interpretation is otherwise not
affected”; if which was not a paperback were a restrictive relative clause,
“the object of bought [were] understood to be the cheapest object in the
intersection of ‘books’ and ‘things’ which are not paperbacks” (Arnold
2007: 272). Often NRRCs modify proper nouns, which have a single
Syntax and prosody of six selected types of parentheticals  49

referent by their very nature, and the function of the NRRC is to add
further information about this referent. The referents of Andy Roberts in
(2.50)a and of Sheedy in (2.50)b, for example, are well known and are not
affected by the respective NRRCs, which only serve to add more informa-
tion. NRRCs are often optional, and they can usually occur as a separate
clause (e.g. Arnold 2007: 272; Quirk et al. 1985: 1258–9). They are used
to modify not only noun phrases, but other categories as well. In (2.49)d,
for example, the NRRC which was a relief modifies the entire main clause
Kim won the race: the relief was not about the race alone, but the fact that
Kim won it. In (2.49)e, the NRRC attaches to the PP on his back. In (2.50)
d, which … is a great pity relates to the assertion that Elgar’s music has
not been played much in Germany. See, for example, Arnold (2007: 274),
Loock (2010: 16) and Lee-Goldman (2012: 577) for further examples of
NRRCs relating to categories other than NP. Some NRRCs display attach-
ment ambiguities. For example, which is great in (2.50)e may relate to a
nominal anchor (dance therapy) or to the fact that people now do things like
that. Burton-Roberts (1999: 39)  takes the fact that a main clause can be
antecedent to the wh-expression in NRRCs as evidence for the syntactic
independence of NRRCs.
NRRCs have often been assumed to be outside the syntactic structure
of their containing host and to be structurally identical to other kinds of
parentheticals. For example, Emonds (1979: 216) emphasizes the “essential
similarity between parentheticals and appositive relatives”. Safir (1986) con-
siders both NRRCs and parenthetical expressions ‘extra’ constituents which
are attached to independently grammatical sentences and are accounted
for at the level of his proposed LF′. Fabb (1990: 75) considers Haegeman’s
(1991) radical orphanage approach correct, according to which a NRRC “is
not syntactically related to the sentence which contains it”. Burton-Roberts
(1999) reviews the properties of NRRCs as representatives of parentheticals.
Sentences (2.49)f and (2.50)c show that NRRCs, like other types of paren-
theticals, may co-occur with other extra-sentential elements. In (2.49)f, the
clausal interpolation most people agree directly follows and adds a comment
on the NRRC who is a good person. In (2.50)c, the vocative Kate is wedged
between the NRRC and the NP other pollutants which it relates to, also
showing that NRRCs do not always have to be immediately adjacent to their
anchor. Another example of occasional non-adjacency of the NRRC and its
nominal anchor is (2.50)f, where the adverb finally intervenes between the
anchor Ron and the NRRC.
It has been observed that NRRCs share structural properties with restrict-
ive relative clauses, but that they behave like main clauses according to
semantic/pragmatic criteria (e.g. Arnold 2007; de Vries 2012b). Accordingly,
both integrated (i.e. adjunct) and non-integrated (i.e. orphanage) syntactic
approaches to NRRCs have been suggested in the literature (for overviews
see de Vries 2006: 231–4; Loock 2010: 21–35; and Heringa 2011: 122–9).
50  The syntax and prosody of parentheticals

Integrated approaches have their justification in the structural similarities


between restrictive relative clauses on the one hand and NRRCs on the other;
the many differences between the two types of relative clause are then con-
sidered to be in the semantics/pragmatics (e.g. Arnold 2007).2 For example,
like restrictive relative clauses and other types of subordinate clauses, NRRCs
are verb final in Dutch and German, languages which are verb second (V2)
in main clauses. Other contexts in which NRRCs behave like subordin-
ate clauses include left dislocation and fronting. However, unlike subordin-
ate clauses, but like other parenthetical clauses, NRRCs have the following
syntactic-semantic properties. Main-clause negation does not license a nega-
tive polarity item in the NRRC (e.g. Burton-Roberts 1999; Arnold 2007; see
(2.51): the grammaticality of (2.51)a shows that the negative polarity item is
licensed in restrictive relative clauses). A pronoun inside a restrictive relative
clause, but not inside a NRRC, can be bound by a quantifier in the host (e.g.
Burton-Roberts 1999: 35; de Vries 2007: 212; see (2.52)). NRRCs express
separate speech acts and the illocutionary force of a NRRC is independent of
that of the host (e.g. Arnold 2007; de Vries 2007; see (2.53)). For example, the
question expressed by the host clause in (2.53)b is a question about linguists
in general, not just those that use the IPA. The NRRC is not under the scope
of a modal operator such as probably in (2.54). Referential expressions in the
NRRC may be co-indexed with those in the host without violating Principle
C of the Binding Theory (Burton-Roberts 1999: 35; see (2.55)).
(2.51) Licensing of negative polarity items (examples from Burton-
Roberts 1999: 35)
a. None of the authors who had any imagination remained with
them. (restrictive)
2
Arnold (2007) argues against the structural identity of parentheticals and NRRCs, suggest-
ing instead that NRRCs are syntactically like restrictive relative clauses and that the differ-
ences are best explained in the semantics. At least one piece of empirical evidence for the
similarities between the two types of relative clauses is, in my view, problematic. He argues
that one property that NRRCs and restrictive relative clauses have in common is the fact
that a “normal parenthetical” (p.  286) may intervene between the relative clause and its
antecedent; see (ia) and (ib), from Arnold (2007: 286).
(ia) On the bridge we saw Horatio – I think – who cried out defiantly. (non-restrictive)
(ib) On the bridge we saw a centurion – I think – that cried out defiantly. (restrictive)
He argues that this is to be expected if the syntactic structures of NRRC and restrict-
ive relative clause are very similar. However, in my view, it does not follow that NRRCs
must have the same syntax as restrictive relative clauses in order to allow the same kind of
intervention. On the contrary, if NRRCs are par-merged parentheticals, it follows that their
syntactic relation with the host is loose enough to permit other parentheticals to co-occur.
As was shown above, NRRCs, like other parentheticals, may co-occur with other kinds of
parentheticals and they do not always have to be adjacent to their anchor. Furthermore,
parentheticals of the kind used by Arnold (i.e. CCs such as I think) occur in many different
positions, e.g. even between a lexical head and its complement (see Chapter 1). It does not
follow from this, for instance, that there is a structural similarity between N and relative
clause on the one hand and N and complement on the other hand.
As de Vries (2012b) shows, NRRCs are best analysed as parentheticals despite their struc-
tural similarities with restrictive relative clauses.
Syntax and prosody of six selected types of parentheticals  51

b. *None of the authors, who had any imagination, remained with


them. (non-restrictive)
(2.52) Pronoun binding (examples from Burton-Roberts 1999: 35)
a. She gave every boyi who/that cleaned hisi teeth well a new
toothbrush. (restrictive)
b. *She gave every boyi, who/that cleaned hisi teeth well, a new
toothbrush. (non-restrictive)
(2.53) Illocutionary force
a. Does Jake, who I met last week, own a car?  (de Vries 2007: 217)
b. Are linguists, who use the IPA, invariably clever people? (Arnold
2007: 276)
(2.54) Scope of a modal operator
Jake probably said that Mary, who is my sister, took a few days off.
(de Vries 2007: 217)
(2.55) Binding (Principle C)
Johni gets on well with those firms, who employ him/Johni fre-
quently. (Burton-Roberts 1999: 35)
Given these properties  – NRRCs sharing certain (structural) properties
with restrictive relative clauses and others with main-clause parentheti-
cals – de Vries (2012b) argues that NRRCs, like non-restrictive appositions,
are parentheticals embedded in an abstract DP (see (2.56)), the D head of
which has specific indefinite semantics. This analysis takes care of and sepa-
rates “the two meaning components involved, namely the relativization part
and the appositive part” (de Vries 2012b: 188). I follow de Vries’ analysis
in this study. Potts (2005) promotes syntactic adjunction. NRRCs, like as-
parentheticals and nominal appositions among others, he argues, have CI
semantics. They are associated with the comma feature, interpreted as inde-
pendent of the at-issue content and pronounced with IP boundaries at their
edges (see Section 2.1 above).
(2.56) NRRCs: syntax according to de Vries (2012b: 188)
[host clause … [ParP [DP antecedent] [Par [DP D [CP relative clause]]]] …]
In writing, NRRCs are typically separated from their host by commas
(Quirk et al. 1985: 1258; Arnold 2007: 272), even if punctuation may not
be a reliable cue (see the discussion in Loock 2010: 12–14 and references
given there). In the prosody, they are commonly assumed to be phrased in
a separate intonation domain (e.g. Schubiger 1958: 103; Rappaport 1983;
Selkirk 1981: 131, 1984: 296; Gussenhoven 2004: 287; Arnold 2007: 272,
280). In some prosodic research, the properties of NRRCs have been tested
experimentally. For example, Astruc-Aguilera (2005: 70–1) compares the
prosodic properties of the NRRC in (2.57)a and the restrictive equivalent in
52  The syntax and prosody of parentheticals

(2.57)b, read by a female speaker. She finds that (2.57)a is phrased accord-
ing to (2.58)a, i.e. the NRRC, the material preceding it, and the material
following it each form a separate domain, separated by high edge tones.
The restrictive relative clause in (2.57)b, on the other hand, was phrased
with the preceding material, resulting in two domains with a boundary fol-
lowing the relative clause (see (2.58)b). In Astruc-Aguilera’s data, NRRCs
are prosodically separate, usually preceded and followed by pauses. From a
semantic perspective, Potts (2005) arrives at a similar result. His “analysis
traces all the semantic differences between [restrictive and non-restric-
tive relative clauses] back to comma intonation”, leaving no option for a
non-restrictive meaning to be realized with prosodic integration, or for a
restrictive meaning with prosodic separation. This ties in with Bolinger’s
(1989: 192)  earlier line of argumentation. He argues that “if intonation
responds to restrictiveness and nonrestrictiveness [this is] because those
categories are associated with some discourse factor that triggers the inton-
ational contrast directly”.
(2.57) NRRC vs. restrictive RC
a. Anna’s friends, who were loyal, supported her.
b. The friends who were loyal supported her.
(2.58) NRRC vs. restrictive RC: intonational phrasing
a. IP[Anna’s friends]IP IP[who were loyal]IP IP[supported her]IP
(non-restrictive)
b. IP[The friends who were loyal]IP IP[supported her]IP
(restrictive)
However, neither prosodic separation of NRRCs nor non-separation
of restrictive relative clauses is obligatory. For example, Bolinger (1989:
193) argues that “prosodic separation does not necessarily signify nonrestric-
tion”. He discusses cases of restrictive relative clauses which are nevertheless
prosodically separate, for example due to the speaker’s planning process (see
his examples in (2.59) and the discussion in Bolinger 1989: 193–4).
(2.59) Relative clauses: restrictive yet prosodically separate (from Bolinger
1989: 193)
a. What man, who is really a man, would do a thing like that?
b. The story I told you  – that you enjoyed so much, remember?  –
was one I made up for our daughter when she was six.
Moreover, psycholinguistic research has shown that the placement of an
intonational boundary before the NRRC is far from obligatory; instead,
it depends on the position in the sentence along with discourse status. In
their experimental reading study, Watson and Gibson (2004) found that
almost 40 per cent of all tested NRRCs were not directly preceded by an
intonational boundary. They speculate that “[t]he absence of commas in
Syntax and prosody of six selected types of parentheticals  53

the stimuli may have biased the speakers against producing more inton-
ational breaks” and that “[i]n spontaneous speech, speakers’ performance
may more closely match” the intuition that NRRCs must be preceded by
an intonational break (p. 749). Watson and Gibson (2004) only report on
the percentages of test items lacking a phrasal boundary in the position
directly preceding the relative pronoun; they do not detail the actual posi-
tions of attested boundaries. Despite Watson and Gibson’s (2004) intu-
ition that spontaneous speech may feature more prosodic breaks before
NRRCs, Selkirk (2005) takes their results as evidence for an asymmetry in
the phrasing of parentheticals such that there is right alignment but no left
alignment of syntactic parenthetical and prosodic IP (see also Chapter  3
below). However, results from corpus studies using natural spontaneous
or semi-spontaneous speech rather seem to support Watson and Gibson’s
(2004) intuition about speaker behaviour in spontaneous speech (see Dehé
2009; Auran and Loock 2011, and Section  4.3.2 below). In their corpus
study, Auran and Loock (2011) found that all 157 NRRCs in their data set
were realized as one or more separate intonation units, with boundaries at
their left and right edges. Long and complex NRRCs in their data set, such
as those reproduced in (2.60) (from Auran and Loock 2011: 184), spanned
more than one intonation unit, but with boundaries at the onset and end
of the NRRC and with additional boundaries in structurally predictable
positions. In (2.60)a, for instance, there was an IP boundary before since, in
(2.60)b between weather and with. Looking at a number of tonal, temporal
and intensity criteria, Auran and Loock found differences within the data
set depending on the discourse function of the NRRCs (three discourse
categories: relevance, subjectivity, continuance; see Loock 2007, 2010).
(2.60) Long and complex NRRCs spanning more than one intonation unit
(Auran and Loock 2011: 184)
a. Israelis have sympathy and liking for Americans which is just as
well since the country is swarming with transatlantic visitors.
b. Northern Scotland will have occasional light rains which will be
followed during the day by colder but still mainly cloudy weather
with a few sleet and snow showers.
Based on this previous work, the default assumption for NRRCs is pros-
odic separation. This follows from the semantic properties of NRRCs and
their respective parenthetical structure (Potts 2005; de Vries 2012b), and
from previous prosodic research. Some of the prosodic literature discusses
the prosodic properties of NRRCs in relation to restrictive relative clauses
such that intonational phrasing is one feature that distinguishes between the
two types and disambiguates between the respective interpretations (see, for
example, Astruc-Aguilera’s 2005 experimental work; Selkirk 2005: 14–15).
Other experimental work did not confirm this strong tendency of separ-
ation (Watson and Gibson 2004). However, according to the same authors,
54  The syntax and prosody of parentheticals

data from spontaneous speech may follow prosodic separation more closely,
an assumption supported by work using corpus data (Dehé 2009; Auran and
Loock 2011). Based on these previous studies taken together, the hypothesis
for the prosodic phrasing of NRRCs is therefore prosodic separation (see
(2.61), illustrated in (2.62)).
(2.61) Prosodic phrasing of NRRCs: hypotheses
The default prosodic phrasing of NRRCs is separation. NRRCs
are phrased in their own intonational domain, preceded and fol-
lowed by the respective boundaries.
(2.62) NRRCs: predicted prosodic phrasing
prosodic separation: (…) IP[…]IP IP[NRRC]IP (IP[…]IP) (…)

2.3.3  Nominal appositions


Nominal apposition is understood here as a relation between two noun
phrases (Quirk et al. 1985: 1300–1), which “have the same syntactic func-
tion … and are offered by the speaker as being in some sense equivalent”
(Burton-Roberts 1999: 25). A prototypical example, taken from Heringa
(2011: 1), is given in (2.63). The appositional construction consists of a nom-
inal anchor in the host (John McClave) and a nominal apposition (N-APP;
my neighbor).
(2.63) Nominal apposition
appositional construction

John McClave, my neighbor, is a nice guy.

anchor apposition (N-APP)

The syntactic and semantic relation between anchor and host has been the
subject of much discussion in the literature (see Heringa 2011 for an over-
view; see Meyer 1992 for an overview and critical discussion of approaches
to apposition until the early 1990s). One distinction that has been made is
between close (or restrictive) apposition on the one hand and loose (or non-
restrictive) apposition on the other. It has often been noted that the dis-
tinction corresponds to the one between restrictive and non-restrictive (or
appositive) relative clauses (see (2.64), from Heringa 2011: 3; see also Quirk
et al. 1985: 1303–4, among others). In (2.64)a, Peter is one among several
brothers; in (2.64)b, there is only one brother and his name is Peter. Heringa
(2011: 2–3) maintains that along with the difference in meaning, there is a
difference in intonation such that loose apposition, unlike close apposition,
Syntax and prosody of six selected types of parentheticals  55

is typically (but not always) set off by comma intonation (see also Quirk et al.
1985: 1303). Heringa further argues that the two types differ in structure,
such that loose appositions, unlike close appositions, may have an indefinite
anchor, that constructions with close appositions but not with loose apposi-
tions allow VP-ellipsis, that the anchor of a loose apposition but not that of
a close apposition may be modified by a PP or a restrictive relative clause,
and that a loose apposition but not a close apposition may be expanded by
an attributive adjective or determiner (see Heringa 2011: 3–5 and references
given there).
(2.64) Close (restrictive) apposition (a) and loose (non-restrictive)
apposition (b)
a. My brother Peter is still at college.
b. My brother, Peter, is still at college.
In addition to the distinction between restrictive and non-restrictive appos-
ition, Quirk et al. (1985: 1302–3) distinguish between full and partial appos-
ition, and between strict and weak apposition. In strict apposition, anchor
and apposition belong “to the same general syntactic class” (Quirk et  al.
1985: 1303), for example two full noun phrases in (2.65)a, while in weak
apposition, anchor and apposition may differ in their syntactic class, for
example a noun phrase as anchor followed by an appositional -ing-clause
in (2.65)b. The data set initially retrieved for the analysis reported on in
Chapter 4 below contained cases of weak apposition such that the constitu-
ent annotated DEFUNC was clausal (e.g. (2.66)); these were discarded and
only strict (nominal) apposition was analysed.
(2.65) Strict (a) and weak (b) apposition
a. Football, his only interest in life, has brought him many friends.
(Quirk et al. 1985: 1303)
b. His only interest in life, playing football, has brought him many
friends. (Quirk et al. 1985: 1301)
(2.66) Weak apposition in the corpus data (excluded from analysis)
a. As for the intriguing question what does she do next Mrs
Thatcher has not apparently given it much thought (ICE-GB:
s2b-003 #107)
b. The most fundamental of all parental wishes to educate our chil-
dren in our own morality is indoctrination and a denial of their
free development (ICE-GB: s2b-029 #147)
Full and partial apposition refers to the difference between omissible and non-
omissible constituents. In full apposition, both anchor and apposition “can be
separately omitted without affecting the acceptability of the sentence” and if
one is omitted, anchor and apposition fulfil “the same syntactic function in
the resultant sentences” and have the same extralinguistic reference (Quirk
56  The syntax and prosody of parentheticals

et al. 1985: 1302; see (2.67) for an example). In partial apposition, this is not
the case. One type of partial apposition occurring in the corpus is represented
in (2.68) and (2.69): the combination of a noun phrase lacking the article, and
a proper noun. In (2.68) the proper noun is the anchor, modified by an appos-
itional NP which lacks the article. Omission of the anchor results in ungram-
maticality. (These cases were included in the prosodic analysis reported on in
Chapter 4.) In (2.69) the anchor, which is a noun phrase lacking a determiner
(British explorer and US Secretary of State in (2.69)a and b, respectively), is fol-
lowed by a proper noun. Omission of the proper noun results in ungrammat-
icality. In cases like these, the apparent anchor becomes a premodifier which
resembles a title (Quirk et  al. 1985: 1317; see also the discussion in Meyer
1992: 47–9). Sequences of premodifier and proper noun such as those given
in (2.69) were excluded from the analysis reported on in Chapter 4. Cases of
“discontinuous full appositions” such as (2.70), also referred to as “a type of
partial apposition” by Quirk et al. (1985: 1302), were included.
(2.67) Full apposition (Quirk et al. 1985: 1302)
a. A neighbour, Fred Brick, is on the telephone.
b. A neighbour is on the telephone.
c. Fred Brick is on the telephone.
(2.68) Partial apposition I
a. Roald Dahl master of the nasty surprise is dead (ICE-GB: s2b-
003 #3)
a1. Roald Dahl is dead.
a2. *Master of the nasty surprise is dead.
b. Captain Richard Sharp editor of Jane’s Fighting Ships is on the
line (ICE-GB: s2b-012 #36)
b1. Captain Richard Sharp is on the line.
b2. *Editor of Jane’s Fighting Ships is on the line
c. Desiree Mills Education Advice Worker is based at the Holmes
Road Centre and provides educational advice and informa-
tion to students and potential students for the whole institute
(ICE-GB: s2b-044 #105)
c1. Desiree Mills is based at the Holmes Road Centre …
c2. *Education Advice Worker is based at the Holmes Road
Centre …
(2.69) Partial apposition II
a. British explorer Sir Ranulph Fiennes reached both Poles with
support in a transglobe expedition of nineteen eighty-one
(ICE-GB: s2b-024 #8)
a1. *British explorer reached both Poles with support in a
transglobe expedition of nineteen eighty-one.
Syntax and prosody of six selected types of parentheticals  57

a2. Sir Ranulph Fiennes reached both Poles with support in


a transglobe expedition of nineteen eighty-one.
b. US Secretary of State James Baker says the American peace
offer is still on the table (ICE-GB: s2b-016 #69)
b1. *US Secretary of State says the American peace offer is
still on the table.
b2. James Baker says the American peace offer is still on the
table.
(2.70) Discontinuous full apposition
a. Graham Taylor has been saying the England manager that this is
perhaps the toughest test that he and his team have faced in his
seven internationals so far (ICE-GB: s2a-001 #2)
b. One allied plane was lost an American Marine Harrier (ICE-GB:
s2b-018 #10)
The semantic relations between (loose) appositions and anchor have been
divided into three main classes (e.g. Quirk et al. 1985; Heringa 2011, and
references given there; see also Meyer 1992: 57–73): equivalence (also:
identification, co-reference; see (2.71)), attribution (see (2.72)) and inclu-
sion (see (2.73)). In (2.71), anchor and apposition have exactly the same
referent in the real world, for example in (2.71)c the lady sitting next to
you actually is Marianne Burr, and in (2.71)d, Tanner and Gore Stansworth
are the co-presenters of the speaker. The N-APPs in (2.72) classify their
anchor ((2.72)a) or ascribe a property to it ((2.72)b and c), i.e. they have
an attributive function. In (2.73), there is a part–whole relationship
between the potential referents of the anchor and the referents of the
apposition. In (2.73)c, for example, the agricultural estates are one asset (of
some), i.e. the reference of the agricultural estates is included in the set of
referents of some assets. Like in (2.73)a (in particular) and (2.73)c (princi-
pally), inclusion is often indicated by expressions marking the semantic
relation.
(2.71) Equivalence
a. The lion, the panther leo, is threatened with extinction.
(Heringa 2011: 26)
b. Paul Jobes, the distinguished art critic, died in his sleep last
night. (Quirk et al. 1985: 1301)
c. Uhm the young lady sitting next to you uhm Marianne Burr
was wearing the T-shirt you see (ICE-GB: s1a-068 #64)
d. Tanner and Gore Stansworth my co-presenters have developed a
four-dimensional model of reactions to torture (ICE-GB: s2a-
034 #70)
58  The syntax and prosody of parentheticals

(2.72) Attribution
a. Jan’s pet, a baboon, shows its teeth when it’s angry. (Heringa
2011: 28)
b. The house, an imposing building, dominates the street. (Quirk
et al. 1985: 1313)
c. Uhm gallate an effective anti-oxidant in the food industry is pro-
duced on multi-term quantities … (ICE-GB: s2a-034 #14)
(2.73) Inclusion
a. An ape, in particular Jan’s gorilla, is a good climber. (Heringa
2011: 29)
b. A neighbour, Fred Brick, is on the telephone. (Quirk et  al.
1985: 1301)
c. Some assets principally the agricultural estates are owned dir-
ectly by the trustees and here more traditional ways of man-
agement remain (ICE-GB: s2a-045 #39)
While semantically, exact or partial equivalence between anchor and host is
common, pragmatically the relation is one of addition: the apposition pro-
vides new information about the anchor (Meyer 1992). In Meyer’s (1992)
corpus, 86 per cent of all appositions (including non-nominal appositions,
across spoken and written genres) consisted entirely of new information
about the anchor. The remaining 14 per cent of appositions in his corpus
provided partially new information. This is also in line with Potts’ (2005)
assumption that N-APPs, like other kinds of supplements, provide discourse-
new information, independent of the main proposition of the host.
Meyer (1992) also notes that appositions can have a full range of syn-
tactic functions such as subject, subject of sentences containing existential
there, object, object of preposition, subject complement, etc. In his corpus
analysis, the most common ones are subject (including non-existential and
existential) and object (direct and indirect objects of verbs and objects of
prepositions). Meyer (1992: 34–7, 123f) argues that the frequency of their
syntactic function is related to their relative heaviness: since appositions are
typically syntactically heavy, they prefer right-peripheral positions, i.e. they
promote end-weight.
It is easy to get the impression from the literature that nominal apposi-
tions have to appear right-adjacent to their anchors in the host clause (e.g.
Potts 2005: 104; Selkirk 2005: 15; de Vries 2007, 2009; Heringa 2011: 113,
2012: 565) and typically they do (see (2.74) through to (2.76)). Heringa (2011:
113) maintains that appositions have to be adjacent to their anchor, that the
position indicated by @ in (2.75) is the only possible position for the anchor
(* indicates ungrammatical positions) and that if the anchor is moved, the
apposition has to move along (see (2.76), from Heringa 2011: 113). Likewise,
Potts (2005: 104) claims that supplements, including N-APPs, NRRCs and
as-parentheticals, “must be immediately adjacent to whatever constituent
Syntax and prosody of six selected types of parentheticals  59

they are dependent upon for their interpretation” and exemplifies this using
the N-APP and NRRC examples given in (2.77) and (2.78), respectively.
However, Quirk et  al. (1985: 1302)  have “discontinuous full apposition”,
and closer inspection of spoken language data confirms that appositions, like
NRRCs (see (2.50)c, f above), do occasionally occur in non-adjacent position
(see (2.79), the anchor is underlined). For example, in (2.79)a, the anchor
Graham Taylor is in subject position and is separated from the N-APP the
England manager by the verbal complex. Similarly, in (2.79)b, the N-APP
Mother Megan is separated from its pronominal anchor in subject position
and surfaces between the adjectival predicate notorious and its complement.
According to Heringa (2011: 113), extraposition “to the right periphery of
the sentence” is the only exception to the adjacency requirement (see (2.80)a
from his work; the corpus example in (2.80)b behaves similarly). However,
in the corpus examples in (2.79)a and (2.79)b, the apposition is positioned
between the argument-taking verb and its clausal complement, and between
the predicative adjective and its clausal complement, respectively, thus does
not occur at the right sentence periphery. Examples of non-adjacency are
too infrequent to call into question the syntactic analysis developed in recent
work (e.g. Heringa 2011; de Vries 2012a; see below), but they will have to be
dealt with in some part of the grammar.
(2.74) N-APPs adjacent to anchor
a. Philippa’s daughter Isabelle married Philip the Good of
Burgundy in fourteen twenty-nine … (ICE-GB: s2b-
043 #39)
b. and I think they were just sort of making up for for the embar-
rassment of having to ask me such an experienced artiste to come
in for those two lines (ICE-GB: s1a-092 #51)
(2.75) (Im-)possible N-APP positions (Apposition: a clever boy; Heringa
2011: 113)
* Pete @ has * won * several * quizzes * today *.
(2.76) Apposition: movement of anchor (and N-APP; from Heringa
2011: 113)
a. Peter met George, his best friend, in primary school.
b. George, his best friend, Peter met in primary school.
c. *George Peter met, his best friend, in primary school.
(2.77) N-APPs non-adjacent to anchor (from Potts 2005: 104)
a. *We spoke with Lance before the race, the famous cyclist, about
the weather.
b. *Jan was the fastest on the course, the famous German sprinter,
yesterday.
c. *Lance has, the famous cyclist, taken the lead.
60  The syntax and prosody of parentheticals

(2.78) NRRCs non-adjacent to anchor (from Potts 2005: 104)


a. *We spoke with Lance before the race, who is a famous cyclist,
about the weather.
b. *Jan was the fastest on the course, who is a famous German
sprinter, yesterday.
c. *Lance has, who is a famous cyclist, taken the lead.
(2.79) N-APPs non-adjacent to anchor
a. Graham Taylor has been saying the England manager that this is
perhaps the toughest test that he and his team have faced in his
seven internationals so far (ICE-GB: s2a-001 #2)
b. And she was notorious Mother Megan for dressing up in
Sunday best at all times (ICE-GB: s1b-014 #86)
(2.80) Extraposition
a. I met John yesterday, a really nice guy. (from Heringa
2011: 15)
b. One allied plane was lost an American Marine Harrier (ICE-GB:
s2b-018 #10)
It has also been observed that it is unusual for appositions to have a per-
sonal pronoun as their anchor (e.g. O’Connor 2008: 23). This is true also
for the data set retrieved from ICE-GB (see Chapter 4 below), but pronom-
inal anchors do occur (see (2.79)b, (2.81) and the respective examples in
Chapter 4; see also Meyer 1992: 20).3
(2.81) N-APPs, pronominal anchor
a. and I think they were just sort of making up for for the embar-
rassment of having to ask me such an experienced artiste to come
in for those two lines (ICE-GB: s1a-092 #51)
b. However those close to the President found they did have
to justify the amount they were costing the taxpayer and
one Mr Bush’s right hand man White House chief of staff John
Sinunu was found to have been acting as if he was the presi-
dent or even the vice president when it came to travel matters
(ICE-GB: s2b-021 #26)
Heringa (2011: 107–14, 2012: 564)  discusses a range of properties which
underline the assumption that nominal appositions are parentheses; some

3
One reason for why pronominal anchors are unusual may be their news value, since pro-
nouns refer to given entities, thus giving additional information about them may not be
necessary. At the same time, however, it is not entirely implausible to provide discourse-new
information about familiar referents, a function which may be served by N-APPs, such as
those in (2.81), for example.
Syntax and prosody of six selected types of parentheticals  61

of them are illustrated in (2.82). Among these properties is the prosodic


separateness of appositions as well as syntactic and interpretational phe-
nomena. For example, there is a lack of agreement between elements in the
matrix and elements in the apposition, appositions are ignored in pronoun
interpretation and the interpretation of elided constituents, the apposition
conveys its own proposition, it is independent of the matrix with respect
to illocutionary force, and appositions can contain the same type of coor-
dinators that can also occur in peripheral adverbial clauses (Heringa 2011:
107–9). Furthermore, appositions cannot be the focus of a cleft sentence
(see (2.3) above, repeated as (2.82)a), they can have a temporal reference
independent from the one in the host, they are not selected as arguments
by the predicate of the host clause and every anchor can therefore relate
to multiple appositions (see (2.82)b); appositions are outside the scope of
quantifiers and other operators in the host (see (2.82)c and (2.9)c above),
and an R-expression in the apposition which is co-referential with one in
the host does not lead to Condition C effects (see (2.82)d) (Heringa 2011:
110–12). The last two properties in particular suggest that, like other types
of parentheticals, appositions fail to be c-commanded by elements in the
host and are thus ‘invisible’ (de Vries 2007) to the syntactic structure of the
host. Heringa (2011: 112)  concludes “that appositions indeed are paren-
theses”, which contain side information and are prosodically separate; see
also Potts (2005), whose analysis was presented in Section 2.1 above.
(2.82) Invisibility of N-APPs (from Heringa 2011: 110–12)
a. Joe asked Bill, a famous trumpet player, to teach him.
*It is a famous trumpet player that Joe asked Bill to teach him.
b.  John, Mary’s boyfriend, a syntactician, is a linguistic celebrity.
c.  *No reporteri believes that Ames, often the subject of hisi col-
umns, is a spy.
d. Johni first met Mary, now Johni’s wife, in the linguistic café.
* Johni first met Johni’s wife in the linguistic café.
Based on his conclusion that appositions are parentheticals, Heringa (2011:
139–43) develops a syntactic analysis along the lines of de Vries’ (2007,
2012a) par(enthetical)-merge (see Section 2.1), reproduced in (2.83) below.
Specifically, Heringa’s (2011: 142–3) analysis works such that the appos-
ition is par-merged as a complement to the parenthetical head Par, resulting
in a ParP. According to Heringa, par-merge (sup-merge in Heringa 2011)
accounts immediately for both the loose syntactic relation between appos-
ition and host on the one hand, and the linear requirements on the other.
Via sup-merge, apposition and anchor can form a constituent, while at the
same time the anchor cannot be c-commanded by any element in the host,
including the anchor.
62  The syntax and prosody of parentheticals

(2.83) The syntax of appositions (Heringa 2011: 143; de Vries


2012a: 159)

…ParP…

Anchor Par'
* *
Par Apposition

Based on the assumptions that anchor and apposition are coordinated


by specifying coordination, and that appositions function as predicate
of their anchor, Heringa elaborates on the analysis given in (2.83) (see
Heringa 2011 and 2012 for details). Most importantly in the present
context, Heringa’s parenthetical analysis accounts for the structural and
semantic relations between apposition and host clause, and it allows – in
any kind of Y-model of the grammar – for the mapping of the complex
sentence (host and apposition) to logical form (LF) and phonetic form
(PF) in the same process, without having to postpone the relation to a
later stage in the derivation. The approach can therefore account not only
for structural/semantic relations between apposition and anchor, but also
for potential prosodic effects of the appositional element on the host.
Alternative syntactic analyses include Potts’ (2005) approach, which sug-
gests regular adjunction and marking of the N-APP with a comma fea-
ture (see Section 2.1 above).
As explicitly assumed by Heringa (2011, 2012) and Potts (2005) among
others, the parenthetical syntax of appositions (par-merge or comma fea-
ture, respectively) suggests prosodic separateness. Like Heringa, Meyer
(1992) and Potts (2005) take prosody as one cue to restrictive vs. non-
restrictive nominal apposition. In their view, non-restrictive, unlike
restrictive apposition is marked by comma intonation, thus by prosodic
separateness of the N-APP. Without comma intonation, Potts argues, there
is no CI semantics. Meyer (1992: 46f) attributes another kind of semantic
disambiguation to the presence vs. absence of a comma and corresponding
tone group boundary; see his example in (2.84). Meyer argues that in (2.84)
a, i.e. without a comma and boundary after Holtom, A Labour stalwart is
interpreted as an adverbial with a causal relation to the main proposition,
as in “because he is a Labour stalwart he slated the previous Tory admin-
istration” (Meyer 1992: 47). Cllr Holtom functions as the subject of the
main clause. The presence of the comma between Cllr Holtom and slated
and the corresponding boundary in (2.84)b, on the other hand, leads to
interpretation of A Labour stalwart as anchor and Cllr Holtom as N-APP in
an appositional construction.
Syntax and prosody of six selected types of parentheticals  63

(2.84) Phrasing of N-APP: effect on interpretation


a. A Labour stalwart, Cllr Holtom slated the council’s previous
Tory administration for causing “untold misery to the least
fortunate”
b. A Labour stalwart, Cllr Holtom, slated the council’s previous
Tory administration for causing “untold misery to the least
fortunate”
Apposition has also been addressed in the prosodic literature. Schubiger
(1958: 104) maintains that, like relative clauses, appositions may be restrict-
ive and non-restrictive; while the former form one tone group with their
anchor, the latter contribute additional information and form a separate
tone group. Based on their experimentally elicited corpus of 462 utter-
ances (29 sentences read twice by 8 speakers of Southern British English;
2 items ­discarded), containing various kinds of parentheticals (“extra-sen-
tential elements” in their terminology), Astruc-Aguilera and Nolan (2007b:
93)  state that NRRCs and appositions are usually phrased in a separate
intonational phrase (88 per cent of cases in their corpus) and often redupli-
cate the tonal contour of their anchor at a lower pitch level and with a com-
pressed pitch range. They are usually set off by pauses from the host and if
only one pause is realized, it is the one following the interpolation.
The idea that appositions reduplicate the tonal contour of the anchor is also
present elsewhere. With respect to the actual intonational contour it has been
noted that appositions and non-restrictive relative clauses are in their own
intonational domain, which repeats the intonational contour of the domain of
their anchor. This has been referred to as tonal reduplication, tonal harmony,
tone concord or tonal parallelism (see Palmer 1922: 89–90; Bing 1985: 68;
Cruttenden 1997: 71; Wichmann 2000: 85–93; Astruc-Aguilera 2005; Wells
2006: 85, and references given there). For example, in (2.85), taken from Bing
(1985: 68), if Smith has a falling tone, then bank also has a falling tone.
(2.85) There’s Mr. Smith, who used to manage the bank.
With regard to prosodic phrasing, the discussion results in the hypoth-
eses given in (2.86) and (2.87), again based on the parenthetical syntax and
semantics of N-APPs, but also on previous prosodic literature. At this stage,
identical predictions are thus made for FPCs, NRRCs and N-APPs.
(2.86) Prosodic phrasing of N-APPs: hypotheses
The default prosodic phrasing of N-APPs is separation. N-APPs
are phrased in their own intonational domain, preceded and fol-
lowed by the respective boundaries.
(2.87) N-APPs: predicted prosodic phrasing
prosodic separation: (…) IP[…]IP IP[N-APP]IP (IP[…]IP) (…)
64  The syntax and prosody of parentheticals

2.3.4  Comment clauses


Brinton (2008: 4–7) defines comment clauses (CCs) as “epistemic/eviden-
tial parentheticals”, which are in a linear but not a hierarchical syntactic
relationship with their host sentence, and which are “positionally mobile”
and “semantically independent … expressing speaker attitude” (Brinton
2008: 18). Quirk et al. (1985: 1112) see comment clauses as “parenthetical
disjuncts” which “generally have a separate tone unit”. They distinguish six
types, which resemble (i) “the matrix clause of a main clause”, (ii) “an adver-
bial finite clause (introduced by as)”, (iii) “a nominal relative clause”, (iv) a
“to-infinite clause as style disjunct”, (v) an “ing-clause as style disjunct” and
(vi) an “ed-clause as style disjunct” (Quirk et  al. 1985: 1112–13). For the
purpose of the present study, the term CC refers to a subset of these elem-
ents, specifically to elements such as I think, I suppose, I imagine, I assume,
I gather, I’m afraid and the like (see Quirk et al.’s 1985: 1112–14, “type (i)”
comment clause), which typically consist of a first-person pronoun and a
verb of knowledge, belief or conjecture or a corresponding adjectival con-
struction. In addition, the CCs as it were and if you like (Brinton 2008:
Chapter 7) were included in the present study.
On the surface, CCs consisting of a pronoun and a verb are “defective
syntactically” (Quirk et al. 1985: 1114) such that the verb lacks a comple-
ment which would be obligatory otherwise. Jackendoff (1972: 99) considers
them “semantically one-place arguments” such that “exactly one argument,
the complement sentence, is missing from the functional structure”, result-
ing in a syntactic structure which has a missing or empty node. CCs occur
sentence-finally and sentence-medially between or within syntactic constit-
uents; see the examples in (2.88) (from Emonds 1973: 333; Quirk et al. 1985:
1113; Nespor and Vogel 1986: 190) and the corpus examples in (2.89).4
(2.88) Comment clauses (CCs)
a. John came later than Sue, I think.
b. John came, I think, later than Sue.
c. John, I think, came later than Sue.
d. There were no other applicants, I believe, for that job.
e. Charles wouldn’t, I imagine, have done such a thing.
(2.89) Comment clauses (CCs)
a. And my mother had a I think quite a romantic attachment to
religion but uh didn’t uh go as far as being a regular church-
goer (ICE-GB: s1b-041 #116)

4
It has been argued that parenthetical CCs (and reporting verbs) may also occur sentence-
initially, where they are identical in morphosyntactic form with main clauses (e.g., Hooper
1975; Aijmer 1997; Kärkkäinen 2003; Kearns 2007; see Dehé and Wichmann 2010b for
an overview). Relating their prosodic form to their interpretation in context, Dehé and
Syntax and prosody of six selected types of parentheticals  65

b. Well they dropped cards in I suppose the doors (ICE-GB: s1a-


020 #177)
c. but if there is a unilateral action it will I’m afraid gravely dam-
age this newly emerging UN consensus and the cooperation
between the superpowers (ICE-GB: s1b-035 #101)
CCs such as I think, I believe, I assume, as well as as it were, along with
interrogative parentheticals, question tags and reporting verbs, have often
been argued to attach an illocutionary commitment to an utterance or to
serve metalinguistic functions rather than serving any descriptive func-
tion or contributing to the truth-conditionality of the host utterance (e.g.
Urmson 1952; Hand 1993). CCs may function as mitigators, i.e. they are
used to modify, correct, reinforce or soften a speech act performed by
the host utterance (e.g. Mittwoch 1979; Fraser 1980; Schneider 2007a, b;
Kaltenböck 2010); in Quirk et  al.’s (1985: 1114)  words: “they express the
speaker’s tentativeness over the truth value of the matrix clause” (e.g. I
believe, I guess, I assume) or “they express the speaker’s certainty” (e.g., I
know, I’m sure, I must say, I have to say). Their function can often also be
achieved and CCs can thus often be replaced by message-oriented adverbs
such as probably, presumably, possibly, certainly, definitely, obviously, or atti-
tudinal adverbs such as unfortunately, luckily, happily or surprisingly, with-
out affecting the meaning of the utterance too much (see e.g. Urmson 1952:
486–9 for discussion). Accordingly, CCs have been treated as epistemic
adverbials, pragmatic markers and discourse markers in the literature (e.g.
Aijmer 1997; Thompson and Mulac 1991; Murphy 1993; Kärkkäinen 2003;
Brinton 2008; Dehé and Wichmann 2010a). Rooryck (2001) lists a range of
evidential meanings expressed by CCs. Dehé and Wichmann (2010a) show
that the range of semantic-pragmatic functions is reflected in prosodic real-
ization. Given their evidential meaning, Scheffler (2009) proposes a two-
dimensional semantics for CCs along the lines of (2.90). The CC (I think in
(2.90)) expresses evidentiality such that it lowers the epistemic threshold.
This threshold determines whether the speaker is sure of the proposition. In
(2.90) the speaker is not absolutely certain that ‘Peter will come today’ but
she nevertheless makes the assertion, albeit weakened by the evidential. The
contribution of the CC is twofold: (i) it affects the epistemic threshold, and
(ii) it contributes a side comment in the domain of conventional implica-
tures (CIs): the actual content of the verb, here think. According to Scheffler
(2009: 190), the effect on the epistemic threshold is the main semantic con-
tribution, the expression of the “attitude content” of the verb the secondary
one. It seems then, that CCs are mixed expressions with respect to their
semantics, on the one hand affecting the ­at-issue ­content and on the other
Wichmann (2010b) show that sentence-initial I think (that) and I believe (that) may func-
tion as main clause, CC or discourse marker. The present study excludes sentence-initial as
well as syntactically ambiguous forms from the analysis.
66  The syntax and prosody of parentheticals

having CI semantics, while full parenthetical clauses, NRRCs and N-APPs


are plain CIs.
(2.90) CCs: two-dimensional semantics
a. Peter, I think, will come today.
b. Assertion: T (Peter will come today)
CI: think (I, Peter will come today)
According to Brinton (2008), CCs are the result of a process of grammat-
icalization. She shows, however, that their development cannot be fully and
invariably accounted for in terms of Thompson and Mulac’s (1991) matrix
clause hypothesis (MCH), which holds that sentence-initial 1st person CCs
“are grammaticized forms of subjects and verbs introducing complement
clauses” starting with that (Thompson and Mulac 1991: 317). In other
words, there is no historically solid evidence for an unambiguous gram-
maticalization path from matrix clause to comment clause. Brinton (2008)
demonstrates that the source construction, a matrix clause with a following
that-clause, is often rare in earlier stages of the language, and if it occurs
in earlier periods, it is often a minority form. Furthermore, the occur-
rence of that-deletion does not necessarily increase over time, as would be
expected if that-deletion indicates grammaticalization in progress. In some
cases, parentheticals pre-date matrix clauses, thus making it implausible
that parentheticals develop from matrix clauses. For some CCs, other types
of verbal complements (interrogative, imperative, phrasal) may outnumber
that-complements.
Due to their nature as syntactically incomplete (or defective) clauses
(e.g. Jackendoff 1972: 99; Quirk et al. 1985: 1114; Knowles 1980: 382) and
due to their often adverbial function, the parenthetical analysis of CCs of
the form (1st person) pronoun plus verb is more controversial than that
of other types of parentheticals. The parenthetical analysis, which goes
back at least to Jespersen (1937), is common in the semantic-pragmatic
literature (e.g. Urmson 1952; Thompson and Mulac 1991; Aijmer 1997),
and has also been promoted by Quirk et  al. (1985) and Brinton (2008).
Recent work has added prosodic evidence for the parenthetical status of
CCs (Dehé and Wichmann 2010a). In the syntactic literature, however,
there has been a debate as to whether CCs in English are best analysed
as base-generated parentheticals (e.g. Jackendoff 1972: 94–100; Emonds
1973, 1976; Peterson 1999) or whether they are derived by some syntac-
tic movement operation from an underlying structure which features the
CC as a main clause taking the rest of the sentence as complement (e.g.
Downing 1973; Ross 1973; Hooper 1975; Newmeyer 2012).5 Arguments in
5
The focus in the present study is on English. However, the syntactic debate is not confined
to English CCs. The related verb-initial construction in German, for example, has also
been subject to discussion along similar lines (see, e.g., Reis 1995, 1996, 2002; Grewendorf
1988; Haider 1993; Tappe 1981; Wagner 2004; Scheffler 2009; Viesel 2011). Kiziak’s (2007)
Syntax and prosody of six selected types of parentheticals  67

favour of the main-clause analysis include the superficial relation between


sentences such as (2.88)a–c on the one hand and (2.91) on the other, which
seems to suggest that (2.91) is the underlying structure and (2.88)a–c are
derived from it by syntactic movement across the CC, for example by top-
icalization (e.g. Taglicht 1998: 197) or focus movement, or by some other
stylistically driven operation.
(2.91) I think (that) John came later than Sue.   (Emonds 1973: 133)
Taglicht (1998: 197) argues that the examples in (2.92) (his syntactic brack-
eting ([ ]) and intonational phrase boundaries (%)) “represent two different
syntactic structures” such that I think is a parenthetical in (2.92)a, but not in
(2.92)b. In (2.92)b, he explains, I think is the main clause and Mary is “the
topicalized subject of the complement clause”. The difference, he claims, is
manifested in the prosodic structure such that I think is followed by an IP
boundary in (2.92)a, but not (2.92)b.6
(2.92)  a.  [[Mary] [I think] % [rather liked this book]]
b. [[Mary] % [I think rather liked this book]]
Other cases that seem structurally ambiguous between a parenthetical and
a structure derived by movement include relative clauses such as the ones
given in (2.93). According to standard syntactic assumptions, the relative
pronoun ends up in the clause-initial position as a result of wh-movement
into the specifier of a functional projection. A structure where the relative
pronoun moves not only across material inside its clause but also across the
CC (as main clause) can therefore not be entirely ruled out.
(2.93)  a. And he said I can well believe that you’ve gone through an
exhaustive search because you’ve made a choice of candidate
which I think is brilliant (ICE-GB: s2a-028 #134)
b. I think it is far better to increase the amount of democracy
rather than to go ahead and reduce it which I believe would be
wrong at this time (ICE-GB: s1b-053 #46)
However, often the string preceding the comment clause is not a syntactic
constituent. For example, in (2.88)b, the CC is preceded by the subject and
main verb, in (2.88)d the CC is positioned inside a noun phrase, in (2.88)e
the CC is preceded by the subject and one of two auxiliary verbs, in (2.89)b

experimental study provides support for the parenthetical analysis suggested by Reis
(1995, 1996, 2002). Adli (2005) argues for a parenthetical analysis for French related elem-
ents; Schelfhout, Coppen and Oostdijk (2004) argue for a parenthetical analysis for Dutch.
Schneider (2007a, b) refers to related elements in Romance as “reduced parenthetical
clauses”.
6
In his account, parentheticals may “group to the left”, i.e. phrase with preceding mater-
ial from the host, “but not to the right” (Taglicht 1998: 196), contra to other prosodic
approaches to parentheticals of the same kind (e.g. Crystal 1969: 235; Dehé and Wichmann
2010a; see below).
68  The syntax and prosody of parentheticals

the CC is positioned between a preposition and its nominal object. Surface


structures like these cannot be the result of a syntactic movement operation
(see also Jackendoff 1972: 96–7), “unless massive remnant-movement-cum-
deletion is marshaled” (Rooryck 2001: 130).
Another point brought up in favour of the main-clause analysis is the fact
that parenthetical verbs lack an object which under the movement analysis is
provided by the main utterance and which it is harder to account for under
the parenthetical analysis (e.g. Rooryck 2001: 129). This has been accounted
for along the lines of a syntactic (re)analysis of the CC as sentence adverbial,
taking into account the corresponding semantic parallels between the two
(e.g. Jackendoff 1972: 99; Dehé and Wichmann 2010b make a similar point
for sentence-initial CCs). More recently, an operator object to the CC has
been suggested, which is co-referential with the host clause (e.g. Kluck and
de Vries 2012; see the discussion of (2.107) below).
Arguments in favour of the parenthetical analysis include the fact that
CCs do not lend themselves to certain main-clause phenomena. In (2.94), for
example, the simple present tense can be used for future reference only when
the future event is highly predictable. This is the case in (2.94)c due to the
presence of the matrix verb hope, but not in (2.94)d and e, where hope is part
of a parenthetical CC (see Peterson 1999). This also holds when the string
preceding the CC is a syntactic constituent, such as the subject in (2.94)e.
(2.94) Tense (a through d from Peterson 1999: 235; examples reordered)
a. The rain will stop before Sunday.
b. *The rain stops before Sunday.
c. I hope the rain stops before Sunday.
d. *The rain stops, I hope, before Sunday.
e. *The rain, I hope, stops before Sunday.
Moreover, CCs can occur with interrogatives and imperatives, but the
respective host clauses cannot occur as subordinate clauses in Standard
English (see the contrasts in (2.95) and (2.96), respectively; examples from
And Rosta, personal email).7 Jackendoff (1972: 97)  notes that “the paren-
thetical must be of ‘positive’ import, and there is no convenient way” to
account for this under a movement analysis (see (2.97), from Jackendoff
1972: 97).
(2.95)  a.  Will she, I wonder, be late?
b. *I wonder will she be late.
(2.96)  a.  Do not, I beg you, be late.
b. *I beg you do not be late.

7
(2.96)b is fine with a comma after you (I beg you, do not be late). In this reading, do not be late
is not a subordinated clause in the sense intended here, i.e. it is not a direct object of beg.
Instead, it must be interpreted as a paratactic construction, as in I beg you: do not be late.
Syntax and prosody of six selected types of parentheticals  69

(2.97) a. John is, I think, a fink.


b. *John is, I don’t think, a fink.
c. *John is, I doubt, a fink.
d. John is, I don’t doubt, a fink.
Addressing the superficial syntactic relation between clauses optionally con-
taining that such as (2.91) and non-initial CCs such as (2.88)a–c, Quirk et al.
(1985: 1113)  offer the example in (2.98) and argue that (2.98)b is not an
exact paraphrase of (2.98)a. While according to Quirk et al. (1985: 1113), the
CC in (2.98)b has a hedging meaning, the main clause in (2.98)a may have
“a more definite meaning”. That the alleged “transformation does not pre-
serve meaning” has also been noted by Jackendoff (1972: 97). Quirk et al.
(1985: 1113)  suggest a “reversal of syntactic roles”, i.e. a reversal of “the
relationship of subordination between the two clauses”. This implies that
(2.98)b cannot have been derived from (2.98)a by a syntactic movement
operation, since movement may not alter the underlying relation between
constituents (see Emonds’ 1976 Structure-Preserving Constraint). A similar
point, related to meaning, has been made by Dixon (1991: 211), who argues
that there is a difference in meaning between sentence (2.99) and sentences
(2.100)a–c such that sentence (2.99) “asserts a suspicion and details what it
is” while sentences (2.100)a–c “assert that John has been visiting the fortune
teller and then qualify the assertion by I suspect, which has a similar func-
tion to an adverb such as allegedly, presumably or probably”. Dixon (1991:
211–12) makes the same point for a reporting verb such as she complains in
(2.101)a as compared to (2.101)b. Any syntactic analysis deriving (2.100)a–c
from (2.99) and (2.101)b from (2.101)a would have to account for these
semantic differences. Note that under a movement analysis (2.101)b would
also involve a Principle C violation in the underlying structure (Maribel
Romero, p.c.): This lady would be c-commanded and thus bound by the co-
referential pronoun she.
(2.98) a. I believe (that) there were no other applicants for that job.
b. There were no other applicants, I believe, for that job.
(2.99) I suspect (that) John has been visiting the fortune teller.
(2.100) a. John, I suspect, has been visiting the fortune teller.
b. John has, I suspect, been visiting the fortune teller.
c. John has been visiting the fortune teller, I suspect.
(2.101) a. This lady complains (that) she has been short-changed.
b. This lady has, she complains, been short-changed.
An important argument against the main-clause analysis for CCs and
in favour of the parenthetical analysis comes from V2-languages such as
German, Dutch and Icelandic. In these languages, main clauses are V2
while sentence-medial CCs are V1 on the surface (see (2.102) through
70  The syntax and prosody of parentheticals

(2.104)), asking for some derivational effort under the main-clause


analysis.8
(2.102) German
a. Ich  glaube,  dass  John  später  als  Sue  angekommen  ist.
I believe that John later than Sue  arrived is
‘I think that John arrived later than Sue.’
b. Ich  glaube,   John   ist   später   als Sue   angekommen.
I believe John is later than  Sue  arrived.
‘I think John arrived later than Sue.’
c. John  ist  später  als  Sue  angekommen,  glaube  ich.
John  is later than  Sue  arrived     believe  I
‘John arrived later than Sue, I think.’
d. John ist, glaube ich, später als Sue angekommen.
(2.103) Dutch
a. Ik geloof dat Jan later aangekomen is dan Sue.
I believe that Jan later arrived is than Sue
‘I think that John arrived later than Sue.’
b. Jan is later dan Sue aangekomen, geloof ik.
Jan  is later than Sue arrived believe I
‘John arrived later than Sue, I think.’
c. Jan is, geloof ik, later dan Sue aangekomen.
John is believe I later than Sue arrived
(2.104) Icelandic
a. Ég  held að Jón hafi  komið  seinna  en María.
I believe  that  John  has come later than  Mary.
‘I think that John arrived later than Mary.’
b. Jón kom seinna  en María,  held ég
John  came  later than  Mary believe  I
‘John arrived later than Mary, I think.’
c. Jón kom held ég  seinna  en María.
John  came  believe  I later than  Mary
‘John arrived I think later than Mary.’
d. Jón, held ég, kom seinna en María.
Based on data such as (2.105) and (2.106), Kluck and de Vries (2012) argue
that all parentheticals of this kind are indeed V2. They argue for a clause-
initial operator which can either be null or spelled out as zo/so (cf. (2.105)
and (2.106)). This operator accounts for the obligatory subject–verb inver-
sion as well as the apparent V1 pattern in the case of a null operator. Kluck

8
Thanks to Janet Grijzenhout and Jóhannes Gísli Jónsson for providing the Dutch and
Icelandic data in (2.103) and (2.104), respectively.
Syntax and prosody of six selected types of parentheticals  71

and de Vries hypothesize that this operator is associated with the object gap
of the verb in the syntactically incomplete CC, i.e. it is base generated in
the object position of the verb and moves to Spec-CP in overt syntax (see
also the discussions in Reis 1995, 2002; Fortmann 2007 and Steinbach 2007
for German). The syntactic analysis is given in (2.107). Unlike Jackendoff
(1972), Kluck and de Vries (2012) thus suggest that the CC-verb has two
arguments (the subject and the operator object) and is thus syntactically
complete. They assume an anaphoric relation between the operator and the
host clause, accounting for the semantic correspondence between host clause
and CC object ‘gap’. Since the CC-verb takes an operator as object, the host
clause cannot at the same time be its object, i.e. the non-initial CC-position
cannot be syntactically derived from an underlying main clause.
(2.105)  Dutch (from Kluck and de Vries 2012)
Bob  is,  (zo)  vermoed  ik,  een  echte  charmeur.
Bob is (so) suspect I a real charmer
(2.106) German (adapted from Kluck and de Vries 2012; see also
Steinbach 2007; Reis 1995, 2002)
Hans,  (so)  glaube ich,  möchte das  Theorem  beweisen.
Hans so believe  I wants.to the theorem prove
(2.107) Syntactic analysis of CCs according to Kluck and de Vries (2012)
Bob,  [  OPi/zoi   vermoed ik ti], kent Bea  goed.
Bob OP/so suspect I knows  Bea  well
In light of all this evidence (and the evidence provided for other languages;
see footnote 8 above for references), intermediate and final CCs are con-
sidered parentheticals in this study; they are par-merged in the syntax.9
However, in order to remove any remaining ambiguity in syntactic structure,
all CCs occurring in a position preceded by a syntactic constituent such as
a subject (see (2.108)), a full clause (see (2.109)), or a relative pronoun (see
(2.93)) were discarded from the analysis. (But note that CCs like these were
included in Dehé and Wichmann’s (2010a) analysis and comparing the two
analyses, no differences between the prosodic properties of syntactically
superficially ambiguous structures on the one hand and non-ambiguous
structures on the other can be detected.)
(2.108)  a. The voters I think just have an opportunity to stick two fingers
up to whoever seems to be on top at the moment (ICE-GB:
s1b-029 #92)
b. The key issue I think is how do we ensure that local govern-
ment survives the next two years (ICE-GB: s1b-034 #116)
9
For yet more evidence, in particular from narrow syntax relations between CC and host,
and for an illustration of par-merged CC, see Griffiths (in press), a paper which was avail-
able too late to be done full justice here.
72  The syntax and prosody of parentheticals

c. This I’m afraid will have to have the light off for this
(ICE-GB: s2a-058 #141)
(2.109)  a.  The rice is marvellous I think (ICE-GB: s1a-022 #92)
b. It was a slightly futuristic production I suppose (ICE-GB: s1b-
023 #48)
c. She’s the first English girl I’ve spoken to for about three or
four years I think (ICE-GB: s1a-020 #28)
A list of all types of 1st person singular pronoun CCs represented in the
original data set retrieved from the ICE-GB is provided in (2.110).10 Not all
types entered the actual analysis. Some CCs, e.g. I assume, if you will, were
infrequent in the corpus and did not survive the various sorting processes,
i.e. the one carried out in order to avoid syntactic ambiguity, and the sorting
processes related to properties of the sound files (see Chapter  4). For the
types that entered the actual analysis and for exact numbers of occurrence,
see Chapter 4 below. Notice that I mean was not included anywhere in the
analysis.
(2.110)  CC types retrieved from the ICE-GB
a. I think, I suppose, I believe, I assume, I guess, I understand, I
reckon, I wonder, I suspect, I fear, I hope, I argue, I imagine,
I expect, I (always) feel, I don’t think, I would/should/’d say,
I’m afraid, (as) I find, so I hear, I don’t know, I might say, I
may say, I must say, I’m glad to say, I’m sure, I would sug-
gest, I would think, I thought, I would have thought, I should
think, I shouldn’t think, I would/’d imagine, I must admit
b. as it were, if you will
In the theoretical prosodic literature, CCs have been argued to be prosodi-
cally phrased separately (e.g. Nespor and Vogel 1986). In the empirical pros-
odic literature, much more variation has been observed and CCs have been
argued to be realized in a variety of ways, ranging from melodic and tem-
poral separateness to full integration, unlike the three types of parentheticals
discussed in the previous three sections, for which the assumption of inton-
ational separateness seems also to be the rule in the empirical literature. For
example, according to Bing (1985: 32–3), epistemic verbs such as think, sup-
pose, know, realize, wonder, hope and imagine behave like sentence adverbi-
als: they may either occur with a Class 0 contour, i.e. without a prominence
tone,11 or they may be prominent, realized with an A-rise (Bolinger’s 1986 A

10
There were three CCs with the first person plural subject we in the data set originally
retrieved from the corpus: one we hope, one we now think and one we would say. These three
tokens were not considered further.
11
Class 0 expressions in Bing’s (1985) terminology form an “outside class”; they are expres-
sions “which seem to be intuitively independent of sentences with which they are associ-
ated” and do “not seem to contribute to the truth value of the sentence” (Bing 1985: 21).
Syntax and prosody of six selected types of parentheticals  73

accent, Ladd’s 1978 “fall”; corresponding to (L+)H* followed by a fall to a


low edge tone; see Chapter 3 below), or they may have no independent con-
tour at all. According to Gussenhoven (1984: 123) CCs of the type I think
“are typically post-nuclear”, i.e. integrated as intonational tail. CCs may be
accentless and may be integrated in another prosodic domain as unstressed
material preceding or following the main prominence of that domain
(Crystal 1969: 235, 268; Wichmann 2001: 186; Dehé 2007: 270–2). Dehé
and Wichmann (2010a), who analyse 156 CCs from the same speech corpus
used here but include syntactically ambiguous CCs, find that CCs may be
integrated in a larger intonation domain or they may be phrased separately.
If separate, they have their own nuclear prominence; if integrated, they may
be unstressed or prominent; if unstressed, they may occur in initial, medial
or final position in their intonation domain. Prosodic realization, they con-
clude, is related to interpretation, such that (i) “[p]rosodic separation and
prominence go along with semantic transparency”, i.e. the CC expresses
speaker attitude such as genuine uncertainty, doubt, etc.; and (ii) “[p]rosodic
integration and deaccentuation go along with semantic bleaching”, i.e. CCs
“are used for discoursal, interactional and interpersonal purposes (polite-
ness, mitigation, narrative cohesion)” or (iii) they may be used as markers
of phases of disfluency and hesitation, may reflect mental planning, or may
be used as floor-holding devices (Dehé and Wichmann 2010a: 24). Also rele-
vant is Peters’ (2006) observation that the length of the parenthetical affects
its prosodic phrasing. Given their length, CCs may lend themselves more
easily to prosodic integration than longer types of parentheticals. On the
other hand, CCs have elsewhere explicitly been treated on a par with other
parentheticals prosodically, assuming that they are intonationally separate
(e.g. Nespor and Vogel 1986: 188–90).
In conclusion, we predict more variation in the prosodic realization of
CCs than that of FPCs, NRRCs and N-APPs; see (2.111) and (2.112).
(2.111)  Prosodic phrasing of CCs: hypotheses
a. As parentheticals, CCs may be prominent and phrased separ-
ately (e.g. Nespor and Vogel 1986; see (2.112)a).
b. CCs may be prosodically integrated (see (2.112)b) as prenu-
clear (e.g. (2.112)b2, b3) or postnuclear (e.g. (2.112)b1, b3)
material (e.g. Crystal 1969; Bing 1985; Gussenhoven 1984).
c. If phrased together with host material (i.e. phrased according
to (2.112)b), CCs may be unstressed or they may be promin-
ent (Dehé and Wichmann 2010a).
They “seem to function as a kind of discourse domain between the speaker and hearer”
(Bing 1985: 49). This semantic/pragmatic independency of Class 0 expressions is reflected
in the prosody. The Class 0 contour “has no prominence tones in the phonological phrase;
the phrase optionally has a rising boundary tone at the end of the sentence. […] [T]he
Class 0 contour merely identifies the utterance on which it occurs as “outside” the root
sentence” (Bing 1985: 49).
74  The syntax and prosody of parentheticals

(2.112)  CCs: predicted prosodic phrasing


a. prosodic separation:
(…) IP[…]IP IP[CC]IP (IP[…]IP) (…)
b. prosodic integration:
b1. (…) (IP[…]IP) IP[… CC]IP (IP[…]IP) (…)
b2. (…) IP[…]IP IP[CC …]IP (IP[…]IP) (…)
b3. (…) (IP[…]IP) IP[… CC …]IP (IP[…]IP) (…)

2.3.5  Reporting verbs
Reporting verbs (RVs) are clausal elements of the form illustrated in (2.113)
(from Quirk et al. 1985: 1022) and (2.114). They consist of a subject and a
verb of speaking or thinking. Like CCs, RVs are syntactically defective on
the surface such that an object, which would otherwise be obligatory with the
verbs involved, is missing (e.g. Quirk et al. 1985: 1022), but similar to CCs,
the missing constituent corresponds semantically to the host clause. In med-
ial and final position of the host utterance inversion of the reported verb and
its subject may be observed (see (2.113)c, (2.114)a). The reported clause (i.e.
the host) may represent direct or indirect speech, with direct speech being
indicated by quotation marks in writing, but also by the use of deictic fea-
tures, such as expressions relating to the time and place of the utterances
or to persons referred to in the utterance (Quirk et  al. 1985: 1025, 1029).
Reporting verbs may be structurally ‘simple’, consisting only of a personal
pronoun and a verb (e.g. (2.113)b, (2.114)b, c), but they may also be syntac-
tically and prosodically more complex; for example, they may consist of a
verb and a full noun (e.g. (2.113)a, c, (2.114)a), the verb may take a prepos-
itional complement (e.g. (2.115)a) or the RV may be modified by an adverb
(e.g. (2.115)b).
(2.113) Reporting verbs (RVs)
a. I wonder, John said, whether I can borrow your bicycle.
b. The radio is too loud, she complained.
c. The radio is too loud, complained Elizabeth.
(2.114) Reporting verbs (RVs)
a. He travels to Manchester for music lessons and his music
teacher thinks so highly of him says Julian that uhm he’s giv-
ing him a whole day instead of half a day on Sunday (ICE-GB:
s1a-032 #237)
b. My learned friend relied on a number of facts of establishing
he says beyond doubt that this vessel had been used to import
prohibited drugs (ICE-GB: s2a-068 #072)
c. Concern for patients’ welfare they claim has never been con-
fined to curing people’s illnesses and ensuring their survival
(ICE-GB: s2b-038 #005)
Syntax and prosody of six selected types of parentheticals  75

(2.115) a. My covenant is with Levi he says to them (ICE-GB: s1b-


001 #055–056)
b. It wasn’t a round table President Gorbachev said later it was a
square table (ICE-GB: s2b-040 #092–093)
Like for CCs, there is a (superficial) structural relationship between utter-
ances containing a reporting verb and a complex sentence which has the
reporting verb as matrix verb taking the reported clause (i.e. the host) as
object; compare (2.113)c with (2.116). On the other hand, Quirk et al. (1985:
1023) argue that reporting verbs behave like adverbials, such that their posi-
tioning is flexible and omission is possible. Moreover, in her discussion of
the relation between (2.117)a and (2.117)b, Reinhart (1983: 172)  argues
“that in spite of the apparent similarities, the two types of sentences dif-
fer radically with respect to point of view”. Reinhart (1983) distinguishes
between sentences containing parentheticals (SCPs) which are subject ori-
ented (e.g. (2.118)a), answering questions of the type “What did the paren-
thetical-subject say or believe?”, and speaker-oriented SCPs (e.g. (2.118)b),
which can answer “questions on the subject matter of the clause” (Reinhart
1983: 176). A direct speech reported clause would always be of the former
type. According to (Reinhart 1983: 176–84), the two types differ in terms of
syntactic, stylistic, pragmatic and intonational properties.12
(2.116)  Main clause occurrences superficially related to (2.113)c
a. Elizabeth complained “The radio is too loud”.
b. Elizabeth complained that the radio was too loud.
(2.117)  a.  A famous actress was going to visit him, Max said.
b. Max said that a famous actress was going to visit him.
(2.118) Reinhart’s (1983) subject-oriented SCPs (a) and speaker-oriented
SCPs (b)
a. Hei would be late, Johni said.
b. Johni will be late, hei said.
Based on examples such as (2.119), Kluck and de Vries (2012) suggest a
syntactic analysis for RVs equivalent to that of CCs (see (2.120); see also
the discussion in Fortmann 2007). Like CC-verbs, RVs take an operator
as object, which can be spelled out as zo/so or can be silent (cf. (2.119)),
and which is anaphoric with respect to the host clause, accounting for the
semantic correspondence between object gap and host clause. Given the
operator object, the host clause cannot be the object of the RV. Like CCs,
RVs are thus clausal parentheticals rather than main clauses ending up in a

12
Griffiths (2012) has recently contested the necessity of the category ‘subject-oriented
SCPs’. Based on a number of syntactic tests comparing speaker-oriented, quotative, and
subject-oriented SCPs, he argues that the category ‘subject-oriented’ is unnecessary and
redundant.
76  The syntax and prosody of parentheticals

non-initial position as the result of syntactic movement. Note also that like
(2.101)b above, (2.118)b would result in a Principle C violation under a non-
parenthetical (= movement) analysis, with he binding John in the under-
lying structure.
(2.119) RVs in Dutch (a), German (b) and English (c); from Kluck and de
Vries (2012)
a. Bob  is, (zo)  zei Bea,  een  echte charmeur.
Bob is so said  Bea a real charmer
b. Bob  ist,  (so)  sagte Bea, ein richtiger  Charmeur.
Bob is  so  said Bea  a real  charmer
c. Bob is, (so) said Bea, a real charmer.
(2.120) Syntactic analysis of RVs according to Kluck and de Vries (2012)
Bob  is,  [  OPi/soi said Bea ti], a real charmer.
Another argument against the main-clause analysis and in favour of the par-
enthetical analysis is the observation that intransitive verbs can occur as RVs
(see (2.121)). Since there is no object position available, the host clause can-
not start out as the object of the RV. According to Kluck and de Vries (2012)
the use of verbs such as explode in (2.121)a and b goes along with a shift
towards a manner interpretation. While this is possible with RVs, it is not
possible with CCs.
(2.121) Use of intransitive verbs as RV (Dutch example from Kluck and
de Vries 2012; English examples from Hanote 2004: 543)
a. Bob is,  zo ontplofte  Bea,  een  gemene  charmeur.
Bob  is so exploded Bea a nasty charmer
b. ‘I hate the blacks, I hate the Kaffirs,’ exploded Ferdi.
c. […] He can’t stand it, I hear him shouting in the kitchen,’ she
laughs, which suggests that she indulges in celluloid infidelity
to even up the relationship.
Verbs such as those in (2.121), i.e. which are not verbs of speaking or think-
ing and which involve a shift towards a manner interpretation, are not part of
the data set analysed in Chapter 4 below. The verbs actually occurring in the
data set are given in (2.122), with actual numbers of occurrences provided in
Chapter 4. Other verbs included in the corpus search but not ending up as
part of the data set analysed prosodically are listed in Section 4.1.5 below.
(2.122) Reporting verbs represented in the data set
(he/she/it/N) says, (I/he/she/they/N) said, (I) argue, (you) ask,
(they) claim, (I) confess, (I) promise
In the prosodic literature (theoretical, empirical and pedagogical), RVs have
often been taken to be (temporally and tonally) separate, but unaccented.
Quirk et  al. (1985: 1023)  note that reporting verbs are usually set off by
Syntax and prosody of six selected types of parentheticals  77

punctuation in writing, but that, similar to comment clauses and adverbi-


als, this separation “is not necessarily reflected intonationally”. Bing (1985)
treats reporting verbs (“parenthetical verbs” in her terminology) on a par
with other ‘Class 0 expressions’ such as vocatives, polite expressions (please,
thank you) and constant polarity tags, all of which, she argues, are typic-
ally realized without prominence. She maintains that “parenthetical verbs
cannot have prominence tones” (Bing 1985: 29)  and that Class 0 expres-
sions are preceded by an optional boundary tone in final position and an
obligatory boundary tone in medial position. Similarly, Gussenhoven (2004:
291) maintains that “[r]eporting clauses are either incorporating or cliticiz-
ing”, i.e. they are never prominent unless they are complex. If incorporated,
they continue the intonation contour of the preceding material; if encliti-
cized, they are unaccented but set off from preceding material by a bound-
ary tone. This is also the gist of Wells’ (2006: 155) descriptive, pedagogical
account: the RV is unstressed and (tonally) integrated, but with an optional
preceding “rhythmic break”. Astruc-Aguilera (2005: 62) claims that report-
ing verbs in English (“quotation markers” in her terminology) “are pro-
sodically detached in any position, and in final position they are deaccented
as well”. Based on her experimentally elicited corpus of 64 reporting verbs
(four examples of complex RVs marking direct speech read twice by eight
speakers), she reports that reporting verbs “are invariably deaccented”
and separated from their hosts by pauses in 65 per cent of cases.13 This is
repeated in Astruc-Aguilera and Nolan (2007b: 91), who claim that English
direct speech markers “always form independent phrases”, that they are
“nearly always deaccented” and “usually set off by pauses”.14 According to
Reinhart (1983: 179), “low, steady pitch” and a preceding pause are typical
of parenthetical-subject-oriented SCPs only, while speaker-oriented RVs are
not typically preceded by a pause but may instead be phonologically assimi-
lated. This is compatible with Astruc-Aguilera’s (2005) results, who tested
direct speech, i.e. parenthetical-subject-oriented SCPs only.
Overall, there seems to be agreement in the literature that RVs are typic-
ally produced without prominence but that they may be set off from the pre-
ceding material by a pause and a boundary tone. Despite the fact that – like
13
Astruc-Aguilera’s corpus of RVs comprises the following four examples, all marking direct
speech (see Astruc-Aguilera 2005: 222). They are complex RVs in the present terminology
because they consist of a full/proper noun and a verb ((ii) and (iii)) and/or include a nom-
inal or prepositional object ((i) and (iv)). All RVs are separated from the reported speech
by punctuation and the reported speech is clearly marked as such by the use of inverted
commas.
(i) ‘Your meal is ready’, I said to Anna.
(ii) ‘The meal is ready’, my mother announced.
(iii) Ramona says, ‘where are you going on your holiday, Anna?’
(iv) ‘How was your trip?’ Alma asked Maria.
14
Notice that in an analysis like this, prominence cannot be a necessary condition for a phrasal
prosodic constituent. Instead, as Wells (2006: 155) remarks, “we would have to say that the
IP [of the RV] was anomalous in having no nuclear tone”.
78  The syntax and prosody of parentheticals

CCs  – RVs are usually considered parentheticals, the idea of them being
phrased separately with a nuclear contour of their own is not prominent
in the literature, except for complex RVs. The hypotheses with respect to
prosodic phrasing are therefore as given in (2.123) and (2.124) (the closing
square bracket preceding the RV in (2.124)b2 indicates the optional bound-
ary observed by Bing 1985 and Gussenhoven 2004).
(2.123) Prosodic phrasing of RVs: Hypotheses
a. As parentheticals, RVs may be prominent and phrased separ-
ately (see (2.124)a).
b. Only complex RVs may be phrased separately with nuclear
prominence; simple RVs are integrated (Bing 1985;
Gussenhoven 2004; see (2.124)b).
c. Integrated RVs (see (2.124)b) are unaccented (Gussenhoven
2004).
d. Unaccented RVs may be set off from the host temporally
(by a pause) and tonally (by a boundary tone) (Gussenhoven
2004; Astruc-Aguilera 2005; Wells 2006; Astruc-Aguilera and
Nolan 2007b; see (2.124)b2).
(2.124) RVs: predicted prosodic phrasing
a. prosodic separation:
(…) IP[…]IP IP[(complex) RV]IP (IP[…]IP) (…)
b. prosodic integration
b1.  (…) IP[… RV]IP (IP[…]IP) (…) (Gussenhoven’s 2004
incorporation)
b2.  (…) IP[IP[…]IP RV]IP (IP[…]IP) (…) (Gussenhoven’s 2004
encliticization)

2.3.6  Question tags
Question tags (QTs; also tag questions) are clausal elements such as the itali-
cized ones in (2.125). Formally, they consist of an operator (an auxiliary
verb that corresponds to the one in the preceding clause or do) and a pro-
nominal subject (in that order), as well as an optional negation (n’t or not);
see, for example, Bald (1980), Gussenhoven (1984: 118), Quirk et al. (1985:
810), McCawley (1998: 501), Biber et  al. (1999: 208), and Huddleston
and Pullum (2002: 893). The pronominal subject corresponds to (i.e. it
is a pronominal copy of) the subject in the preceding clause; the subject
and verb in the QT thus repeat those in the main clause as proforms. The
part preceding or surrounding the QT has been referred to as anchor (e.g.
Huddleston and Pullum 2002: 891; Tottie and Hoffmann 2006) or host
clause (e.g. Cattell 1973), among other names. In the present context I will
refer to it as the host (clause), parallel to the material surrounding other
types of parentheticals.
Syntax and prosody of six selected types of parentheticals  79

(2.125) Question tags (QTs)


a. John is a genius, isn’t he, Peter
b. Paula didn’t go to the party, did she, last night
c. So you missed another class, did you

QTs are often reversed polarity tags, i.e. a negative tag follows a positive
clause or vice versa; see (2.125)a, b, and the corpus examples in (2.126).
In the header of (2.126), pos-neg refers to a positive verb in the host and a
negation in the QT, neg-pos to the reverse. As shown in examples (2.125)c
and (2.127), constant polarity tags (reduplicative tags in McCawley’s 1998:
501 terminology) occur, too, but according to Huddleston and Pullum
(2002: 892), reversed polarity tags are much more frequent than constant
polarity tags. If the tag has the same polarity as the host clause, this is usu-
ally positive (cf. the pos-pos examples in (2.125)c and (2.127)a–d). This
tendency was confirmed for both British and American English in Tottie
and Hoffmann’s (2006) corpus study. Investigating the differences in the
use of QTs in British and American English, Tottie and Hoffmann (2006)
based their study on almost 5,000 QTs taken from the British National
Corpus (BNC) and about 2,300 QTs from the Longman Spoken American
Corpus (LSAC). The same tendency is also found for the set of data
retrieved from the ICE-GB and investigated below. Constant negative
polarity tags are infrequent in general and only the two examples given
in (2.127)e and f were found in the present data set (see also Tottie and
Hoffmann 2006: 290 for some examples from the BNC and the LSAC).
Given their infrequency even in actual spoken language, it is perhaps not
surprising that constant negative polarity tags have sometimes been argued
to be ungrammatical or non-existing/uncommon (e.g. Lakoff 1969: 142;
König 1977: 46 for American English). A reviewer for Dehé and Braun
(2013) notes that (2.127)e is ungrammatical to the native ear, but that the
use of the constant polarity tag may be the effect of the distance between
the QT and the anchor in the host, “analogous to concord of proximity in
number agreement”. The example in (2.127)f is unusual in two respects:
first, it features a constant negative polarity tag, which is uncommon in
general; second, the host is interrogative in form, although it has previ-
ously been assumed that tags can only be attached to declarative host sen-
tences (e.g. Gussenhoven 1984: 118; but see also Biber et  al. 1999: 210
for examples of QTs attached to interrogatives). Anne Wichmann (p.c.)
notes that the interrogative reinforces the meaning of the host – a stylistic
device in courtroom interaction. The example is indeed taken from a legal
cross-examination.
In the data set studied here and reported on in Chapter 4 below, pos-neg
reversed polarity tags are by far most frequent, followed by neg-pos reversed
polarity tags, and constant polarity tags (see Table  4–2 in Section  4.1.6
below).
80  The syntax and prosody of parentheticals

(2.126) Reversed polarity tags (a–d: pos-neg; e–h: neg-pos)


a. The state of the wall uh the state in which you found the wall
was the result of the work that you had directed to be done to
the wall wasn’t it (ICE-GB: s1b-069 #174)
b. We had a big chat about this before didn’t we (ICE-GB: s1a-
006 #172)
c. I mean your mother there was a large picture of your moth-
er’s mother wasn’t there in a sort of wig looking as fierce as
anything (ICE-GB: s1a-007 #167)
d. … it’s surely quite difficult these days to persuade an actor
or actress to commit themselves for what six to eight months
when there is always the the possibility isn’t there of lucrative
television work (ICE-GB: s1b-050 #72)
e. But the Labour Party’s not going to abolish private medicine
is it Donald (ICE-GB: s1b-039 #88)
f. You can’t blame her for that really can you (ICE-GB: s1a-
007 #017)
g. it’s not like a French writer is it I don’t see a French writer
voluntarily writing in English (ICE-GB: s1b-026 #112–113)
h. You don’t know do you whether the hotel had any system at all
for checking the fittings round the swimming pool (ICE-GB:
s1b-067 #072)
(2.127) Constant polarity tags (a–d: pos-pos; e, f: neg-neg)
a. But then there’s another audience is there that goes to regional
playhouses a rather more stolid audience (ICE-GB: s1b-
050 #87)
b. But you were there for other reasons then were you (ICE-GB:
s1b-042 #080)
c. So that’s really unrealistic is it wanting to do to teach English
(ICE-GB: s1a-033 #176)
d. They’re called Gasser the people next door are they (ICE-GB:
s1a-057 #30)
e. Well you aren’t supposed to record the systolic when you you
can hear more than one sound occurring you know succes-
sively aren’t you (ICE-GB: s1b-004 #150)
f. Uh didn’t Mr Hook say to you that at this point that he was
interested in further expansion of the uh Ferndale business
did he not (ICE-GB: s1b-064 #118)
Huddleston (1970) postulates an underlying paratactic construction to
account for the relation between QT and host. More generally speak-
ing, QTs have been considered parentheticals based on their (syntactic)
form and function (e.g. Knowles 1980; Quirk et al. 1985: 919; Ziv 1985).
Syntax and prosody of six selected types of parentheticals  81

Huddleston and Pullum (2002: 891)  introduce a distinction between


interrogative tag and interrogative parenthetical, but “[n]o great signifi-
cance” (p.  892) is then ascribed to this terminological distinction. In
his ­­syntactic-semantic approach to clausal parentheticals, Potts (2002b:
651) sees QTs as clausal expressions and treats them on a par with other
types of parentheticals such as sentence adverbs, CCs and other types of
phrasal and clausal parentheticals. Following Emonds (1976), he argues
that they have comma intonation, i.e. they are produced with an inton-
ation break immediately before and after. Similar to CCs and RVs, QTs
are syntactically defective (or elliptical; Culicover 1992: 195) on the sur-
face, consisting only of a subject and an auxiliary. The missing VP can be
reconstructed from the host clause; for example, in (2.125)b, the tag did
she can be continued go to the party, resulting in the syntactically com-
plete question Did she go to the party? (see e.g. Knowles 1980: 396–401 for
discussion).
While Ziv (1985: 189) claims that tags are formally different from other
types of parentheticals in that they are “restricted to sentence-final pos-
ition”, the present corpus data do not support this claim: although QTs do
indeed often occur utterance- or sentence-finally, it is not unusual for them
to occur clause-medially between syntactic constituents and even within
syntactic constituents (see e.g. (2.126) c, d, h, and (2.128) for examples of
non-final QTs; see also Knowles 1980; Bald 1980; Tottie and Hoffmann
2006 and Section 4.3.6 below).
(2.128) Non-clause-final QTs
a. You don’t do you want us to think of faith as a synonym for
tradition (ICE-GB: s1b-028 #103)
b. I mean in in that piece we’ve just heard from The Revenger’s
Tragedy it’s a mixture isn’t it of original instruments and kind
of what sound to me like modern trumpets (ICE-GB: s1b-
023 #140)
c. And uh we’ve learnt haven’t we that books are all right but
there is nothing like uh the help you get from society mem-
bers as the team have said and doing it yourself (ICE-GB:
s1b-025 #166)
Assuming that QTs are parentheticals in the syntax, the prediction is that they
should prosodically behave like other types of parentheticals and form a sep-
arate prosodic domain by default (see e.g. Selkirk 1984: 295; Nespor and Vogel
1986: 189–90; Potts 2002b: 651). On the other hand, it has been shown that
prosodic separation of other short and prosodically light parentheticals such
as CCs is far from obligatory (e.g. Dehé and Wichmann 2010a; Dehé 2009 and
references given in the previous two sections), and this finding will be con-
firmed for CCs and RVs in Chapter 4 below. Approaches specifically dealing
with the prosody of QTs are varied and their results on occasion conflicting.
82  The syntax and prosody of parentheticals

With respect to prosodic phrasing, for example, Bolinger (1989) notes


that it is generally common for QTs containing modals such as will and can,
along with other “appended material” such as RVs, to be unaccented and
integrated into a prosodic domain as intonational tail, i.e. as postnuclear
unstressed material. However, he also notes that “[i]t is difficult to general-
ize about tag intonation because once we look beyond the central auxiliaries
will and can we find that the intonation may depend on the choice of the
tag” (Bolinger 1989: 164). Given that at least be and do occur in QTs more
frequently than modals (see Tottie and Hoffmann 2006 and the sections on
QTs in Chapter 4 below), this is an important qualification. Other authors
maintain that QTs should be phrased separately in general: Knowles (1980)
and Potts (2002b), for instance, maintain that QTs have comma intonation,
i.e. that they are set off intonationally and temporally from preceding and
following host material. Selkirk (1984: 295) includes QTs with other types
of parentheticals which “should be fated to constitute IPs on their own”.
Other authors have linked phrasing to polarity. According to Gussenhoven
(1984: 118–20, 2004: 292), reversed polarity tags consist of an accented aux-
iliary and a pronominal subject and have to be phrased separately, while
constant polarity tags incorporate obligatorily: they are “always included in
the tail of the nuclear tone in the host sentence” (Gussenhoven 1984: 120).
Ladd’s (1981) basic distinction is between nuclear and postnuclear tags
and the way they relate to phrasing. (Note incidentally that Knowles 1980
and Quirk et al. 1985 observe that within the tag, it is always the auxiliary,
never the pronoun, that is accented.) While nuclear tags have a main accent
(nucleus) of their own and are generally preceded by a phrasal prosodic
boundary, Ladd (1981) argues, postnuclear tags do not have a nucleus or
preceding boundary; instead, the pitch contour continues from the preced-
ing nucleus in the same intonational domain, which is associated with a syl-
lable in the host. This distinction is then related to general syntactic and
pragmatic distinctions. In the syntax, nuclear and postnuclear tags are said
to behave differently, for instance, with regard to the distribution of adverbs
(e.g. even) and negative polarity items, among others (Ladd 1981: 168). This
goes back, for example, to Sadock (1974: 133), who argues that the difference
in intonation and interpretation goes along with differences in the syntac-
tic behaviour, such as the occurrence of negative polarity items in the host,
and the co-presence of non-restrictive relative clauses, conjoined questions,
and the like (see also Rando 1980: 246–7 for grammatical restrictions on
the use of tags). In pragmatics, the nuclear tag states or asserts the speaker’s
assumption and signals a hedge to the proposition expressed by the host,
while the postnuclear tag is closer to a true interrogative and expresses real
doubt or uncertainty. In conclusion (Ladd 1981: 168–9), the nuclear tag is
separate in its meaning and prosody: it expresses a separate speech act and
has its own intonation domain and contour; the postnuclear tag, on the other
hand, is integrated in meaning and intonation. If both Ladd (1981) and
Gussenhoven (2004) are right, then constant polarity tags can never express
Syntax and prosody of six selected types of parentheticals  83

a separate speech act, because according to Gussenhoven (2004), they are


always integrated. Reversed polarity tags, on the other hand, always express
separate speech acts because they have to be phrased separately.
With regard to the intonational realization of QTs, a number of research-
ers have noted that pitch accent choice (i.e. rising (L*+H) or falling (H*
L-) accent on the auxiliary) is related to its polarity and interpretation (e.g.
Crystal 1969: 273; Sadock 1974: 126–34; Bing 1985: 36; Rando 1980: 245–7;
Gussenhoven 1984: 118f; Quirk et al. 1985: 811–12; Huddleston and Pullum
2002: 894–5), though researchers do not necessarily agree on the exact rela-
tion between polarity, interpretation and intonation (see Dehé and Braun
2013: 132 for a brief overview). Based on the same data set reported on in
the present study, Dehé and Braun (2013) found that intonational contours
were affected by polarity and position such that (i) falling contours are more
frequent in reversed polarity QTs than in constant polarity QTs and that
within reversed QTs, falling contours were more frequent in pos-neg than
in neg-pos QTs; and (ii) falling contours were more frequent in sentence-
final (but not turn-final) position than in utterance (turn-) final position,
suggesting that within a speaker’s turn, QTs express certainty while in turn-
final position they are used to ask for an immediate reply. According to Ladd
(1986: 330), the pitch contours of tags are also dependent on the choice of
nucleus in the host. However, Dehé and Braun (2013) found no effect of the
nuclear tune preceding the QT on the contour associated with the QT. See
Dehé and Braun (2013) for more details about the intonational realization of
QTs. The present study focuses on their prosodic phrasing.
Overall, based on previous claims in the literature, the following predic-
tions can be made with respect to the prosodic phrasing of QTs. Like for
CCs and RVs, we predict some variation in the prosodic phrasing of QTs,
depending on several factors, among them the lexical factor verb type and
syntactic factors such as polarity. The predictions are spelled out in (2.129)
and (2.130). In Chapter 4, the factor syntactic position will also be tested.
(2.129) Prosodic phrasing of QTs: Hypotheses
a. As parentheticals, QTs may be prominent and phrased separ-
ately (Knowles 1980; Potts 2002b; Selkirk 1984; see (2.130)a).
b. If the QT has nuclear prominence, it is always preceded by
a phrasal prosodic boundary (Ladd 1981; see phrasing in
(2.130)a and b1), i.e. it cannot be phrased together with pre-
ceding host material (i.e. (2.130)b2 and b3 are impossible with
nuclear QTs).
c. Prosodic phrasing is determined by polarity such that
reversed polarity tags are obligatorily phrased separately
(i.e. they follow (2.130)a), while constant polarity tags must
incorporate as postnuclear material, i.e. phrased according to
(2.130)b3 (or possibly b2, if preceded by main prominence)
(see Gussenhoven 1984, 2004).
84  The syntax and prosody of parentheticals

d. Prosodic phrasing is affected by verb type (Bolinger 1989),


such that QTs containing modals are typically integrated
as postnuclear material, i.e. they are phrased according to
(2.130)b3 (or possibly b2, if preceded by main prominence).
(2.130) QTs: predicted prosodic phrasing
a. prosodic separation:
(…) IP[…]IP IP[QT]IP (IP[…]IP) (…)
b. prosodic integration:
b1. (…) IP[…]IP IP[QT …]IP (IP[…]IP) (…)
b2. (…) (IP[…]IP) IP[… QT …]IP (IP[…]IP) (…)
b3. (…) (IP[…]IP) IP[… QT]IP (IP[…]IP) (…)

2.4  Concluding remarks


Following de Vries (2012a and elsewhere), parentheticals are syntactic
adjuncts, whose internal structure is invisible to the host due to par-merge.
Based on previous literature, this chapter established the predictions with
respect to prosodic phrasing for the six types of parentheticals under inves-
tigation here. Superimposed are general constraints governing ­prosodic
phrasing (see Chapter  3 below). Based on the previous sections, the six
types of parentheticals can be divided into two groups, according to the
predictions with regard to their prosodic phrasing (see Table 2.1). For the
first group, comprising FPCs, NRRCs and N-APPs, the prediction is pros-
odic separation. For the second group, consisting of CCs, RVs and QTs, the
predictions as to their phrasing are more varied, allowing for phrasing of
the parenthetical with host material (= prosodic integration). The factors
potentially affecting phrasing of this latter group include prosodic proper-
ties (e.g. prosodic weight/length), syntactic factors (e.g. QT polarity, RV
complexity) and their discourse contribution/interpretation and function
in the given context. Within this group, RVs differ from the other two in
that not all patterns of prosodic integration have been previously observed.
In particular, RVs have been argued to be part of the postnuclear mater-
ial within a prosodic domain, but have not been observed to be followed in
their domain by host material. The predictions developed in the previous
sections are summarized in Table 2.1 (compare (2.47)/(2.48), (2.61)/(2.62),
(2.86)/(2.87), (2.111)/(2.112), (2.123)/(2.124) and (2.129)/(2.130) above).
A question that comes to mind is whether the six types of parentheticals
investigated here form a natural class in the first place (thanks to Lisa Selkirk
for pointing this out to me). I assume that they do insofar as they are paren-
theticals in the syntax, but that they may not be a natural class semantically.
It is striking that the second group (CCs, RVs, QTs) consists of elements
whose parenthetical status has been doubted in the literature and which are
syntactically defective on the surface. In particular, CCs and RVs have been
Concluding remarks  85

Table 2.1  Prosodic phrasing of parentheticals: predictions

FPCs, NRRCs,
Type N-APPs CCs, QTs RVs
Predicted prosodic separation: prosodic separation:
phrasing (…) IP[…]IP IP[par]IP (…) IP[…]IP IP[par]IP (IP[…]IP) (…)
(IP[…]IP) (…)
prosodic integration: prosodic
integration:
(…) IP[…]IP IP[par …]IP (…) IP[… RV]IP
(IP[…]IP) (…) (IP[…]IP) (…)

(…) (IP[…]IP) IP[… par…]IP (…) IP[IP[…]IP RV]IP


(IP[…]IP) (…) (IP[…]IP) (…)

(…) (IP[…]IP) IP[… par]IP


(IP[…]IP) (…)

argued to be main clauses rather than parentheticals (see Sections 2.3.4 and
2.3.5 above). Naturally, if they are main clauses ending up in their surface
position as a result of movement, the predictions for the prosodic structure
will change. However, the sections above have explained why CCs, RVs and
QTs are considered parentheticals syntactically, and more evidence can be
found following up the references given. Moreover, it has been explained
how the apparent syntactic deficiency can be accounted for syntactically
and that, as a consequence, the host clause cannot be the syntactic object of
the verb involved in the CC or RV. Syntactically ambiguous CCs have been
removed from the analysis.
I therefore consider it more likely that the different predictions between
the two groups result from the semantic-pragmatic contribution to their
host sentences. Parentheticals such as FPCs, NRRCs and N-APPs are typ-
ically speaker-oriented additions, outside the at-issue entailments, whose
primary function is to provide discourse-new information which is not part
of the host proposition. They have CI semantics (Potts 2005). This accounts
for prosodic prominence as well as for a certain length or syntactic com-
plexity of the parenthetical (i.e. FPCs, NRRCs and N-APPs are typically
longer and more complex than CCs, RVs and QTs). Syntactic parentheti-
cals such as CCs, RVs and QTs, on the other hand, may serve these func-
tions, too, but they are also commonly used for discoursal, interactional and
interpersonal purposes, e.g. as politeness markers, to mitigate the speech act
performed by the host clause or to ask for confirmation. They are not typ-
ically used to provide new information to the discourse which adds to the
content of the main proposition, thus this function may be secondary, and
they may instead affect the main proposition by implying that it may not be
86  The syntax and prosody of parentheticals

true. This has been explicitly formulated by Scheffler (2009) for CCs and
may be extended – in slightly adapted form – to RVs and QTs. The primary
functions of this group of parentheticals may not require or allow pros-
odic prominence, thus excluding prosodic separation, or they may require
prosodic integration for scope reasons. It is therefore conceivable that the
syntax–prosody interface constraints predicting prosodic boundaries at the
edge(s) of the parenthetical are overridden by other constraints, accounting
for semantic or discourse requirements and resulting in prosodic restruc-
turing. Other factors such as the length of the parenthetical may corrobor-
ate these effects.
The following chapter will take a closer look at the syntax–prosody
interface theory, before the analysis of the corpus data is reported on in
Chapter 4.
3 Parentheticals, intonational phrasing
and prosodic theory

As shown in the previous chapter, parentheticals have generally been


assumed to form prosodic constituents of their own, an assumption reflected
also in the hypotheses formulated for each of the parenthetical types, sum-
marized at the end of Chapter 2, and in particular for the first three types of
parentheticals (full parenthetical clauses, NRRCs and N-APPs). However,
not much has been said so far about prosodic theory and how the relation
between syntactic parentheticalhood and prosodic constituency can be
accounted for in, and follows from, a respective interface theory without
resorting to parenthetical-specific assumptions. Prosodic constituents have
been defined and prosodic structure has been described in a number of dif-
ferent approaches (see e.g. Shattuck-Hufnagel and Turk 1996; Scheer 2011;
Frota 2012 for overviews). In my view, it is important to keep in mind that
the various approaches to the specific prosodic properties of parentheticals
have been couched in different frameworks of the description and analysis
of intonation. In particular, while some relevant work (e.g. Crystal 1969;
Bolinger 1989; Wichmann 2000, 2001) is based on the Standard British sys-
tem of intonation description (e.g. Palmer 1922; Kingdon 1958; Schubiger
1958; Crystal 1969; O’Connor and Arnold 1971) and variants thereof,
others are best interpreted in the Autosegmental-Metrical framework (e.g.
Pierrehumbert 1980; Beckman and Pierrehumbert 1986; Pierrehumbert
and Hirschberg 1990), now widely seen as the “‘standard theory’ of pros-
odic structure” (Selkirk 2011: 437). In particular, the end-based (e.g. Selkirk
1986, 1995) and relation-based (Nespor and Vogel 1986)  accounts to the
syntax–prosody relationship have been influential. The present chapter will
serve as a brief introduction to these frameworks and it will emphasize simi-
larities and differences between them to the extent that they are relevant in
the present context (see e.g. Ladd 2008 for a more comprehensive review).
The chapter will provide background information on relevant aspects of
intonational phrasing, prosodic theory and the syntax–prosody interface,
and it will define the notion of ‘prosodic separation’ for the present pur-
poses. As we go along, the phonetic parameters for boundary identification
will be introduced as relevant for the analysis reported on in Chapter 4.

87
88  Intonational phrasing and prosodic theory

3.1 Intonational phrasing: the Tone Group and the


Intonational Phrase
Parentheticals have been argued to be phrased in a separate prosodic con-
stituent, specifically the Tone Group (also known as tone unit or inton-
ation group) in the British tradition and the Intonational Phrase in the
Autosegmental-Metrical (AM) framework. The term intonation domain (ID)
is used throughout this book to neutrally refer to both the Tone Group and
the Intonational Phrase.1
A Tone Group (TG) in the British tradition has been defined in terms
of positions of prominence in speech. Palmer (1922: 7)  defines it as “a
word or a series of words in connected speech containing one and only
one maximum of prominence”. The TG consists of an obligatory nucleus
and up to three optional constituents: the prehead, head and tail (see
(3.1), optional components in parentheses; see e.g. Crystal 1969: 207f,
1972: 112).
(3.1) Structure of the Tone Group:
(prehead) (head) nucleus (tail)
The nucleus, i.e. the “stressed syllable of the most prominent word in the
Tone-Group” (Palmer 1922: 7), bears the nuclear tone. The nuclear tone
is described according to its shape. It may be realized on a single syllable
or on a longer stretch, if the nuclear syllable is followed by more syllables
in the same TG; these syllables constitute the tail (e.g. Palmer 1922: 10).
Different authors have varying numbers of nuclear tones (see e.g. Crystal
1969: 210–11 for an overview). For example, O’Connor and Arnold (1971:
7) have six tones: two kinds of falls, specifically the low fall, i.e. pitch fall-
ing “from a medium to a very low pitch” and the high fall, i.e. pitch falling
“from a high to a very low pitch”; two kinds of rises, specifically the low
rise, i.e. a pitch rising “from a low to a medium pitch or a little above” and
the high rise, i.e. pitch rising “from a medium to a high pitch”; the rise-fall,
i.e. a pitch rise “from a fairly low to a high” level quickly followed by a
fall “to a very low pitch”; and the fall-rise, i.e. pitch falling “from a fairly
high to a rather low pitch” and subsequently rising to a medium pitch.
Palmer (1922: 8) describes four tones for Southern British English: a fall,
the two rises, and the fall-rise. TG-internally, there must therefore be pitch
movement on, to or from at least one accented syllable. However, as Crystal
(1969: 210) notes, Palmer implicitly adds level nuclear tones to the tones
characterized by pitch movement. Crystal (1969: 211–20) distinguishes

1
The two categories are specifically discussed here because they are employed in much of the
literature on parentheticals. The selection does not a priori suggest a phonological or phon-
etic distinction between the two, nor does it a priori suggest that they are identical or that
the terms may be used as synonyms of each other.
Tone Group and Intonational Phrase  89

three main types of nuclear tones: simple tones involve unidirectional pitch
movement (rising, falling or level); complex tones involve “a change in the
direction of the pitch movement … within a syllable, and only one max-
imum of prominence” (Crystal 1969: 217), e.g. fall-rise and rise-fall, but
also rise-fall-rise and fall-rise-fall; compound tones “are combinations of
two kinetic elements of different major phonetic types acting as a single
tonal unit” (Crystal 1969: 218), one kinetic element being more prominent
than the other (see also O’Connor and Arnold 1971: 25–8). Gussenhoven
(2004: 296–302) reviews the nuclear contours occurring in English, com-
bining British-style terminology such as fall-rise, high rise, low rise, etc.,
with ToBI-style annotation as introduced below, and relating nuclear tones
to meaning.
The nucleus is optionally preceded by the head. The head reaches from
the first prominent syllable preceding the nucleus, also known as the onset,
up to but not including the nuclear syllable (e.g. Crystal 1969: 207, 226;
O’Connor and Arnold 1971: 15–25). The head in turn is optionally preceded
by the prehead, which comprises any unstressed syllables at the beginning
of the TG up to but not including the first stressed syllable, i.e. the head
(e.g. O’Connor and Arnold 1971: 15–16). The nucleus is optionally followed
by the (nuclear) tail, which comprises any syllables following the nucleus
before the end of the TG. These syllables may be stressed (but less prom-
inent than the nuclear syllable), but are more typically unstressed and they
continue the pitch movement until the end of the TG (see Crystal 1969: 207
and 223–5 for illustration; O’Connor and Arnold 1971: 11–15).2
To determine the exact location of a TG boundary after the nuclear
tone, in particular if the nucleus is followed by unstressed syllables which
could either be the tail of the nucleus or belong to the prehead of the fol-
lowing TG, the change in pitch level and/or direction on these syllables
has been seen as a “fairly clear boundary marker” (Cruttenden 1997: 34;
see also Crystal 1969: 205). According to Cruttenden (1997: 34), changes
in pitch level and direction most frequently occur on accented syllables.
If a pitch change occurs on unaccented syllables, it can be taken as an
indicator of a boundary. Specifically, after a falling tone followed by a
low unaccented syllable there will be a step-up to the pitch level of the
unaccented syllables at the beginning of a new domain, while after a rising
tone, there is “a step-down to the pitch level of any unaccented syllables
at the beginning of the following [domain]” (Cruttenden 1997: 34). For
example, the sentence in (3.2)a, from Cruttenden (1997: 37), is ambigu-
ous such that of course may be phrased with the first clause (He went to the
States; see (3.2)b), or with the second clause (he didn’t stay very long; see

2
In Crystal’s data “[l]ess than 10 per cent of all nuclei have tails with stressed syllables on fol-
lowing words” (Crystal 1969: 224).
90  Intonational phrasing and prosodic theory

(3.2)c), and may be interpreted accordingly. Suppose States is the nuclear


syllable associated with a falling tone. One cue to the position of the TG
boundary is the pitch level on the series of unaccented syllables following
the accented syllable States. The pitch level on of course is low for both
phrasing options. However, a boundary between of course and he, as indi-
cated in (3.2)b, will involve a pitch discontinuity and a step-up in pitch on
he following of course, while a TG boundary before of course as indicated
in (3.2)c will instead involve a pitch discontinuity following States (see
Cruttenden 1997: 37 for illustration).
(3.2) a. He went to the States of course he didn’t stay very long.
b. [He went to the States of course] [he didn’t stay very long]
c. [He went to the States] [of course he didn’t stay very long]
In Palmer’s (1922) system, the diacritic || indicates a boundary between
two TGs. However, the diacritic “does not imply any break or pause, except
when preceded by a comma or other sign of punctuation”, i.e. the diacritic
signals the end of the nuclear tune of one TG, but it does not represent any
other phonetic boundary markers. In some of the literature, two different
kinds of TG have been suggested such that a major TG is composed of a
number of minor TGs, corresponding in notation to single bar (minor TG
boundary) and double bar (major TG boundary) (e.g. Trim 1959).
Like the TG, the Intonational Phrase (IP) in the AM framework has been
intonationally defined (e.g. Pierrehumbert 1980; Selkirk 1981; Beckman
and Pierrehumbert 1986; Pierrehumbert and Hirschberg 1990 and Nespor
and Vogel 1986: 188). Unlike in the British system, in the AM framework it
is not the shape of the F0 contour that is described but the tonal structure
consists of a string of local targets, referred to as tones, specifically high
(H) and low (L) tones; pitch movement between these targets is accounted
for by rules of interpolation. Tones are of two types: pitch accents and edge
tones. Pitch accents are local tonal events associated with a prominent syl-
lable. There are monotonal and bitonal pitch accents; a star indicates the
association with the accented syllable. For example, a prominent syllable
associated with a local pitch maximum has a monotonal H* pitch accent;
L+H* refers to a local event such that the peak target is immediately pre-
ceded by a sharp rise from a local minimum (or valley) in the contour.
The tone preceding the starred tone in a bitonal pitch accent (e.g. L in
the L+H* pitch accent) is referred to as the leading tone; a tone following
the starred tone in a bitonal accent (e.g. a low target associated with the
prominent syllable immediately followed by a sharp rise: L*+H) is called
the trailing tone. The ToBI (Tone and Break Indices) transcription system
for English (Silverman et  al. 1992; Beckman and Elam 1993; Beckman,
Hirschberg and Shattuck-Hufnagel 2005) has five pitch accent types: H*
(!H*), L*, L*+H (L*+!H), L+H* (L+!H*) and H+!H*; the exclamation
mark ‘!’ is the diacritic for tones downstepped relative to a preceding H
Tone Group and Intonational Phrase  91

tone.3 Edge tones are of two types: phrase accents and boundary tones.
Boundary tones are associated with the (right) edge of IPs; they are high
(H%) or low (L%) targets terminating the IP. Phrase accents (L-, H-) are
edge tones of intermediate phrases (see below).4 The IP is the domain of a
complete intonational contour (CTC), which means that it has at least one
nuclear pitch accent and is terminated by a sequence of phrase accent and
boundary tone (T-T%). English has two boundary tones: high (H%) and
low (L%) (Beckman et al. 2005; Jun 2005, and much previous work). H%
serves to indicate continuation as well as the end of yes-no questions, while
L%, if in sentence-final position, typically terminates declaratives and wh-
questions (for details, see Pierrehumbert and Hirschberg 1990; Hirschberg
2004; Nolan 2006 among others).
Since the British tradition focuses on the shape of the nuclear contour
rather than tonal targets, there is no equivalent for T% in the British trad-
ition. Instead, “British nuclear tones such as fall-rise conflate the pitch
movement on the last accent (e.g. fall) with the pitch movement at the
end of the phrase (e.g. rise)” (Ladd 1996: 88). At the end of an intonation
domain, a fall-rise would thus correspond most closely to H* L-H%, and
a rise-fall to L*+H L-L% or L* H-L%. In sentence-medial position, the
fall-rise (H* L-H%) is also known as the “continuation rise” (see (3.3),
from Selkirk 1984: 288; see also Selkirk 2005), indicating that there is more
to come. In (3.3), the fall-rise (continuation rise; H*L-H%) is realized on
lunch, i.e. pitch accent, phrase accent and boundary tone are all realized on
one syllable. If more syllables follow the prominent one, the same contour
will stretch over more syllables, as shown in (3.4).
(3.3) Continuation rise
After  lunch,  we think we’ll go for a drive.
H* L- H%
(3.4) Continuation rise
After lunch with them,  we went for a drive.
H* L- H%

3
ToBI is a transcription system. Its aim is to provide a framework “for developing com-
munity-wide conventions for transcribing the intonation and prosodic structure of spo-
ken utterances in a language variety” (www.ling.ohio-state.edu/~tobi/; last accessed on 9
September 2013). Complete and nearly complete ToBI systems have been developed for a
variety of languages, among them English, Catalan, German, Greek, Japanese, Korean and
Portuguese (see www.ling.ohio-state.edu/~tobi/ and links provided there).
4
The phrase accent T- is the edge tone terminating the intermediate phrase (ip). In
Pierrehumbert (1980), the assumption of the phrase accent T- was mainly motivated by the
observation that two tonal targets can be identified after the nuclear pitch accent, i.e. that
there is a tone after the last pitch accent but before the phrase boundary T%. In Beckman
and Pierrehumbert (1986: 256) the phrase accent in English is defined as “a tone that fills
the space after the last pitch accent in a phrase”. (See also Pierrehumbert and Hirschberg
1990; Grice, Ladd and Arvaniti 2000, among others.)
92  Intonational phrasing and prosodic theory

Ladd (2008: 91, Table 3.1) offers a summary of the possible combinations


of pitch accents and edge tones in the AM system and possible British-style
equivalents (see also Roach 1994 for an attempt to convert British nuclear
tones to ToBI transcription equivalents, and Cruttenden 1997: 61–63 for
simultaneous interlinear British-style and ToBI notation), but Ladd (2008:
90)  also notes that a complete correspondence between the two systems
cannot be established. I agree with Ladd (2008: 45) in that the distinction
between tonal events related to prominence (pitch accents) and those related
to domain edges (i.e. edge tones) is a useful one. Consider Ladd’s (2008:
45–6) example in (3.5) below. Both B1’s and B2’s utterances are realized
with a rising-falling-rising tune, which is associated with the syllable Sue in
B1 but spans driving instructor in B2. The nuclear syllables are Sue in B1 and
dri- in B2; thus while in B1 the nuclear syllable is associated with the entire
nuclear tune, in B2 the falling-rising part is associated with unstressed syl-
lables and the rising event is associated with the edge of the domain. Also,
the different parts of the tunes may serve distinct functions. For example,
in (3.3) and (3.4), it is reasonable to assume that the pitch accent (i.e. the
high peak with subsequent fall) conveys the news value of lunch, while the
rise towards H% at the end of the prosodic constituent signals non-finality.
Similarly, the rise terminating the utterances in (3.5) signals speaker atti-
tude (surprise, disbelief).

(3.5) A:  I hear Sue’s taking a course to become a driving instructor.

B1: Sue!?

B2: A driving instructor!?

The IP is assumed to be made up of one or more prosodic constitu-


ents one level down in a hierarchy of prosodic constituent categories. In
Pierrehumbert’s and related work, this is the intermediate phrase (ip; also
known as phonological phrase or prosodic phrase, among others; see Frota
2012: 257 for an overview), which in turn is made up of one or more pros-
odic words. The ip contains at least one pitch accent and is terminated by a
tonal event, the phrase accent (T-) but not a boundary tone (T%). While not
all authors agree on the number of prosodic category types and on termin-
ology, there is general agreement among researchers working in this frame-
work that there is more than one level above the prosodic word (see e.g.
Frota 2012 for discussion). The figure of the prosodic hierarchy and relation
to tones in (3.6) is borrowed from Krivokapić (2007: 164), who adopted it
from a course reader edited by Mary E. Beckman (see also Selkirk 1981;
Truckenbrodt 2007: 436 and Selkirk 2011: 437, among many others).
Tone Group and Intonational Phrase  93

(3.6) Prosodic hierarchy and association with tones in the AM


framework5
IP Intonation Phrase

ip ip Intermediate Phrase

ω ω ω ω Word

σ σ σ σ σ σ σ σ Syllable

T* T* T– T* T–

T%

Along with the complete tonal contour (CTC) and prosodic constituency,
the IP has been defined in terms of its edges. According to Nespor and Vogel
(1986: 188), the right edge of an IP is the place where a pause may poten-
tially occur. Pauses may be silent or they may be filled (e.g. uh, uhm), with
filled pauses occurring in spontaneous speech more than in read speech
(Stenström 1990). However, since pauses are not obligatory and both their
occurrence and length depend on a number of factors such as speaker, speech
rate, prosodic structure, segmental make-up, phrase length, syntax and dis-
course (see e.g. Krivokapić 2007 for an overview), pauses cannot always be
considered a reliable cue to IP edges. Pauses may be absent at IP boundaries
and pauses owing to performance factors may be present within an IP. It is
therefore important to distinguish between structure-related (breath) pauses
on the one hand, and hesitational or performance-related stretches on the
other (see also Stenström 1990 on pauses in monologues and dialogues
in ten texts from the London-Lund Corpus, and Cruttenden 1997: 30–2
for pauses within intonation groups vs. pauses at their edges). Moreover,
experimental work by Steinhauer, Alter and Friederici (1999) shows that in
the absence of pauses, listeners will still reliably perceive prosodic bound-
aries if other prosodic cues are present, thus perception of boundaries does
not depend on pauses. Similarly, based on data from the Boston University
Radio Speech corpus, Yoon, Cole and Hasegawa-Johnson (2007) report that
the perception of a prosodic boundary (ip or IP) does not depend on the
presence of a silent pause.
IP edges are furthermore defined along the lines of segmental rule appli-
cation (Nespor and Vogel 1986: 205ff). For example, Nespor and Vogel (1986:
205–11) show for the Tuscan variety of Italian that the sandhi rules Gorgia
Toscana and Intervocalic Spirantization apply within the IP but not across
5
© Elsevier; reproduced with permission.
94  Intonational phrasing and prosodic theory

IP boundaries. Furthermore, the rules of Nasal Assimilation in Spanish and


s-Voicing in Greek are sensitive to IP boundaries in this way (Nespor and
Vogel 1986: 211–16), as are fricative voicing, syllable degemination and vowel
adjacency resolution processes in European Portuguese (Frota 2000). More
generally speaking, segmental rule application has been seen as an important
cue to prosodic boundaries of different categories and strengths (e.g. pros-
odic word, intermediate phrase, Intonational Phrase, phonological utter-
ance) for various languages. For (non-rhotic varieties of) British English,
the prosodic utterance has been argued to be the domain of two segmental
processes involving /r/, i.e. these two processes apply across IP boundaries:
the ‘Linking-r’ (see (3.7)) and the ‘Intrusive-r’ (see (3.8)) (Nespor and Vogel
1986: 226–9). In (3.7) and (3.8), the comma indicates the position of the IP
boundary, underlining indicates the target position for the two phenomena.
In (3.7), the linking-r may be produced to connect year and over despite the
IP boundary. Similarly, in (3.8), the intrusive-r may be pronounced to con-
nect law and a. See also the examples in (3.10) and (3.11) below.
(3.7) Linking-r (from Nespor and Vogel 1986: 228)
Just last year, over a hundred dinosaur tracks were discovered in the
Arizona desert.
(3.8) Intrusive-r (from Nespor and Vogel 1986: 229)
Even though they’re protected by law, a lot of migratory birds are
killed by hunters every year.
Another cue to constituent edges is declination and pitch reset, i.e. the
interruption of a downward pitch trend (see e.g. Gussenhoven 2004 for
phonetic declination and phonological downstep). For English, the lev-
els of IP (e.g. Pierrehumbert 1980; Gussenhoven 2004: 113)  and ip (e.g.
Beckman and Pierrehumbert 1986: 298–300) have been argued to be the
prosodic domain across which declination/downstep applies, such that later
pitch peaks within domains are realized with lower pitch compared to earl-
ier ones; i.e. pitch reset occurs after a phrasal IP boundary indicating a new
domain. Across subsequent domains, there may be an overall downtrend
such that pitch reset in a later domain is not to the same maximum level as
pitch peaks in earlier domains. For example, within an utterance consisting
of two IPs, declination applies across each IP and there is pitch reset at the
boundary between them, but superimposed is an overall downward trend
across the utterance (Ladd 1986). Moreover, specific events such as focal
pitch peaks may interrupt the overall downtrend. Parenthetical phrases
may also interrupt this overall downtrend such that their overall pitch is
lower (compression) or higher (expansion) than the trend associated with
the host utterance. After the parenthetical, the pitch level may be reset to
a level higher than before the parenthetical, indicating separate IPs before
and after the parenthetical (Figure 3.1a; parenthetical at compressed pitch
level) or it may return to where it was left before the parenthetical insertion,
Tone Group and Intonational Phrase  95

(a)

IP1[host]IP1 IP2[parenthetical]IP2 IP3[host]IP3

(b)

IP1[host IP2[parenthetical]IP2 host]IP1

Figure 3.1  Non-recursive and recursive pitch structures. Panel a: non-


recursive structure: pitch reset in IP3 as compared to IP1; panel b:
recursive structure: no pitch reset after the parenthetical as compared
to before the parenthetical; solid lines represent the superimposed
overall downtrend of the utterance; dashed lines in panel a represent
the downtrend within one IP

suggesting a recursive prosodic structure (Figure  3.1b; Ladd 1986, 1996,


2008; see also Wichmann 2000, 2001).
Another cue to prosodic boundaries is domain-final (pre-boundary)
lengthening of segments and syllables such that phonological units located
at the end of a prosodic constituent are longer than the same units located in
constituent-medial position (Lehiste 1973; Gussenhoven and Rietveld 1992;
Vaissière 1983; Wightman et  al. 1992; Ferreira 1993; Turk and Shattuck-
Hufnagel 2007, among many others; see Turk 2012 for a review of the tem-
poral implementation of prosodic structure). In (3.9), for example, taken
from Ferreira (1993: 234), the word black is presumably longer in (3.9)a than
in (3.9)b, because in (3.9)a it is located at a syntactic and prosodic boundary,
while in (3.9)b it is located within a syntactic phrase (the subject NP) and
within a phrasal prosodic constituent.6 In the ToBI system lengthening is
taken as a cue to either intermediate phrase or intonational phrase boundary
(see Beckman et al. 2005).
(3.9) Phrase-final lengthening
a. The table that I thought was black tempted me.
b. The black table tempted me.

6
In her study, Ferreira (1993) separates the syntactic and prosodic structure in order to test
whether phrase-final lengthening is due to the effect of prosodic or syntactic structure.
96  Intonational phrasing and prosodic theory

Given the nature of the spoken corpus data analysed in Chapter 4, the work
documenting the prosodic transcription and labelling in the Lancaster/
IBM Spoken English Corpus (see Knowles 1991) is also of interest in the
present context. The boundary cues fall into three categories: (i) temporal
discontinuities, which essentially include silent and filled pauses and final
lengthening, and combinations of these; (ii) pitch discontinuities, which
essentially correspond to Cruttenden’s (1997) changes in pitch level as out-
lined above; and (iii) segmental discontinuities, which refer to the block-
ing of processes of connected speech such as assimilation, elision, r-linking,
[j, w] glides after close vowels, gemination of stop phases, and contraction
(Knowles 1991: 151–4). Consider the examples in (3.10) and (3.11).
(3.10) Contraction
A: What is your profession?
a. B: We’re linguists and he’s a mathematician.
b. B: Well, WE are linguistics, but HE is a mathematician.
H*L H%
c. B: *Well, WE are linguistics, but   HE  ’s a mathematician.
H*L H%
(3.11) Gemination of stop phases
a. Many people are afraid of black cats.
b. While many cats are black, cats may also be grey.
(3.10) is an example of contraction. In (3.10)a, the subject and copula verb
are in the same intonation domain and are contracted. In (3.10)b, with con-
trastive emphasis on the subject pronouns, it is possible to separate the sub-
ject pronoun he from the remainder of the sentence by a domain boundary.
Contraction would then be blocked (see the ungrammaticality of (3.10)c).
In (3.11), there are two adjacent velar plosives in the sequence black cats,
which may be geminated in connected speech. In (3.11)a, no boundary
above the level of prosodic word is predicted between black and cats, and the
first of the two stops will not be released. In (3.11)b, on the other hand, the
comma will likely coincide with the boundary of an intonational domain, and
gemination will therefore be blocked and the final stop on black is expected
to be released.
Based on the analysis of the corpus data, Knowles (1991: 156ff) also
suggests a distinction between ‘major’ and ‘minor’ tone groups within the
British school. While major TGs are always separated by pauses, pauses may
or may not occur between minor TGs, and pauses between major TGs would
be longer. Functionally, “a major tone group corresponds to a segment of
text of the same order as a sentence”, while “the minor tone group [corres-
ponds] to part of a sentence” (Knowles 1991: 157). This division is related
to the distinction between the IP and the ip in the AM system. However, the
IP/ip distinction relies on tonal as well as temporal cues. Knowles (1991)
Prosodic separation  97

only touches on variables other than pauses, including pitch discontinuities.


Due to their varying nature as structure-related or hesitation pauses, pauses
alone seem too unreliable a cue to be taken as the key feature for the dis-
tinction between prosodic constituents of different hierarchical levels. As
Knowles (1991: 158) puts it, “[t]here is no one-to-one relationship between
prosodic breaks and prosodic group boundaries”. Roach (1994: 95) argues
that “on the basis of experience so far” the distinction made by Knowles in
the corpus transcriptions does not correspond to the ip/IP distinction and
that instead, both the minor and the major TG “require the assignment of a
boundary tone”, i.e. correspond to IP.
The following section will look more closely at the correspondence
between TG and IP and will define prosodic separation for the present
purpose.

3.2 Prosodic separation and the correspondence between the Tone


Group and the Intonational Phrase
It is obvious that the Tone Group and the Intonational Phrase have vari-
ous properties in common, among them the fact that they have been into-
nationally defined. As Ladd (1986: 311)  puts it, “[t]hey have a specifiable
intonational structure, including … a single most prominent point (primary
stress, tonic, nucleus)”. To this, Ladd (ibid.) adds two more properties they
have in common: (i) both TG and IP “are the largest phonological chunk
into which utterances are divided”; (ii) TG and IP “are phonological units
which are … assumed … to match up in some poorly understood way with
elements of syntactic or discourse-level structure”. Section 3.3 below will
outline and compare the ways in which TG and IP have been related to syn-
tactic structure.
If a systematic correspondence could be established between the British-
style nuclear tone on the one hand and combinations of pitch accent, phrase
accent and boundary tone in the AM system on the other, then this would
result in a more general correspondence between the TG in the British trad-
ition and the IP in the AM framework. Attempts have been made to relate
the two intonational systems (see e.g. Roach 1994: 96 and Ladd 2008: 91;
e.g. a (rise-) fall corresponds to (L+)H* L-L%; see also the previous sec-
tion). According to Beckman and Pierrehumbert (1986: 286), “the sequence
of nuclear pitch accent, phrase accent and boundary tone is a phonological
decomposition of what would be referred to as the ‘nuclear tone’ in the
British tradition”. However, including T% in each case of correspondence
between British-style nuclear tone and AM sequence, the nucleus in the
British tradition systematically corresponds to the last pitch accent in a series
of pitch accents in an IP, even if the IP is made up of a sequence of inter-
mediate phrases, each of which has at least one pitch accent, i.e. the compari-
son holds for IP-final pitch accents, only. Roach (1994), on the other hand,
98  Intonational phrasing and prosodic theory

generally allows for the correspondence between a British-style nucleus on


the one hand and a combination of non-final pitch accent and phrase accent,
i.e. intermediate phrase edge tone, on the other, thus allowing for the cor-
respondence between TG and ip (but see Roach 1994: 96f for problems this
causes). In the AM framework, the nuclear accent is referred to as the pitch
accent “on the word with most prominence … in each intermediate phrase”
(Pitrelli, Beckman and Hirschberg 1994: 123), and “the last accent in the
intermediate phrase, not the intonation phrase” (Ladd 2008: 133), i.e. nuclear
accent refers to main prominence in the ip. Consequently, the British-style
nucleus may correspond to the last accent in either an ip or an IP; and the
British-style TG may thus correspond to either the ip or the IP.
In my view, the distinction is important here because parentheticals have
been argued to be separate TGs in the British tradition (e.g. Schubiger 1958
and subsequent work), and separate IPs in the AM tradition (e.g. Selkirk
1981, 1984, among many others). However, if a TG may in principle corres-
pond to an ip as well as to an IP, a parenthetical that is separate in an analysis
couched in the British system may count as integrated (i.e. not phrased in a
separate IP but in a separate ip contained in an IP) in an analysis couched in
the AM tradition. A definition of ‘prosodic separation’ is therefore in order.
One example to illustrate the need for the distinction between the two
frameworks and, more importantly, the definition of the notion ‘prosodic
separation’ in the present context is Beckman and Pierrehumbert’s ana-
lysis of question tags (QTs). Beckman and Pierrehumbert (1986: 295) take
the boundary between host clause and QT as one piece of evidence for the
existence of the intermediate phrase (ip) in English. They discuss two pos-
sible realizations of the accented sentence-final QT in (3.12), as plotted in
Figure 3.2 (from ibid.). In each of these realizations, there is a nuclear fall
on win, which is not continued on the QT. In the first realization (top panel
in Figure 3.2), the fall from the local pitch peak associated with win is com-
plete by the end of win; the QT is associated with a rise terminated by H%.
In the second realization (bottom panel in Figure 3.2), the peak and subse-
quent fall associated with win is followed by another pitch peak associated
with won’t and a subsequent fall towards L%. In both occurrences, win is
assumed to be ip-final and associated with a nuclear accent. Beckman and
Pierrehumbert (1986) argue that if in the second version (bottom panel
in Figure  3.2) the whole sentence Mary will win, won’t she were phrased
together in one IP, the fall on the QT, given its rightmost position, would
be the nuclear accent in the phrase. However, they argue that intuitively,
and given its lower pitch range and peak, it is subordinate to the one in the
host. They maintain that an “intermediate phrase boundary before the QT
provides a way to handle the subordination” because “each intermediate
phrase can have its own pitch range” (Beckman and Pierrehumbert 1986:
195). Both the rising (top panel of Figure 3.2) and the falling (bottom panel)
QT are analysed as separate ips, which, being IP-final, are associated with
L*L-H% and H*L-L%, respectively. This analysis reflects the intuition
Prosodic separation  99

that tags may be prominent, but also that if they are, they are less promin-
ent than and subordinated to the nuclear accent in the host (Beckman and
Pierrehumbert 1986 do not discuss cases of unaccented QTs). According
to the common assumption in the literature that parentheticals are separate
IPs (see Chapter 2 above), a QT which is phrased in an ip of its own but is
not preceded by an IP edge would have to be analysed as prosodically inte-
grated (i.e. not separate), because it is not phrased in its own IP.
(3.12) Mary will win, won’t she
Let us compare Beckman and Pierrehumbert’s AM analysis to a British-
style analysis of the QT in (3.12). The utterance in (3.12), pronounced as
in the top panel of Figure 3.2, i.e. with the fall completed at the end of win
and a subsequent rise from a prominent low syllable (won’t), could perhaps
be analysed as a compound tone in Crystal’s (1969) sense, with a falling tone
followed by a rising one and the former one being more prominent than
the latter. The utterance would then be phrased in a single TG with a com-
pound nuclear tone. However, for two kinetic elements to form a compound
tone “there must be no evidence of a tone-unit boundary between the tones”
and there must be “an evenness of pitch pattern” between the two events
(Crystal 1969: 218). This is not the case here: Beckman and Pierrehumbert
(1986) argue that win is phrase final, thus identifying some kind of boundary
cue between the two tonal events. If the string Mary will win and won’t she
are analysed as two separate TGs, the latter cannot be analysed as subordin-
ate to the first in Crystal’s (1969) sense, because in order for one unit (here:
TG2, won’t she) to be subordinate to another one (here: TG1, Mary will
win), “[t]he nuclear type postulated as subordinate must repeat the direction
of the nucleus in [TG1]” (Crystal 1969: 245), which is not the case here.
Following Crystal (1969), the two TGs in the top panel of Figure 3.2 must
therefore be treated as independent. Likewise, an analysis along the lines
of a compound tone is impossible for the utterance in (3.12) pronounced
with rising-falling pitch on the tag after a completed fall associated with
win (bottom panel of Figure  3.2) for at least two reasons. First, Beckman
and Pierrehumbert (1986) identify a boundary after win, and second the two
nuclear tones do not have an endocentric relationship (Crystal 1969: 218).7
They thus have to be interpreted as separate. Here it is possible to analyse
the second TG as subordinate to the first along Crystal’s lines – the second
nuclear tone repeats the first and the width of the nuclear pitch movement
is greater in TG1 than in TG2 (see Crystal 1969: 245) – but there are still
two nuclear tones, thus two TGs, with two separate pitch contours and two
places of prominence.
7
An endocentric relationship between tones is one between tones of ‘opposite’ pitch direc-
tion, e.g. combinations such as rise+fall, rise-fall+rise, or fall+rise, but not repetitions of
tones of identical pitch direction, such as fall+fall, rise+rise or fall-rise+rise (which are
exocentric sequences). In compound tones, the combined tone units are in an endocentric
relation; exocentric sequences are either separate or subordinate; see Crystal (1969).
100  Intonational phrasing and prosodic theory

(a)

400

300

200

100
Mary will win, won’t she

Rising question tag


(b)
400

300

200

100
Mary will win, won’t she

Falling question tag


Figure 3.2  Example (3.12), from Beckman and Pierrehumbert (1986:
295):* two possible realizations of an accented sentence-final QT. Top
panel: QT associated with L* L-H%; bottom panel: QT associated
with H* L-L%
* © Cambridge University Press; reproduced with permission.

Note also that the local prominence associated with won’t cannot be ana-
lysed as a postnuclear prominence or phrase accent (see the discussion in
Ladd 2008: 142–7). The two pitch accents in the bottom panel of Figure 3.2
are clearly phrased in separate phrases. While the second one (associated
with won’t) may be intuitively subordinate, i.e. weaker than the one asso-
ciated with win, it is nevertheless independent and it conveys a separate
meaning. Question tags associated with intonational prominence have gen-
erally been noted to convey speech acts independent from those of their
Prosodic separation  101

host utterances (e.g. Ladd 1981); they serve a specific function with respect
to the host utterance, for example expressing certainty on the part of the
speaker about the truth of the proposition expressed by the host or seeking
confirmation (see Dehé and Braun 2013 for an overview and references).
This is not true for the parts of speech associated with postnuclear prom-
inence in examples such as (3.13) (adapted from Ladd 2008: 144; postnu-
clear prominence indicated by underlining). The postnuclear prominence in
(3.13), i.e. the prominence associated with the first syllable of dancing, is not
related to a separate speech act or discourse function. In an AM analysis of
(3.13), the postnuclear prominence associated with the first syllable of dan-
cing is analysed as a phrase accent (L-), which has a secondary association
with a postnuclear accent following the nuclear stressed syllable in the same
intermediate phrase. An analysis along these lines, however, is not feasible
for (3.12)/bottom panel of Figure 3.2.
(3.13) I thought she was  dancing tonight.
H*       L- H%
In sum, the utterance in the bottom panel of Figure 3.2 has to be analysed
as consisting of two separate intonation domains: ips according to Beckman
and Pierrehumbert (1986), and TGs in the British system. Following stand-
ard assumptions in the literature on parentheticals (see Chapter  2 above),
the QT would count as prosodically integrated (i.e. not phrased in a sep-
arate IP) in the AM system and as prosodically separate (i.e. phrased in its
own TG) in the British framework.
The inaccuracy with respect to prosodic ‘integration’ vs. ‘separation’
results from the insufficient definition of these terms, along with differences
related to the hierarchy of prosodic categories between different systems,
as well as the indeterminacy of prosodic boundaries as already discussed
by Ladd (1986). In my own analysis in Chapter 4 below, examples like the
second realization of (3.12) (bottom panel of Figure 3.2) have been analysed
as prosodically separate. Crucially, prosodically separate parentheticals have
nuclear prominence and are set off from preceding and following material
by one or more clear boundary markers. Prosodic separation in English can
thus mean phrased in a separate ip or IP, extending the more common use
of prosodic separation in the literature on parentheticals from ‘phrased in
its own IP’ to also include the ip; however, it always means that the sepa-
rated material has nuclear prominence and is set off from surrounding host
material by edge tones (T- or T-T%). This analysis was guided by the fol-
lowing reasoning. First, cases such as (3.12)/bottom panel of Figure  3.2
are set off from their host semantically and phonetically. There is a tonal
boundary before the QT which, according to Beckman and Pierrehumbert,
is associated with L-; host and QT have their own nuclear accent, even if the
second one sounds subordinate. Second, as Beckman and Pierrehumbert
(1986: 289–98) explain, many cases previously (e.g. by Pierrehumbert 1980)
102  Intonational phrasing and prosodic theory

analysed as IP boundaries are reanalysed by Beckman and Pierrehumbert


(1986) as ip boundaries. This includes cases such as listed NPs, sequences
of modifiers, the attachment of vocatives, and other tags. The difference
between ip and IP boundary may be, for example, the distinct articulation of
specific phonetic parameters. For example, in Beckman and Pierrehumbert’s
(1986: 290) listing example, the listed NPs all end in a rise, “but the final
rise is by far the largest”, thus only the last one is interpreted as a rise to
H%, while preceding rises at the ends of listed NPs are interpreted as rises
to H-. Third, “[p]hrasing in English is highly facultative” (Beckman and
Pierrehumbert 1986: 288): ip boundaries, for instance, can easily be pro-
moted to IP boundaries, and IP boundaries can be weakened into ip bound-
aries, depending on factors such as speech rate or constituent length (see
also Nespor and Vogel 1986; Selkirk 2005, among others). For example, the
QT in (3.12)/bottom panel of Figure 3.2 can easily be phrased in its own IP,
preceded by L% and with its own non-subordinate accent. However, fully
integrating it, for example as unstressed or little stressed postnuclear mater-
ial (i.e. British-style tail) would change its meaning, and thus its contribu-
tion to the discourse (e.g. Ladd 1981 on the distinction between nuclear and
postnuclear tags and their interpretation). It is conceivable that prosodic
integration (i.e. downgrading an IP or ip boundary to a word boundary)
can therefore not occur as readily, perhaps indicating the relevance of the
semantic-pragmatic contribution, as marked, for example, by Potts’ (2005)
comma feature. As Potts notes, if the comma intonation disappears, so does
the CI semantics. Finally, from the annotator’s perspective, it has been dem-
onstrated that the intermediate phrase level is harder to annotate reliably
(Syrdal and McGory 2000), especially for spontaneous spoken language
data, for which temporal and articulatory cues to boundary strength com-
plementing tonal criteria, such as domain-final lengthening and segmental
phenomena, cannot be reliably taken into account. For example, Yoon et al.
(2004: 2732) compare their own results with previous research in inter-tran-
scriber reliability and conclude that “decisions about boundary strength are
more difficult for spontaneous speech than for read speech”, which is rele-
vant in the present context because the spoken part of ICE-GB includes
a large percentage of spontaneous and semi-spontaneous speech. Based on
data from a perception study using the Buckeye corpus of spontaneous, con-
versational speech (Pitt et al. 2007), Cole, Mo and Baek (2010) report that
higher-level prosodic boundaries (i.e. IP boundaries) are more frequently
perceived by untrained listeners than lower-level boundaries (such as ip
boundaries). Cole et al. (2010: 1170) note that “ip-boundaries (L- and H-)
are potentially ambiguous with the IP boundaries (L-L% and H-L% […])”
and that ip-level boundaries are distinguished from IP-level boundaries by
“the degree of F0 lowering, final lengthening or decreased intensity, and by
the likelihood of creaky voicing” as opposed to the presence or absence of
these criteria (their italics). The “inherent ambiguity of IP boundaries”,
they continue, “is lesser, since in at least some cases there are distinctive F0
Prosodic separation  103

contours that derive from the sequence of ip+IP tones, which do not arise
with the single tones of the ip boundaries”. This hypothesis, they argue, “is
supported by the observation that even expert prosody transcribers struggle
to annotate lower-level boundaries” (Cole et al. 2010: 1170–1). It is there-
fore certainly possible that parentheticals analysed as prosodically separate
in Chapter 4 below are phrased in their own ip, rather than IP. However, this
is only rarely the case for full parenthetical clauses, non-restrictive relative
clauses and nominal appositions, which are typically tonally and temporally
set off from surrounding material.
A domain boundary in the present context is one clearly marked by an
edge tone and preceded by a nuclear accent. Separation involves boundaries
to the left and right of the parenthetical as well as nuclear prominence in the
parenthetical. A definition of ‘prosodic separation’ is given in (3.14).
(3.14) Prosodic separation
A segmental string (e.g. a parenthetical) is prosodically separate if
it has a nuclear accent and is immediately preceded and followed
by a phrasal prosodic boundary marked by edge tones along with
other phonetic parameters marking the edge of a phrasal prosodic
constituent.
This notion of prosodic separation allows us to keep up what I consider an
important distinction: the distinction between parentheticals that are pro-
sodically integrated (phrased together with host material) and those that are
prosodically separate. Note that the definition of prosodic separation allows
for prosodically integrated parentheticals to be prominent, because separ-
ation demands nuclear prominence and boundaries to the left and right of
the parenthetical. As shown in Chapter 4 below, integrated parentheticals of
type CC, RV and QT may indeed be unstressed or they may be associated
with nuclear or non-nuclear prominence; parentheticals of type full paren-
thetical clause, NRRC and N-APP are never unstressed, although N-APPs
in particular may phrase together with host material.8

8
Note that ‘prosodic separation’ might have to be defined differently for other languages.
For example, based on a production experiment using parentheticals of the kind bence (‘I
think’, lit: ‘for me’) and yanılmıyorsam (‘if I am not mistaken’) Güneş (2012) concludes that
parentheticals in Turkish are integrated as intermediate phrases. Interestingly, Güneş (2012)
shows that parentheticals in Turkish do not behave in any other ways than as syntactic
adjuncts and arguments, i.e. they do seem prosodically more integrated than parentheticals
in languages such as English. Güneş (2012) takes her results as evidence for Turkish being a
phrase language (see Féry 2010), i.e. a language which has no pitch accents or lexical tones
but phrasal tones only, adding to the evidence for Féry’s (2010) typology on intonation sys-
tems. Féry (2010) posits a typology of intonation systems in which the focus is on sentence
melodies rather than lexical properties such as lexical tone or stress. Her typology consists
of three categories: (i) intonation languages, which have pitch accents as well as (pragmat-
ically triggered) phrasal tones (e.g. English, German, Italian, French, Swedish), (ii)  tone
languages, which have lexical tone but few phrasal tones, and (iii) phrase languages, which
do not have pitch accents or lexical tones, but phrasal tones only (e.g. Indian languages such
as Hindi, Bengali, Tamil and Malayalam). According to Güneş (2012), Turkish, too, has the
properties of a phrase language.
104  Intonational phrasing and prosodic theory

3.3 The syntax–prosody interface: the relation between syntactic


and prosodic constituents
Having defined prosodic separation, this section focuses more closely
on the relation between prosody and syntax, which is necessary to better
understand the relation between syntactic parenthesis and prosodic con-
stituency. The syntax–prosody interface has been the topic of much work;
various options of relating syntactic structures and phonological processes
have been discussed (see Scheer 2011 and Selkirk 2011 for recent over-
views). Based on a brief revision of several decades of research, Selkirk
(2011) argues “for a prosodic structure representation of phonological
domains, a representation that is independent of syntactic constituency
but related to it by a module of syntactic-prosodic constituency corres-
pondence constraints”. I follow here this kind of approach to the syntax–
prosody interface, an approach which has been developed in the last three
decades or more. The prosodic hierarchy as given in (3.6) above has been
related to syntactic structure in much previous work (see e.g. Truckenbrodt
2007; Selkirk 2011 for overviews). According to Selkirk’s (1986, 1995)
cross-linguistic theory of edge-alignment, edges of prosodic constituents
are aligned with edges of syntactic constituents such that the edge of a (lex-
ical) syntactic phrase (XP) aligns with the edge of an intermediate phrase
(ip), and the edge of a clause aligns with the edge of an Intonational Phrase
(IP). The correspondence between syntactic and prosodic phrase edges has
been formulated as a violable constraint, summarized as the edge-alignment
constraint in (3.15). For English, it has been argued that right alignment is
relevant; see the constraints in (3.16); for the IP in English, Selkirk (2005)
argues that the alignment constraint in (3.17) holds, relating the IP to Potts’
(2005) comma feature (see also below). Kawahara (2012) adds evidence for
this constraint from Japanese. More specifically, he takes the results of a
reading study designed to test the prosodic properties of Japanese nominal
parentheticals as evidence for the existence of the level of IP as part of the
Japanese prosodic hierarchy and towards the universality  of this level in
the prosodic hierarchy. Kawahara further takes the results as evidence for
an alignment constraint at work in Japanese that aligns the left edge of a
Comma Phrase with the left edge of an IP.
(3.15) Edge-alignment: Align R/L (Σ1, πΣ1) (from Selkirk 2005: 18)
Align the R/L edge of a constituent of type Σ1 in syntactic (PF)
representation with the R/L edge of a corresponding constituent
of type πΣ1 in phonological (PR) representation.
(3.16) Edge-alignment (English; adapted from Gussenhoven 2004: 167)
a. Align(XP, ip): Align the right edge of every XP with the right
edge of ip.
The syntax-prosody interface  105

b. Align(S, IP): Align the right edge of every S with the right edge
of IP.
(3.17) Interface constraint for the Intonational Phrase in English (Selkirk
2005: 19):
Align R (CommaP, IP)
Align the R edge of a constituent of type Comma Phrase in syntactic
(PF) representation with the R edge of a corresponding constituent
of type πCommaP (= Intonational Phrase, IP) in phonological (PR)
representation.
In addition to the alignment constraints, Truckenbrodt (1995, 1999, 2007)
suggests the wrapping constraint in (3.18), which “demands that each syn-
tactic XP be contained in a phonological phrase” (Truckenbrodt 1999: 219).
For Complementizer Phrases (CPs), Truckenbrodt’s (2005) Wrap-CP con-
straint demands that each CP is contained in a single IP (see (3.19)). It is
tied to an AlignCP constraint which demands for the right edge of a CP to
coincide with the right edge of an IP. This pair of constraints thus predicts
phrasing of CPs in one IP, but allows for additional material in the same
domain at its left periphery.
(3.18) Wrap-XP (Truckenbrodt 1999: 228)
Each XP is contained in a phonological phrase.
(3.19) Wrap-CP (Truckenbrodt 2005: 14)
Each CP is contained in a single intonation phrase.
Together, the syntax–prosody interface theory based on Align and Wrap
constraints may be referred to as the Align/Wrap theory (Selkirk 2005: 18,
among others). According to Truckenbrodt (1995, 1999) the Align/Wrap
theory is defined in terms of syntactic categories, not segments. For example,
the category XP in (3.20) has two segments: XP1 and XP2 (e.g. Chomsky
1995: 177). UP is adjoined to XP; it is dominated by only one segment of
XP, thus not by the category XP. ZP, X′, X and YP, on the other hand, are
dominated by category XP: they are dominated by all segments of XP. Edge-
alignment and Wrap-XP refers to the syntactic category (i.e. XP in (3.16)
a and (3.18)), not a segment of that category. The constraints thus apply to
material dominated by the lowest segment of a category. In (3.20), UP is not
dominated by all segments of the category XP but only by XP1. It will there-
fore not be seen by the syntax–prosody mapping constraints relating to XP.
Edge-alignment will align the edge of XP2 with the edge of a prosodic con-
stituent, thus phrasing the adjunct UP separately (see Truckenbrodt 1995:
Section 3.3 for details). Following any kind of adjunction analysis for paren-
theticals (see Chapter 2 above) the mapping constraints will result in separate
prosodic phrasing of parentheticals regardless of the specific category of XP.
106  Intonational phrasing and prosodic theory

(3.20) Category–segment distinction

(a) XP1 (b) XP1

UP XP2 XP2 UP

ZP X' X' ZP

X YP X YP

Taken together, the Align/Wrap theory predicts prosodic separation for


clausal parentheticals (i.e. the assumed default phrasing of a parenthetical in
its own IP; see (3.21)a below) as well as phrasing of the parenthetical clause
together with preceding host material (see (3.21)b and c), but it does not
predict phrasing of the parenthetical clause together with following host
material (see (3.21)d), and it does not predict an IP boundary inside the par-
enthetical (see (3.21)e).

(3.21) Predictions of the Align/Wrap theory for the prosodic phrasing of


parenthetical clauses (√ predicted, # not predicted)
a. √ IP[…]IP IP[parenthetical clause]IP IP[…]IP
b. √ IP[…]IP IP[… parenthetical clause]IP IP[…]IP
c. √ IP[… IP[(…) parenthetical clause]IP …]IP
d. # IP[…]IP IP[(…) parenthetical clause …]IP IP[…]IP
e. # IP[(…) part of parenthetical clause]IP IP[part of parenthetical
clause]IP IP[…]IP
Notice, incidentally, that the interface constraints as outlined here can only
account for a relation between syntactic parentheticals and their prosodic
phrasing if parentheticals are syntactically integrated, i.e. if they follow some
kind of adjunction, non-orphan approach (see also Dehé 2009: 576). De Vries
(2007: 220) arrives at the same conclusion. He argues that since parentheti-
cals “are interpreted as well as pronounced … they must be present at the LF
interface and at the PF interface” and that “[a]ccording to standard assump-
tions about the organisation of the grammar, there is only one way to get at
these interfaces, namely via the overt syntax”. He concludes that “parataxis
must be represented in syntax” (see also de Vries 2012a: 154). Only under the
assumption of syntactic integration are parentheticals subject to the interface
constraints of prosodic theory.
It has been critically remarked that within the edge-based theory, “phono-
logical phrasing is essentially determined by (surface) syntactic structure”
and that this view has been “widely accepted as self-evidently correct” (Lahiri
and Plank 2010: 372). However, establishing the prosodic hierarchy with
categories syllable, foot, prosodic word, phonological phrase, intonational
The syntax-prosody interface  107

phrase, and utterance, and while relating syntactic and prosodic structure,
Selkirk (1981) explicitly argues against isomorphism between prosodic and
syntactic structure. Moreover, Selkirk (1984: 285) states that “the syntactic
structure of a sentence cannot be said to determine its intonational phrasing”
(italics in original) and that “the relation between syntactic structure and
all aspects of intonational structure can be depicted as a one-to-many map-
ping”. This is again reinforced in Selkirk (2011: 437): “no inherent relation is
assumed to exist between the prosodic category types found in phonological
representations and the category types of syntactic representation”. In fact, a
lot of evidence has been provided against isomorphism between syntactic and
prosodic structure. Alignment and Wrap constraints may thus be violated
due to other factors also affecting prosodic phrasing, among them syntac-
tic and prosodic length, complexity and weight, balanced prosodic constitu-
ent size, performance factors such as speech rate and style, focus, contrastive
prominence, eurhythmy, and semantic coherence (see references given in
Chapter 2). We will see in Chapter 4 below that the prosodic phrasing of par-
entheticals in spontaneous and semi-spontaneous spoken corpus data serves
as further evidence for a one-to-many mapping of syntactic structure onto
prosodic constituents and for the violability of the interface constraints.
Most research in the tradition of the Align/Wrap theory has in fact
focused on the relation between lexical phrases (XPs) in syntax and inter-
mediate (or: phonological) phrases in prosodic structure (i.e. constraints
(3.16)a and (3.18)). What is particularly relevant in the present context,
however, is the relation between the IP or the TG and syntactic domains,
already touched upon above in the shape of the alignment constraints in
(3.17) and (3.19). In the British framework, a tendency has been observed
for a TG to correspond to a clause in syntax (e.g. Quirk et al. 1964). At the
same time, it has often been argued that this observation can indeed not be
more than a tendency and that TGs may instead correspond to a range of
structures and their extension affected by a range of factors (e.g. Quirk et al.
1964; Crystal 1969: 257–63). Crystal (1969: 259f), for example, observes a
particularly frequent correspondence between a TG and the following nine
elements of clause structure: adverbial, subject plus verbal complex, ver-
bal complex plus complement, subject, complement, verbal complex, nom-
inal group, vocative and postmodification within a nominal group. Crystal
(1969: 260)  therefore suggests correspondence between TGs in prosody
and “elements of structure” (his italics) in syntax. Given his nine elements
just listed, the “elements of structure” together comprised by a TG do
not necessarily form syntactic constituents. However, according to Crystal
(1969: 263) “breaks within elements of clause structure”, e.g. within a sub-
ject or a complement, are rare. It has also been observed that the tendency
for a clause to correspond to a TG is in conflict with other factors, among
them speech rate, the position of the nucleus and the overall length of the
TG in terms of number of syllables or words (e.g. Quirk et al. 1964; Crystal
108  Intonational phrasing and prosodic theory

1969: 256). For example, a relatively long NP subject of a clause is more


likely to be followed by a TG boundary, splitting the clause in two TGs, than
a relatively short subject (see Quirk et al. 1964).
As to the relation between the IP and syntactic constituents in prosodic
theory, in early work on intonational phrasing, Selkirk (1981: 131–4) assumes
that some syntactic constituents, such as parentheticals, non-restrictive rela-
tive clauses and preposed adverbials, must correspond to an IP, and that all
other IP structure is made up without direct reference to syntax in terms of
(a sequence of) phonological phrases. Selkirk (1984: 286)  has a Syntactic-
Prosodic Correspondence Rule for Intonational Phrasing, which demands that
a “matrix sentence must be exhaustively parsed into a sequence of (one or
more) intonational phrases”. Elsewhere, a root clause in syntax has often
been assumed to correspond to an IP in prosody (e.g. Downing 1970;
Emonds 1976, 1979; Nespor and Vogel 1986; (3.22) is from Truckenbrodt
2005: 275).

(3.22) Correspondence between root clause and IP


Root clauses (but not embedded clauses) are bounded by obligatory
intonation phrases boundaries.

In some of the same work, as well as elsewhere, the parentheticals inves-


tigated here have also been conceived as root clauses (e.g. Downing 1970;
Emonds 1979; Nespor and Vogel 1986; Fabb 1990 for NRRCs; Espinal 1991;
Safir 1986). Selkirk (2005: 14) maintains that the status of parentheticals “as
root sentences is disputable”. However, based on her discussion of Potts’
(2005) Comma Phrase and comma feature, Selkirk (2005: 17)  generalizes
that root sentences and parentheticals have in common that they are both
characterized by the feature [+comma], and that it is the Comma Phrase
rather than the root clause that corresponds to the IP. Semantically this
makes sense, she argues, because both root sentences and parentheticals
perform separate speech acts, and their content is independent of the at-
issue entailment, i.e. of the regular content of the host or any other root
clause; prosodically, they form separate IPs. Even recent work which goes
against a close correspondence between syntactic and prosodic constituency
acknowledges that “phonological phrasing … respects … clause boundaries
(reflecting major planning units)” (Lahiri and Plank 2010: 374).
A different approach was suggested by Selkirk (1984: 286), who argues
that the semantic well-formedness conditions on intonational phrasing
may be “more substantive” than syntactic–prosodic correspondences. This
is expressed in terms of the Sense Unit Condition (SUC) given in (3.23)
(see Selkirk 1984: 290–6 for definitions and illustrations of ‘immediate
constituent of IP’ and ‘sense unit’). This is reminiscent of Schubiger’s
(1958: 9) assumption (in the British framework) that “[e]ach sense-group
corresponds to a tone-group”. Selkirk’s sense unit is defined in terms of
The syntax-prosody interface  109

modifier-head and argument-head relations (see (3.25)). Note that a sense


unit as defined in (3.25)b includes syntactic non-constituents; see (3.26).
In (3.26)a, the subject is phrased together with the verb; in (3.26)b the verb
is phrased together with the first of two objects.
(3.23) Sense Unit Condition on Intonational Phrasing (Selkirk
1984: 286)
The immediate constituents of an intonational phrase must
together form a sense unit.
(3.24) Immediate Constituent of an IP (Selkirk 1984: 290)
An immediate constituent of an IPi is a syntactic constituent con-
tained entirely within (‘dominated’ exclusively by) IPi and not
dominated by any other syntactic constituent contained entirely
within IPi.
(3.25) Sense Unit (Selkirk 1984: 291)
Two constituents Ci, Cj form a sense unit if (a) or (b) is true of the
semantic interpretation of the sentence:
a. Ci modifies Cj (a head)
b. Ci is an argument of Cj (a head)
(3.26) IPs containing syntactic non-constituents (Selkirk 1984)
a. IP(Mary prefers)IP IP(corduroy)IP.
b. IP(Jane)IP IP(gave the book)IP IP(to Mary)IP.
Selkirk (1984: 290) argues explicitly against her own earlier account (Selkirk
1981; see above) and also against related accounts, including Downing
(1970), which had claimed that there are some syntactic configurations,
among them parentheticals, NRRCs and preposed adverbials, which require
phrasing in an IP. While the semantically driven explanation will still
account for separate phrasing of parentheticals, which are semantically sep-
arate from the host, it also accounts for variable phrasing.
In later work, Selkirk (2005: 42–7) emphasizes that the SUC, building on
semantic notions rather than syntactic representation, cannot in fact be part
of a grammar that allows for an interface between phonology and syntax but
not between phonology and semantics; i.e. the SUC cannot be a principle
of an Align/Wrap kind of prosodic theory. Selkirk (2005: 46)  claims that
“any and all desirable effects that have been ascribed to the SUC are instead
understandable as the consequence of constraints on the interface between
syntactic and phonological representations” such as put forward in Selkirk
(2005) and related work. For evidence against phrasing patterns such as
(3.26)a, she refers to Schafer et al. (2000), who, using a cooperative-game
language-production task to elicit spontaneous language data designed to
test prosodic disambiguation of early vs. late closure ambiguities as shown
in (3.27), find that speakers do not produce phrase breaks between a lexical
110  Intonational phrasing and prosodic theory

head and a complement (e.g. between a verb and its complement). (3.26)a
would thus be predicted not to exist or, if found, would have to be accounted
for along other kinds of constraints. Moreover, Selkirk (2005) argues that
the SUC cannot account for the asymmetry of the phrasing of CommaP,
whose left edge does often not coincide with an IP boundary: for the IP in
English, she argues, Align R (CommaP, IP) (see (3.17)) holds, but there is
no Align L (CommaP, IP). According to Selkirk (2005: 46), the SUC pre-
dicts that phrasing of a CommaP with material to its left should not exist,
because “no possible head-argument or modifier relation exists between”
the head (= anchor) of the NRRC (e.g. Romans in (3.28)) and the NRRC;
right alignment, on the other hand, makes the right predictions, because no
CommaP or XP-edge follows the head of the NRRC, thus no IP boundary is
predicted. In her earlier account, Selkirk (1984: 296) mentions that prosodic
separation of non-restrictive modifiers such as NRRCs cannot be accounted
for along the lines of the SUC, because they are “not included in the same
IP with the constituent they modify”. However, she speculates that “for one
constituent to form a sense unit with another, the two must form part of the
same assertion”, while NRRCs and related expressions “function as separate
assertions”. Parenthetical Comma Phrases, while being modifiers in the syn-
tax, have CI semantics, i.e. they are independent of the main assertion. With
the CI semantics for parenthetical clauses established, prosodic separation
will then not violate the SUC. However, both the phrasing in (3.28)b and
that in (3.28)c of the NRRC in (3.28)a can also be accounted for along the
lines of Selkirk’s (2005) version of edge-alignment: (3.28)b follows straight-
forwardly from Align R (CommaP, IP); (3.28)c is the result of the stylistic
promotion (Selkirk 2005: 49; see also Nespor and Vogel 1986) of the lower-
level prosodic boundary separating Romans and who to an IP boundary.
(3.27) Early (a) vs. late (b) closure ambiguity (Schafer et al. 2000: 171; my
square brackets)
a. When that moves [the square will encounter a cookie]
b. When that moves the square [it should land in a good spot]

(3.28) Phrasing of NRRCs (Selkirk 2005: 20)


a. The Romans, who arrived early, found a land of wooded hills.
b. (The Romans, who arrived early)IP (found a land of wooded
hills)IP
c. (The Romans)IP (who arrived early)IP (found a land of wooded
hills)IP
For the IP in English, Selkirk (2005) argues that root sentences as well as
supplementary Comma Phrases such as as-parentheticals, NRRCs and
N-APPs (Potts 2005), which have in common that they perform separate
speech acts of their own, are marked by the comma feature and that the IP
in the phonological representation is grounded in the Comma Phrase. Taken
The syntax-prosody interface  111

together with right alignment of the Comma Phrase with IP (cf. (3.17)), the
result is separate IP-phrasing of all right-peripheral adjuncts, and separ-
ate IP-phrasing of all comma-marked left-peripheral adjuncts (cf. (3.29)–
(3.30)). Assuming that parentheticals are syntactic adjuncts (Potts 2005; de
Vries 2012a, among others), right-peripheral parentheticals are then pre-
dicted to be phrased in a separate IP. Given that parentheticals are comma-
marked (Potts 2005), all parentheticals will be followed by an IP boundary.
However, parentheticals are predicted to phrase together with material to
their left unless preceded by the edge of another Comma Phrase (i.e. root
clause or another comma adjunct).
(3.29) Right-peripheral sentence adjuncts and their intonational
phrasing (Selkirk 2005: 22); IP-phrasing according to Align R
(CommaP, IP)
a1. b1.

Root Adjunct Root Adjunct


Comma Comma Comma

a2. (……)IP (……)IP b2. (……)IP (……)IP

(3.30) Left-peripheral sentence adjuncts and their intonational phras-


ing (Selkirk 2005: 22); IP-phrasing according to Align R
(CommaP, IP)
a1. b1.
Adjunct Root Adjunct Root
Comma Comma Comma

a2. (……)IP (……)IP b2. (……………………)IP

As Selkirk (2005: 19–21) argues, there is an asymmetry in the prosodic


phrasing of a sentence-medial Comma Phrase such that its right edge but
not its left edge coincides with an IP boundary in prosody. Selkirk maintains
that phonetic parameters interpreted as cues to IP boundaries, such as con-
tinuation rise and pause, are often absent at the left edge of a Comma Phrase.
She interprets IP boundaries preceding parentheticals as promoted from a
boundary at a lower level in the prosodic hierarchy. This is in line with an
earlier suggestion by Taglicht (1998: 196), who suggests that in intonational
phrasing, parentheticals “may group to the left … but not to the right”.
Any IP boundary to the left of parentheticals would have to be the result
of a promotion of an ip-level boundary as resulting from Align(XP, ip) (see
(3.16)a) to IP-level boundary, or of the right boundary of a left-adjacent
Comma Phrase, or else it would be hard to derive in a theory such as the one
112  Intonational phrasing and prosodic theory

put forward in Selkirk (2005). In particular, note that at least one case is not
accounted for: parentheticals which are positioned such that they follow syn-
tactic strings which are not constituents of type XP or CommaP, but which
are nevertheless preceded by an IP boundary, would be difficult to account
for along Selkirk’s (2005) lines. Yet corresponding examples are attested, see
the utterance in (3.31) with equally strong prosodic boundaries to the right
and to the left of the N-APP, and Peterson’s (1999: 239) example in (3.32).
As indicated by the orthographic commas put by Peterson, an IP boundary
manifested by a continuation rise and possible pause may be realized, set-
ting the parenthetical off from the preceding subject-verb sequence. An IP
boundary in this position is not predicted by Align R (CommaP, IP), nor is
an ip boundary – which could then be promoted to an IP boundary – pre-
dicted by Align(XP, ip), since there is no XP-edge preceding the parenthet-
ical. We will therefore have to either allow relevant constraints (and their
ranking) in the interface theory such that they result in a boundary to the
left of the parenthetical, or derive phrasing instances like (3.31) and (3.32)
in the phonological component of the grammar such that even lower-level
prosodic boundaries such as prosodic word boundaries may easily be pro-
moted to IP boundaries in contexts like these.

(3.31) N-APP following a non-constituent, but produced with IP-boundary


to its left
a. Graham Taylor has been saying the England manager that this is
perhaps the toughest test that he and his team have faced in his
seven internationals so far (ICE-GB: s2a-001 #2)
b. … has been saying)IP IP(the England manager)IP …

(3.32) N-APP following a non-constituent (Peterson 1999: 239)


John likes, it must be admitted, icecream with peanut butter
topping.
Yet more recently, Selkirk (2009, 2011) argues for a Match theory, which
demands that syntactic and prosodic constituency structure correspond
in the way given in (3.33). Edge-alignment and Match share the idea that
edges of syntactic constituents coincide with edges of prosodic constituents.
However, while Match requires both the left and right edge of a syntac-
tic constituent to coincide with a prosodic one, edge-alignment constraints
refer to either the left or the right edge. Match constraint (3.33)(i) refers
to the relation between the clause in syntactic structure and the IP in the
phonological representation. Like edge-alignment, Match theory “predicts
a strong tendency for phonological domains to mirror syntactic constitu-
ents” (Selkirk 2011: 439–40), but like alignment constraints, Match con-
straints are violable and violation results in non-isomorphic prosodic and
syntactic structures. Match theory therefore allows for mismatch between
The syntax-prosody interface  113

prosodic and syntactic constituents, including mismatch between the clause


in syntax and the IP in prosody. High-ranked phonological markedness con-
straints, Selkirk (2011) argues, may lead to violations of Match constraints
and produce instances of non-isomorphism between syntactic constituency
and phonological domain structure (Selkirk 2011: 440). Phonological mark-
edness constraints include size and eurhythmy constraints (see Selkirk 2011:
468–72 for review and discussion).
(3.33) Match theory (from Selkirk 2009: 40, 2011: 439)
(i) Match clause
A clause in syntactic constituent structure must be matched by a
corresponding prosodic constituent (i.e. the IP) in phonological
representation.
(ii) Match phrase
A phrase in syntactic constituent structure must be matched by a
corresponding prosodic constituent (i.e. the intermediate phrase)
in phonological representation.
(iii)  Match word
A word in syntactic constituent structure must be matched by a
corresponding prosodic constituent (i.e. the prosodic word) in
phonological representation.
In the context of Match clause (see (3.33)i), Selkirk (2009, 2011) discusses
two notions of clause: (i) the ‘standard clause’, i.e. the complement of the
functional head C° in syntactic theory, containing subject, predicate and
tense marking, and (ii) the ‘illocutionary clause’. The illocutionary clause
“is the highest syntactic projection of the sentence and carries its illocut-
ionary force, which determines its appropriateness in a discourse context”
(Selkirk 2011: 452). According to Selkirk, this is Emonds’ (1976) root clause,
Rizzi’s (1997) Force Phrase and Potts’ (2005) Comma phrase, thus includ-
ing parentheticals. Illocutionary clauses, Selkirk (2011: 452) mentions, “are
commonly observed to correspond to intonational phrases”, and there is a
stronger tendency for embedded illocutionary clauses to be phrased in a
separate IP than for embedded standard clauses (e.g. the difference between
restrictive and non-restrictive relative clauses). Accordingly, Selkirk (2009:
47–50) develops and distinguishes two Match clause constraints, a more
general one phrased here as in (3.34)(i) below, and a more specific one (see
(3.34)(ii), the latter outranking the former “perhaps universally”; Selkirk
2011: 453). Note that Match clause as given in (3.34)(ii) will also account for
the prosodic boundaries to the left and right of the parenthetical in (3.31)
and (3.32). It is also consistent with footnote 12 in Kratzer and Selkirk
(2007: 125), which suggests that there might be full correspondence between
CommaP and IP.
114  Intonational phrasing and prosodic theory

(3.34) Match clause (based on Selkirk 2009, 2011)


(i) Match (standard clause, IP)
A standard clause in syntactic constituent structure, specifically the
complement of C°, must be matched by a corresponding prosodic
constituent (i.e. the IP) in phonological representation.
(ii)  Match (illocutionary clause, IP)
An illocutionary clause in syntactic constituent structure, spe-
cifically the complement of Force°, must be matched by a cor-
responding prosodic constituent (i.e. the IP) in phonological
representation.
Taken together, Align/Wrap theory and Match theory make the following
predictions given in (3.35) and (3.36) with respect to the prosodic phrasing
of parentheticals/Comma Phrase. Prosodic separation, including phrasing
of the parenthetical in a compound prosodic domain (CPD), satisfies Align
R, Wrap and Match clause (see (3.35)a–b and immediately below (3.35)
for an important qualification); phrasing of the parenthetical with preced-
ing host material satisfies both Align R and Wrap, but violates Match (see
(3.35)c), because the parenthetical clause is not matched by an IP; phras-
ing of the parenthetical with host material to the right results in a violation
of both Align R and Match (see (3.35)d); and phrasing of the parenthet-
ical such that an IP boundary is inserted within the parenthetical constitu-
ent violates Wrap and Match (see (3.35)e). The order of the constraints in
(3.35) is not intended to represent any ranking between them. At least the
first four of these patterns (i.e. (3.35)a–d) have regularly been attested in
the literature on the prosody of parentheticals (see Chapter 2 above; in par-
ticular the hypotheses formulated for CCs and QTs). Given that all inter-
face constraints of the Align/Wrap/Match families are violable, this may be
accounted for by overriding constraints.
(3.35) Predictions of the Align/Wrap and Match constraints for the pros-
odic phrasing of parentheticals (par)/Comma Phrases
a. IP[…]IP IP[par]IP IP[…]IP √Align R, √Wrap, √Match
b. IP[… IP[par]IP …]IP √Align R, √Wrap, √Match
c. IP[…]IP IP[… par]IP IP[…]IP √Align R, √Wrap, *Match
d. IP[…]IP IP[(…) par …] IP IP[…]IP *Align R, √Wrap, *Match
e. # IP[… part of par]IP IP[part
of par]IP IP[…]IP √Align R, *Wrap, *Match
Note that with regard to the closing IP boundary preceding the paren-
thetical in (3.35)a, it depends on the constituency of the string preceding
the parenthetical whether that boundary violates Match or not. If the par-
enthetical is preceded by a clause, the IP boundary is predicted by Align,
The syntax-prosody interface  115

Wrap and Match (see (3.36)a). If the parenthetical is preceded by an XP,


an ip boundary is predicted by all three constraints, but no IP boundary
(see (3.36)b); if it is a word boundary, a prosodic word boundary is pre-
dicted. The IP boundary in this position would then have to be the result
of the promotion of lower-level boundaries to an IP boundary. The recur-
sive structure in (3.35)b, on the other hand, violates Match clause as well
as Align R if the parenthetical is preceded by a clause (see (3.36)c). In other
words, Match predicts a recursive prosodic structure for parentheticals in
clause-medial position (see (3.36)d), but not for parentheticals between two
clauses (see (3.36)c).
(3.36) Prosodic separation: predictions of the Align/Wrap and Match
theories for the prosodic separation of parentheticals (par); non-
recursive (a, b) and recursive structures (c, d)
a. IP[clause1]IP IP[par]IP IP[clause2]IP √Align R (S/IP), √Wrap-CP,
√Match clause
b. IP[XP]IP IP[par]IP IP[XP]IP *Match phrase9
c. IP[clause1 IP[par]IP clause2]IP *Align R (S/IP), *Wrap-CP,
*Match clause
d. IP[XP IP[par]IP XP]IP Match phrase (if XP
matched by ip)
Finally, prosodic separation as predicted based on previous research on par-
entheticals (see Table 2.1 at the end of Chapter 2), including phrasing in a
CPD structure, corresponds to (3.35)a–b. Variation in the prosodic phrasing
of parentheticals such as prosodic integration, as predicted for CCs, RVs
and QTs on the basis of previous research (see Table 2.1), corresponds to
(3.35)c and d. (3.35)e has not been regularly observed except for complex
parentheticals consisting of more than one clause (see (2.60) in Chapter 2).
The following chapters will report on the analysis of the prosodic phras-
ing of six types of parentheticals as retrieved from the ICE-GB and will
interpret the results against the descriptive and theoretical background
established in Chapters 2 and 3.

9
One thing to keep in mind, of course, is that a syntactic phrase followed by an IP boundary
(or phrased in an IP) is also followed by an ip boundary (or phrased in an ip). This fol-
lows from the prosodic hierarchy. Therefore Match phrase is strictly speaking not violated.
However, Match Theory predicts for XPs to be phrased in ips rather than IPs (which in
turn are reserved for clauses), thus the asterisk indicates violation of this idea.
4 Data analysis, results and discussion

In order to find out more about the intonational phrasing of parentheti-


cals in spontaneous and semi-spontaneous spoken language and to test the
predictions based on previous research as outlined in Chapters  2 and 3
above, a set of data from actual spoken British English retrieved from the
ICE-GB was investigated. This chapter provides details on the nature of
the data used here, their source, their treatment, and the results of the ana-
lysis. As anticipated in previous sections, a prosodic analysis of the follow-
ing six types of parenthetical elements as used in actual spoken language
was carried out: full parenthetical clauses (FPCs), non-restrictive relative
clauses (NRRCs), nominal appositions (N-APPs), comment clauses (CCs),
reporting verbs (RVs) and question tags (QTs). Approaches to their syntax
and aspects of their prosodic behaviour as previously observed were intro-
duced in Chapter 2. Some subsections of the present chapter are based on
Dehé (2009) and Dehé and Braun (2013).1 The data analysed in Dehé (2009)
were carefully analysed once again, with very little deviation from the earlier
results. The three data sets reported on in Dehé (2009), i.e. FPCs (sentential
parentheticals in Dehé 2009), NRRCs and CCs, have been extended such
that sound files which were unavailable in ICE-GB Version 1 and were there-
fore not analysed in Dehé (2009) have now been included. Furthermore, the
types of parentheticals analysed for Dehé (2009) have been complemented
by N-APPs, RVs, and QTs; the analysis of as it were was added to the set of
CCs. The results for QTs are also reported in Dehé and Braun (2013), but
the focus in that paper is a broader one, also including more details on the
intonational realization of QTs along with prosodic phrasing. The present
discussion in light of current prosodic theory is also new as compared to
Dehé (2009) and Dehé and Braun (2013).
Sections 4.1 and 4.2 of the present chapter provide information about the
source, retrieval and treatment of the data analysed here. Section 4.3 will be
concerned with the data analysis and report on the results for each of the
types under investigation. The results will briefly be discussed in light of

1
All material reproduced from Dehé (2009) and Dehé and Braun (2013) is reproduced with
permission; © Cambridge University Press.

116
Data source and data retrieval  117

the hypotheses at the end of each subsection. Section 4.4 will summarize
the results and compare them across types of parentheticals.

4.1  Data source and data retrieval


The data for this study were retrieved from the spoken part of the ICE-GB
(introduced in Chapter  1 above), using an automatic search followed by
manual sorting. The ICE-GB grammar contains a number of functions and
categories, among them Detached Function (DEFUNC), which includes
parenthetical clauses along with vocatives (Nelson, Wallis and Aarts 2002:
45). The automatic search was for this category. It yielded 4,969 hits (dis-
tributed across 4,319 corpus items). From this data set, the following sub-
sets were extracted for analysis: FPCs, NRRCs, N-APPs, and CCs. RVs and
QTs were retrieved separately (see Sections 4.1.5 and 4.1.6 below). All items
retrieved from the corpus entered an initial manual sorting procedure dur-
ing which items were generally excluded from the data set if they met one or
more of the following criteria: (i) the author did not agree with the corpus
annotators as to the syntactic status of a target element; (ii) the sound file
was unavailable; and (iii) the quality of the sound file was not good enough
for auditory analysis. Bad recording quality embraced various factors,
among them the following: incompleteness of the sound file, in particular
if parts of the parenthetical or of elements in its immediate vicinity were
missing due to the editing of the file as retrieved from the corpus (e.g. items
were trimmed at locations relevant in the present context); extremely quiet
recordings; overlapping speech, laughter, music or noise in the target area,
which made a conclusive analysis of the target impossible; and unclear words
in the critical region. In addition to these general criteria, there were type-
specific criteria, e.g. syntactic ambiguity between parenthetical and main-
clause analysis for CCs (see the discussion in Section 2.3.4 above), syntactic
mismatches between tag and host for QTs (see Section 4.1.6 below), among
others (further type-specific criteria will be given in the individual sections
below). Criteria that did not lead to immediate exclusion from the analysis
included features typical of natural spoken language such as restarts, hesita-
tions, silent and filled pauses, and sniffing, coughing or laughing anywhere
in the utterance, unless they made the analysis of the target element impos-
sible. For example, hesitations and non-structure-related silent and filled
pauses were frequently observed within and around FPCs and NRRCs, but
did not affect intonational phrasing; restarts frequently occur after complex
parenthetical insertions (see also the discussion of backtracking in Chapter 1
above); CCs may be used as part of hesitant phases (Dehé and Wichmann
2010a), thus exclusion of these items from the analysis would have been
counterintuitive. Table 4.1 sums up the numbers of items retrieved from the
original data set and the numbers of items that survived manual sorting to be
considered in the analysis for each type of parenthetical under investigation.
The following sections address the six types of parentheticals separately.
118  Data analysis, results and discussion

Table 4.1  Summary of items

FPC NRRC N-APP CC RV QT overall

No. of items 77 62 549 402 54 722


retrieved from (21/50/6)
initial data set

No. of items after 72 60 516 147 (plus 34 383


initial manual (19/47/6) 36 as it
sorting procedure were)

No. of items 72 60 466 147 32 383 1,160


remaining after
additional manual
sorting procedure
and finally analysed

Numbers in parentheses for FPC relate to complete independent sentences/and-


parentheticals/as-parentheticals in that order.

4.1.1  Full parenthetical clauses


Seventy-four corpus items containing 77 sentential parentheticals were con-
tained in the initial subset. More specifically, the 77 parentheticals were 21
complete independent sentences (16 with declarative syntax, 4 with interroga-
tive syntax, and 1 with imperative syntax), 50 and-parentheticals, and 6 as-par-
entheticals. Of these 77 parentheticals, 4 had to be discarded for the following
reasons: overlapping speech or noise in the target area (2 complete independent
sentences), erroneous corpus annotation (1 and-parenthetical) and unclear syn-
tax (1 and-parenthetical). One further and-parenthetical was discarded because
it was phrasal rather than clausal. The remaining 72 parentheticals contained in
69 utterances (19 complete independent sentences, 47 and-parentheticals, and
6 as-parentheticals) entered the final analysis. Only 2 of these 72 parentheticals,
both of them and-parentheticals, were from the private domain of the corpus
(ICE-GB: s1a, private direct conversations and telephone calls), perhaps indi-
cating a preference for complex insertions to occur in more formal registers of
spoken language. Examples of the analysed parenthetical clauses are given in
(4.1); (4.1)b contains two parenthetical clauses.
(4.1) Corpus items containing full parenthetical clauses
a. What I’ve done here I hope you don’t entirely disapprove is try
and limit the time taken on this item by putting it in writing
(ICE-GB: s1b-075 #180)
b. And as I’ve said on a number of occasions for example in the
uprating may I just finish this uh this sentence before the honourable
lady uh comes in again uh that does I think reflect and it’s some-
thing which in my view is right and proper that there is a growing
acceptance by employers of much greater responsibility to cover
Data source and data retrieval  119

short-term sickness among their employees (ICE-GB: s1b-


058 #47)
c. In that case summary it is made plain and you could see a copy
if you wish at paragraph fifty-four that what crown counsel were
saying was that the Defiant appeared to have been largely aban-
doned (ICE-GB: s1b-063 #269)
d. Uhm the title of the talk as you’ll see on the handout is Renewing
the Connection a phrase of Firth’s (ICE-GB: s2a-030 #4)

4.1.2  Non-restrictive relative clauses


Fifty-seven corpus items containing 62 NRRCs, 2 of which were coordi-
nated NRRCs counted here as one (see (4.4)), were initially considered.
Two of them were discarded because of unclear structure or unclear
words in the target area. The remaining 55 utterances containing 60
NRRCs entered the analysis, examples of which are given in (4.2) through
(4.4) below. Of these, 14 were from the private domain and 46 from the
public domain, 15 of them from scripted speech. Of the 60 NRRCs
which entered the analysis, 51 related to a nominal anchor (see examples
(4.2)a–c), and 9 related to the whole sentence or verb phrase (see (4.3)).
Example (4.3)a contains two NRRCs. All NRRCs in the data set followed
their anchor (see Lee-Goldman 2012 for a discussion of relative clauses
which precede their anchor). In (4.2)b, the NRRCs is followed by a free
sentential parenthetical (we checked it up earlier on), which relates to the
content of the NRRC.
(4.2) Corpus items containing NRRCs relating to a nominal anchor
a. And they the English formed up on Harraden Hill which is on
the the Berwickshire side of Berwick in in a star-shaped formation
(ICE-GB: s1a-065 #343)
b. Walker gets a hand on it but the ball in fact runs out to long-on
where Andy Roberts who is by the way the smallest first-class crick-
eter we checked it up earlier on does the fielding and that’s the
end of Worcestershire’s innings (ICE-GB: s2a-013 #137)
c. White House officials say Mr Bush who’s spending the weekend
at the presidential retreat at Camp David in Maryland was woken
by his National Security Adviser General Brent Schocroft and
briefed on the latest developments in Israel (ICE-GB: s2b-
015 #33)
(4.3) Corpus items containing NRRCs relating to VP or sentence
a. Now you see if Nell went somewhere like that which after all
wouldn’t be for a year and a half so you’re not forcing her off yet she
wouldn’t have to think all the time about you and Bernard and
Gavin which is so unhealthy … (ICE-GB: s1a-054 #130)
120  Data analysis, results and discussion

b. I think it is far better to increase the amount of democracy rather


than to go ahead and reduce it which I believe would be wrong at
this time (ICE-GB: s1b-053 #46)
(4.4) Corpus items containing coordinated NRRCs
a. So David Robertson got the recollections of one such child
Rachel Pearce who’s now seventeen and whose parents divorced when
she was eight (ICE-GB: s2b-019 #64)
b. Above all Paul whose job it was originally to expose the Christian
racket and who must have heard and investigated all stories about it
Paul was more convinced than anyone of its truth (ICE-GB: s2b-
028 #65)

4.1.3  Nominal appositions


Five hundred and twenty-six corpus items containing 549 appositions were
originally retrieved from the DEFUNC data set. Of the 526 utterances, only
19 were from the private domain of the spoken part of the corpus, indicating
a preference for N-APPs to occur in more formal registers of spoken lan-
guage.2 In a first manual sorting process and before listening to the sound
files, they were organized into the following categories (see also Meyer
1992): (i) the anchor in the host is a proper noun modified by a N-APP
(120 N-APPs; see examples in (4.5)); (ii) neither the N/NP anchor in the
host nor the N-APP is a proper noun (100 N-APPs; see examples in (4.6));
(iii) the N-APP but not its anchor is a proper N (286 N-APPs; see examples
in (4.7)); and (iv) a pronominal anchor in the host is modified by a N-APP
(19 N-APPs; see examples in (4.8)).
(4.5) Nominal anchor = proper noun, N-APP ≠ proper N
a. Simon doesn’t pay but Laura the student does (ICE-GB: s1a-
007 #231)
b. He’s played it back to Svaba the sweeper and he plays it again
square across his own eighteen yard line and eventually inevitably
it goes back to Michalichenko (ICE-GB: s2a-001 #54)
(4.6) Nominal anchor ≠ proper N, N-APP ≠ proper N
a. I’m Simon James from the education service and today’s lec-
ture the very last lecture before Christmas uh is The Ancient Celts
Through Caesar’s Eyes (ICE-GB: s2a-022 #2)

2
Within the public domain, 50 utterances were from spoken dialogues (ICE-GB: s1b), 126
from unscripted public monologues (ICE-GB: s2a) and the majority of 331 utterances from
scripted monologues held in public (ICE-GB: s2b).
Data source and data retrieval  121

b. You know that guy who died the rower and he was uhm very pally
with Jim the third year classicist and another girl a classicist
called Camilla a third year (ICE-GB: s1a-093 #188)
(4.7) Nominal anchor ≠ proper N, N-APP = proper N
a. Europe was the subject of a major speech by the Foreign
Secretary Mr Douglas Hurd this morning (ICE-GB: s2b-
007 #13)
b. In fact my mate at work Anne was in Scotland about a couple of
months ago and she was just talking about uhm with some friend
of hers about London and it was the first place they said they
went to (ICE-GB: s1a-098 #196)
(4.8) Pronominal anchor
a. and I think they were just sort of making up for for the embar-
rassment of having to ask me such an experienced artiste to come in
for those two lines (ICE-GB: s1a-092 #51)
b. They the owners were demanding payment of installments as they
fell due or became due up to the date of the award (ICE-GB: s2a-
065 #68)
Immediately excluded from the analysis were the following items: (i) all
utterances which contained sequences of a premodifier/title followed by a
proper noun, as explained in Chapter 2 above (13 N-APPs; see (4.9) below
and (2.69) in Chapter 2 above); (ii) all utterances whose syntax was unclear
or for which the author disagreed with the corpus annotators as to the cat-
egory of the constituent annotated nominal apposition, including cases of
weak apposition along the lines of Quirk et al. (1985: 1303) (11 N-APPs;
exemplified by the clausal appositions in (4.10)); and (iii) utterances in
which the constituent annotated N-APP was in fact a right-dislocated
noun phrase (9 N-APPs, 8 of which had a pronominal anchor; see (4.11)
for examples; the anchor is underlined). Unlike N-APPs, right-dislocated
constituents do not contribute additional information about the nominal
referent. Instead, they represent discourse-old and hearer-old information
(e.g. Ward and Birner 1996, 2001 and references given there). Therefore,
they are predicted to bear less prominence, which in turn may have an
effect on phrasing. Astruc-Aguilera (2005) finds a prosodic difference
between Catalan right-dislocated phrases and appositions such that the
former lack accentuation. Astruc-Aguilera and Nolan (2007b: 91) report
that in their English data, all of the final right-dislocated phrases were
deaccented.3

3
Astruc-Aguilera and Nolan (2007b: 91) also claim that the majority of final right-dislocated
phrases “form independent phrases”, meaning that they are temporally set off from the
preceding material or preceded by an edge tone.
122  Data analysis, results and discussion

(4.9) Premodifier/title (utterance ungrammatical without apposition);


discarded from analysis
a. That’s a question P H D student Sally Dawson has been trying
to answer (ICE-GB: s2b-038 #39)
b. Labour M P Tony Benn has studied the constitution throughout
his political life and is currently drafting his own Government
of Britain Bill in which he will be proposing to end the exist-
ing powers and influence of the monarchy (ICE-GB: s2b-
032 #73)
(4.10) Weak apposition; discarded from analysis
a. As for the intriguing question what does she do next Mrs
Thatcher has not apparently given it much thought (ICE-GB:
s2b-003 #107)
b. The most fundamental of all parental wishes to educate our chil-
dren in our own morality is indoctrination and a denial of their
free development (ICE-GB: s2b-029 #147)
(4.11) Right dislocation; discarded from analysis
a. What I meant was to uh get today some uhm abstract or brief
summary of the uh part that we would contribute Jan Aarts and
I to your uhm your project with LSE (ICE-GB: s1a-024 #1)
b. They’ve got a pet rabbit Laura and her boyfriend Simon
(ICE-GB: s1a-017 #128:1)
c. But it’s an official thing the wedding or or they’re just doing
(ICE-GB: s1a-071 #183)
The remaining 516 appositional constructions entered the final analysis. For
the purpose of the statistical analysis, they were further coded for syntactic
function, semantics and position (see Section 4.3.3 below).

4.1.4  Comment clauses


Three hundred and ninety-seven corpus items containing 402 CCs in med-
ial or final position were retrieved from the DEFUNC data set. Of these,
226 utterances (229 CCs) were discarded because they were syntactically
ambiguous between a parenthetical and a structure derived by movement
from an underlying main clause. Of the remaining 171 utterances (173
CCs), the sound files of 25 utterances (26 CCs) were either unavailable or
defective, or useless due to their bad recording quality.4 The remaining 146
utterances containing 147 CCs entered the analysis. Their types are listed
in (4.12); numbers in parentheses represent token numbers in the analysis.
Types not listed here but listed in (2.110)a in Chapter 2 above fell victim to
4
The number of files with bad recording quality is higher than for FPC, NRRCs and N-APPs
because more CCs than parentheticals of those types are from the private domain, for which
the recording quality is generally less good.
Data source and data retrieval  123

the sorting process. Corpus examples are given in (4.13) below; (4.13)a con-
tains one CC, (4.13)b contains two CCs.5
(4.12) CCs entering the analysis
I think (72), I suppose (18), I believe (12), I don’t think (10), I don’t
know (7), I would/should/’d say (5), I guess (4), I’m sure (4), I’m
afraid (3), I fear (2), I understand (2), I wonder (2), I reckon (1), I sus-
pect (1), I argue (1), I may say (1), I must say (1), I’m glad to say (1)
(4.13) Corpus items containing CCs
a. She is wearing a lime green suit carrying a dark blue I think
handbag white gloves and a pale hat with a rather pretty lime
green bobble in it (ICE-GB: s2a-019 #63)
b. and that was the very controversial Honours List that I think
came out finally on May the twenty-seventh i e some time after
uh six weeks or so after he’d resigned the Prime Ministership
but had been trailed I think in a number of newspapers before-
hand including a fairly full lead Joe Haines in the Sunday
Times (ICE-GB: s1b-040 #42)
In addition, following a manual search, 46 utterances containing as it were
and one utterance containing if you will were retrieved from the spoken part
of the ICE-GB. The instance of if you will turned out not to be a CC and
was thus discarded. All 46 sound files for as it were were available, but 10 had
to be excluded, mostly because of a combination of overlapping speech in
the target area and bad recording quality. The remaining 36 items entered
the analysis. An example is given in (4.14).
(4.14) Corpus item containing CC as it were
I know many people who are as it were not going to church very
much because they’re embarrassed by the fact that they cannot
actually say the Creed (ICE-GB: s1b-028 #110)

4.1.5  Reporting verbs
For reporting verbs, the automatic search for DEFUNC did not yield the
desired results. Therefore, a manual search was done for the 43 verbs given
in (4.15), which, according to Quirk et  al. (1985: 1024–5), may be used
as RVs with both direct and indirect reported speech, with say being the
most frequent. For each of these verbs, three forms were retrieved from the

5
Note that the first CC in example (4.13)b is located at the beginning of a relative clause.
However, unlike the examples in (2.93) in Section 2.3.4 above, which were discarded from
the analysis due to their potential syntactic ambiguity, this item was included in the ana-
lysis. Unlike wh-pronouns, that has been analysed as a complementizer rather than a relative
pronoun in the syntactic literature. See, for example, Radford (2004: 228.34) and references
given there for discussion. Movement of a relative pronoun across the CC can therefore not
be involved here.
124  Data analysis, results and discussion

corpus: uninflected form (e.g. say), third person singular present tense (e.g.
says) and past tense/participle form (e.g. said).
(4.15) Verbs included in the manual search for RVs
add, admit, announce, answer, argue, ask, assert, beg, boast, claim,
comment, conclude, confess, cry, declare, exclaim, explain, insist,
maintain, mumble, murmur, mutter, note, object, observe, order,
promise, protest, recall, remark, repeat, reply, report, say, snap,
snort, state, tell, urge, warn, whisper, wonder, write
From the data set retrieved in this way, all apparent RVs were extracted. For
example, the search for the verb form said yielded 1,054 hits (in 1,029 utter-
ances), but only 23 items seemed to qualify as RVs. The verb form admitted
yielded 15 hits (14 utterances), but none of them was a RV. Furthermore,
three items were discarded because of unclear syntactic structure or unclear
words in the target area. Overall, 54 items survived the initial manual sort-
ing process. These 54 items were inspected in their corpus contexts to verify
their status as RVs (see also the discussion of (4.22) in Section 4.1.7 below).
Thirty-four items survived this second sorting process and entered the ana-
lysis (see (4.16) for types and token numbers; see (4.17) for examples). They
were in medial (25 RVs) and final (9 RVs) positions in their host sentences.
Two of these items had undergone inversion such that the verb preceded
the subject (see (4.17)c and d). Five RVs were structurally complex such
that the subject was a full noun (all four cases given in (4.17)c–f) and/or
there was an adverbial modifier (see (4.17)f). Unlike for CCs, RVs preceded
by syntactic constituents were not excluded from the analysis.
(4.16) RVs entering the analysis
said (19), says (10), argue (1), ask (1), claim (1), confess (1),
promise (1)
(4.17) Corpus items containing RVs
a. Allied attacks on bridges were he said having a devastating effect
on Iraq’s ability to supply its forces (ICE-GB: s2b-005 #30)
b. Proper testing hasn’t been carried out she says and there could
be an increased risk of cancer (ICE-GB: s2b-023 #50)
c. You don’t understand the Japanese says Watanabe (ICE-GB:
s2b-033 #55)
d. He travels to Manchester for music lessons and his music
teacher thinks so highly of him says Julian that uhm he’s giv-
ing him a whole day instead of half a day on Sunday (ICE-GB:
s1a-032 #237)
e. This the spokesman said was not in line with the third Geneva
Convention which merely calls on combatants to perform their
humanitarian duty to release prisoners of war once hostilities
have ceased (ICE-GB: s2b-004 #24)
Data source and data retrieval  125

f. It wasn’t a round table President Gorbachev said later it was a


square table (ICE-GB: s2b-040 #92–93)

4.1.6  Question tags
Like the RVs, the question tags analysed here were not part of the out-
put of the DEFUNC search. Instead, the spoken part of the ICE-GB was
automatically searched for the function TAGQ (tag question; see Nelson
et  al. 2002), yielding 722 question tags in 718 utterances. Of these, 338
utterances had to be discarded in the manual sorting process. They were
either unavailable or the quality of their sound files was not good enough
for prosodic analysis. Furthermore, QTs were disregarded if they did not
correspond to the syntactic description given in Section 2.3.6, i.e. if the
pronominal subject or the auxiliary in the QT did not correspond to sub-
ject or verb in the host, respectively (e.g. (4.18) and (4.19)), if the tags
were of a different syntactic form (e.g. the interrogative in (4.20)) or if
they had been misannotated as QTs. The remaining 380 utterances con-
taining 383 QTs (202 containing be, 105 containing do, 31 containing have,
and 45 containing a modal verb) entered the analysis. Table 4–2 shows the
distribution of all 383 tags that entered the analysis according to auxiliary
verb (be, do, have, modal) and polarity (reversed polarity: positive clause/
negative tag (pos-neg) and negative clause/positive tag (neg-pos); con-
stant polarity: neg-neg and pos-pos). The table shows that in the data set
analysed here, reversed polarity tags with negation in the QT (pos-neg)
are most frequent across all verbs, followed by reversed polarity tags with
negation in the host. Furthermore, be is the verb most frequently found in
QTs, followed by the verb do, while QTs with have and modal verbs are
less frequent. This distribution is consistent with Tottie and Hoffmann’s
(2006: 290–1) findings for British and American English. Examples from
the present data set are given in (4.21); (4.21)g is one of the utterances
containing two QTs.
(4.18) Example QTs excluded from analysis because of lack of corres-
pondence between the subjects in the QT and the host
a. I bet there’s a lo load of old rubbish put on these forms don’t
you (ICE-GB: s1a-007 #251)
b. She’s planning a wild party aren’t you (ICE-GB: s1a-019 #351)
c. Brandon’s going to lend me loads of money aren’t you cos my
cash and money are stolen (ICE-GB: s1a-040 #368)
(4.19) Example QTs excluded from analysis because of lack of corres-
pondence between the verbs in the QT and the host
a. Oh that’ll be handy wouldn’t it (ICE-GB: s1A-019 #241)
b. She looks she looks Puerto Rican or something is it (ICE-GB:
s1a-058 #8)
126  Data analysis, results and discussion

Table 4.2  Distribution numbers of QTs according to


auxiliary verb in the tag (be, do, have, modal) and type
(reversed polarity tag: pos-neg, neg-pos; constant polarity
tag: pos-pos, neg-neg)

be do have modal overall

pos-neg 149 68 26 28 271


neg-pos 26 25 5 12 68
pos-pos 26 11 0 5 42
neg-neg 1 1 0 0 2
overall 202 105 31 45 383

c. Because he just switched again over to the search of Beowulf


and how uhm we could actually see uhm what he was doing at
the end with Wiglaf wasn’t it and why why he was then why was
he so keen (ICE-GB: s1a-090 #71)
(4.20) Example QT excluded from analysis because of syntactic form
And you suffer from mild asthma is that right (ICE-GB: s1a-
051 #106)
(4.21) Corpus items containing QTs
a. The state of the wall uh the state in which you found the wall
was the result of the work that you had directed to be done to
the wall wasn’t it (ICE-GB: s1b-069 #174)
b. You don’t like coffee particularly strong do you (ICE-GB: s1a-
046 #269)
c. You can hear him all right can’t you (ICE-GB: s1a-004 #24)
d. I mean in in that piece we’ve just heard from The Revenger’s
Tragedy it’s a mixture isn’t it of original instruments and kind
of what sound to me like modern trumpets (ICE-GB: s1b-
023 #140)
e. That’s a fantastic investment each year in that and in my belief
and one watches these things doesn’t one as one goes about uh it
pays off (ICE-GB: s2a-031 #80)
f. And uh we’ve learnt haven’t we that books are all right but there
is nothing like uh the help you get from society members as the
team have said and doing it yourself (ICE-GB: s1b-025 #166)
g. And you were satisfied with those weren’t you because you
already had the earlier accounts didn’t you (ICE-GB: s1b-
065 #127)

4.1.7  General notes on data retrieval


While working with the ICE-GB, and presumably with other annotated
corpora, it has to be kept in mind that the output of the corpus search may
Data source and data retrieval  127

sometimes be misleading. Across all data, it was therefore important to


inspect for each individual item the immediate environment in the corpus.
Two examples of why this was necessary are given in (4.22) and (4.23) below.
The first potential source of error is to do with corpus editing. It is illus-
trated by (4.22)a, which was part of the output of the search for say, i.e. the
manual search for RVs. Considered in isolation, it seems as if we say in (4.22)
a might be a sentence-final reporting verb. However, upon closer inspection
of the immediate context, it turns out that we say is directly followed by this is
American English here (see the fuller context given in (4.22)b). Simultaneous
inspection of the context and the sound file reveals that this is an example of
direct reported speech, where we say in fact immediately precedes the object
of say (see (4.22)c). The item was therefore excluded from the analysis.
(4.22) a. That’s the that is an area of difference isn’t it because we look
at a lexical item we say (ICE-GB: s1b-076 #060)
b. Speaker A: That’s the that is an area of difference isn’t it because
we look at a lexical item we say (ICE-GB: s1b-076 #060)
Speaker A: this is American English here (ICE-GB: s1b-
076 #061)
c. That’s the that is an area of difference isn’t it because we look
at a lexical item. We say: this is American English here.
A second possible source of error is the fact that the ICE-GB contains some
“ignored text”, which may be hidden or displayed on screen, depending on
the user’s choice. For example, Kaltenböck (2009a: 39–40, 2009b: 52–3)
discusses (4.23)a as an example of sentence-initial I think with zero that.
In this particular case, upon closer inspection including “ignored text”, the
full corpus item turns out to be as shown in (4.23)b: What’s gone wrong is
“ignored text” in <ICE-GB: s1b-022 #19>. The string it would be silly just to
sling mud around is in fact a parenthetical clause and I think is a hedging CC
related to What’s gone wrong. The string What’s gone wrong is then repeated
after the parenthetical insertion, a phenomenon referred to as backtracking
(see Chapter 1). The whole complex sentence uttered by Speaker B is given
in (4.23)c. The prosodic analysis confirms this interpretation (Wichmann
and Dehé 2011).
(4.23) a. I think it would be silly just to sling mud around (ICE-GB:
s1b-022 #019)
b. Speaker B: What’s gone wrong I think it would be silly just to
sling mud around
(ICE-GB: s1b-022 #019)
Speaker B: What’s gone wrong is a general breakdown of cen-
tral investment
(ICE-GB: s1b-022 #020)
c. What’s gone wrong I think – it would be silly just to sling mud
around – what’s gone wrong is a general breakdown of central
investment.
128  Data analysis, results and discussion

And finally, there is no claim to exhaustiveness for the extracted paren-


theticals. Different search and sorting methods may retrieve more or fewer
tokens. However, I believe that the analysed data set is representative.

4.2  Data treatment


All items that survived the initial sorting processes were analysed audito-
rily, and, if possible, instrumentally. Items whose sound files were not of
good enough quality for instrumental analysis were excluded only if the
auditory analysis was inconclusive. In the auditory analysis, the relevant
locations and types of pitch accents, and intonation domain boundaries
and boundary tones were identified, and the overall prosodic structure of
the utterance was described. For the purpose of the instrumental analysis,
the sound files retrieved from the corpus materials were edited into indi-
vidual files containing the parenthetical, its host, and as much preceding
and following material as necessary to determine the relevant boundaries.
The instrumental analysis was done in Praat (Boersma 2001; Boersma and
Weenink 2012). All sound files were annotated on a segmental and a tonal
tier, identifying the locations and types of pitch accents and intonation
domain boundaries in the parenthetical, the syntactic host of the paren-
thetical and, if necessary, in other parts of the immediate environment (e.g.
in the domain spanning linguistic material following a sentence-final par-
enthetical). This was done following ToBI ( Silverman et al. 1992; Beckman
and Elam 1993). The criteria for intonation domain boundaries followed
those discussed in the literature (see Chapter 3 above). They are summa-
rized in (4.24) and addressed in turn below. If domain boundaries could not
be identified immediately before and/or after the parenthetical, their real
position was identified and the occurring pattern was described. Prosodic
separation refers to phrasing such that the parenthetical has nuclear prom-
inence and is immediately preceded and followed by an IP boundary and/
or ip boundary (see Section 3.2 above for a definition of prosodic separ-
ation and reasons for combining these levels). Prosodic integration refers
to phrasing of the parenthetical with preceding and/or following host
material. According to this definition, prosodically separate parentheticals
always have main prominence, while prosodically integrated parentheti-
cals may have nuclear or non-nuclear prominence or may be unstressed. It
turns out that FPCs, NRRCs and N-APPs are always prominent regardless
of their phrasing, while CCs, RVs and QTs may be unstressed (see Section
4.3 below).
(4.24) Criteria for the identification of an IP/intonation domain
a. complete tonal contour (CTC) (nuclear pitch accent and
edge tones)
b. domain across which declination applies
Data treatment  129

c. creaky voice
d. presence and nature of pauses (structure-related vs.
hesitational)
e. pitch on unaccented syllables following a nuclear tone
f. domain-final lengthening
g. presence or absence of segmental processes
The criteria in (4.24)a–c are tonal and laryngeal features which apply in the
target domain including its periphery. As outlined in Chapter  3, the IP/
intonation domain is defined intonationally in terms of a complete tonal
contour (CTC), i.e. a nuclear tone according to the British tradition (e.g.
fall, fall-rise, rise, rise-fall), or a combination of nuclear pitch accent and
edge tone in the Autosegmental-Metrical model (i.e. ToBI). The so-called
continuation rise (e.g. Selkirk 1984: 288, 2005: 12–13) is one clear indication
of an IP boundary within a sentence or utterance. Another phonetic mani-
festation of a right IP edge is L%, i.e. a fall to the base-line of the speaker’s
pitch range, e.g. following a H* pitch accent. The IP/intonation domain is
also the domain across which declination applies, such that later pitch peaks
within an IP are realized with a lower pitch compared to earlier ones, and
pitch reset indicates a new domain (see Ladd 1986, 1996/2008). This has
to be seen in context: parentheticals have often been argued to be realized
at lower pitch (register); moreover, and quite generally, pitch peaks in later
domains may not be reset to the same maximum level as pitch peaks in earl-
ier domains (see Chapter 3 above). Furthermore, the use of creaky voice has
been considered a boundary marker at the end of an IP (e.g. Lehiste 1973;
Kreiman 1982).
The criteria listed in (4.24)d–g apply at boundaries between intonation
domains. First, domains may be separated by pauses (e.g. Nespor and
Vogel 1986: 188). However, pauses are not obligatory and their occurrence,
length and position depend on a number of factors (e.g. Stenström 1990;
Krivokapić 2007); therefore, they constitute neither a sufficient nor a neces-
sary condition for an IP boundary. On the other hand, Swerts (1997), for
example, finds a significant correlation between boundary strength and
pause durations for spontaneous Dutch speech. In the present context, all
pauses were identified and carefully studied as to their nature. In particu-
lar, structure-related breath pauses, but not pauses due to performance fac-
tors such as hesitational stretches, were interpreted as boundary markers.
Second, the pitch on unaccented syllables following a nuclear tone was stud-
ied. Pitch discontinuities and a step-up in pitch on an unaccented syllable
after a nuclear fall, and a step-down in pitch on an unaccented syllable fol-
lowing a rise, were seen as evidence for a boundary (Cruttenden 1997: 34).
Pitch continuation, i.e. a continuous rise or fall or no change in pitch level/
direction in a target position, was taken as evidence against a boundary.
Third, final lengthening, in particular lengthening of the very last syllable
130  Data analysis, results and discussion

of a domain, lengthening of the last stressed syllable, and lengthening of


the entire last word in a phrase, was interpreted as a signal for a boundary
(Lehiste 1973; Vaissière 1983; Gussenhoven and Rietveld 1992; Wightman
et al. 1992; Ferreira 1993; Turk and Shattuck-Hufnagel 2007, among many
others). However, final lengthening may only be an impressionistic fac-
tor. A systematic analysis was impossible due to the (uncontrolled) nature
of the data.6 Finally, the presence or absence of segmental processes was
taken into account, but not analysed systematically, since due to the nature
of the corpus data, it was impossible to control for environments in which
either the blocking or the application of certain segmental processes would
be predicted. Similar problems were faced by Knowles (1991) in his work
on the Lancaster/IBM Spoken English Corpus. He notes that “segmental
discontinuities … can only occur in a subset of cases anyway, e.g. assimila-
tion and elision require an alveolar consonant before the boundary, and a
linking /r/ can be replaced by a glottal stop only if the conditions for /r/ or
glottal stop exist in the first place” (Knowles 1991: 155). Furthermore, the
absence of segmental processes can be taken for granted at the site of a long
pause: “it would be surprising to find assimilation over a pause of more than
a second!” (Knowles 1991: 158).
Naturally, not all criteria were always present. One reason for this is the
well-known fact of inter- and intra-speaker variation in the phonetic encod-
ing of prosodic features such as pitch accents and the marking of boundaries
(see Cole et al. 2010: 1143 for references). There is also, of course, listener
(and transcriber) variability in the perception/transcription of prosody.
As noted in Chapter 1, all items in the present study were analysed by the
author. However, all cases that caused uncertainty were given to at least
one other experienced linguist trained in intonational analysis.7 Items for
which a final analysis could not be agreed upon or which were indistinct
were coded ‘unclear’ and were not analysed further. The analytic procedure

6
See, for example, Yoon et al. (2007) for the analysis of normalized segment duration in data
taken from the Boston University Radio Speech corpus. However, that corpus consists of
read paragraphs (approximately 10,000 words) from only five speakers. The spoken ICE-GB
data is much more varied, containing data from various text types (including read and spon-
taneous speech) and register types from 1,193 speakers overall. Yoon et al. (2007) restrict
their analysis to domain-final syllables and take the position of word stress into account as
an additional factor. Mo (2008) and Cole et al. (2010) study z-normalized segmental dura-
tions as a cue to prosodic boundaries in the Buckeye Corpus of spontaneous, conversational
speech (Pitt et al. 2007), using data from 18 speakers. However, this corpus is controlled for
speaker variety (all speakers are from Columbus, OH) and text type (interviews about every-
day topics such as politics, sports, traffic, schools). In the present context, using data from
the ICE-GB, in order to systematically take pre-boundary lengthening into consideration as
a cue to prosodic boundaries, all factors contained in the data would have to be taken into
account, including text and register type, speaker, word stress and syllable position, among
several others. While this might be a desirable topic for future research, it was impossible as
part of the present study.
7
Special thanks to Anne Wichmann and Jill House for their intonational analysis of numer-
ous sound files.
Data treatment  131

is exemplified in (4.25)/Figure 4.1 (corpus example in (4.25)a, intonational


phrasing in (4.25)b). The example contains a separately phrased and-par-
enthetical, as well as a prosodically integrated (IP-final) CC (it seems to me).8
(4.25)  a. One crucial difference it seems to me and this is where the parallel
between Yiddish and English breaks down is that whereas English
speakers have been overwhelmingly monolingual since the dis-
appearance of Anglo-Norman at least Yiddish speakers like
Jews everywhere since the Babylonian exile have been predom-
inantly bilingual if not polylingual (ICE-GB: s2b-042#32)
b.  IP1[One crucial difference it seems to me]IP1 IP2[and this is where
the parallel between Yiddish and English breaks down]IP2 IP3[is
that whereas English speakers have been overwhelmingly
monolingual]IP3 …
The auditory analysis suggested a series of IPs as indicated in (4.25)b and
panel a of Figure 4.1, with the and-parenthetical phrased separately and the
CC it seems to me integrated in IP1. This analysis was based predominantly
on perceived pauses before and after the and-parenthetical, and CTCs with
H% terminating IP1 and IP2. The instrumental analysis confirmed these
findings:
1. Cf. (4.24)a: All three IPs are the domain of a CTC of their own. IP1 has
prenuclear pitch accents associated with one and crucial and a nuclear
pitch peak associated with the first syllable of difference, followed by a
falling-rising pitch contour; IP1 is terminated by L-H%, i.e. it ends
in a continuation rise. IP2, spanning the parenthetical, has a series of
prenuclear accents associated with this, parallel and Yiddish, followed
by a nuclear pitch peak (H*) associated with down and terminated by
L-H% (continuation rise). IP3 has prenuclear pitch accents followed
by a nuclear L+!H* associated with the stressed syllable of monolingual,
followed by a falling pitch contour terminating in L-L%.
Furthermore, within IP1, the CC it seems to me is not prominent.
The falling pitch movement from !H* associated with the first syllable
of difference continues on it seems, then the pitch rises towards H%, the
rise being associated with to me.
2. Cf. (4.24)b: While pitch peaks in the parenthetical do not reach the high
level of peaks in the preceding and following domains, each IP is a domain
across which declination applies (except that within IP2 the final, nuclear
pitch peak associated with down is highest due to focus). We observe pitch
reset in the domain following the parenthetical (see the local peak on
where in whereas as compared to pitch peaks in IP1 and IP2).
8
This CC was discarded from the analysis due to its syntactic position following a subject,
resulting in structural ambiguity (see Section  2.3.4 above). However, I will use it here to
illustrate the analytical procedure.
(a)

350

300
Pitch (Hz)

200

100

50
L+H* L+H* !H* L-H% H* !H* L+!H* H* L-H% L+H* !H* L+H* L+!H* L-L%

CC FPC (and-parenthetical)

one crucial difference it seems to me and this is where the parallel between Yiddish and English breaks down is that whereas English speakers have been overwhelmingly monolingual

0 9.37
Time (s)

Figure 4.1  Example (4.25): example analytic procedure. Panel a: three adjacent IPs as given in (4.25)b; panel b: zooming in on
the and-parenthetical, the preceding domain and the IP boundary between the two; panel c: zooming in on the and-parenthetical,
the following IP and the IP boundary between the two
(b)

350
300
Pitch (Hz)

200

100
50
L+H* L+H* !H* L-H% H* !H* L+!H* H* L-H%

CC FPC (and-parenthetical)

one crucial difference it seems to me <SIL> and this is where the parallel between Yiddish ‘n’ English breaks down <SIL>

0 5.667
Time (s)

Figure 4.1  (cont.)


(c)

350

300
Pitch (Hz)

200

100
50
H* !H* L+!H* H* L-H% L+H* !H* L+H* L+!H* L-L%

FPC (and-parenthetical)

and this is where the parallel between Yiddish ‘n’ English breaks down <SIL> is that whereas English speakers have been overwhelmingly monolingual

0 7.537
Time (s)

Figure 4.1  (cont.)


Data treatment  135

3. Cf. (4.24)d: There are structure-related pauses of approximately 300 ms


before and 500 ms after the parenthetical, coinciding with the beginning
and end of the and-parenthetical and with the end of the CC. There is
no pause at the left edge of the CC.
4. Cf. (4.24)e: There are pitch discontinuities such that there is (i) a step-
down in pitch on unstressed and as the first syllable of IP2 following
the H% terminating IP1 (see Figure 4.1b), and (ii) a slight step-down
in pitch associated with the first syllable of IP3 (i.e. unstressed is) fol-
lowing the H% terminating IP2 (see Figure 4.1c). There is continuous
pitch movement between difference and it, not separating the CC from
the preceding material.
5. Cf. (4.24)f: The (prominent) final syllable of the IP spanning the and-
parenthetical (down) is impressionistically lengthened, accommodating
three tonal targets (H* L-H%).
6. Cf. (4.24)g: The presence of silent pauses blocks the application of
any segmental processes between segments on either side of the IP
boundaries.

Overall, the analysis suggests prosodic separation of the and-parenthetical


(and this is where the parallel between Yiddish and English breaks down) and
prosodic integration of the CC it seems to me (see (4.25)b).
One more word is in order regarding prosodic separation. As explained
above, parentheticals are prosodically separate if they have nuclear promin-
ence and are preceded and followed by a boundary. Given this definition, par-
entheticals in both recursive prosodic structures and non-recursive prosodic
structures may count as separate. However, the interface constraints make clear
predictions about when to expect a recursive and when a non-recursive struc-
ture (see end of Chapter  3). Phonetic parameters reflecting the distinction
include F0 declination across the utterance and pitch reset at domain bound-
aries. However, while these parameters were taken into account as phonetic
parameters cueing IP boundaries, and were interpreted whenever possible, the
nature of the corpus data did not always allow for a clear distinction between
non-recursive structures as in (3.36)a–b, and recursive structures as in (3.36)
c–d. It is possible, therefore, that overall more recursive structures occur than
are reported below. The F0 contour in Figure 4.1 represents a clear example
of post-parenthetical pitch reset to the level of the beginning of IP1, thus a
non-recursive prosodic structure, corresponding to (3.36)b.
Finally, note that within parenthetical types a statistical analysis was
appropriate for N-APPs and QTs only. This was due to there being too
few data points in all other groups. For the purpose of the statistical ana-
lysis, N-APPs and QTs were coded not only for prosodic phrasing, but also
according to the factors tested in the statistical analysis (see Sections 4.3.3
and 4.3.6 below). Ultimately, a statistical analysis including all six types
would be desirable, including parenthetical type as one factor. However, this
136  Data analysis, results and discussion

is impossible on the basis of the present data, not only because there are too
few data points but also because there is too much variation in the variables
and factors relevant for each type of parenthetical. Moreover, the descrip-
tive results clearly show the differences between the types, without submit-
ting them to a statistical analysis.

4.3  Analysis and results


This section reports on the results of the analysis of the six parenthetical
types in turn. The results are reported immediately below. For long and syn-
tactically complex parentheticals including coordinated parentheticals (see,
for example, the coordinated parentheticals in (4.26)), it was sometimes found
that they spanned more than one IP. The parenthetical-­internal IP boundaries
occurred in positions predicted by prosodic theory, e.g. clause boundaries such
as between employers and but in (4.26)a, and between racket and and in (4.26)b.
If the syntactic left and right edges of these parentheticals coincided with an
IP boundary, they were simply counted as prosodically separate. No separate
phrasing category was assumed for these examples. See also the discussion of
(4.49)/(4.52) in Section 4.3.2.2 below (for more examples of NRRCs span-
ning more than one IP see, for example, Auran and Loock 2011).
(4.26) Parentheticals spanning more than one IP
a. Moreover the development of software packages for those
employers with computer pay-rolls and I accept that this is not
true of all small employers but it’s true of a growing number have
further eased any operational difficulties that may originally
have been experienced (ICE-GB: s1b-058 #34)
b. Above all Paul whose job it was originally to expose the Christian
racket and who must have heard and investigated all stories about
it Paul was more convinced than anyone of its truth (ICE-GB:
s2b-028 #65)
Unless otherwise stated, the prosodic phrasing indicated in the examples is
the attested, not the predicted one.
At the beginning of each of the following sections reporting on one type
of parenthetical, the hypotheses developed with respect to their prosodic
phrasing on the basis of literature reviewed in Chapter  2 are repeated.
Examples (4.27) and (4.28) below repeat the hypotheses based on current
prosodic theory developed in Chapter  3 (see (3.35) and (3.36) and recall
their discussion at the end of Chapter 3).
(4.27) Predictions of the Align/Wrap and Match theories for the prosodic
phrasing of parentheticals (par)/Comma Phrases
a. IP[…]IP IP[par]IP IP[…]IP √Align R, √Wrap, √Match
b. IP[… IP[par]IP…]IP √Align R, √Wrap, √Match
Analysis and results  137

c. IP[…]IP IP[… par]IP IP[…]IP √Align R, √Wrap, *Match


d. IP[…]IP IP[(…) par …]IP IP[…]IP *Align R, √Wrap, *Match
e. # IP[… part of par] IP IP[part
of par]IP IP[…]IP √Align R, *Wrap, *Match
(4.28) Prosodic separation: predictions of the Align/Wrap and Match
theories for the prosodic separation of parentheticals (par); non-
recursive (a–b) and recursive structures (c–d)
a. IP[clause1]IP IP[par]IP IP[clause2]IP √Align R (S/IP), √Wrap-CP,
√Match clause
b. IP[XP]IP IP[par]IP IP[XP]IP *Match phrase
c. IP[clause1 IP[par]IP clause2]IP   *Align R (S/IP), *Wrap-CP,
  *Match clause
d. IP[XP IP[par]IP XP]IP         √Match phrase (if XP
matched by ip)

4.3.1  Full parenthetical clauses


Of the 72 sentential parentheticals that entered the acoustic analysis, five
underwent an auditory analysis only, because an instrumental analysis was
impossible due to the quality of the sound file. However, since the phrasing
of these parentheticals was clear, they were not discarded from the ana-
lysis. Remember that for full parenthetical clauses, the prediction is that
they are prosodically phrased separately by default (see the hypothesis and
predicted phrasing in (4.29) and (4.30), respectively, repeated from (2.47)
and (2.48) in Section 2.3.1 above). The phrasing patterns found for the 72
sentential parentheticals analysed here are given in (4.31) and detailed in
Sections 4.3.1.1 and 4.3.1.2, followed by a discussion in Section 4.3.1.3.
(4.29) Prosodic phrasing of full parenthetical clauses: hypothesis
The default prosodic phrasing of full parenthetical clauses is sep-
aration. Parenthetical clauses are phrased in their own intonational
domain, preceded and followed by the respective boundaries.
(4.30) Full parenthetical clauses (FPC): predicted prosodic phrasing
prosodic separation: (…) IP[…]IP IP[FPC]IP (IP[…]IP) (…)
(4.31) Observed patterns of prosodic phrasing of full parenthetical
clauses
a. FPC: prosodic separation
a1. … IP[…]IP IP[FPC]IP IP[…]IP …  67 FPCs  (93%)
a2. … IP[… IP[FPC]IP …]IP … 4 FPCs   (6%)
b. FPC: prosodic integration; FPC phrased together with pre-
ceding host material
… IP[… FPC]IP IP[…]IP … 1 FPC (1%)
Total: 72 FPCs
138  Data analysis, results and discussion

4.3.1.1  Pattern (4.31)a: prosodic separation


Overall, 67 full parenthetical clauses (93%; 17 out of 19 full parenthetical
clauses, 45 out of 47 and-parentheticals, and 5 out of 6 as-parentheticals)
were phrased separately in a non-recursive structure (see (4.30)/(4.31)a1).
Typical cues to IP boundaries included a CTC for each IP, boundary tones
terminating IPs, structure-related pauses to the left and right of the par-
entheticals, creaky voice at the end of an IP, and changes in pitch level
and/or direction associated with unstressed syllables at the beginning of
an IP, i.e. specifically associated with and and as for and-parentheticals
and as-parentheticals, respectively, the unstressed first syllable of an inde-
pendent parenthetical clause, and the unstressed first syllable of the fol-
lowing IP. One example of a separately phrased and-parenthetical was
explained in Section  4.2 above (see the discussion of (4.25)/Figure  4.1).
Further examples of the three types of parenthetical clauses are provided
in (4.32)/Figure 4.2, (4.33) and (4.34) below.
In (4.32)/Figure 4.2 the IP spanning the parenthetical (IP2) is preceded
by a continuation rise (L-H%) terminating IP1. The parenthetical is real-
ized at lower overall pitch than the preceding and following IPs. It has a
prenuclear accent on just and a nuclear accent associated with off, which is
also (impressionistically) lengthened. IP2 is followed by a silent pause of
almost one second, which the speaker presumably uses to turn the slide pro-
jector off, before the beginning of IP3. The fact that IP2 is realized with
pitch compression can be accounted for in terms of discourse function. It
is a detached parenthetical, a side remark having to do with the situational
context (the slide projector is still running), but not making any semantic
contribution to or comment on the host utterance. Note that there is a pitch
fall on off to L-L%, which is masked here due to background noise.
(4.32) a. So what we can do in fact I’ll just turn it off is to use that sig-
nal to train people’s ability to perceive voicing distinctions in
speech and their ability to actually produce them (ICE-GB:
s2a-056 #87)
b. IP1[So  what we can  do    in fact]IP1  IP2[I’ll  just turn it  off  ]IP2
H* L+H*    L-H% H* !H* L-L%
The and-parenthetical in (4.33) is preceded by a continuation rise L-H%
terminating IP2 and a silent pause of approximately 450 ms, then a step-
down in pitch on and. IP3, the IP spanning the parenthetical, has its own
CTC with the nuclear accent associated with wish and terminating in H%;
it is followed by a silent pause of > 400 ms, then a step-down in pitch on at.
Similarly, the as-parenthetical in (4.34) is set off from the preceding IP by
(impressionistic) lengthening of nuclear talk and a continuation rise L-H%
terminating IP1, as well as a break of approximately 150 ms between IP1 and
IP2. IP2, the IP spanning the parenthetical, has a nuclear accent associated
350
300
Pitch (Hz)

200

80
H* L+H* L-H% H* !H* L-L% L+H*

FPC

So what we can do in fact I just turn it off 0.982 is to use that signal

0 4.774
Time (s)

Figure 4.2  Example (4.32): full parenthetical clause, prosodic separation


140  Data analysis, results and discussion

with see and subsequent falling-rising pitch movement terminating in H%.


(The handout was mentioned immediately before: “There’s a handout uh at
the top; You may have them”; ICE-GB: s2a-030 #2–3, thus it is deaccented).
IP2 is followed by a pause of just over two seconds before the beginning
of IP3.
(4.33)  a. In that case summary it is made plain and you could see a copy if
you wish at paragraph fifty-four that what crown counsel were
saying was that the Defiant appeared to have been largely aban-
doned (ICE-GB: s1b-063 #269)
b. IP1[in that case   summary]IP1 IP2[it is  made plain]IP2
H*  L-H%      H*  !H* L-H%
IP3 [and you could see a copy if you wish]IP3
       L+H* L+!H* H-H%
IP4[at paragraph fifty-four] …
L+!H* L+H* H-H%
(4.34) a. Uhm the title of the talk as you’ll see on the handout is Renewing
the Connection a phrase of Firth’s (ICE-GB: s2a-030 #004)
b. IP1[Uhm the title of the talk]IP1
L+H*    !H* L-H%
IP2[as you’ll see on the handout]IP2  IP3[is …
H*      L-H%
These three examples illustrate the default prosodic phrasing of full paren-
thetical clauses: prosodic separation. All 67 prosodically separate full par-
enthetical clauses are phrased in their own IPs (not ips). Note that in this
phrasing an XP (e.g. subject XP) may well be matched by an IP; e.g. (4.34)
is phrased according to (4.28)b. Of the remaining five parentheticals, four
(two full parenthetical clauses and two and-parentheticals) were also phrased
separately, but in a CPD structure (Ladd 1986, 1996/2008; see (4.31)a2);
they are given in (4.35) through to (4.38)). The parenthetical in (4.35) is an
example of a semantically detached parenthetical, and of an interrogative
interpolation in a declarative host; it is related to its host only via the situ-
ational context. The speaker pauses during the production of the utterance
to directly address the audience and make sure he is speaking loud enough,
using a parenthetical. This parenthetical, however, does not affect the inton-
ational pattern of the host, reflecting that parenthetical and host are inde-
pendently planned units (see also Wichmann 2001). The parentheticals in
(4.36) and (4.37) are free/floating parentheticals, related to their hosts via
the reference of the pronouns it and this, respectively. The parenthetical how
shall I put it in (4.36) is a reflection on the part of the speaker on how to find
the best wording for what he wants to convey; the interpolation and Jimmy’s
put this in in (4.37) is a meta-comment on what is being expressed in the
host clause.
Analysis and results  141

(4.35) a. Well esterases are able in organic solvents to carry out a number
of useful can you hear me all right now organic uhm processes to
produce things like food products (ICE-GB: s2a-034 #13)
b. IP1[to carry out a number of useful IP2[can you hear me all
right now]IP2 organic processes to produce things like food
products]IP1
(4.36) a. That’s a little bit of uh how shall I put it uh uhm uh arrogance
that has still got to be eliminated uh in my life (ICE-GB: s1b-
041 #13)
b. IP1[That’s a  little bit of IP2[how shall I put it]IP2
L+H* H* !H* L-L%
arrogance]IP1      IP3[that …
L+!H*  L-L%
(4.37) a. In all across the world we employ directly some six hundred
and fifty people in our various activities including and Jimmy’s
put this in fifty chartered surveyors (ICE-GB: s2a-045 #62)
b. … IP1[including IP2[and Jimmy’s put this in]IP2 fifty chartered
surveyors]IP1
(4.38) a. He was and I think you would agree with me at the outset looking
at expanding his business at that time not selling it (ICE-GB:
s1b-064 #172)
b. IP1[He was IP2[and I think you would agree with me at the outset]IP2
looking at expanding his business]IP1 […]
These parenthetical clauses phrased in a CPD structure all have their own
CTC, while another contour spreads across the surrounding material mak-
ing up IP1. Crucially, the first part of IP1, i.e. the part preceding the par-
enthetical, is not terminated by a boundary tone, and while there may be
prenuclear accents in the first part of IP1, the nuclear accent is located
in the second part. We may also find pauses before and/or after the par-
enthetical. For example, the IP spanning the parenthetical how shall I put
it in (4.36) has a prenuclear peak (H*) associated with how and a nuclear
accent (!H*) associated with put, followed by L-L% terminating this IP.
IP1, spanning the host material that’s a little bit of arrogance, has a prenu-
clear peak (L+H*) associated with the first syllable of little and a nuclear
accent (L+!H*) on arrogance, downstepped with respect to the accent on
little. The first part of IP1, i.e. the part preceding the parenthetical, is not
terminated by a boundary tone. The second part of IP1 is terminated by
L-L%, which is followed by a step-up in pitch on that, i.e. on the first,
unaccented syllable of IP3. The interpolated parenthetical is also part of
a hesitant phase with a reduced pitch range and reduced speech rate. It is
preceded by a filled pause (uh) of approximately 250 ms, and followed by
142  Data analysis, results and discussion

a silent stretch of > 150 ms and filled pause (uh uhm uh) of > 1.7 seconds.
However, even though it functions as a kind of word-searching phrase, the
parenthetical cannot be seen as part of the filled pause, since it contains
clearly prominent material.
The host utterance in (4.35) is phrased in one large IP with several pre-
nuclear pitch accents and a nuclear accent associated with food. The paren-
thetical has its own CTC with a nuclear accent associated with hear. The
speech rate is higher in the parenthetical than in the host utterance. In
(4.37), IP1 is preceded by H% terminating the preceding IP. The nuclear
accent of IP1 is associated with the stressed syllable of surveyors. The par-
enthetical has its own CTC with a nuclear accent associated with Jimmy. In
(4.38), He was starts an IP which is interrupted by the parenthetical, but
continues after the parenthetical. (See also Dehé 2009: 591–2 for an analysis
of this example.) Note that none of the parentheticals in (4.35)–(4.38) is
preceded by a syntactic clause boundary. In fact, all four parentheticals are
XP-internal. Based on current prosodic theory, recursive prosodic phrasing
is therefore predicted (see end of Chapter 3).

4.3.1.2  Pattern (4.31)b: prosodic integration


Only one out of 72 parentheticals was phrased according to (4.31)b, i.e. in such
a way that material from the parenthetical formed an IP with material from
the host. It was the as-parenthetical given in (4.39)/Figure 4.3.9 The paren-
thetical as my right honourable friend the Prime Minister did two weeks ago forms
one long IP together with by proclaiming from the host. This IP (IP3 in (4.39)
b) is preceded and followed by silent pauses of approximately 730 ms and 430
ms, respectively, not plotted in Figure 4.3 for space reasons. Other boundary
markers include the presence of a boundary tone (L%) terminating the pre-
ceding IP (not plotted here), a step-up in pitch on unaccented by at the begin-
ning of IP3, the CTC in IP3 with a nuclear peak associated with ago inside
the parenthetical and terminated by L-L%, with subsequent pitch reset, and
the fact that declination applies across the target IP. There is no pitch discon-
tinuity or any other prosodic cue before or inside the parenthetical that would
justify the assumption of an additional IP boundary. In particular, there is no
reason to assume an IP boundary between proclaiming and as.
(4.39) a. We commit a serious error if we think always in terms of sur-
rendering sovereignty and seek to stand pat for all time on a
given deal by proclaiming as my right honourable friend the
Prime Minister did two weeks ago that we have surrendered
enough (ICE-GB: s2b-050 #46)

9
One case of integration reported in Dehé (2009) was reanalysed as separation (see example
(34) in Dehé 2009: 591).
220
200
Pitch (Hz)

150

100

50
H* !H* !H* !H* L-L%

FPC (as-parenthetical)

by pro clai ming as my right honourable friend the Prime Mi nister did two weeks a go

0 3.062
Time (s)

Figure 4.3  Example (4.39): full parenthetical clause, prosodic integration


144  Data analysis, results and discussion

b. IP1[We commit a serious error]IP IP[if we think always in terms


of surrendering sovereignty]IP1 IP2[and seek to stand pat for all
time on a given deal]IP2
IP3[by pro  claiming as my right honourable  friend
H* !H*
the Prime Minister did two weeks  ago]IP3
!H* !H* L-L%
IP4 [that we have surrendered enough] IP4

4.3.1.3  General remarks and discussion


Of the 72 parenthetical clauses analysed here, 7 involved backtracking. Two
examples are given in (4.40) and (4.41); the repeated material is underlined.
In (4.40), the speaker begins a new IP after the vocative Mr Lehrer, which
he then immediately interrupts in order to give the reason for his question,
using a free/floating parenthetical (I want to be clear about this), which spans
its own IP. This is overall lower in pitch than the following IP, but has its own
CTC with a nuclear accent associated with clear. The speaker then restarts
his question and with it the IP begun before the insertion. This parenthet-
ical clause was analysed as prosodically separate in a non-recursive structure
because of the restart. In (4.41) the phrasing is such that the repeated mater-
ial ((that/but) if there is a unilateral action) forms its own IP both before and
after the parenthetical. The parenthetical insertion consists of two complex
sentences and is phrased separately.
(4.40) a. Mr Lehrer are you I want to be clear about this are you telling
us that this is your interpretation of what he was telling you or
what he was actually telling you (ICE-GB: s1b-064 #167)
b. [are you IP[I want to be clear about this]IP IP[are you telling
us]IP …
(4.41) a. And the trouble is that if there is a unilateral action and I don’t
really think now that Mr Bush is heading for that I think he wants
to work through the United Nations but if there is a unilateral
action it will I’m afraid gravely damage this newly emerging
UN consensus and the cooperation between the superpowers
(ICE-GB: s1B-035 #98–101)
b. IP1[And the trouble is]IP1 IP2[that if there is a unilateral action]IP2
IP3[and I don’t really think now that Mr Bush is heading for that]IP3
IP4[I think he wants to work through the United Nations]IP4 IP5[but
if there is a unilateral action]IP5 [it …
It follows from the results reported in the previous sections that prosodic sep-
aration is the default for the prosodic phrasing of full parenthetical clauses,
in line with hypothesis (4.29) and the predicted phrasing in (4.30) and in
accordance with the assumptions made in previous literature on the pros-
odic phrasing of parentheticals. Ninety-nine per cent of observed cases follow
Analysis and results  145

patterns (3.35)a–b. Note once again that parentheticals preceded by a closing


IP boundary but not by a syntactic clause boundary (e.g. (4.34)) are not directly
predicted by Match or Align R. The IP boundary terminating the string pre-
ceding the parenthetical must be the result of the promotion of a lower-level
prosodic boundary to an IP boundary. The only example deviating from the
pattern of prosodic separation, presented in (4.39)/Figure 4.3 above, follows
pattern (3.35)c, satisfying Align/Wrap, but violating Match. It thus displays
the left-right asymmetry predicted by Selkirk’s (2005) Align R (CommaP, IP)
constraint, which demands an IP boundary to the right of the parenthetical,
but not to its left. This example was explained in Dehé (2009: 600–1) as fol-
lows. The corpus item is taken from a political speech delivered to the House of
Commons in London. The parenthetical is a variation of the phrase as my right
honourable friend the Prime Minister plus a VP. The as-phrase is thus a set phrase
which is a routinized part of parliamentary debates and speeches. Intonation
therefore does not have to serve the function of indicating a true aside, a func-
tion which is commonly corroborated by prosodic separateness. Furthermore,
the speaker produces the utterance at a high speech rate. As Nespor and Vogel
(1986: 195) note, “the faster the rate of speech, the longer the [IPs] of a given
utterance tend to be”. Any potential IP boundary to the left of the parenthet-
ical thus falls victim to the high speech rate. In (4.39), the boundary termin-
ating IP2, i.e. the IP preceding the IP spanning the parenthetical, is predicted
by syntactic structure, since by proclaiming that … is a complex adjunct. The
remaining host material preceding the parenthetical, i.e. by proclaiming, would
then form a small IP on its own. The speaker might wish to avoid this, given
that all other IPs are considerably longer; previous research has shown that
speakers tend to prefer a series of prosodic constituents that are balanced in
size (e.g. Gee and Grosjean 1983; Nespor and Vogel 1986; Ghini 1993; Frota
2000). Furthermore, an IP boundary in the position following proclaiming,
i.e. between a verb and its (clausal) complement, would not be predicted in
the absence of the parenthetical. In other words, there are factors, including
speech style, discourse function and size constraints, which override syntactic
structure in determining the prosodic structure of this example.

4.3.2  Non-restrictive relative clauses


Among the 60 NRRCs that entered the analysis, the auditory analysis of
one NRRC was not conclusive and the quality of the sound file was not
good enough to allow for instrumental analysis. It was therefore coded as
‘phrasing unclear’. The relative clause in (4.42) is ambiguous between a
restrictive interpretation such that there are several colleagues one of which
the speaker already knew and the non-restrictive meaning such that there
was exactly one colleague and the speaker knew him in advance. This item
was phrased as shown in (4.42)b, with IPs 1–4 each terminated by H%,
and pauses of > 400 ms after IPs 2, 3 and 4. It was counted here as a NRRC
phrased according to prosodic separation. All other NRRCs included in
146  Data analysis, results and discussion

the analysis were unambiguously non-restrictive. Eight NRRCs clearly


referred to categories other than NPs.10 Remember that, like for full par-
enthetical clauses, the prediction for NRRCs is that they are prosodically
phrased separately by default (see the hypothesis and predicted phrasing
in (4.43) and (4.44), respectively, repeated from Section 2.3.2 above), fol-
lowing pattern (3.35)a. The phrasing results for NRRCs are summarized
in (4.45) and detailed in Sections 4.3.2.1 and 4.3.2.2 below, followed by a
discussion.
(4.42) Relative clause potentially ambiguous between restrictive and non-
restrictive meaning
a. The only colleague whom I knew already was the social and
physical geographer Doctor Dudley Stamp who like myself had
long been on the Council for the Protection of Rural England
(ICE-GB: s2b-025 #75)
b. IP1[The only colleague]IP1 IP2[whom I knew already]IP2 [was the
social and physical geographer]IP3 IP4[Doctor Dudley Stamp]IP4
IP5[who …

(4.43) Prosodic phrasing of NRRCs: hypothesis


The default prosodic phrasing of NRRCs is separation. NRRCs are
phrased in their own intonational domain, preceded and followed
by the respective boundaries.
(4.44) NRRCs: predicted prosodic phrasing
prosodic separation: (…) IP[…]IP IP[NRRC]IP (IP[…]IP) (…)
(4.45) Observed patterns of prosodic phrasing of non-restrictive relative
clauses (NRRCs)
a. NRRC: prosodic separation
… IP[…]IP IP[which NRRC]IP IP[…]IP … 55 NRRCs  (92%)
b. NRRC: relative pronoun phrased with preceding host material,
NRRC-remnant separate
… IP[… (P) which]IP IP[NRRC]IP IP[…]IP … 4 NRRCs  (7%)
c phrasing unclear 1 NRRC   (1%)
Total: 60 NRRCs

4.3.2.1  Pattern (4.45)a: prosodic separation (NRRC)


Fifty-five NRRCs (92%) followed pattern (4.45)a, i.e. they were clearly
separate, and were preceded and followed by separate IPs. All utterances

10
Of the original nine NRRCs relating to the whole sentence or the VP (see Section 4.1.2
above), one was the item whose phrasing was unclear. Of the remaining eight items, seven
were phrased according to (4.44) (non-recursive structure, prosodic separation), and one
was phrased as given in (4.45)b (this is the example illustrated in (4.47)/Figure 4.5).
Analysis and results  147

containing NRRCs were phrased in a non-recursive structure. All NRRCs


were phrased in separate IPs (not ips). Typical cues to intonational separ-
ateness were the same as for FPCs; changes in pitch level/direction at the
beginning of an IP were associated with the relative pronoun of the NRRC
and the first syllable of the following IP. This is illustrated immediately
below for example (4.46)/Figure 4.4.
(4.46) a. At the same time Iran’s President Rafsanjani who’s strongly
condemned the bombing of Baghdad says he’ll meet President
Saddam and renew contacts with America if it could lead to a
peace process (ICE-GB: s2b-001 #19)
b. … IP1[Iran’s President]IP1 IP2[Rafsanjani]IP2
L+H* L-H% L+H* L-L%
IP3 [who’s  strongly condemned the  bombing of Baghdad]IP3
    L+H*  L+!H*    !H*      L+H* H-H%
IP4[says he’ll  meet President Saddam]IP4 …
        H*      L*+H H-H%
The IPs indicated in (4.46)b and the respective boundaries can clearly be
identified on the basis of the following criteria (see Figure 4.4). First, all
IPs have their own CTCs. IP1 has a nuclear pitch accent associated with
the second syllable of Iran’s and is terminated by H%; IP2 has a nuclear
accent associated with the final syllable of Rafsanjani and is terminated
by L%. IP3, the IP spanning the NRRC, has prenuclear pitch accents
associated with the stressed syllables of strongly, condemned and bombing,
followed by a nuclear accent on Baghdad with a rising terminal (L*+H
H-H%). IP4 has a prenuclear accent on meet and a nuclear accent associ-
ated with Saddam, terminated by H%. Second, the IPs identified here are
each domains across which declination applies. We find pitch reset such
that the first local peak in IP3 (the NRRC), associated with the first syl-
lable of strongly, is higher than the preceding one on the last syllable of
Rafsanjani, and approximates the height of the first peak in IP1. The first
peak in IP3 is followed by downstepped pitch peaks in the same domain.
Pitch reset can again be observed in IP4 (see the local pitch peak on meet).
Third, the NRRC is preceded and followed by structure-related pauses of
0.432 seconds and 0.32 seconds, respectively. Fourth, the L% terminating
IP2 is followed by a step-up in pitch level on the relative pronoun who.
While we do not find a step-down from H% between IP3 and IP4, there
is a pitch discontinuity related to the pause. Fifth, segmental processes
are absent at the boundaries, also because of the presence of the silent
pauses. Finally, it follows from the previous points that the phrasing is
non-recursive.
The NRRC relates to the subject, which is an appositional construc-
tion, consisting of the anchor Iran’s President and the N-APP Rafsanjani,
350

300
Pitch (Hz)

200

100

50

L+H* L-H% L-H* L-L% L+H* L+!H* !H* L+H*H-H% H* L*+H H-H%

NRRC

Iran’s President Rafsanja ni <SIL> who’s strongly condemned thebombing of Baghdad <SIL> says he’ll meet President Saddam

0.6212 8.335
Time (s)

Figure 4.4  Example (4.46): NRRC, prosodic separation


Analysis and results  149

which is also phrased separately. The IP boundary preceding the N-APP is


marked by tonal features (H% and a step-down in pitch on the first syllable
of Rafsanjani).11

4.3.2.2 Pattern (4.45)b: relative pronoun phrased with preceding host


material, NRRC-remnant separate
Only 4 out of 59 analysed NRRCs (7%) turned out not to be phrased accord-
ing to (4.45)a. They are given in (4.47) through to (4.50).12
(4.47) a. I was programming in Pascal which really wasn’t very exciting
I’m afraid (ICE-GB: s1a-008#1)
b. IP1[I was programming in Pascal which]IP1  IP2[IP2[really wasn’t
very exciting]IP2 I’m afraid]IP2
(4.48) a. And I put forward the argument which could certainly be con-
tested uh but which I have some faith in that some residue
remains some flickering ember that would allow for some reli-
gious renewal within our culture (ICE-GB: s1b-028 #30)
b. IP1[And I put forward the argument which]IP1 IP2[could certainly
be contested]IP2 IP3[but which I have some faith in]IP3 …
(4.49) a. And the first which which I have to say sounds to those who don’t
share this point of view always sounds the noisier one uh is domi-
nated by a view which I can express as follows (ICE-GB: s2a-
021 #19)
b. IP1[And the first which]IP1 IP2[which I have to say sounds to those
who don’t share this point of view]IP2 IP3[always sounds the noisier
one]IP3 …
(4.50) a. First then the context the wider context within which discus-
sion on education of which uh teacher education’s only a part but
I can’t disentangle it is taking place (ICE-GB: s2a-021 #17)
b. … IP1[the wider context]IP1 IP2[within which discussion on edu-
cation of which]IP2 IP3[uh teacher education’s only a part]IP3 IP4[but I
can’t disentangle it]IP4 IP5[is taking place]IP5
All four NRRCs show the same pattern: while the right syntactic edge of the
NRRC coincides with an IP boundary, the left edge does not. Specifically,

11
The F0 contour associated with the appositional construction Iran’s President Rafsanjani
is reminiscent of Beckman and Pierrehumbert’s (1986) example of tags discussed in
Section  3.2 above. Looking at the contour, it seems that declination applies across the
whole appositional construction: the peak associated with Rafsanjani is lower than the one
associated with Iran’s. Moreover, no pause can be observed before Rafsanjani. An analysis
with two subsequent ips instead of IPs therefore seems possible. However, both anchor and
apposition are associated with nuclear accents and there is a clear fall-rise from the accen-
tual peak associated with Iran’s, thus the apposition is clearly prosodically separate.
12
As compared to Dehé (2009) one item was reanalysed as being separate.
250

200
Pitch (Hz)

150

100

50

H* L-L% H* L-L% L L%

NRRC

I was programming in Pas cal which <SIL> really wasn’t very ex ci ting <SIL> I’m afraid

0 3.914
Time (s)

Figure 4.5  Example (4.47), phrasing as in (4.51)


Analysis and results  151

the relative pronoun which, or a pied-piped preposition along with which


(see (4.50)), is part of unstressed material, which is tonally integrated with
the preceding domain and followed by an IP boundary separating the rela-
tive pronoun from the rest of the NRRC. The pattern is repeated from (4.45)
b in (4.51) (parentheses indicate optional material; P = preposition). By way
of illustration, consider example (4.47), plotted in Figure 4.5.13
(4.51) IP[… (P) which]IP IP[NRRC] IP (IP […]IP)
The nuclear syllable of the first IP is the second syllable of Pascal. The pitch
fall from H* associated with -cal continues on the relative pronoun which.
No pitch discontinuity or change in pitch level or direction can be found on
the relative pronoun. The IP is terminated by L%, which is followed by a
pause of 185 ms. There is a step-up in pitch on the first syllable of really at
the beginning of the second IP, which has a nuclear pitch accent associated
with exciting. The two domains show tonal parallelism (e.g. Bolinger 1989:
205ff; Wichmann 2000). They begin at approximately the same mid pitch
level, which rises towards the peak on the nuclear syllable (-scal [ˈskæl] and
-ci- [ˈsaɪ], respectively). The fall from the nuclear peak continues on the
immediately following syllable (which and -ting, respectively). Both domains
are terminated by L%.
Example (4.49) is different from the others in that the relative pronoun
which is repeated by the speaker and the second relative pronoun is followed
by a kind of comment clause (I have to say) before the NRRC continues. The
syntactically complex NRRC, which in turn contains a restrictive relative
clause, is phrased in more than one IP (see (4.52)). The phrasing is such that
the first which phrases with the preceding domain, followed by an IP bound-
ary between the two pronouns. IP1 has its own CTC with its own nuclear
accent and boundary tone; the F0 peak is associated with the nuclear syl-
lable first and the rise is completed on which, followed by a break of approxi-
mately 180 ms and a step-down in pitch on the second which, marking the
boundary. IP2 also has its own CTC: the nuclear accent is associated with
the syllable share, followed by the unstressed material this point of view and
terminated by H%. This is then followed by a structure-related pause of
> 800 ms. IP3, which spans the second part of the NRRC, has a nuclear
L+H* followed by L-H%; the local peak is associated with the first syllable
of noisier. IP3 is followed by a silent and filled pause of approximately 440
ms overall before the beginning of IP4.

13
The phrasing of I’m afraid indicated here follows Gussenhoven’s (2004: 291) encliticization
and tone copy analysis. I’m afraid is unaccented, at low level pitch, but it is preceded by a
boundary tone associated with the end of the domain ending in exciting and by a consid-
erable pause. The encliticized material receives a copy of the tones after the last stressed
syllable. Since exciting is associated with a nuclear peak followed by low edge tones (H*
L-L%), the sequence of L tones is copied onto the CC (Gussenhoven 1990, 2004). Note
that I’m afraid is proceeded by a full clause. It was therefore discarded from the analysis
of CCs.
152  Data analysis, results and discussion

(4.52) Example (4.49): prosodic phrasing and tonal annotation


IP1[And the  first which]IP1
H*  L-H%
IP2[which I  have to say sounds to those who don’t  share this point
H* !H* !H*
of view]IP2
L-H%
IP3[always sounds the noisier one]IP3  IP4[is dominated …
L+H*  L-H%
In (4.50), another relative clause (of which teacher education is only a part but
I can’t disentangle it) has been inserted into the NRRC within which discussion
on education is taking place. The first NRRC (within which discussion …) is
preceded by an IP boundary (see (4.53)). The interpolated NRRC is com-
plex in itself in that it contains a coordinated clause (but I can’t disentangle
it). The intonational phrasing is such that of which joins the preceding IP.
All three IPs spanning target material have nuclear H* peaks and L-H%
edge tones (see (4.53)). The complex interpolation is split up into two IPs,
but the IP boundary between them coincides with a clause boundary.
(4.53) Example (4.50): prosodic phrasing and tonal annotation
IP1[the wider context]IP1
IP2[within which dis  cussion on edu  cation of which]IP2
H*  L- H* L-H%
IP2[uh teacher education’s  only a part]IP2
H* L-H%
IP3[but I  can’t disen  tangle it]IP3
H* !H*  L-H%
IP4[is taking  place]IP4
H+   !H* L-L%

4.3.2.3  General remarks and discussion


Two of the 60 NRRCs that entered the analysis involved backtracking. They
are given in (4.54) and (4.55); repeated material is underlined. (4.54) is phrased
such that the repeated material forms its own IP preceding the NRRC (IP1 in
(4.54)b), ending in H%. The NRRC is overall lower in pitch but separate with
its own nuclear accent associated with won’t and L% terminating IP2. The
repeated material is then part of the following IP, with a prenuclear accent on
were emphasizing the contrast between which we won’t be and but if we were,
and a nuclear accent on Mercedes. In (4.55), Paul is associated with a nuclear
H*L-H% tune. In its second occurrence Paul is integrated in IP4.
(4.54)  a. but I mean just off the top of your head if we were jammy
enough which we won’t be but if we were jammy enough to keep
the Mercedes which I would be the main if you like you could
put it down as the only driver of and we carried on keeping the
Escort as well (ICE-GB: s1b-080 #333)
Analysis and results  153

b. … IP1[if we were jammy enough]IP1 IP2[which we won’t be]IP2


IP3[but if we were jammy enough to keep the Mercedes]IP3
IP4[which …

(4.55) a. Above all Paul whose job it was originally to expose the Christian
racket and who must have heard and investigated all stories about
it Paul was more convinced than anyone of its truth (ICE-GB:
s2b-028 #65)
b. … Paul]IP1 IP2[whose job it was originally to expose the Christian
racket]IP2 IP3[who must have heard and investigated all stories about
it]IP3 IP4[Paul was more convinced than anyone]IP4 IP5[…
Overall, NRRCs, like full parenthetical clauses, follow pattern (3.35)a/(4.45)
a by default (92% of observed cases), i.e. they are prosodically phrased sep-
arately in an IP of their own. The hypothesis in (4.43) and the predicted
prosodic phrasing in (4.44) are thus corroborated by the results of the ana-
lysis of the corpus data. In the corpus data analysed here, the phrasing of
the relative pronoun with the preceding IP, occurring in four cases, i.e. less
than 10% of the data, is the only exception.
The results are in line with previous experimental work. For example,
Astruc-Aguilera (2005) identified separation as the default for NRRCs (see
Section 2.3.2 above), except that she found strict prosodic separation only.
However, the NRRCs in her experimental material, which was read by the
participants, were all set off from the host by commas (see Astruc-Aguilera
2005: 222). I assume that the participants placed prosodic boundaries in the
comma positions, in particular to delimit them from the sentences contain-
ing restrictive relative clauses presented in her experiment. In a sense, the
experimental materials did thus not allow for variation in terms of boundary
position. The results of the present study are furthermore in line with Auran
and Loock’s (2011) corpus study, which is expected because they used the
same corpus (ICE-GB) along with the Aix-Marsec corpus (see Auran and
Loock 2011: 186–7 for details). The difference between their analysis and
the present one is that all NRRCs in their data set were analysed as pro-
sodically separate, i.e. they did not find the phrasing in which the relative
pronoun joins the preceding domain. The cause of this difference is a mat-
ter of speculation. One possible reason is guidance by syntactic structure.
For example, Cole et al. (2010: 1162) report that there is a strong tendency
for (albeit untrained) listeners to hear prosodic boundaries preceding rather
than following coordinating conjunctions, which, by extension, could also
hold for relative pronouns introducing NRRCs. The present findings also
support Watson and Gibson’s (2004: 749)  intuition about NRRCs being
phrased separately in non-laboratory speech.
The default phrasing for NRRCs being (3.35)a/(4.45)a, the surface
phrasing must be the result of the promotion of a lower-level prosodic
boundary to an IP boundary terminating the string preceding the NRRC
unless that string is a clause. If the NRRC relates to an NP (e.g. (4.42))
154  Data analysis, results and discussion

or other lexical phrase, current prosodic theory predicts an ip boundary


at the right edge of that constituent and thus a recursive prosodic struc-
ture. Like in (4.42), however, NRRCs relating to lexical phrases are by
default preceded by an IP boundary terminating the phrase of the XP.
The non-default cases follow (3.35)e, satisfying Align R, but not Wrap and
Match. However, all four cases involve the relative pronoun (plus a pied-
piped preposition where appropriate) joining the preceding IP spanning
host material (see (4.45)b/(4.51)). In no case is the IP boundary placed
beyond the right edge of the relative pronoun. In Dehé (2009) I employ a
constraint “Shift IP boundary”, which allows a left edge IP boundary to
shift to the right of a clause-initial functional element (relative pronoun,
coordinating conjunction, connective), but not any further. This con-
straint operates on the input default pattern, i.e. taking prosodic separ-
ation as input. Boundary shift of this kind may be driven by factors other
than syntax, for example prosodic length in terms of number of syllables
or tonal parallelism. However, the factor length would predict prosodic
separation for the NRRC in (4.48). It is also possible that boundary shift is
a discourse effect, for example such that it is used as a floor-holding device
(Dehé 2009: 602), thus a secondary effect dependent on the individual
discourse situation.
No cases were found in which preceding host material joins the IP of the
NRRC, thus no evidence for a left-right asymmetry as shown in (3.35)c was
found in this subset of the data.

4.3.3  Nominal appositions


Overall, 516 N-APPs had entered the prosodic analysis (see Section 4.1.3).
Of these, 50 had to be discarded upon closer inspection for one of the fol-
lowing reasons, which had not been detected in the initial sorting process:
(i) any acoustic analysis was impossible due to the quality of the sound file;
(ii) the sound file was good enough for auditory but not for instrumental
analysis and the auditory analysis was inconclusive; (iii) the corpus anno-
tation did not correspond to the syntactic structure of the utterance (i.e.
constituents annotated N-APP were not noun phrases); (iv) the wording in
the sound file did not correspond to the annotation; (v) the syntactic struc-
ture of the utterance was unclear; or (vi) the sound file was cut in the middle
of the utterance, thus obscuring the syntactic or prosodic structure. The
remaining 466 tokens entered the analysis. Remember that like for full par-
enthetical clauses and NRRCs, the prediction for N-APPs is prosodic separ-
ation by default (see hypothesis (4.56) and the predicted prosodic phrasing
in (4.57), repeated from Section 2.3.3 above). The phrasing results obtained
for N-APPs are summarized in (4.58). Table 4.3 lists the results separately
for each type of N-APP as given in (4.5) through to (4.8).
Analysis and results  155

(4.56) Prosodic phrasing of N-APPs: hypothesis


The default prosodic phrasing of N-APPs is separation. N-APPs
are phrased in their own intonational domain, preceded and fol-
lowed by the respective boundaries.
(4.57) N-APPs: predicted prosodic phrasing
Prosodic separation: (…) IP[…]IP IP[N-APP]IP (IP[…]IP) (…)
(4.58) Observed patterns of prosodic phrasing of nominal appositions
(N-APPs)
a. N-APP: prosodic separation
a1. … IP[…]IP IP[N-APP]IP IP[…]IP … 375 N-APPs  (80%)
a2. … IP[… IP[N-APP]IP …]IP … 2 N-APPs (< 1%)
b. N-APP: prosodic integration; N-APPs phrased together with
preceding and/or following host material
b1. … IP[…]IP IP[… N-APP]IP IP[…]IP … 65 N-APPs (14%)
b2. … IP[…]IP IP[N-APP …]IP IP[…]IP … 3 N-APPs (< 1%)
b3. … IP[…]IP IP[… N-APP …]IP IP[…]IP … 9 N-APPs  (2%)
c. phrasing unclear 12 N-APPs (2.5%)
Total: 466 N-APPs

4.3.3.1  Pattern (4.58)a: prosodic separation (N-APP)


Overall, 375 N-APPs (80%) were phrased separately in a non-recursive
structure as given in (4.58)a1 (see Table  4.3a1 for distribution among the
types). Typical cues to IP boundaries included a CTC for each IP, bound-
ary tones at the end of the IP spanning the N-APP and at the end of the
preceding IP, structure-related pauses in target position, and changes in
pitch level/direction associated with unstressed syllables at the beginning
of an IP. Examples of non-recursive prosodic separateness are given in
(4.59)/Figure 4.6 and (4.60)/Figure 4.7.
In (4.59)/Figure 4.6, IP1, which precedes the IP spanning the N-APP,
has a CTC: a prenuclear peak and nuclear L+!H* accent, terminated by
L-H%. The nuclear tune in IP2, spanning the N-APP, is L+H* L-L%;
the nuclear syllable is the second syllable of presenters. The N-APP (IP2) is
realized with compressed pitch, following a pattern often observed for par-
enthetical insertions. It is preceded and followed by pauses of 800 and 544
ms, respectively. In IP3 the pitch is reset almost to the level of IP1.
(4.59) N-APP: prosodic separation
a. Tanner and Gore Stansworth my co-presenters have developed a
four-dimensional model of reactions to torture (ICE-GB: s2a-
034 #70)
b. IP1[Tanner and Gore Stansworth]IP1 IP2[my co-presenters]IP2
H* L+!H* L-H% L+H* L-L%
IP3[have developed …
156  Data analysis, results and discussion

Table 4.3  Patterns of prosodic phrasing of nominal appositions (N-APPs)

anchor
= no
proper proper N-APP = Pronominal
N (see N (see proper N anchor (see
phrasing (4.5)) (4.6)) (see (4.7)) (4.8)) overall
a. a1. […]IP IP[N-
IP 95 73 198 9 375 (80%)
APP]IP IP[…]IP

a2. [… IP[N-APP
IP 0 2 0 0 2 (< 1%)
(…)]IP ]…IP

b. b1. IP[…]IP IP[… 5 5 54 1 65 (14%)


N-APP]IP
IP[…]IP

b2. IP[…]IP IP[N- 2 0 1 0 3 (< 1%)


APP …]IP
IP[…]IP

b3. IP[…]IP IP[… 2 0 6 1 9 (2%)


N-APP …]IP
IP[…]IP

c. phrasing 3 1 8 0 12 (3%)
unclear
107 81 267 11 466

The example in (4.60)/Figure 4.7 illustrates a N-APP modifying a pronom-


inal anchor and phrased in a separate IP. The N-APP the rest of the world
is preceded and followed by pauses (311 ms and 496 ms, respectively) and
it has its own CTC with nuclear prominence associated with world (!H*
L-H%). The preceding and following IPs also have their own CTCs. The
subject pronoun we (the anchor) would not normally be predicted to form
its own prosodic domain, but it is phrased separately here with a H* L-H%
nuclear tune due to the presence of the nominal apposition and discourse
factors (in the broadcast talk it is taken from we contrasts with he, referring
to Saddam Hussein).
(4.60) N-APP: pronominal anchor; prosodic separation
a. For our part we the rest of the world have acted with enormous
restraint (ICE-GB: s2b-030 #32)
b. IP1[for  our part]IP1 IP2[we]IP2    IP3[the  rest of the  world]IP3
H* L-H% H* L-H%    L+H*   !H* L-H%
IP4[have acted with enormous restraint]IP4

4.3.3.2 Pattern (4.58)b1: N-APP phrased together with preceding host


material (… IP[…]IP IP[… N-APP]IP IP[…]IP …)
The second most frequent prosodic pattern is the one given in (4.58)
b1/Table  4.3b1. The N-APP typically bears nuclear prominence but is
350

300
Pitch (Hz)

250

100

50
H* L+!H* L-H% L+H% L-L% L+H* !H* L+H* L-L%

N-APP

Tanner and Gore Stansworth <SIL> my co- presenters <SIL> had developed a four- dimensional model of reactions to torture

0 7.77
Time (s)

Figure 4.6  Example (4.59): N-APP, prosodic separation


200
Pitch (Hz)

150

100

80
H* L-H% H* L-H% L+H* !H* L-H% L+H* L+!H* !H* L-L%

N-APP

for our part we <SIL> the rest of the world <SIL> have acted with enormous restraint

0 4.899

Time (s)

Figure 4.7  Example (4.60): N-APP with pronominal anchor, prosodic separation


250

200
Pitch (Hz)

150

100

60

H% L+H* L+H* L-H% L+H* L-H%

N-APP

last month The Defence Secretary Mr Cheney said it was conceivable <SIL> that

0 3.795
Time (s)

Figure 4.8  Example (4.61): N-APP phrased together with anchor


160  Data analysis, results and discussion

phrased together with preceding unstressed material from the syntactic host,
which includes the anchor of the N-APP, i.e. the appositional construction
is phrased together in one IP. Examples are given in (4.61)/Figure 4.8 and
(4.62) through (4.64).
In (4.61)/Figure  4.8, there is no pitch discontinuity between Secretary
and Mister, nor is there any other reason to assume a boundary before the
N-APP, thus the N-APP phrases together with its preceding anchor. There
is a prenuclear prominence associated with the second syllable of Defence.
A nuclear L+H* pitch accent is associated with Cheney and the IP (IP2 in
(4.61)b) is terminated by L-H%. (The apparent interruption of the F0 con-
tour before Cheney is due to the voiceless /tʃ/ onset).
(4.61) N-APP phrased together with preceding host material
a. Last month the Defence Secretary Mr Cheney said it was
conceivable that a further one hundred thousand American
soldiers could be sent to the region and it’s believed that the
reserve units would be part of that additional force (ICE-GB:
s2b-017 #18)
b. IP1[Last month]IP1 IP2[the Defence Secretary Mr Cheney]IP2
H%     L+H* L+H* L-H%
IP3[said it was conceivable]IP3 IP4[that …
L+H* L-H%
Example (4.62) represents a typical pattern in this group. It is taken from
a parliamentary debate. The constituent annotated as N-APP denotes a
state function (here: the Secretary of State for Trade and Industry) following
the expression my right honourable friend. The frequent use of this phrase
in parliamentary debates and its nature of being a fixed expression often
results in fast speech rate and the formation of large IPs, and specifically
in the lack of an IP boundary following my right honourable friend (see also
the discussion of (4.39)b/Figure 4.3 in Section 4.3.1 above). In (4.62), the
main prominence in IP2 (spanning the syntactic adjunct containing the
appositional construction) is associated with the N-APP, specifically with
Industry; the nuclear contour is L+!H* L-H%. As compared to IP1 and
IP3, IP2 is realized at compressed pitch. It is preceded by L% and by a
break between the release of the final plosive on account and the onset of by
and followed by a combination of silent and filled pauses of 580 ms overall.
(4.62) N-APP phrased together with preceding host material
a. of course those matters are taken into account by my right hon-
ourable friend the Secretary of State for Trade and Industry uh
in the negotiations to which the honourable gentleman refers
(ICE-GB: s1b-055#35)
b. … into account]IP1 IP2[by my right honourable friend the
H* L-L%
Analysis and results  161

Secretary of  State for  Trade and  Industry]IP2


L+H* L+H* L+H* L+!H* L-H%
IP3 [in the negotiations …
In (4.63), the N-APP the Cabinet cat forms one IP together with the preced-
ing string Even Humphrey, thus again the appositional construction phrases
together. The nuclear accent of this IP (IP1 in (4.63)b) is associated with cat.
IP1 is terminated by L% and followed by a pause of 174 ms after the release
of the final plosive. In (4.64), the N-APP the able-bodied relates to the pro-
nominal anchor we. The string and we the able-bodied forms one IP (IP1 in
(4.64)b) with main prominence associated with the N-APP, specifically with
bodied. Unlike in (4.60) above, the presence of the N-APP does not go along
with separate phrasing of the subject pronoun we.
(4.63) N-APP phrased together with preceding host material
a. Even Humphrey the Cabinet cat seems affected by the immi-
nent change of occupier (ICE-GB: s2b-003 #101)
b. IP1[Even Humphrey the Cabinet cat]IP1 IP2[seems affected by the
imminent change of occupier]IP2
(4.64) N-APP phrased together with preceding host material
a. Uhm so one of the things that that the group offers is is a meet-
ing of people who have been in some way segregated and falsely
segregated uh that these people with disabilities are somehow
placed to one side and we the able-bodied get on with our phys-
ical activities which are separate (ICE-GB: s1a-001 #124)
b. … IP1[and we the able-bodied]IP1 IP2[get on …

4.3.3.3 Pattern (4.58)b2/b3: N-APP phrased together with following or


with preceding and following host material
The two relatively infrequent non-separate phrasing patterns in (4.58)b2
and b3 have in common that the N-APP may either have nuclear promin-
ence, or it may bear non-nuclear prominence, but no unstressed N-APPs
were found in the data. Examples of these patterns are given in (4.65),
(4.66)/Figure 4.9 and (4.67).
The N-APP in (4.65), taken from a spontaneous radio commentary, is
phrased according to (4.58)b2, representing a particularly infrequent pat-
tern. The N-APP Peter Brooke has main prominence in its IP (IP2 in
(4.65)b) and is followed in its IP by unstressed material from the syntactic
host (is there). There is no reason to assume an IP boundary after Brooke.
The cues to the boundary separating IP2 from the following IP include the
low boundary tone and a step-up in pitch on and at the beginning of IP3.
The anchor is located in IP1. This phrasing is probably best explained in
terms of length and information structure (see also Section 4.3.3.5 below).
If the speaker was planning an IP boundary before the conjunction and, then
300

250
Pitch (Hz)

200

150

100

50

L% L+H* !H* !H* L-L%

N-APP

goes chasing after it Kuznecov the substitute gets there first and Garcia

0 3.981
Time (s)

Figure 4.9  Example (4.66): N-APP phrased together with anchor and following host material
Analysis and results  163

separate phrasing of the N-APP, i.e. a boundary immediately after Brooke,


would have resulted in two short IPs. Moreover, the news value of the string
is there does not justify prominence, which it would have received had it been
phrased in its own IP. In the attested phrasing, it is realized as unstressed
material following the nuclear accent.
(4.65) N-APP phrased together with following host material
a. The Northern Ireland Secretary Peter Brooke is there and
other defence ministers Archy Hamilton the Earl of Arran
Kenneth Carlisle (ICE-GB: s2a-019 #14–15)
b. IP1[Northern Ireland Secretary]IP1  IP2[Peter   Brooke is there]IP2
H* L- L%
IP3 [and …
The example in (4.66)/Figure  4.9 illustrates phrasing (4.58)b3: it repre-
sents a prosodically integrated N-APP with non-nuclear prominence, which
is preceded in its IP (IP2 in (4.66)b) by its anchor and followed by more
material from the syntactic host. IP2 spans a complete clause. The mono-
syllabic verb gets is associated with main prominence; both the appositional
anchor Kuznecov and the N-APP noun substitute have prenuclear accents.
Declination applies across IP2. There is a step-up in pitch on the first
(unaccented) syllable of Kuznecov from the preceding low boundary tone,
and on and at the beginning of IP3 from L% terminating L2; IP2 is fol-
lowed by a break of approximately 295 ms. There is no reason to assume
an IP boundary anywhere within IP2. The utterance is taken from a spon-
taneous, unscripted radio BBC 5 sports (football) commentary. The speech
rate is very high and the prosodic domains are correspondingly large. In this
particular context, at least two players, Kuznecov among them, are chasing
after the ball, which, presumably together with rhythmical factors, explains
the prominence on the verb.
(4.66) N-APP phrased together with anchor and following host material
a. Garcia the blond-haired figure goes chasing after it (ICE-GB:
s2a-014 #140)
Kuznecov the substitute gets there first and Garcia was being
over-aggressive and he mouthed his discontent to the referee
for the decision (ICE-GB: s2a-014 #141)
b. … chasing after it]IP1 IP2[Kuz  necov the  substitute  gets there
L% L+H* !H* !H*
first ]IP2 IP3[and …
L-L%
Similarly to (4.66), the N-APP Mr Daniel in (4.67) is phrased such that it
is preceded and followed in its prosodic domain (IP2 in (4.67)b) by mater-
ial from its syntactic host, the preceding anchor the debtor included. Both
the anchor and Mr Daniel are associated with prenuclear prominence. The
main prominence in IP2 follows the N-APP: it is associated with one. Unlike
164  Data analysis, results and discussion

(4.66), the IP containing the N-APP in (4.67) also includes host material
preceding the appositional anchor.
(4.67) N-APP phrased together with preceding and following host
material
a. The uh background uh to the uh matter which has been argued
a and uh relied on in the Registrar’s judgement uh is a confus-
ing state of cross claims between the debtor Mr Daniel on the
one hand a and Mr Sigrani uh with whom he had dealings on
the other (ICE-GB: s2a-069 #13)
b. … cross claims]IP1 IP2[between the   debtor Mr Daniel on the
H* !H* L-L% H* !H*
one hand]IP2    IP3[…
L+H* L-H%
Overall, like for full parenthetical clauses and NRRCs there is a strong ten-
dency for N-APPs to be phrased separately, even if this tendency is not quite
as strong for this type as for the other two. Including the two items phrased
in a CPD, just over 80 per cent of observed cases were phrased separately (see
(4.58)), and recursive phrasing is the exception rather than the rule. The only
other pattern which reaches some degree of frequency is (4.58)b1, which is
such that the N-APP is followed by an IP boundary but is joined by the pre-
ceding anchor and possibly further host material to form one IP. Of the 65
N-APPs phrased in this way, 54 are of type (4.7), i.e. a proper-noun N-APP
modifies a non-proper N anchor. Unlike other types of parentheticals such
as CCs, RVs and QTs, which if domain-final are often unstressed and fol-
low the nuclear prominence (see Sections 4.3.4 through to 4.3.6 below),
IP-final N-APPs in the present data set always have nuclear prominence (see
(4.61)/Figure 4.8, (4.62) and (4.63) for examples). Furthermore, there are no
unstressed N-APPs in the data set; all integrated N-APPs have either prenu-
clear or nuclear prominence. This fits in with the assumption that they pro-
vide discourse-new information (e.g. Meyer 1992; Potts 2005).
Notice incidentally that of the only 19 N-APPs from the private domain
that originally entered the analysis, a further 9 had to be excluded either
because of the quality of the sound file or because they were right-dislo-
cations rather than appositions. The results reported here may thus not be
representative of this register.

4.3.3.4  Statistical analysis


Given the considerable number of N-APPs in the data set, it was possible
also to perform a statistical analysis.14 The analysis was carried out in order
to test whether semantic and syntactic factors have an effect on phrasing.
In particular, it was tested whether the nature of the anchor, the syntactic
14
I am grateful to Bettina Braun for indispensable help with the statistical analysis and for
discussion.
Analysis and results  165

function of the appositional construction, its syntactic position and the


semantic relation between anchor and host affect whether the N-APP is
phrased separately or together with host material. In addition to their pros-
odic phrasing, the data were therefore coded according to their types as
given in (4.5) through to (4.8) above (4 levels: anchor = proper N; no proper
N; N-APP = proper N; pronominal anchor), according to their syntactic
function, i.e. whether they related to a subject; subject in existential there-
construction, cleft construction or inverted construction; direct object;
object of preposition; predicate; or adjunct (= 6 levels; see (4.68)), accord-
ing to their position within the sentence/utterance (2 levels: final, non-final;
see (4.69)), and according to their semantics (3 levels: equivalence, attribu-
tion, inclusion; see (2.71)–(2.73) in Section 2.3.3 above).
(4.68) N-APPs: syntactic function
a. N-APP relates to subject
Dr Morris the Minister of Glasgow Cathedral is one of these and
very soon in a few days’ time indeed he will become Dean of
the Chapel Royal … (ICE-GB: s2a-020 #114)
b. N-APP relates to subject in existential there-constructions (b1),
clefts (b2) or inverted (b3) constructions
b1. There’s also of course Charter Eighty-Eight a new non-
party constitutional campaign with now twenty-five thou-
sand subscribers which is taking up the whole subject of
citizenship as indeed are a number of institutions such
as Leicester University and many individuals (ICE-GB:
s2a-039 #24)
b2. It was Britain’s ambassador Sir David Hannay who took
on the thankless job of explaining why Washington and
London would have nothing to do with the proposal
from Paris (ICE-GB: s2b-010 #42)
b3. And listening with us here in the studio is our diplomatic
correspondent Paul Reynolds (ICE-GB: s2b-012 #121)
c. N-APP relates to object
I asked Patrick Seal a writer and expert on Arab affairs about the
importance of such a summit if it goes ahead (ICE-GB: s2b-
012 #103)
d. N-APP relates to object of preposition (in non-adjunct
position)
Well tonight I talked to that elder statesman of Mrs
Thatcher’s first two governments Lord Hailsham (ICE-GB:
s2b-017 #79)
e. N-APP relates to predicate in a copula construction
The leader incidentally is Nigel Adams the twenty-nine-year-
old from Swansea who was seventh in the United Kingdom ten
thousand metres championships (ICE-GB: s2a-007 #11)
166  Data analysis, results and discussion

f. N-APP relates to adjunct


The reason that this is such a sacred place is because the two
Marys remained in the Camargue with Sarah the servant and it
is she who has become the most important person for the gyp-
sies (ICE-GB: s2b-027 #115)
(4.69) N-APPs: position
a. Final
There have been six federal presidents since the birth of the
modern federal German state and none has been as popular
as Richard von Weizsäcker the present president (ICE-GB: s2b-
021 #102)
b. Non-final
Lord Whitelaw the influential Tory elder statesman said Douglas
Hurd had great leadership qualities (ICE-GB: s2b-009 #9)
All items were discarded from the statistical analysis whose phrasing was
coded ‘unclear’ (12 N-APPs; see (4.58) and Table 4.3). Of the remaining 454
N-APPs, 4 had to be discarded because their syntactic function or position
was unclear due to hesitant stretches or self-corrections on the part of the
speaker. The dependent variable was prosodic phrasing (2 levels: prosodic
separation, see (4.58)a; prosodic integration, see (4.58)b1: N-APP phrased
together with preceding host material IP[… N-APP]IP). The 12 N-APPs
phrased according to (4.58)b2 and b3, i.e. phrased together with following
or with preceding and following host material, were excluded from the stat-
istical analysis because this phrasing pattern only occurred in non-final pos-
ition. The remaining 438 cases entered the statistical analysis. A binomial
logistic regression model with phrasing as the dependent variable and the
above-mentioned factors as fixed factors (function glm with a binary linking
function, see Baayen 2008: Chapter 6.3.1) was calculated.
Regarding the type of N-APP, there was no difference with respect to
prosodic phrasing between N-APPs with pronominal anchors and any other
type (all p-values > 0.4). However, this result is impossible to interpret and
no generalization is possible due to there being too few observations (10
pronominal cases overall, only one integrated). More data will be needed
to draw safer conclusions about the effect of the pronominal status of the
anchor on the phrasing pattern. The remaining three groups show a clear
pattern such that proper-noun N-APPs behave differently from N-APPs
which are not proper nouns. N-APPs were significantly less frequently
phrased separately (i.e. more integrated) when the N-APP was a proper N
than when the anchor was a proper N (beta = −1.8, SE = 0.5, p < 0.0005),
and when the N-APP was a proper N than when the appositional con-
struction did not involve a proper N (beta = −1.6, SE = 0.5, p < 0.005).
Distribution numbers are given in Table 4.4.
Analysis and results  167

Table 4.4  Cross-tabulation of N-APPs according to phrasing and type

Anchor = proper No proper N-APP = proper Pronominal


Phrasing N (see (4.5)) N (see (4.6)) N (see (4.7)) anchor (see (4.8))

separate 93 75 196 9
integrated 5 5 54 1
ratio 18.6 15 3.6 9

Table 4.5  Cross-tabulation of N-APPs according to phrasing and syntactic function

N-APP N-APP N-APP


N-APP N-APP relating N-APP relating relating to
relating relating to to relating to there-, cleft
to object predicate subject to object adjunct or inverted
(see (see (4.68) (see of P (see (see construction
Phrasing (4.68)c) e) (4.68)a) (4.68)d) (4.68)f) (see (4.68)b)

separate 36 31 196 41 57 12
integrated 3 3 30 9 15 5
ratio 12 10.3 6.5 4.6 3.8 2.4

Regarding the syntactic function, the analysis reveals two main groups
such that (i) N-APPs relating to subjects, N-APPs relating to objects and
N-APPs relating to predicates do not differ significantly from each other
with respect to prosodic phrasing (all p-values > 0.4), and (ii) N-APPs
relating to subjects in existential there-constructions, clefts or inverted con-
structions, N-APPs relating to objects of prepositions and N-APPs relating
to adjuncts do not differ significantly from each other with respect to pros-
odic phrasing (all p-values > 0.2) (see Table 4.5 for ratios). Moreover, there
is an overlap between the two groups such that N-APPs relating to subjects
and those relating to objects of prepositions do not differ from each other
(p > 0.4). All other comparisons resulted in significant effects (all p-values
< 0.05). N-APPs are most often prosodically separate when relating to an
object and most often prosodically integrated when relating to the thematic
subject in an existential there-construction, cleft- or inverted construction.
However, this latter group will not be further considered because of its het-
erogeneous make-up paired with the fact that there were only a very few
occurrences (17 overall).
Regarding the semantic relation between anchor and N-APP, N-APPs
are more frequently integrated when the relation is one of equivalence than
when it is one of attribution or inclusion (see Table 4.6 for distribution num-
bers and ratios). Type inclusion was discarded from the analysis because
they were never integrated. The factor semantics is significant (beta = 2.37,
168  Data analysis, results and discussion

Table 4.6  Cross-tabulation of N-APPs according to phrasing and


semantic relation between anchor and N-APP

Equivalence Attribution Inclusion


Phrasing (see (2.71)) (see (2.72)) (see (2.73))

separate 300 52 21
integrated 64 1 0
ratio 4.7 52 n/a

Table 4.7  Cross-tabulation of N-APPs of semantic type equivalence according to


phrasing and type of N-APP

Equivalence

Pronominal
Phrasing Anchor = proper N No proper N N-APP = proper N anchor

separate 62 46 187 5
integrated 4 5 54 1
ratio 15.5 9.2 3.5 5

Table 4.8  Cross-tabulation of N-APPs according to phrasing and position

Phrasing N-APP: final (see (4.69)a) N-APP: non-final (see (4.69)b)

separate 81 292
integrated 13 52
ratio 6.2 5.6

SE = 1.02, p < 0.05) only when the factor type of N-APP is removed from
the model. This is because both factors affect phrasing in the same direc-
tion. For example, within N-APPs of semantic type equivalence, proper-
noun N-APPs were integrated more often than other types of N-APPs (see
Table 4.7).
Finally, there was no significant effect of the factor position on the pros-
odic phrasing of N-APPs (p > 0.2; see Table 4.8 for figures). There were no
interactions between any of the factors.

4.3.3.5  Remarks and discussion


The hypothesis for N-APPs is prosodic separation as default prosodic
phrasing (see (4.56) and (4.57)). This hypothesis was borne out: 80 per cent
of all 466 prosodically analysed N-APPs were phrased in their own IP (see
(4.58)a/(4.70)a), following (3.35)a (and in two cases (3.35)b). Prosodically
Analysis and results  169

separate N-APPs are in line with Potts’ (2005) and Heringa’s (2011)
accounts, who argue that the syntactic-semantic features of N-APPs (comma
feature or par-merge, respectively) result in prosodic separateness. They are
furthermore in line with Astruc-Aguilera and Nolan’s (2007b) experimental
results. However, the non-recursive structure is not fully in line with Match
Theory, which would predict an ip boundary but no IP boundary terminat-
ing the string preceding the N-APP, unless it is a full clause. In particular,
Match Theory predicts that N-APPs relating to the subject (see (4.68)a) will
be preceded by an ip boundary rather than an IP boundary, but the presence
of an IP boundary is the rule in the present data. The only other prosodic
phrasing pattern that reached considerable frequency was (4.58)b1/(4.70)b,
i.e. the pattern in which the N-APP bears nuclear prominence but is phrased
together with preceding host material, which is or includes the anchor of
the N-APP, i.e. the appositional construction is phrased together in one
IP. This pattern violates Match, but satisfies Align R and Wrap (see (3.35)
c). Together, the two phrasing patterns account for 94 per cent of all ana-
lysed appositional constructions investigated here. Compared with full par-
enthetical clauses and NRRCs, more variation is found at the left edge of
N-APPs.
(4.70) N-APPs: observed prosodic phrasing; most frequent patterns
a. … IP[…]IP IP[N-APP]IP IP[…]IP … 375 N-APPs  (80%)
b. … IP[…]IP IP[… N-APP]IP IP[…]IP …  65 N-APPs (14%)
The statistical analysis shows that the syntactic function of the N-APP and
its syntactic type affect the prosodic phrasing of N-APPs, but the position of
the appositional construction within the utterance does not. First, N-APPs
which are proper nouns were more often phrased together with host material
than N-APPs which are not proper nouns. One possible explanation of this
proper-noun effect is that anchors of proper N appositions are occasionally
interpreted as premodifiers even if preceded by a determiner. As explained
in Section 2.3.3, cases of premodifiers resembling titles (see (4.71), repeated
from Section 2.3.3) were excluded from the analysis; they were identified by
the lack of a determiner in the anchor and by the fact that omission of the
N-APP resulted in ungrammaticality (see (4.71)b), while omission of the
apparent anchor did not (see (4.71)c). Adding a determiner to the apparent
anchor will rescue (4.71)b (see (4.71)d). With regard to prosody, the idea is
that a premodifier would be phrased together with the proper-noun N-APP,
while an anchor in an appositional construction would not, reflecting the
semantic difference between the two interpretations. If the NP preceding
the proper noun is interpreted as a premodifier even in the presence of a
determiner, this may therefore result in prosodic integration. This line of
argumentation would account for example (4.61) above, repeated as (4.72)
below; more examples are given in (4.73) (all N-APPs phrased together with
their anchor).
170  Data analysis, results and discussion

(4.71) Partial apposition; premodifier resembling title


a. British explorer Sir Ranulph Fiennes reached both Poles with
support in a transglobe expedition of nineteen eighty-one
(ICE-GB: s2b-024 #8)
b. *British explorer reached both Poles with support in a trans-
globe expedition of nineteen eighty-one.
c. Sir Ranulph Fiennes reached both Poles with support in a
transglobe expedition of nineteen eighty-one.
d. The British explorer reached both Poles with support in a
transglobe expedition of nineteen eighty-one.
(4.72) Proper-noun N-APP phrased together with anchor
a. Last month the Defence Secretary Mr Cheney said it was
conceivable that a further one hundred thousand American
soldiers could be sent to the region and it’s believed that the
reserve units would be part of that additional force (ICE-GB:
s2b-017 #18)
b. IP1[Last month]IP1 IP2[the Defence Secretary Mr Cheney]IP2
IP3[said it was conceivable]IP3 IP4[that …

(4.73) Proper-noun N-APP phrased together with anchor


a. The Labour leader Neil Kinnock has been in Westminster this
morning and he gave his reaction to our political correspond-
ent Stephen Richards (ICE-GB: s2b-018 #80)
b. Here the Environment Secretary Michael Heseltine has said
that Britain will play a full part in dealing with the spillage in
the Gulf which he described as an abomination (ICE-GB: s2b-
018 #31)
c. The Prime Minister John Major speaking to reporters out-
side Chequers this morning said it was a massive encounter
which would not be finished until the Iraqis were driven out of
Kuwait (ICE-GB: s2b-014 #15)
Notice also that the information conveyed by a proper-noun N-APP does
not have to be hearer-new in the sense that it contributes unknown infor-
mation about the anchor. It is (or was at the time) common knowledge, for
example, that the leader of the Labour Party was Neil Kinnock (see (4.73)a)
and that the Prime Minister’s name was John Major (see (4.73)c). However,
they may still be discourse-new.
Another possible explanation that might come to mind is that N-APPs
phrased together with their anchors may have been interpreted as seman-
tically restrictive (and therefore, in Potts’ 2005 account, not marked by
the comma feature). For example, Isabelle in (4.74) may be one of several
daughters of Philippa. However, I do not see any reason to assume that
this should be more often the case with proper-noun N-APPs than with
Analysis and results  171

other types of N-APPs, and a look at the data does not provide any sup-
port for this assumption either, mostly because a restrictive interpretation
was unlikely for most of the proper-noun N-APPs phrased in this way. For
example, there is only one Environment Secretary, and there is only one
Prime Minister (see (4.73)b and c, respectively).
(4.74) Proper-noun N-APP phrased together with anchor
Philippa’s daughter Isabelle married Philip the Good of Burgundy
in fourteen twenty-nine … (ICE-GB: s2b-043 #39)
Second, the syntactic function of the appositional construction affected
prosodic phrasing such that the analysis revealed two main groups: group
1 comprises N-APPs relating to subjects, N-APPs relating to objects, and
N-APPs relating to predicates; the second group is made up of N-APPs
relating to subjects in existential there-constructions, clefts or inverted con-
structions, N-APPs relating to objects of prepositions, and N-APPs relating
to adjuncts. Moreover, there was an overlap between the two groups such
that N-APPs relating to subjects and those relating to objects of prepositions
did not differ from each other. Members of the first group were more fre-
quently prosodically separate than members of the second group; N-APPs
were most often prosodically separate when relating to an object. At present,
I do not have an explanation for this syntactic effect.
Third, the analysis further shows that the semantic relation between
anchor and N-APP has an effect on prosodic phrasing such that N-APPs
of type equivalence were integrated more often than N-APPs of type attri-
bution and inclusion. Within equivalence, more proper-noun N-APPs were
integrated than other types, repeating the proper-noun effect. The fact that
the factors semantics and type of N-APP go in the same direction with
respect to prosodic phrasing makes sense if the N-APP is the name of the
referent of the anchor (e.g. (4.61)/(4.72), (4.73) and (4.74)). In this case, the
relation between the two is one of equivalence and identification. Proper-
noun N-APPs occur in other semantic types, as well (see Quirk et al.’s 1985
example for inclusion, repeated in (4.75) from Section 2.3.3), but presum-
ably not as frequently. The two factors would best be separated in a con-
trolled experiment.
(4.75) Inclusion (Quirk et al. 1985: 1301)
A neighbour, Fred Brick, is on the telephone.
Another possible factor affecting phrasing, briefly touched upon above but
not tested statistically, is the news value of the position such that N-APPs
which contribute new, important information are associated with nuclear
prominence and phrased separately, while N-APPs which contribute
generally known information are less likely to have nuclear prominence
and therefore do not lend themselves to separate IPs. Given that it is the
primary function of N-APPs to provide new information about a referent
172  Data analysis, results and discussion

in the discourse (see, e.g. Meyer 1992; Potts 2005 and Section 2.3.3 above),
prominence is the default. This corresponds to the results reported on in
the previous sections: there are no unstressed N-APPs in the data set; how-
ever, there are N-APPs which are not discourse-new. It seems that N-APPs
contributing discourse-old information may nevertheless be prominent
(at least receive non-nuclear prominence) due to either rhythmical or dis-
course-related reasons. For example, in (4.66) above, the fact that Kuznecov
is the substitute is known from the previous commentary (it was mentioned
by the same speaker when the player was allowed on to the pitch), but the
N-APP still receives prenuclear prominence. One reason is that in the given
discourse situation, it may be necessary to repeat the information that the
respective player is a substitute, perhaps to remind the listeners; it is there-
fore relevant to the narrative in the given context. Another may be that sub-
stitute is made prosodically strong in order not to allow too many unstressed
syllables between two stressed ones. A systematic analysis of the contexts in
which N-APPs have non-nuclear prominence will have to be left to future
research.
Recall from Section 2.3.3 that Meyer (1992: 46f) attributes a difference
in interpretation to the presence vs. absence of a comma and corresponding
tone group boundary in examples such as (4.76), such that in (4.76)a, i.e.
without a comma and boundary after Holtom, A Labour stalwart is inter-
preted as an adverbial with a causal relation to the main proposition, while
the presence of the comma and corresponding boundary in (4.76)b leads
to interpretation of A Labour stalwart as the first unit in an appositional
construction.
(4.76) Phrasing of N-APP: effect on interpretation
a. A Labour stalwart, Cllr Holtom slated the council’s previous
Tory administration for causing “untold misery to the least
fortunate”
b. A Labour stalwart, Cllr Holtom, slated the council’s previous
Tory administration for causing “untold misery to the least
fortunate”
The present data reveal that the phrasing pattern assumed for (4.76)a by
Meyer (1992) occurs with N-APPs, too, even if it is extremely infrequent
(less than 1% of the analysed data set). The relevant prosodic pattern corres-
ponds to (4.58)b2, i.e. phrasing of a NP labelled N-APP with following host
material. This suggests that while prosodic phrasing may serve the func-
tion of disambiguation between meanings, as illustrated in (4.76), the lack
of a prosodic boundary after the second NP does not exclude the possibility
that this NP functions as a true N-APP. For example, the sentence given in
(4.77)a, phrased as in (4.77)b (repeated from (4.65) for convenience), cannot
be interpreted in the way suggested by Meyer (1992) for (4.76). More specif-
ically, The Northern Ireland Secretary is not best interpreted as functioning
Analysis and results  173

as an adverbial as paraphrased in (4.77)c. The example is taken from a spon-


taneous BBC 4 radio commentary entitled The Gulf Ceremony. The com-
mentator lists the public figures present at the event. The Northern Ireland
Secretary is there as one among many others, but there is no direct causal
relation between him being the Northern Ireland Secretary and being pre-
sent at the event. Instead, the semantic relation between anchor and N-APP
is one of equivalence and identification. It is nevertheless conceivable that
the difference in phrasing suggested by Meyer (1992) and illustrated in
(4.76) is used when there is a need for semantic disambiguation, an assump-
tion that would perhaps best be tested experimentally. Another possible
explanation to account for (4.77) is the length or balanced size of the two
adjacent IPs. With the anchor Northern Ireland Secretary (8 syllables) being
phrased in one IP, separate phrasing of the N-APP Peter Brooke (3 sylla-
bles) would have resulted in a considerably shorter IP. The IP boundary to
the right of Brooke may therefore be deleted, resulting in a series of IPs
of more balanced size. Moreover, separate phrasing of Peter Brooke would
presumably also result in the following material is there being phrased in a
separate IP. This would, in turn, have required nuclear prominence on the
string is there (see (4.77)d), which would have been infelicitous in the given
discourse context, because is there refers to discourse-old material and the
more important part of the discourse is the listing of the names of people
present on the occasion. Thus discourse factors as well as length require-
ments account for the prosodic phrasing of (4.77)b.
(4.77) a. The Northern Ireland Secretary Peter Brooke is there and other
defence ministers Archy Hamilton the Earl of Arran Kenneth
Carlisle (ICE-GB: s2a-019 #14–15)
b. IP1[Northern Ireland Secretary]IP1 IP2[Peter Brooke is there]IP2
IP3[and …
c. # Because he is the Northern Ireland Secretary Peter Brooke is
there …
d. # IP1[Northern Ireland Secretary]IP1 IP2[Peter Brooke]IP2 IP3[is
there]IP3 IP4[and …
Meyer (1992: 35)  also observes that appositions “had functions associated
with positions … that promoted the principle of end-weight”. Since appos-
itional constructions, he argues, are typically heavy, they are less likely to
occur in subject position than in positions such as direct object, object of
preposition and as subject in sentences containing existential there. In his
corpus, 65 per cent of all appositions had functions like these. However, 31
per cent still occurred in preverbal subject position. He reports that 49 per
cent of these occurred “in the journalistic style, a style containing a high
number of relatively short and non-complex appositions” such as the two
instances in (4.78) (Meyer 1992: 36). Meyer (1992) follows Quirk et  al.
(1985: 1361–2) in their notion of end-weight, but the principle goes back at
174  Data analysis, results and discussion

least to Behaghel’s (1909, 1930) Gesetz der wachsenden Glieder (‘Law of the
Growing Elements’), which states that of two constituents of different size,
the larger one follows the smaller one. However, both Behaghel’s law and
Quirk et al.’s principle of end-weight make reference not only to the length,
complexity and weight of the relevant constituents, but also to their news
value (Quirk et al.’s 1985: 1362 principle of end-focus). It is the information
structure of the complete utterance that affects the positioning of individ-
ual constituents rather than weight alone. Consider the example in (4.79),
taken from a broadcast scripted talk. The appositional construction, a long,
syntactically complex and prosodically heavy anchor followed by a similarly
long, complex and heavy N-APP, functions as the subject of the sentence.
Given that it is a passive sentence, it would have been possible syntactic-
ally to follow the principle of end-weight and place the appositional con-
struction at the end of the utterance. However, this is not what the speaker
did, presumably because both the referent of the subject and the informa-
tion conveyed in the VP are new information to the hearer and emphasis has
been put on the fact that the referent of the subject was removed from his
job. At the same time, there are prosodic means to deal with the complexity
and weight of the subject; the phrasing in this case is such that the complex
subject spans five IPs (see (4.79)b).
In the present data set, N-APPs in subject position are in fact very fre-
quent. Of the 454 N-APPs whose prosodic phrasing was analysed (exclud-
ing the ones whose phrasing was unclear), 237 (52%) N-APPs were in
subject position (examples can be found in (4.59), (4.61), (4.63), (4.65),
(4.66) and (4.68) above). This distribution may be accounted for along the
lines of information-structure categories. The subject position is often a
topic position (e.g. Gundel 1988; Lambrecht 1994), introducing the ref-
erent about which the further utterance makes a comment. The N-APP is
used to provide additional information about this referent in adjacent pos-
ition and preceding the comment. Most of the subject occurrences (126 of
238)  were found in news broadcasts (s2b-001–s2b-020), a genre in which
the topic of a news item is typically introduced in initial position as subject
(e.g. Moya Guijarro 2005 for written news). Rather than associating the pos-
ition of appositional constructions with their weight, it may be more fruitful
to relate them to the information structure of the context or the discourse
type. A systematic analysis will have to be left to future research.
(4.78) Comic Gary Morton signed to play the Living Room here Dec. 18,
because that’s the only time his heart, Lucille Ball, can come along.
(4.79) a. The one Community official who did try to compete with the
Washington machine before the world’s media the first and pos-
sibly last Commission press spokesman to engage in the politics of
the twenty second soundbite has since been removed from his job
(ICE-GB: s2b-040 #104)
Analysis and results  175

b. IP1[The one Community official]IP1 IP2[who did try to compete


with the Washington machine]IP2 IP3[before the world’s media]IP3
IP4[the first and possibly last Commission press spokesman]IP4 IP5[to
engage in the politics of the twenty second soundbite]IP5 IP6[has since
been removed from his job]IP6
NRRCs and N-APPs have often been assumed to be similar in form and
function, which has been accounted for, for example, by the assumption that
N-APPs are reduced relative clauses; at the same time, there are some dif-
ferences between the two types (see Heringa 2011: 14–16 for a short review
and for references). Based on the present results for NRRCs and N-APPs, a
further similarity between the two is that both types are prosodically separate
by default; however, adding to the differences, NRRCs follow the pattern of
prosodic separation even more systematically than N-APPs. It is conceivable
that this is due to factors other than the parenthetical (+comma) status of the
NRRC/N-APP. One such factor may be the length of the N-APP, which is
often shorter than a full clause, and may therefore be more readily phrased
together with host material. The statistical effect regarding the syntactic type
of the N-APP may also be relevant here: N-APPs which were proper nouns
were phrased together with host material more often than other N-APPs.
Proper nouns are less easily extended to full clauses (compare: Mrs Smith,
(who is) my neighbour, works at the ministry. vs. My neighbour, (*who is) Mrs
Smith, works at the ministry.). Moreover, anchors will not as readily be inter-
preted as premodifiers of NRRCs as they may be interpreted as premodi-
fiers of a proper N or a title preceding a proper N. The semantic relation
of equivalence, another factor leading to more prosodic integration with the
anchor, may also be more frequent with N-APPs than NRRCs.

4.3.4  Comment clauses


For CCs, the predictions based on previous research were more varied than
for full parentheticals, NRRCs and N-APPs. They are repeated in (4.80)/
(4.81) from Section 2.3.4 above. The results, too, are much more varied than
those for the previous three groups. The intonational phrasings observed
with all 147 CCs analysed are summarized in (4.82). The phrasing of five
CCs remained unclear. Table 4.9 lists the results separately for the most fre-
quent type I think, and all remaining types of CC (see (4.12) above). As will
be outlined below, the phrasings in (4.82)b each correspond to more than one
intonational pattern. Unlike the three types of parentheticals discussed in the
previous sections, CCs may be unstressed. It is clear immediately from the
summary in (4.82) that CCs do not “generally form a separate tone unit”, as
has been claimed by Brinton (2008: 5). Rather, the phrasing results corres-
pond to the variation predicted in Chapter 2 above such that CCs may be pro-
sodically separate or integrated and they may be prominent or unstressed.
176  Data analysis, results and discussion

Table 4.9  Patterns of prosodic phrasing of comment clauses (CCs)

CCs other than


Phrasing CC I think I think Overall

a. a1. IP[…]IP IP[CC]IP IP[…]IP 7 (10%) 20 (27%) 27 (18%)


a2. IP[… IP[CC (…)]IP …]IP 2 (3%) 6 (8%) 8 (5%)
b. b1. IP[…]IP IP[… CC]IP 23 (32%) 22 (29%) 45 (31%)
IP[…]IP
b2. IP[…]IP IP[CC …]IP 5 (7%) 4 (5%) 9 (6%)
IP[…]IP
b3. IP[…]IP IP[… CC …]IP 35 (48%) 18 (24%) 53 (36%)
IP[…]IP
c. phrasing unclear 0 5 (7%) 5 (4%)
72 (100%) 75 (100%) 147 (100%)

(4.80) Prosodic phrasing of CCs: hypotheses


a. As parentheticals, CCs may be prominent and phrased separ-
ately (e.g. Nespor and Vogel 1986; see (4.81)a).
b. CCs may be prosodically integrated (see (4.81)b) as prenuclear
(e.g. (4.81)b2, b3) or postnuclear (e.g. (4.81)b1, b3) material
(e.g. Crystal 1969; Bing 1985; Gussenhoven 1984).
c. If phrased together with host material (i.e. phrased according
to (4.81)b), CCs may be unstressed or they may be prominent
(Dehé and Wichmann 2010a).
(4.81) CCs: predicted prosodic phrasing
a. prosodic separation:
(…) IP[…]IP IP[CC]IP (IP[…]IP) (…)
b. prosodic integration:
b1. (…) (IP[…]IP) IP[… CC]IP (IP[…]IP) (…)
b2. (…) IP[…]IP IP[CC …]IP (IP[…]IP) (…)
b3. (…) (IP[…]IP) IP[… CC …]IP (IP[…]IP) (…)
(4.82) Observed patterns of prosodic phrasing of comment clauses (CCs)
a. CC: prosodic separation
a1.  … IP[…]IP IP[CC]IP IP[…]IP … 27 CCs (18%)
a2. … IP[… IP[CC]IP …]IP … 8 CCs (5%)
b. CC: prosodic integration; CC phrased together with preceding
and/or following host material
b1.  … IP[…]IP IP[… CC]IP IP[…]IP … 45 CCs (31%)
b2. … IP[…]IP IP[CC …]IP IP[…]IP … 9 CCs (6%)
b3. … IP[…]IP IP[… CC …]IP IP[…]IP … 53 CCs (36%)
c. phrasing unclear 5 CCs (4%)
Total: 147 CCs
Analysis and results  177

4.3.4.1  Pattern (4.82)a: prosodic separation (CC)


Of the 147 CCs that entered the analysis, only 27 (18%) were phrased sep-
arately in a non-recursive structure, following pattern (4.82)a1. They have
their own CTC with a nuclear accent either on the pronoun (6 CCs) or the
predicate (21 CCs), and an IP boundary before and after the CC. The pre-
ceding and following host material is also phrased in separate domains. An
example is given in (4.83)/Figure  4.10. IP1, i.e. the IP preceding the CC
has final nuclear prominence associated with effectively; it is terminated
by L-H%. The boundary between IP1 and IP2 is marked by a pitch dis-
continuity and break (but no audible pause) before the pronoun I; a glottal
stop separates the final vowel of effectively and the pronoun I. The pronoun
bears the nuclear accent of the IP spanning I believe (IP2 in (4.83)b), which
has the same combination of pitch accent and boundary tone as IP1 (L+H*
L-H%). I believe is followed by a pause of > 500 ms, which, together with
H%, marks the boundary between IP2 and IP3. IP3 has a prenuclear peak
associated with the first syllable of underestimating and a nuclear L+H*
L-L% associated with scale. IP4 shows a similar pattern. The pitch is lower
in the IP spanning the CC than in the other IPs, but the CC clearly has a
nuclear accent and is separate.
(4.83) CC: prosodic separation
a. But we have lost a tremendous amount of time time in which
the Western powers simply failed to respond time that went on
through the whole of nineteen ninety and nineteen ninety-one
and where we are still effectively I believe underestimating the
scale of what is required for us (ICE-GB: s2b-047 #30)
b. … IP1[and where we are  still ef  fectively]IP1 
H* L+H* L-H%
IP2[I believe]IP2 IP3[underestimating the scale  ]IP3
H*  L-H% H* L+H* L-L%
IP4[of what is required for   us]IP4
L+H* !H* L-L%
Eight CCs (5%) were phrased in a CPD pattern, i.e. a recursive prosodic
structure (see (4.82)a2). An example is given in (4.84)/Figure 4.11. The
host clause this would be my fourth year is the domain of one CTC with a
prenuclear H* associated with be and a nuclear L*+H H-H% at the end
of the domain. The CC I believe, which is joined in its IP by the repeated
possessive pronoun my, has its own CTC with a nuclear L+H* accent
associated with the verb and terminating in L-L%. This does not affect
the contour of the host clause. Discourse-functionally, the CC is inserted
into a hesitant phase signalling the speaker’s uncertainty about the num-
ber of years, which also explains the lengthening of the first occurrence
of my.
330
Pitch (Hz)

200

100

20

H* L+H* L-H% H* L-H% H* L+H* L-L%

CC

and where we are still effectively ? I believe <SIL> underestimating the scale <SIL>

0 5.016
Time (s)

Figure 4.10  Example (4.83): CC, prosodic separation


250

200
Pitch (Hz)

150

100

50
H* !H* L-H% H* L+H* L-L% L*-H H-H%

CC

for the same bike this would be my I belive my <SIL> fourth year

0 4.195
Time (s)

Figure 4.11  Example (4.84): CC phrased in a CPD


180  Data analysis, results and discussion

(4.84) CC phrased in a CPD


a. For the same bike this would be my I believe my fourth year
(ICE-GB: s1b-074 #221)
b. IP1[For the  same   bike]IP1  IP2[this would  be my IP3[I
H* !H* L-H% H*
believe my]IP3 fourth year]IP2
L+H* L-L%  L*+H H-H%

4.3.4.2 Pattern (4.82)b1: CC phrased together with preceding host material


(… IP[…]IP IP[… CC]IP IP[…]IP …)
The remaining 107 CCs (73%) each follow one of the three phrasing pat-
terns in (4.82)b. These CCs are phrased together with host material, regard-
less of whether or not they are prominent. Very few of these CCs have main
prominence in their domains. If they are prominent at all, they have prenu-
clear prominence and are followed by the nuclear material in their domain.
Otherwise they are unstressed and either precede or follow the prominent
material in their domain. In (4.82)b1, the CC is phrased such that it is pre-
ceded but not followed in its IP by host material. The CC may either have
nuclear prominence and be preceded by unstressed or prenuclear material
from the host clause, or it may be unstressed. The CC in (4.85)/Figure 4.12
is an example of the former, i.e. with nuclear prominence associated with the
verb in the CC. The IP spanning which produces I would say (IP2 in (4.85)b)
is set off from the preceding and following IPs by structure-related pauses
of 0.45 and 0.34 seconds, respectively. IP1 has a nuclear peak associated with
discretion and is terminated by L-L%. There is no reason to assume another
IP boundary anywhere within the target IP; in particular, there is no reason
to assume a boundary between produces and the pronoun I of the CC, i.e. at
the left edge of the CC.
(4.85) CC phrased according to (4.82)b1, realized with nuclear
prominence
a. The restructured and reformed training systems which seek to
inculcate at an early stage the confidence and the exercise of
discretion which produces I would say a more thoughtful and
caring officer attuned to the needs of his changing community
(ICE-GB: s2b-031 #75)
b. IP[…]IP IP1[… the confidence and the   exercise of
L*H L*+!H
discretion]IP1 IP2[which produces I would  say]IP2 
!H*  L-L% L+H* !H* L-L%
IP3[a more thoughtful …
H*
200

150
Pitch (Hz)

100

50
!H* L-L% L+H* !H* L-L%

CC

discretion <SIL> which produces I would say <SIL>

0 2.879
Time (s)

Figure 4.12  Example (4.85): CC with nuclear prominence, phrased together with preceding host material
182  Data analysis, results and discussion

The examples in (4.86) and (4.87) also represent phrasing pattern (4.82)b1,
i.e. the CC is preceded in its IP by material from the host, but here the CC
is postnuclear and unstressed; preceding host material is associated with
the nuclear accent. In (4.86), for example, the nuclear accent is a down-
stepped L+H* associated with the first syllable of motorway, followed by
a transition towards L-H% terminating IP1; the CC I think is part of this
transition. This is the pattern referred to by Gussenhoven (2004: 291) as
incorporation. In (4.87), the CC is also unstressed and follows the nuclear
accent in its IP, but it is also preceded by a downward pitch movement to a
very low level in the speaker’s range and perceived final lengthening, indi-
cating an IP boundary (L%). This pattern represents Bing’s (1985) 0 con-
tour and has been analysed as encliticization by Gussenhoven (2004: 291).
The encliticized material, i.e. the CC I don’t think, receives a copy of the
tones that directly follow the last stressed syllable on; since the preceding
tonal targets are the low edge tones L-L%, these are copied onto I don’t
think. Encliticization of this kind was very infrequent in the data and was
therefore combined with incorporation in (4.82)b1 and Table 4.9b1.15
(4.86) CC phrased according to (4.82)b1, realized as unstressed, postnu-
clear material
a. But we managed to find somewhere on the motorway I think at
this service station (ICE-GB: s1a-021 #45)
b. IP1[But we managed to  find somewhere on the motorway
L+H* L+!H* L+H*
I think]IP1 IP2[at this  service station]IP2
L-H%
(4.87) CC encliticized
a. I wouldn’t put anything on it I don’t think because two feet isn’t
very high (ICE-GB: s1b-025 #12)
b. IP1[IP1[I wouldn’t put anything on  it]IP1   I don’t think]IP1
L+H* !H* L- L%  L      L%
IP2[because …

4.3.4.3 Pattern (4.82)b2: CC phrased together with following host material


(… IP[…]IP IP[CC …]IP IP[…]IP …)
In (4.82)b2, the CC is followed but not preceded in its IP by material from
the syntactic host. Only nine CCs (6%) were phrased in this way. Again, the

15
Note that the CC I don’t think in (4.87) was not discarded from the analysis despite being pre-
ceded by a full clause. This is because I don’t think as a main clause would be ungrammatical
under the intended reading, and a movement analysis is therefore implausible. See (x):
(x)  a.  *I don’t think I wouldn’t put anything on it because …
b. I don’t think I would put anything on it because …
Analysis and results  183

CC may in principle have nuclear prominence, followed by unstressed host


material, or it may be unstressed preceding any prominent material, or it
may bear a prenuclear accent. An example of an unstressed CC in this pos-
ition is given in (4.88)/Figure 4.13. IP1 terminates with L-L%, followed by
a pause of almost a second. (Note that the pitch level at L% is in fact lower
than it appears from the contour in Figure 4.13. The continued fall is invis-
ible in the plotted contour due to the overlapping noise.) The CC I think at
the beginning of IP2 is unstressed and reduced phonetically to think. On
think, there is a step-up in pitch from the level of the preceding L%. The
apparent pitch fall on think is due to the consonantal influence on the F0
contour and background noise.

(4.88) CC unstressed and phrased together with following host material


a. But my friend got it I think about twelve years ago (ICE-GB:
s1a-071 #90)
b. IP1[But my friend got it]IP1 IP2[I think about   twelve years
H* L-L% H*
ago]IP2
  !H* L-L%

4.3.4.4 Pattern (4.82)b3: CC phrased together with preceding and following


host material (… IP[…]IP IP[… CC …]IP IP[…]IP …)
In (4.82)b3 the CC is neither directly preceded nor directly followed by
an IP boundary but is phrased together with preceding and following host
material. Fifty-three CCs (36%) were phrased in this way. The CC may be
unstressed or it may have prenuclear or nuclear prominence. If unstressed,
it may also be part of a hesitation stretch inside an IP. Two examples are
given in (4.89)/Figure 4.14 and (4.90)/Figure 4.15. In (4.89)/Figure 4.14,
the CC I think in IP1 is part of the unstressed transition between two pre-
nuclear pitch accents associated with Arabs and little, respectively, preceding
the nuclear accent associated with slow. Declination applies across IP1. IP1
ends in a continuation rise. The following IP starts after the pause, which,
along with H% and a step-down in pitch on with at the beginning of IP2,
cues the IP boundary between IP1 and IP2.

(4.89) CC phrased together with preceding and following host material,


unstressed
a. the Arabs have I think been a little bit slow with the sole
exception of Syria of President Assad of Syria (ICE-GB: s2b-
012 #109)
b. IP1[the   Arabs have I think been a   little bit slow]IP1
H* L+!H*  !H* L-H%
IP2[with the sole exception of Syria]IP2 IP3[…
300

250
Pitch (Hz)

200

150

100
H* L-L% H* !H* L-L%

CC

but my friend got it <SIL> think about twelve years ago

<noise>

0 Time (s) 2.845

Figure 4.13  Example (4.88): CC unstressed and phrased together with following host material
250

200
Pitch (Hz)

150

100

50
H* L+!H* !H* L-H%

cc

well <SIL> the Arabs have I think been a little bit slow <SIL>

0 Time (s) 3.909

Figure 4.14  Example (4.89): CC phrased together with preceding and following host material, unstressed
186  Data analysis, results and discussion

The CC I suppose in (4.90)/Figure 4.15 is part of a hesitation stretch (see


also Dehé and Wichmann 2010a: 14f). The whole utterance is phrased in
one IP with a performance-related, hesitational pause between in and under-
taking. Specifically, we find a sequence of filled pause (uhm), CC (I suppose),
and then a silent pause of approximately 670 ms. The CC is part of this hesi-
tation stretch. The pitch on uhm I suppose is maintained at the level reached
on in and changes only after the transitional phrase.16
(4.90) CC phrased together with preceding and following host material,
part of hesitant stretch
a. There’s no point in uhm I suppose undertaking experiments if
people can’t read them (ICE-GB: s1a-059 #286)
b. IP1[There’s   no   point in uhm I suppose undertaking
H* !H*
experiments if people can’t read them]IP1
L+!H* !H* L-L%
The logically possible pattern in which the CC phrases as in (4.82)b3 (pre-
ceded and followed by material from the host clause in the same IP) but
with nuclear prominence on the CC was not observed in the data.

4.3.4.5  Phrasing of as it were


The results for the 36 tokens of CC as it were that entered the analysis are
summarized in (4.91).
(4.91) Patterns of prosodic phrasing of CC as it were
a. CC: prosodic separation
… IP[…]IP IP[CC]IP IP[…]IP … 1 CC  (3%)
b. CC: prosodic integration; CC phrased together with preceding
and/or following host material
b1. … IP[…]IP IP[… CC]IP IP[…]IP … 17 CCs (47%)
b2. … IP[…]IP IP[CC …]IP IP[…]IP … 1 CC (3%)
b3. … IP[…]IP IP[… CC …]IP IP[…]IP … 17 CCs (47%)
Total: 36 CCs
Only one as it were was phrased according to (4.91)a, i.e. in a separate domain
with nuclear prominence and a CTC. All other cases of as it were were pro-
sodically integrated with material from the host. Seventeen tokens of as it
were were phrased as in (4.91)b1. They were final und unstressed in their
domain, following nuclear host material, either terminated by H% (5 cases)
or L% (12 cases). One as it were was phrased as in (4.91)b2. It was realized
as unstressed material in initial position in its IP, followed by host material

16
Note that there is no audible pitch movement on the second syllable of suppose. What might
look like pitch movement on the second syllable (see Figure 4.15) is due to microprosodic
effects, specifically the plosive syllable onset.
150
140

120
Pitch (Hz)

100

80

60
50
H* !H* L+!H* !H* L-L%

CC

There’s no point in uhm I suppose <SIL> undertaking experiments if people can’t read them

0 6.005
Time (s)

Figure 4.15  Example (4.90): CC phrased together with preceding and following host material, part of hesitant stretch
188  Data analysis, results and discussion

carrying the nuclear accent. Of the 17 cases of as it were which were pre-
ceded and followed in their domain by host material (i.e. phrasing as in (4.91)
b3), one followed the nuclear material, while 16 preceded it, i.e. no CC had
nuclear prominence. Two of the CCs phrased in this way were part of a hesi-
tant, word-searching stretch. An example is given in (4.92). In this case, as
it were is preceded by a stretch of silent (indicated in (4.92)) and filled (uh)
pauses and repetition of they, and followed by the word carrying the nuclear
accent (gain). The contour associated with gain is H*L-H% and it is length-
ened in accordance with nuclear prominence and domain-final position.
(4.92) And I would say to the honourable lady uh as I’ve said to my hon-
ourable friend that I hope uh later in my speech when I’ll give some
examples of what might be the effect on uh uh some representative
smaller employees with a a a representative incidence of sickness
and then looking at what they uh they [HESITANT PAUSE > 350
ms] would uh uh uh as it were gain in the reduction of the third
National Insurance contributions for employers which have been
specially focused on those lower earning ba earnings bands below
about a hundred and seventy-five or a hundred and eighty-five
pounds a week (ICE-GB: s1b-058 #010)

4.3.4.6  Remarks and discussion


Overall, these examples serve to illustrate the variety of phrasing and inton-
ational patterns found with CCs. What is important in the present context is
that the majority of 1st person CCs (73%) and 35 out of 36 cases of as it were
failed to be phrased in their own separate IP. This result is in contrast to the
findings for the three types of parentheticals reported on in the previous
sections (i.e. FPCs, NRRCs and N-APPs), for which prosodic separation
was the default. However, the results are similar to those for reporting verbs
(RVs; see Section 4.3.5 below). Table 4.9 suggests that there is overall less
separation (patterns a1 and a2) and more integration (patterns b1–3) with
I think than with other CC types. A statistical analysis taking into account
verb type as a factor would need more data points to be meaningful. The
current figures suggest an interaction between phrasing and verb type such
that integration is more frequent with think than other types. In particu-
lar, phrasing of the CC together with preceding and following host material
(pattern (4.82)b3) is considerably more frequent with I think than with other
CC types, while prosodic separation (pattern (4.82)a) is considerably less
frequent with I think than with other CC types. The pattern is reminiscent
of what Dehé and Wichmann (2010b) observe for sentence-initial I think
(that) vs. I believe (that), where, in their data set, I think is unstressed and
integrated more often than I believe. They argue that this finding reflects
the semantic properties of I think as well as its relative frequency. However,
Analysis and results  189

even if verb type does have an effect, only 35 per cent of CCs involving
predicates other than think were phrased according to prosodic separation,
which is still in contrast with some claims in the literature.
Given that overall prosodic separation was observed in 23 per cent of all
cases, hypothesis (4.80)a, stating that CCs may be prominent and phrased
separately following patterns (3.35)a/b, is confirmed. However, prosodic
integration (see hypothesis (4.80)b) is more frequent. A considerable num-
ber of CCs (31% overall) phrase together with preceding host material, fol-
lowing pattern (4.81)b1, satisfying Align R and Wrap, but violating Match.
The remaining two integrated patterns, corresponding to (4.81)b2/b3,
account for 42 per cent of all observed cases of CCs, and 55 per cent of
CC I think. These patterns violate Match and Align R, serving as evidence
for other factors being at play. The observed integrated patterns confirm
all aspects of hypothesis (4.80)b. Hypothesis (4.80)c is also confirmed: inte-
grated CCs may either have non-nuclear or nuclear prominence (e.g. (4.85)),
or they may be unstressed (e.g. (4.86)–(4.90)).
Given that clause-final CCs were not considered in this study (see
Section  2.3.4 above), CCs phrased according to (4.81)a/(4.82)a1 are not
fully in line with Match Theory. The IP boundary terminating the string
preceding the parenthetical must be the result of the promotion of a lower-
level boundary, sometimes a prosodic word boundary, to an IP boundary.
For example, Match Theory would not predict an IP boundary between
effectively and I in (4.83).
One important factor affecting the prosodic phrasing of a CC is its mean-
ing and contribution to the discourse, which is reflected in prosody. The
meaning of CCs ranges from propositional to formulaic, i.e. from seman-
tic transparency (e.g. expression of genuine uncertainty, doubt, etc.) to dis-
coursal, interactional and interpersonal meaning (e.g. politeness, mitigation,
narrative cohesion) and the marking of phases of disfluency and hesitation
(see Section 2.3.4 above). These shades of meaning are reflected in the pros-
ody such that prosodic separation goes along with semantic transparency,
and prosodic integration and deaccentuation goes along with semantic
bleaching (Dehé and Wichmann 2010a). The prosodic phrasing of CCs may
also be seen in relation to their CI semantics. Remember from Section 2.3.4
Scheffler’s (2009) suggestion that the contribution of the CC is twofold: (i)
it affects the epistemic threshold, and (ii) it contributes a side comment in
the domain of conventional implicatures: the actual content of the CC-verb.
The latter, i.e. the CI semantics, is related by Potts (2005) to prosodic separ-
ation. However, the first, i.e. the effect on the epistemic threshold, Scheffler
(2009) argues, is the main semantic contribution to the utterance as a whole,
while the side comment is the secondary one. On the other hand, the pre-
sent data suggest that it is conceivable that according to context and dis-
course function, the actual content of the verb, e.g. the expression of genuine
speaker attitude, may actually serve a primary function, resulting in prosodic
190  Data analysis, results and discussion

prominence and separation. In (4.84), for example, I believe is prominent


and separate, emphasizing the actual semantic content of the verb. In (4.89),
on the other hand, think has a polite, mitigating effect, toning down the
force of the utterance rather than expressing true speaker opinion; this goes
along with integration (see also Dehé and Wichmann 2010a). We will return
to this point in Chapter 5 below.
Another factor affecting the prosodic phrasing of CCs may be their length
and respective prosodic weight (Dehé 2009). It has long been known that
constituent length affects prosodic phrasing in general (see Chapters 2 and
3 above); according to Peters (2006), shorter parentheticals are more likely
to be prosodically integrated than longer ones. A CC typically consists of a
combination of pronoun and verb (I think, I believe, etc.), or pronoun, auxil-
iary and verb or adjective (I would say, I’m afraid, etc.), i.e. it constitutes one
prosodic word (and up to four syllables) only. Moreover, it has already been
shown in Section  2.2 above that prosodic phrasing may help avoid scope
ambiguities. A CC phrasing with preceding host material may be interpreted
as having semantic scope over preceding host material, while CCs phrasing
with following host material have scope accordingly. These constraints on
prosodic phrasing may override the syntax–prosody interface constraints,
resulting in patterns of prosodic phrasing which violate interface constraints
such as Match, Align R and Wrap, or combinations of them.

4.3.4.7  Excursus: comparison with Kaltenböck’s recent work


In the remainder of this section, I will briefly compare the present results
with recent work by Kaltenböck (2007, 2008, 2010), which, like the present
section, focuses on the prosody of CCs and its relation to meaning, using
ICE-GB data (see also Dehé and Wichmann 2010a: 15–17 for a critical
discussion of Kaltenböck 2007). Kaltenböck essentially identifies four dif-
ferent prosodic patterns for CCs: (i) left-binding, which corresponds to inte-
gration as British-style tail (i.e. as postnuclear material); (ii) right-binding,
which corresponds to integration as unstressed British-style prehead; (iii)
left-right-binding, which corresponds to integration as tail except that the
CC is followed by more unstressed material in the same IP; and (iv) prosodic
independence, which corresponds to prosodic separation.
As argued in Dehé and Wichmann (2010a), the introduction of a new
four-way classification and new terminology does not seem justified on the
basis of the data Kaltenböck provides. All patterns described have previ-
ously been observed in the literature (e.g. Crystal 1969; Wichmann 2001).
The term binding lacks a prosodic definition and thus remains unclear. In
particular, while left-binding makes reference to a preceding nucleus such
that left-bound material constitutes the British-style tail of that nucleus, and
while right-binding makes reference to a following nucleus such that right-
Analysis and results  191

bound material forms the British-style prehead of its domain, left-right-bind-


ing makes reference to both a preceding nucleus and to following unstressed
material. Right-binding on its own thus means ‘bound’ to following nuclear
material, while right-binding in left-right-binding means ‘bound’ to following
unstressed material in the postnuclear area.
A few remarks about Kaltenböck’s prosodic analyses of specific examples
are in order. To begin with, cues to IP boundaries in Kaltenböck’s analysis
often remain unclear or imprecise. For example, consider (4.93) (discussion
repeated from Dehé and Wichmann 2010a: 16), which Kaltenböck (2007: 22,
2008: 95) takes as an example for right-binding, i.e. integration as unstressed
British-style prehead. Kaltenböck (2007: 22) argues that I think “has to be
classified as initial as it is part of the following tone unit” which, he con-
tinues, is “indicated by the considerable step up in pitch (from around 100
Hz on you to around 180 Hz on think) …”. Here, Kaltenböck seems to make
use of the pitch of unaccented syllables as a boundary criterion (Cruttenden
1997: 34). However, the step-up in pitch should be at the boundary, i.e. on
the unstressed first syllable of the new domain (cf. Cruttenden’s 1997: 37
example). In Kaltenböck’s example, if the CC is preceded by a boundary, the
step-up should be on I rather than on think. Strictly speaking, if the step-up
is on think, the boundary could be between I and think. From Kaltenböck’s
analysis it is also unclear whether this step-up in pitch is in fact pitch move-
ment to a stressed syllable (think) or whether think is unstressed. Moreover,
Kaltenböck fails to provide evidence for the lack of boundary following I
think. On the basis of a preceding boundary alone, an intonational domain is
not sufficiently established.
(4.93) Speaker A: But these features (ICE-GB: s2a-021 #97)
Speaker A: and they’ll be familiar to you (ICE-GB: s2a-021 #98)
Speaker A: I think (ICE-GB: s2a-021 #99)
Speaker A: they include such things as uh uh a a a certain distrust
of fact (ICE-GB: s2a-021 #100)
Another example is given in (4.94)/Figure  4.16. Kaltenböck (2010:
252)  claims that Monday I think forms one domain, which “is separated
from the preceding tone unit by a pitch change … and a pause”. My own
analysis, confirmed by Anne Wichmann and Jill House (p.c.), is such that
there is an IP boundary between tomorrow and and, and that the pause pre-
ceding Monday, and interrupting the IP and again on uh Monday I think, is
hesitational rather than structure-related. The filled pause uh accompanying
the silent stretch of approximately 377 ms immediately preceding Monday
(see Figure  4.16) confirms the hesitation status. Along with this pause,
Kaltenböck takes the “pitch change (step up by about 100 Hz)” on Monday
as another cue to a boundary. However, Monday carries the nuclear accent,
300

250

200
Pitch (Hz)

150

100

50

L+H* L+!H* L-L% H* L+H* L-H%

CC

specifically tomorrow ? and again on uh <SIL> Monday I think

0 3.4
Time (s)

Figure 4.16  Example (4.94): CC, an alternative analysis to that of Kaltenböck (2010: 252)


Analysis and results  193

i.e. the local peak is associated with the first syllable of Monday, explaining
the pitch rise from the preceding syllable. The boundary between tomorrow
and and, on the other hand, is indicated by a step-up in pitch on unstressed
and from the preceding low boundary tone, along with the discontinuity and
glottal stop preceding and.
(4.94) The House knows that this matter may be debated on the Queen’s
speech specifically tomorrow and again on uh Monday I think
(ICE-GB: s1b-052 #3–4)
A final example is given in (4.95) (see Figure  11 in Dehé and Wichmann
2010a: 16 for illustration).
(4.95) Uh or you could have depressive illness or schizophrenia I think
(ICE-GB: s1b-016 #18)
Kaltenböck (2007: 12, 2008: 104) argues that the CC I think is “prosodically
a separate intonation unit”. Auditory and visual inspection of the corre-
sponding sound file reveals that I think is unstressed with no pitch move-
ment from the high level reached on the last syllable of schizophrenia. In
the framework Kaltenböck uses (i.e. Cruttenden 1997), an intonation group
minimally contains one stressed syllable and “there must be a pitch move-
ment to or from at least one syllable” (Cruttenden 1997: 34). I think in (4.95)
has neither of the two, thus does not qualify as a separate unit in this sense.
It is true that I think is preceded by a short silent pause (approximately 175
ms) and that the high pitch level reached at the end of schizophrenia can be
analysed as H%. I think receives a copy of the tones after the last stressed
syllable, i.e. of H-H%, and is realized on high pitch; it is encliticized to the
preceding domain (Gussenhoven 1990, 2004).

4.3.5  Reporting verbs
Of the 34 reporting verbs which survived the sorting processes explained
in Section 4.1.5 above, 2 items had to be discarded. For one item, no agree-
ment could be reached between the author and the second expert as to its
intonational phrasing; the second item was excluded because the bad qual-
ity of the sound file did not allow for a conclusive analysis. Of the remain-
ing 32 RVs, 27 were structurally simple, 5 were more complex. Nine were
in clause-final position; 23 RVs were in non-clause-final position. There
were no clause-initial RVs. The hypotheses with regard to prosodic phrasing
of RVs are repeated in (4.96)/(4.97) from Section 2.3.5 above. The phras-
ing patterns observed for the 32 items are summarized in (4.98), detailed
in Sections 4.3.5.1 and 4.3.5.2 and discussed in Section 4.3.5.3. The most
frequently observed pattern is phrasing with preceding host material (inte-
gration as given in (4.97)b1/(4.98)b1), followed by prosodic separation (see
(4.97)a and (4.98)a1).
194  Data analysis, results and discussion

(4.96) Prosodic phrasing of RVs: hypotheses


a. As parentheticals, RVs may be prominent and phrased separ-
ately (see (4.97)a).
b. Only complex RVs may be phrased separately with nuclear
prominence; simple RVs are integrated (Bing 1985;
Gussenhoven 2004; see (4.97)b).
c. Integrated RVs (see (4.97)b) are unaccented (Gussenhoven
2004).
d. Unaccented RVs may be set off from the host temporally (by a
pause) and tonally (by a boundary tone) (Gussenhoven 2004;
Astruc-Aguilera 2005; Wells 2006; Astruc-Aguilera and Nolan
2007b; see (4.97)b2).
(4.97) RVs: predicted prosodic phrasing
a. prosodic separation:
(…) IP[…]IP IP[(complex) RV]IP (IP[…]IP) (…)
b. prosodic integration
b1. (…) IP[… RV]IP (IP[…]IP) (…)
b2. (…) IP[IP[…]IP RV]IP (IP[…]IP) (…)
(4.98) Observed patterns of prosodic phrasing of reporting verbs (RVs)
a. RV: prosodic separation
a1. … IP[…]IP IP[RV]IP ( IP[…]IP ) … 6 RVs (19%)
a2. … IP[… IP[RV]IP …]IP … 1 RV (3%)
b. RV: prosodic integration; RV phrased together with preceding
and/or following host material
b1. … IP[…]IP IP[… RV]IP ( IP[…]IP ) … 21 RVs (66%)
b2. … IP[…]IP IP[IP[…]IP RV]IP ( IP[…]IP ) … 2 RVs (6%)
b3. … IP[…]IP IP[… RV …]IP ( IP[…]IP ) … 2 RVs (6%)
Total: 32 RVs
Tables  4.10 and 4.11 plot the phrasing and accent patterns according to
position of the RV (clause-final, non-clause-final) and complexity (simple,
complex), respectively. The RV phrased in a CPD is combined here with the
separately phrased ones.

4.3.5.1  Pattern (4.98)a: prosodic separation (RV)


Six out of 32 RVs (19%) were phrased separately in a non-recursive struc-
ture, following phrasing pattern (4.98)a1. Two examples, one complex and
one simple, are given in (4.99) and (4.100)/Figure  4.17, respectively. In
(4.99), there is a contrast between the two adjectives round and square in the
host, which is interrupted by the RV. The contrast results in prominence
(nuclear H* peaks) associated with the two adjectives and in lengthening.
The downward pitch movement on round is continued on table and IP1 is
terminated by L-L%. IP2, spanning the complex RV, has a L*+H pitch
accent on the final adverb later. The IP is terminated by H-H% and followed
Analysis and results  195

Table 4.10  Phrasing/accent patterns of RVs according to position (final, non-final)

Final Non-final Overall

separation: 2 5 7
[…][RV]([…]) / […[RV]…] ((4.98)a)

integration: [… RV] 7 15 22
unstressed / (incorporated or encliticized; (4.98)b1/b2)

integration: [… RV], nuclear ((4.98)b1) 0 1 1

integration: [… RV …] 0 2 2
unstressed or prenuclear prominence ((4.98)b3)

overall 9 23 32

Table 4.11  Phrasing/accent patterns of RVs according to complexity (simple, complex)

Simple Complex Overall

separation: 4 3 7
[…][RV]([…]) / […[RV]…] ((4.98)a)

integration: [… RV] 20 2 22
unstressed / (incorporated or encliticized; (4.98)b1/b2)

integration: [… RV], nuclear ((4.98)b1) 1 0 1

integration: [… RV …] 2 0 2
unstressed or prenuclear prominence ((4.98)b3)

overall 27 5 32

by a structure-related breath pause (which cannot be measured since the


sound file was cut in this position by the corpus builders). The separation of
he says in (4.100)/Figure 4.17 is very clear. There is a nuclear H* associated
with off in IP1 with the pitch falling towards L-L%. Moreover, off at the
end of IP1 is subject to final lengthening due to its prominence and phrase-
final position. It is followed by a silent pause of approximately 340 ms and a
filled pause of 360 ms. Within IP2, spanning the RV, says has a peak accent.
The IP boundary at the right edge of the RV is marked by L-L% and creaky
voice. There is a step-up in pitch from L% on the first syllable of because at
the beginning of IP3. The nuclear accent in the following IP is associated
with communicate, not plotted in Figure 4.17.
(4.99) Complex RV: prosodic separation
a. It wasn’t a round table President Gorbachev said later it was a
square table (ICE-GB: s2b-040 #92–93)
b. IP1[It wasn’t a round table]IP1 IP2[President Gorbachev said
H* L- L%
180

160
Pitch (Hz)

140

120

100

80
H* !H* L-L% H* L-L%

RV

it is <SIL> paying off uhm he says because for example the uh

0 Time (s) 4.872

Figure 4.17  Example (4.100): simple RV, prosodic separation


Analysis and results  197

later]IP2    IP3[it was a   square table]IP3


L*+H H-H%      H*   L- L%
(4.100) Simple RV: prosodic separation
a. Now it is paying off uhm he says because for example the the
air force can’t communicate on its primary sources of com-
mand and control and therefore even if it were to be up in the
air it couldn’t actually be coordinated properly and therefore
couldn’t uh uh fly as a as a consolidated force (ICE-GB: s1b-
038 #23)
b. IP1[Now it is paying off]IP1 IP2[he says]IP2 IP3[because …
H* !H*L-L% H* L-L%
The RV phrased in a CPD, following phrasing pattern (4.98)a2, is given in
(4.101). The string of establishing beyond doubt, IP2 in (4.101)b, is clearly set
off from the preceding domain by L% and a silent pause of more than one
second between facts and of. IP2 is terminated by L% and followed by a
silent pause of > 500 ms. Declination applies across IP2, with the H* asso-
ciated with doubt downstepped in relation to the H* peak associated with
establishing. IP3, the embedded IP spanning the RV, has its own CTC (H*
L-H%), which is overall lower in pitch than IP2. IP3 is furthermore aud-
ibly set off from IP2 by short breaks. The speaker interrupts IP2 in order to
insert the RV and resumes the pitch contour at the point where it was inter-
rupted in order to complete it.
(4.101) RV phrased in a CPD
a. My learned friend relied on a number of facts of establishing
he says beyond doubt that this vessel had been used to import
prohibited drugs (ICE-GB: s2a-068 #72)
b. … facts]IP1 IP2[of es tablishing IP3[ he says]IP3
L% L+H* H* L-H%
beyond doubt]IP2   IP4[that …
!H* L-L%

4.3.5.2 Pattern (4.98)b: RV phrased together with preceding and/or


following host material
The phrasing pattern most frequently observed with RVs is such that they
are phrased in one domain with preceding host material and immediately
followed by a phrase boundary (see (4.98)b1). Twenty-three RVs (72% of
analysed items) followed this pattern. Of these, 20 were integrated as post-
nuclear unstressed material, either associated with a rising pitch contour
and terminated by H% (12 RVs) or with falling/low pitch and terminated
by L% (8 RVs). Two RVs were encliticized in Gussenhoven’s (1990, 2004)
sense (see (4.98)b2). One complex RV was phrased together with the pro-
nominal subject of the host, was final in its domain and had the nuclear
accent of the domain.
198  Data analysis, results and discussion

One frequent pattern (8 observed cases) within the set of RVs phrased
according to (4.98)b1 is such that the RV phrases together with the preced-
ing (phrasal or clausal) subject of the sentence. While the subject is asso-
ciated with main prominence, the RV, placed between subject and finite
verb of the host sentence, is unstressed and is associated with a rising pitch
contour terminated by L-H% (continuation rise). One example is given in
(4.102). There is no reason to assume an IP boundary anywhere within IP1;
in particular, there is no cue to a boundary preceding he said. The falling
pitch from H* associated with the nuclear syllable clear is continued on he,
then followed by a rise towards H%, which terminates IP1. H% is followed
by a step-down in pitch on whether at the beginning of IP2.
(4.102) Simple RV phrased with preceding host material
a. It was unclear he said whether the fire had been caused by
enemy action or mechanical failure (ICE-GB: s2b-008 #112)
b. IP[It was un clear he said]IP1 IP2[whether the fire had been
H* L- H%
caused …
Example (4.103) is another illustration of pattern (4.98)b1. In IP1 table is
most prominent. It is associated with a nuclear peak; the downward trend
is continued on they said and terminated by L-L%. There is no pitch dis-
continuity or any other cue to a boundary between table and the RV. After
they said, there is a step-up in pitch on when. IP2 has a nuclear accent on
Secretary.
(4.103) Simple RV phrased with preceding host material
a. can you sit across the table they said when I became Chief
Secretary with Willie Whitelaw and others sitting the other
side and negotiate a spending round satisfactorily (ICE-GB:
s1b-043 #073)
b. IP1[can you sit across the  table they said ]IP1
H* L- L%
IP2[when I became Chief Secretary]IP2

The two encliticized RVs are given in (4.104)/Figure  4.18 and


(4.105)/Figure 4.19. In (4.104)/Figure 4.18 the RV he said is realized with-
out prominence at low pitch level and set off from the preceding material by
a short break. As is common for listings, there is a sequence of IPs termin-
ating in H% followed by an IP (and one culture) terminating in L-L%. The
final low tones of this IP are copied onto he said. The RV is then followed by
a pause of > 600 ms (not plotted here).
200

150
Pitch (Hz)

100

50
L* H-H% L+H* L-H% H* L-H% H* H+!H* L-L% L L%

RV

Somalis share one language one religion and one culture he said

0 3.993
Time (s)

Figure 4.18  Example (4.104): simple RV, encliticized


200  Data analysis, results and discussion

(4.104) Simple RV encliticized


a. Somalis share one language one religion and one culture he
said but Siad Barre is still dividing our people (ICE-GB: s2b-
023 #78)
b. IP1[So  malis]IP1   IP2[share  one  language]IP2
L*  H-H% L+H*    L-H%
IP3[one religion]IP3  [
IP4 IP4 [and   one   culture]    he said]
H* L-H% H*   H+!H* L-L%  L  L%
Similarly, in (4.105)/Figure 4.19 the complex RV says Watanabe is encliti-
cized to the preceding domain. The verb understand is most prominent in
its IP. The IP boundary following Japanese is marked by final lengthening of
the last syllable of Japanese, creaky voice and L%. The RV is realized with
low level pitch with audible lexical stress on the third syllable of Watanabe,
but not with enough prominence to qualify as a post-lexical accent.
(4.105) Complex RV encliticized
a. You don’t understand the Japanese says Watanabe (ICE-GB:
s2b-033 #055)
b. IP1[IP1[You don’t under  stand the Japanese]IP1  says Watanabe]IP1
L+H* L-L%    L L%
Finally, two non-final RVs were phrased such they were preceded and fol-
lowed by material from the host in their domain (see (4.98)d). One of the
two was unstressed; the other one had a prenuclear accent on the verb.

4.3.5.3  Remarks and discussion


Because of the low overall figures the distributional results cannot be inter-
preted as more than a tendency and a statistical analysis with position and
complexity as factors would be inappropriate due to there being too few
data points. Table 4.10 shows that in the given data set, non-final RVs are
more frequent than clause-final RVs and there are more unstressed and pro-
sodically integrated RVs than prominent/separate ones. Overall, the results
resemble those obtained for CCs, with the same interpretation regarding
prosodic theory. With regard to the hypotheses and predicted phrasing given
in (3.35) and in (4.96) and (4.97), the following conclusions can be drawn.
First, with more than two-thirds of RVs in the present data set being
phrased in such a way that they are prosodically integrated, prosodic sep-
aration as given in (3.35)a/b and (4.97)a cannot be argued to be the default
phrasing for RVs, but separation is nevertheless a frequent pattern, con-
firming hypothesis (4.96)a. Second, according to Gussenhoven (2004; see
also Bing 1985), RVs cannot be separate with nuclear prominence unless
they are complex: see hypothesis (4.96)b. This is not borne out. As illus-
trated, for example, by (4.100)/Figure 4.17, RVs consisting of a pronoun
and a verb may be prominent and phrased separately. These findings also
200

150
Pitch (Hz)

100

50

L+H* L-L% L L%

RV

You don’t understand the Japanese says Watanabe

0 Time (s) 2.836

Figure 4.19  Example (4.105): complex RV, encliticized


202  Data analysis, results and discussion

suggest that the assumption that RVs are generally unaccented and may
be set off from the host, as based on the experimental results of Astruc-
Aguilera (2005) and Astruc-Aguilera and Nolan (2007b) among others, is
too strong. Instead, RVs may be produced with a nuclear accent and phrased
separately. Moreover, according to hypothesis (4.96)c, integrated RVs are
unaccented, i.e. phrasing of RVs bearing nuclear or prenuclear promin-
ence with host material is not predicted. As the present results suggest, this
hypothesis may be too strong, but phrasing of prominent RVs with host
material is rare (see Table 4.10). Finally, according to (4.96)d, unaccented
RVs may be encliticized (see (4.97)b2). This hypothesis is borne out: 6 per
cent of RVs followed this pattern (see examples (4.104)/Figure  4.18 and
(4.105)/Figure  4.19). Taken together, prosodic incorporation and encliti-
cization in Gussenhoven’s (2004) sense, i.e. phrasing according to (3.35)c/
(4.98)b1–b2, account for 72 per cent of RVs, with incorporation (i.e. (4.98)
b1) being the most frequent.
The most important difference between the present results and previous
findings (e.g. Bing 1985; Gussenhoven 2004; Astruc-Aguilera 2005; Astruc-
Aguilera and Nolan 2007b) is that simple RVs, consisting of a pronoun and a
verb, may indeed be prosodically separate with a nuclear accent of their own,
following assumptions in the theoretical literature about the prosodic phras-
ing of parentheticals in general. Moreover, accented RVs may occasionally
be phrased together with host material (see Table 4.10). How can these dif-
ferences between the present results and results from previous experimental
work and from work based on introspective data be explained? First, the
nature of the data is a crucial factor. For example, Astruc-Aguilera (2005)
and Astruc-Aguilera and Nolan (2007b) tested four sentences, read by eight
speakers (see footnote 13 in Chapter 2). All RVs in their data set are sepa-
rated from the reported speech by commas, which may account for the fact
that they are temporally set off from the host in read speech. Furthermore,
to the best of my knowledge (see data list in Astruc-Aguilera 2005: 222),
these sentences were read without context, thus the reported speech was
most likely considered the main message, with the RVs merely appended.
This may account for the fact that the RVs were produced without promin-
ence. Finally, no utterance-medial RVs were tested. The ICE-GB examples,
on the other hand, are all utterances produced within a discourse context,
in utterance-medial or final position, which may or may not require prom-
inence on the RV. Following Bing (1985) and Gussenhoven (2004), a simple
RV like he says would not be expected ever to be prominent.
To show that prominence may indeed be required on the RV, I will begin
by looking more closely at the critical RV in (4.100) and its discourse con-
text. It is taken from a broadcast discussion between three speakers about
aspects of the 1990/1991 Gulf War, broadcast in January 1991 on BBC
Radio 4 News; see (4.106) for the first 25 lines of the corpus text (some
material omitted). Unfortunately, the corpus text does not start at the
beginning of the discussion; therefore we cannot be sure who he in the RV
Analysis and results  203

refers to, but it is likely from the context to refer to a commander of the
Allied troops. To Speaker A, when uttering the critical sentence, more than
one aspect of the utterance is important: first, that the effort is paying off;
second, that it is paying off according to the referent of he and that it is
not necessarily the speaker’s belief or opinion nor any general opinion or
uncontroversial truth. This is a comment made by the speaker using the RV
and resulting in emphasis on says and lack of reduction of he. The promin-
ence almost indicates a contrast between what the person referred to by he
says on the one hand, and what is the truth on the other: He says so but we
don’t know whether this is the case. In this particular case, the function of the
RV is not merely to mark the host as reported speech, but it also expresses
a strong degree of reservation and perhaps disagreement on the part of
the speaker, which results in the prominence on the RV. In order to bring
both messages across (first what was said, and second that this may not be
uncontroversial), the strings expressing the two messages are prosodically
phrased separately and the RV, like the previous chunk, is produced with a
nuclear accent.

(4.106) Broadcast discussion (ICE-GB: s1b-038 #1–25)


Speaker A: Yes
He had three areas of his objectives
He said first the uhm the Allies had started off to try and
hit and destroy the leadership command and control

Then he said they wanted to try and get to the air
defence system’s command

and then hit at the Republican Guard

That was phase one
Speaker B: Now just in that phase one
Can you describe in phase one what the purpose of
those objectives was
Speaker A: Well very simple
We get back to this idea of hitting command and con-
trol uhm and air defence
If you can get at the command and control of any force
whether it’s a naval an air force or an army then you’re
getting at its brain
If you get at its brain then it can’t operate successfully
And that is why they pounded day after day after day at
those sort of things
→ Now it is paying off uhm he says because for example
the the air force can’t communicate on its primary
sources of command and control and therefore even if
204  Data analysis, results and discussion

it were to be up in the air it couldn’t actually be coordi-


nated properly and therefore couldn’t uh uh fly as a as a
consolidated force
Speaker B: And then there was phase two
Speaker A: Yes

The RV he says in (4.101) is phrased separately in a CPD with a nuclear accent
associated with the pronoun he. The item is taken from an unscripted mono-
logue, specifically from a legal presentation made in court by the plaintiff ’s
counsel (Speaker A). The accent on he establishes a link between the preced-
ing referent My learned friend and the subject of the RV and at the same time
puts emphasis on the fact that it is the referent of these noun phrases who
has no doubt, but not necessarily any other person, and, in particular, not
the speaker. In what follows, the speaker outlines the evidence brought up in
favour of the assumption that “this vessel had been used to import prohibited
drugs”, in order to then argue that this evidence is inconclusive; see (4.107).
By accenting he in the RV, the speaker makes it clear that the proposition of
his utterance is the opinion of the “learned friend”, not his own.
(4.107) Legal presentation (ICE-GB: s2a-068 #71–83)
Speaker A: My Lord shall I <unclear-word> then and make sub-
missions as to the issues before you firstly under the
heading of the use to which this vessel was put
→ My learned friend relied on a number of facts of estab-
lishing he says beyond doubt that this vessel had been
used to import prohibited drugs
My Lord he relied on the forensic evidence
In my submission the forensic evidence that is and was
available is inconclusive
I rely upon Dr Williamson’s evidence
I accept that there were traces of cannabis found in the
samples that were recovered from the vessel
but as to whether or not those traces of cannabis and
those samples are evidence from which a court could
infer importation the only piece of evidence Your
Lordship has to rely on is really the appearance of the
stem as Mr Neal conceded in cross examination
It is a point which might suggest
but My Lord that vessel was for a length of time the
court knows on the port at the port of Fort Henry and
didn’t anywhere
People could have boarded it people could have used it
to conceal drugs
That is not wild speculation
Analysis and results  205

Speaker Z: (unclear)
Speaker A: Well My Lord there isn’t any evidence to support it

With respect to the relation between prosody and meaning, RVs behave in
a similar way to CCs (see Section 4.3.4 above). There are more prosodically
integrated than separate RVs, and the prosodic phrasing and the position
of the accent reflect not only the syntactic structure but the function and
meaning of the RV in discourse. RVs, like CCs, may be unaccented and
integrated, in which case they may simply convey the source of the content
of the host, or they may have interactional or interpersonal function (see
also Reinhart 1983 for a discussion of possible meanings). The numbers
of occurrences in the corpus reported here and the results of the previous
studies reported in Chapter  2 suggest that this may be the default case.
On the other hand, RVs may make an important contribution of their own
to the discourse and emphasis may be put on either the pronoun or the
verb for reasons arising from the discourse context, in which case they may
be prominent and phrased separately even if structurally simple. This is
expected if RVs, like other types of parentheticals, convey meaning separ-
ate from that of the host. Like CCs, RVs function as evidentials (Rooryck
2001) beyond their function of marking reported speech, thus like CCs,
they may be argued to be mixed expressions semantically in Scheffler’s
(2009) sense, with similar consequences for the relation between seman-
tic contribution and prosodic realization. Differences between the prosodic
phrasing of CCs and RVs can be observed such that within the pattern of
prosodic integration, CCs seem to be more frequently phrased together
with preceding and following host material than RVs (patterns (4.82)b3 and
(4.98)b3: 36% vs. 6%, respectively), and CCs seem to integrate less fre-
quently with preceding material while followed by a boundary than RVs
(patterns (4.82)b1 and (4.98)b1: 31% vs. 66%, respectively). However, the
results for RVs are based on a data set which is perhaps too small to allow
safe conclusions about differences between CCs and RVs with respect to
their prosodic behaviour.

4.3.6  Question tags
Of the 383 QTs that entered the analysis, the phrasing of 13 QTs (12 be,
1 do) remained unclear even after consulting another experienced linguist
trained in intonational analysis. The hypotheses with respect to the prosodic
phrasing of QTs based on the literature and the predicted prosodic phrasing
are repeated in (4.108) and (4.109), respectively. The intonational phrasings
found with all QTs are summarized in (4.110). Details are given in Sections
4.3.6.1 through to 4.3.6.3, followed by a statistical analysis in Section 4.3.6.4
and a discussion in Section 4.3.6.5.
206  Data analysis, results and discussion

(4.108) Prosodic phrasing of QTs: hypotheses


a. As parentheticals, QTs may be prominent and phrased separ-
ately (Knowles 1980; Selkirk 1984; Potts 2002b; see (4.109)a).
b. If the QT has nuclear prominence, it is always preceded by
a phrasal prosodic boundary (Ladd 1981; see phrasing in
(4.109)a and b1), i.e. it cannot be phrased together with pre-
ceding host material (i.e. (4.109)b2 and b3 are impossible with
nuclear QTs).
c. Prosodic phrasing is determined by polarity such that reversed
polarity tags are obligatorily phrased separately (i.e. they follow
(4.109)a), while constant polarity tags must incorporate as post-
nuclear material, i.e. phrased according to (4.109)b3 (or pos-
sibly b2, if preceded by main prominence) (see Gussenhoven
1984, 2004).
d. Prosodic phrasing is affected by verb type (Bolinger 1989),
such that QTs containing modals are typically integrated as
postnuclear material, i.e. they are phrased according to (4.109)
b3 (or possibly b2, if preceded by main prominence).
(4.109) QTs: predicted prosodic phrasing
a. QT: prosodic separation
(…) IP[…]IP IP[QT]IP (IP[…]IP) (…)
b. QT: prosodic integration
b1. (…) IP[…]IP IP[QT …]IP (IP[…]IP) (…)
b2. (…) (IP[…]IP) IP[… QT …]IP (IP[…]IP) (…)
b3. (…) (IP[…]IP) IP[… QT]IP (IP[…]IP) (…)
(4.110) Observed patterns of prosodic phrasing of question tags (QTs)
a. QT: prosodic separation
a1. … IP[…]IP IP[QT]IP ( IP[…]IP ) … 278 QTs (73%)
a2. … IP[… IP[QT]IP …]IP … 2 QTs (< 1%)
b. QT: prosodic integration; QT phrased together with preced-
ing and/or following host material
b1. … IP[…]IP IP[… QT]IP ( IP[…]IP ) … 79 QTs (21%)
b2. … IP[…]IP IP[QT …]IP ( IP[…]IP ) … 6 QTs (< 2%)
b3. … IP[…]IP IP[… QT …]IP ( IP[…]IP ) … 5 QTs (< 2%)
c. phrasing unclear 13 QTs (3%)
Total: 383 QTs

4.3.6.1  Pattern (4.110)a: Prosodic separation (QT)


Of the 370 QTs whose phrasing was conclusively analysed, a majority of 278
(73%) follow phrasing pattern (4.110)a1, i.e. prosodic separation in a non-
recursive structure. They have their own nuclear accent and a boundary
before and after the QT. Of the 278 QTs phrased in this way, the majority of
Analysis and results  207

250 QTs had a falling (H* L-L%) accent (predominantly preceded by an IP


terminated by L%). Nineteen QTs had a rising accent (L*(+H) H-H%) and
six had a falling-rising accent (H* L-H%). Examples of separately phrased
QTs with falling accents are given in (4.111) and (4.112)/Figure 4.20; a sep-
arate QT with a rising accent is given in (4.113)/Figure 4.21, a QT with a
falling-rising accent in (4.114).
In (4.111), all IPs have their own CTCs: a nuclear fall-rise (H* L-H%) in
IP1 preceding the QT, and nuclear falls on the QT (IP2) and the following
domain (IP3). The QT is preceded and followed by pauses of 105 ms and
172 ms, respectively. It is followed by a step-up in pitch on the unstressed
first syllable of the following IP (the). The F0 peak associated with the QT is
lower than the previous but not the following one on messenger, which is of
approximately the same height. See Dehé and Braun (2013: 139) for an illus-
trating figure (their Figure 1).
(4.111) QT non-final, prosodic separation
a. It’s also the sign of Hermes isn’t it the messenger uhm who
uhm is a guide to the occult world (ICE-GB: s1a-067 #234)
b. IP1[it’s also the sign of   Hermes]IP1    IP2[isn’t it]IP2
H* L-H% H* L-L%
IP3[the messenger]IP3 …
H* L-L%
In (4.112)/Figure 4.20 (Dehé and Braun 2013: 138–9), IP1 is terminated by
L% and there is (impressionistic) lengthening of down. Declination applies
across IP1. There is pitch reset such that the nuclear F0 peak associated
with don’t in IP2 (i.e. the IP spanning the QT don’t you) is higher than the
previous one on down. The QT is followed by a pause of approximately 480
ms, after which the interlocutor takes his turn.
(4.112) QT final, prosodic separation
a. God you really know how to put someone down don’t you
(ICE-GB: s1a-038 #252)
b. IP1[you really know how to put someone down]IP1
H* !H* L+H* L-L%
IP2[ don’t you]IP2
H* L-L%
In (4.113)/Figure 4.21 (Dehé and Braun 2013: 139–40), IP1 and IP2 have
complete CTCs: a falling accent (H+!H* L-L%) in IP1 preceding the QT,
and a rising accent (L*+H H-H%) in IP2 (the QT hasn’t it). The fall to L%
in IP1 reaches the lowest point in the speaker’s pitch range in this utterance.
The pitch fall is accompanied by (impressionistic) domain-final lengthen-
ing and creaky voice and there is a discontinuity before the beginning of the
350

300
Pitch (Hz)

200

100

50
H* !H* L+H* L-L% H* L-L%

QT

you really know how to put someone down don’t you

0 2.278
Time (s)

Figure 4.20  Example (4.112): QT, prosodic separation


250

200
Pitch (Hz)

150

100

50

H* !H* H+!H* L-L%L*+H H-H%

QT

has become a national organisation for police policy hasn’t it

0.318 4.321
Time (s)

Figure 4.21  Example (4.113): QT, prosodic separation


210  Data analysis, results and discussion

QT. The prominence and rising accent of the QT is presumably due to the
information-seeking nature of the tag. It is truly interrogative and the inter-
locutor steps in immediately (see (4.113)c).
(4.113) QT final, prosodic separation
a. Speaker A: uh ACPO as it’s called uh accountable to nobody
has become a national organisation for police policy hasn’t it
(ICE-GB: s1b-033 #69)
b. IP1[has become a national organisation for po  lice 
H* !H*
policy]IP1    IP2[  hasn’t it]IP2
H+!H* L-L% L*+H  H-H%
c. Speaker C: I wouldn’t accept that ACPO is not specifically
accountable to to nobody as to use your phrase
The reality of the situation is that […] (ICE-GB: s1b-
033 #70–71)
In (4.114), the QT has a falling-rising pitch accent and its domain (IP2)
is terminated by H%; it is preceded by a short break of about 100 ms and
followed by a pause of > 500 ms. IP1 has a falling nuclear accent associated
with model and terminates in L%. The F0 peak on the QT is slightly lower
than the final peak in IP1. See Dehé and Braun (2013: 141) for an illustrat-
ing figure (their Figure 4).
(4.114) QT non-final, prosodic separation
a. I mean actually you’ve therefore taken a model haven’t you which
you produced <unclear-words> (ICE-GB: s1a-064 #47)
b. IP1[I mean actually you’ve therefore  taken a
H* !H* L+H*
model]IP1    IP2[  haven’t you]IP2
L+H* L- L% H*    L- H%
According to previous literature (e.g. Quirk et al. 1985: 810; Knowles 1980:
393), if the QT is prominent, the accent is associated with the auxiliary,
never with the pronoun. This is also true for 275 of the 278 QTs in the cur-
rent data set which are prominent and phrased separately. The remaining
three QTs all have a negative tag (one neg-neg, two pos-neg; see (4.115))
and the accent is associated with the negation not. Example (4.115)a is
from a legal cross-examination. The QT is preceded by L% and a silent
pause of approximately 800 ms. It has a rising accent associated with not.
Semantically, the QT is interrogative and a reply is given by the interlocu-
tor (see the excerpt of the dialogue in (4.116)). Similarly, (4.115)b, from
the same legal cross-examination just a few utterances later, has a rising
accent on not asking for confirmation, and the interlocutor immediately
replies “Yes”, before the speaker continues. (4.115)c, taken from a broadcast
Analysis and results  211

discussion, has a falling accent on not and the same speaker continues, not
asking for a reply. (It is possible, of course, that there is a non-verbal reply
on the part of the interlocutors.) No items were found with a nuclear accent
on the pronoun.
(4.115) QTs, prosodic separation, main prominence associated with not
a. At that point Mr Hook on behalf of Ferndale had indicated to
you that he was interested in further expansion of his business
had he not (ICE-GB: s1b-064 #113)
b. Uh didn’t Mr Hook say to you that at this point that he was
interested in further expansion of the uh Ferndale business did
he not (ICE-GB: s1b-064 #118)
c. Now in practice there was censorship was there not by which it
was I I a suggested and b very readily and compliantly agreed
that nothing that might uh disturb the patriotic mood over the
conflicts should be allowed to appear on patriotic <unclear-
word> newspapers (ICE-GB: s1b-031 #55)
(4.116) Speaker A: At that point Mr Hook on behalf of Ferndale had
indicated to you that he was interested in further
expansion of his business had he not (ICE-GB: s1b-
064 #113)
Speaker B: Well Hook said one thing and uh the other guy said
something else. They both were at almost at conflict
with each other (ICE-GB: s1b-064 #114–115)
Speaker A: Well let me concentrate on what Mr Hook said
(ICE-GB: s1b-064 #116)
Two QTs with do (one pos-neg, one neg-pos) were phrased in a CPD
pattern (see (4.110)a2). An example is given in (4.117). There is no rea-
son to assume an IP boundary after don’t, preceding the QT. IP1 is the
domain of declination and its contour continues after the insertion of
the QT (and the L+!H* on want is therefore downstepped with regard
to the H* on don’t). The QT do you is clearly prominent and has its own
CTC: a falling-rising pitch accent associated with do. Its pitch peak is
lower than in the surrounding IP, but it is perceptibly not subordinated
to the main accent in IP1. It is conceivable that the phrasing pattern
is related to the syntactic position of the QT: it occurs within the ver-
bal complex between auxiliary and main verb, which is not a position at
which a strong prosodic boundary would be expected or predicted by
prosodic theory.
(4.117) QT phrased in a CPD
a. You don’t do you want us to think of faith as a synonym for
tradition (ICE-GB: s1b-028 #103)
212  Data analysis, results and discussion

b. IP1[you  don’t  IP2[do    you]IP2   want us to think of 


  H*    L+H*  L-H%   L+!H*
faith]IP1  IP3[…]IP3
!H* L-L%

4.3.6.2 Pattern (4.110)b2: QT followed (but not preceded) by material from


the host in its IP (… IP[…]IP IP[QT …]IP ( IP[…]IP ) …)
The remaining 90 QTs were phrased together with material from the syn-
tactic host. Of these, six QTs were also prominent but were joined in their
IP by host material, corresponding to phrasing pattern (4.110)b2. Of these
six QTs, two were followed in their domain by unstressed vocatives (e.g.
(4.118)), two by adverbs (one here, see (4.119)a, one really, see (4.119)b) and
two by hesitant stretches or discourse markers such as I mean (e.g. (4.120)).
All six QTs have a falling accent with the nuclear peak associated with the
verb and the falling/low contour being continued on the material follow-
ing the QT in the same domain (see (4.118)b and (4.120)b). In (4.120),
escape is lengthened and IP1 is terminated by L%. I mean continues the
fall/low pitch on the QT and is followed by a step-up in pitch on that’s.
After the second that’s the interlocutor steps in (Oh yes. Yeah I mean you’re
escaping into the book in uh you know everything around you; ICE-GB: s1a-
016 #325–327).
(4.118) QT phrased according to (4.110)b2; QT followed by vocative in
its IP
a. and then I say well next week is a full week’s block teaching
isn’t it Paul (ICE-GB: s1a-082 #42)
b. IP1[… block teaching]IP1 IP2[  isn’t it Paul]IP2 IP3[…]IP3
H* L* L- H% H* L-L%
(4.119) QTs phrased according to (4.110)b2; QT followed by adverb in
its IP
a. a lot happens at the edges of paintings doesn’t it and in a
way that’s taken away from you isn’t it here (ICE-GB: s1b-
018 #75 )
b. I mean it it sounds <laugh> sounds a bit holy doesn’t it really
<laugh> (ICE-GB: s1a-003 #120)
(4.120) QT phrased according to (4.110)b2; QT followed by hesitant
stretch in its IP
a. But y you still need the book to escape don’t you I mean that’s
that’s (ICE-GB: s1a-016 #323–324)
b. IP1[… you still need the book to  escape]IP1
H* L+!H* L-L% 
IP2 [ don’t you I mean]IP2

H* L-L%
Analysis and results  213

4.3.6.3 Pattern (4.110)b1/b3: QT phrased together with preceding or with


preceding and following host material
Seventy-nine QTs (21%) were integrated as unstressed postnuclear mater-
ial, following phrasing pattern (4.110)b1, i.e. they were phrased together
with preceding host material and were followed by an IP boundary. Of
these, 61 had rising contours terminated by H% (e.g. (4.121)/Figure 4.22;
(Dehé and Braun 2013: 143–4), and 18 were produced with falling or low
pitch terminated by L% (see Dehé and Braun’s 2013: 144 example (25)).
No domain-final QTs were prominent. In (4.121)/Figure  4.22, the fall
from the F0 peak associated with well is continued on the QT, with the
elbow in the QT at the beginning of you and the IP being terminated by
H%; the QT is unstressed. (Note that the interruption of the F0 contour
does not signal a pause but is due to the closure of the initial plosive of
can.) Comparing a QT phrased separately with a rising nuclear accent such
as (4.113)/Figure  4.21 to the integrated one in (4.121)/Figure  4.22, it is
obvious that the fall in (4.113)/Figure  4.21 (towards L%) is completed
before the start of the QT and is followed by a separate rise associated
with the QT, while in (4.121)/Figure  4.22 the tag continues the nuclear
falling-rising contour starting at the local peak associated with the nuclear
syllable well.
(4.121) QT realized as unstressed, postnuclear material
a. Oh really, can’t fit me in as well can you (ICE-GB: s1a-
039 #360–361)
b. IP1[can’t fit me in as well can you]IP1
H* !H* L-H%
Finally, five QTs were phrased according to (4.110)b3, i.e. they were pre-
ceded and followed by material from the host in the same domain. One of
the five was unstressed and preceded the nuclear accent in its domain (see
(4.122)/Figure 4.23; Dehé and Braun 2013: 144–5); this is the only excep-
tion to Ladd’s (1981) claim that QTs never precede the nuclear accent, and
is presumably due to the high speech rate and corresponding length of
the IP. An IP boundary between the QT and the following relative clause
(who was …) can easily be imagined. The utterance in (4.122) is divided
into three rather long IPs (see (4.122)b), all of which have a CTC, are the
domain of declination and are terminated by L%. The tonal analysis of the
IP containing the QT (IP2) is given in (4.122)c/Figure 4.23. The QT wasn’t
it is unstressed and reduced to /wɒzənt/, i.e. the pronoun disappears.17 It

17
Wichmann (2007: 356) notes that QTs might be most likely to be reduced if they “have no
informational or discourse-organising function”, but have an attitudinal function instead.
Attitudinal QTs, she assumes, are most likely to occur utterance-medially (Wichmann
2007: 350). The QT in (4.122) is compatible with Wichmann’s analysis. However, a thor-
ough analysis of all QTs in the corpus relating their prosody to interpretation and function
in discourse has yet to be done.
400

300
Pitch (Hz)

200

100
H* !H* L- H%

QT

can’t fit me in as well can you

0 1.24
Time (s)

Figure 4.22  Example (4.121): QT unstressed, prosodic integration with preceding host material
300

250
Pitch (Hz)

200

150

100

H* L+!H* L+!H* L+!H* L-L%

QT

it was your mother wasn’t it who was the e the driving force behind all of this

2.239 6.145
Time (s)

Figure 4.23  Example (4.122): QT unstressed, prosodic integration with preceding and following host material
216  Data analysis, results and discussion

belongs to the transition between two prenuclear accents associated with


mother and driving, respectively. The nuclear accent of this IP is associated
with behind.
(4.122)  a. So on going back to your to your childhood it was your mother
wasn’t it who was the the driving force behind all of this behind
this sort of intellectual rigour (ICE-GB: s1b-046 #057)
b. IP1[So on going back to your to your childhood]IP1 IP2[it was
your mother wasn’t it who was the the driving force behind all
of this]IP2 IP3[behind this sort of intellectual rigour]IP3
c. IP2[it was your mother wasn’t it who was the the
H*
driving  force be  hind     all of this]IP2
L+!H* L+!H* L+!H* L-L%
The other four QTs phrased together with preceding and following host
material were unstressed and followed the nuclear syllable. Similar to
(4.118), the QT in (4.123)/Figure  4.24 (Dehé and Braun 2013: 145–6) is
followed by a vocative, but unlike the QT in (4.118), the QT in (4.123) is not
prominent. In a large IP, which spans the whole utterance, the QT is part of
the unusually long stretch of postnuclear material, which has a falling-rising
F0 contour; the nuclear pitch peak is associated with a syllable in the host
(i.e. with the first syllable of Labour). Auditory and visual inspection of the
data gives no reason to assume an IP boundary anywhere else.
(4.123)  a. But the Labour Party’s not going to abolish private medicine is
it Donald (ICE-GB: s1b-039 #88)
b. IP1[But the   Labour Party’s not going to abolish private
H* L*
medicine is it Donald]IP1
H-H%

4.3.6.4  Statistical analysis


Like the N-APP data set, the QT data set was large enough to allow for stat-
istical analysis.18 In light of the variation observed in the prosodic phrasing
and intonational realization of the QTs, the statistical analysis was carried
out in order to test which factors may affect their phrasing. To this end, the
data were coded according to the factors polarity (3 levels: pos-neg, neg-
pos, pos-pos; see (4.124)),19 position (4 levels: utterance-final, sentence-

18
I owe this statistical analysis to Bettina Braun; see also Dehé and Braun (2013). In addition
to prosodic phrasing, Dehé and Braun (2013) focus on the intonational realization of the
QTs in this data set.
19
Polarity had four levels originally: two reversed (pos-neg, neg-pos) and two constant
(pos-pos, neg-neg); see Table 4.2 in Section 4.1.6 above. However, the neg-neg QTs were
excluded from the statistical analysis because there were too few data points (only two
occurrences overall).
300

250
Pitch (Hz)

200

150

100

50
H* L* H-H%

QT

But the Labour party’s not going to abolish private medicine is it Donald

0 2.871
Time (s)

Figure 4.24  Example (4.123): QT unstressed, prosodic integration with preceding and following host material
218  Data analysis, results and discussion

final, phrase (XP)-final, XP-medial; see (4.125)), and verb type (4 levels: be,
do, have, modal; see (4.126)). With regard to position, a QT was coded as
utterance-final at the end of a syntactic sentence and simultaneous end of
the speaker’s turn (e.g. (4.125)a); it was coded as sentence-final if it was
at the end of a syntactic sentence but the same speaker continued in the
discourse (e.g. (4.125)b). A QT was coded phrase (XP)-final if inserted at
the right edge of a syntactic XP (e.g. a NP in (4.125)c1 and c2), and it was
coded XP-medial if no XP-boundary preceded the QT (e.g. between a noun
and its syntactic complement in (4.125)d1 or within the verbal complex in
(4.125)d2). The data were also coded for intonational contour, i.e. whether
they were falling/low (e.g. in separate QTs with H* L-L%, in integrated
QTs with a low or falling pitch leading to a low boundary tone) or rising
(e.g. in separate QTs with L*(+H) H-H% or L* L-H%, in integrated QTs
with a rising pitch leading to H%).
(4.124) Polarity
a. Reversed: pos-neg
The state of the wall uh the state in which you found the wall was
the result of the work that you had directed to be done to the wall
wasn’t it (ICE-GB: s1b-069 #174)
b. Reversed: neg-pos
You can’t blame her for that really can you (ICE-GB: s1a-007 #17)
c. Constant: pos-pos
But then there’s another audience is there that goes to regional play-
houses a rather more stolid audience (ICE-GB: s1b-050 #87)
(4.125) Position
a. Utterance (U)-final:
a1.  Speaker A: That’s the trouble if you’re out a bit isn’t it
Speaker C: Yeah. I mean through the Blackwall Tunnel
it’ll probably take about half an hour or an hour (ICE-GB:
s1a-019 #358–360)
a2.  Speaker A: but you’ve never touched that form have you
Speaker B: No because I never liked it. I never liked read-
ing it I mean (ICE-GB: s1b-048 #125–127)
b. Sentence (S)-final
b1. Well it’s a very boring sort of cheese isn’t it. I mean it’s
all right thinly sliced with a with an apple I suppose.
(ICE-GB: s1a-061 #354–355)
b2. Well we’re going to have a we’re going to have endless dis-
cussions over what kind of film to get out aren’t we. Can
you set it up on your uhm uhm sort of stereo thingum-
mybob that you (ICE-GB: s1a-006 #79)
c. Phrase (XP)-final
c1. You were in in America uh in Washington weren’t you at
the time of the Gulf deadline (ICE-GB: s1b-042 #112)
Analysis and results  219

c2. He suffered great mental distress didn’t he after the war
(ICE-GB: s1b-032 #164)
d. XP-medial
d1. but isn’t that at the root of it because it’s surely quite dif-
ficult these days to persuade an actor or actress to com-
mit themselves for what six to eight months when there is
always the the possibility isn’t there of lucrative television
work (ICE-GB: s1b-050 #72)
d2. You don’t do you want us to think of faith as a synonym
for tradition (ICE-GB: s1b-028 #103)
(4.126) Verb type
a. be
The state of the wall uh the state in which you found the wall
was the result of the work that you had directed to be done to
the wall wasn’t it (ICE-GB: s1b-069 #174)
b. do
You don’t like coffee particularly strong do you (ICE-GB: s1a-
046 #269)
c. have
And uh we’ve learnt haven’t we that books are all right but there
is nothing like uh the help you get from society members as the
team have said and doing it yourself (ICE-GB: s1b-025 #166)
d. modal
You can hear him all right can’t you (ICE-GB: s1a-004 #24)
Of the 370 QTs which were conclusively analysed and included in the
descriptive analysis reported above, 41 had to be discarded from the stat-
istical analysis for the following reasons: the only two cases with constant
negative polarity tags (neg-neg) and six cases with a nuclear accent on the
QT (phrased as in (4.110)b2/Section 4.3.6.2) were excluded because there
were too few data points. The same was true for 24 cases with infrequent
intonation contours on the tag (8 with falling-rising, 15 with low and 1
unaccented).20 Finally, nine cases in which the position was unclear, mostly
due to hesitant stretches or self-corrections on the part of the speaker, were
not considered. The remaining 329 cases entered the statistical analysis. The
dependent variable relevant in the present context was prosodic phrasing
(two levels: separation, see (4.110)a/Section 4.3.6.1; integration, see (4.110)
b1 and (4.110)b3/Section 4.3.6.3).

20
This was done for the sake of the statistical analysis addressing the intonational realization
of QTs (reported on in Dehé and Braun 2013) and the respective QTs were consequently
also excluded from the analysis for phrasing in order to allow for maximal comparability
between the analyses.
220  Data analysis, results and discussion

Table 4.12  Cross-tabulation of QTs according to phrasing and intonation


contour

Phrasing Falling contour on QT Rising contour on QT

separate (see (4.110)a) 243 20


integrated (see (4.110)b) 5 61

Table 4.13  Cross-tabulation of QTs according to phrasing and polarity

Phrasing neg-pos pos-neg pos-pos

separate (see (4.110)a) 46 203 14


integrated (see (4.110)b) 18 28 20
ratio 2.6 7.3 0.7

The prosodic annotation suggested a correlation between prosodic phras-


ing (i.e. integration vs. separation) and intonational realization of the QT
(rising vs. falling intonation contour). As can be seen in Table 4.12, separate
QTs are generally realized with a falling contour (see (4.113)/Figure 4.21
for an exception), and integrated QTs with a rising one (more than 90%
of the cases each). This correlation is corroborated by a non-parametric
Kendall’s tau test (R version 2.12.2). Not surprisingly, the results showed a
highly significant correlation between phrasing and intonation contour (tau
= 0.79, z = 14.3, p < 0.0001).
Subsequently, the effects of the factors polarity and verb type on the
phrasing of the QTs were investigated, testing hypotheses (4.108)c–d, as
well as the factor position. Intonation contour was not included as a factor
because of its strong correlation with phrasing. A binomial logistic regres-
sion model with phrasing as the dependent variable and the above-men-
tioned factors as fixed factors (function glm with a binary linking function,
see Baayen 2008: chapter 6.3.1) was calculated.21 To test the validity of the
model, data points with residuals greater than 2.5 standard deviations were
removed and the model was refitted (removing not more than 7 QTs, i.e.
less than 2% of the data). The statistical analysis reported here is based on
this final model.
Regarding polarity, QTs with reversed polarity have a higher proportion
of separate compared to integrated QTs, while QTs with constant (posi-
tive) polarity are more often integrated rather than separate (neg-pos vs.
pos-pos: beta = 1.0, z = 2.0, p < 0.05; pos-neg vs. pos-pos: beta = 2.3,

21
Note that the distribution of the data points across factors did not allow us to test for inter-
actions between factors.
Analysis and results  221

Table 4.14  Cross-tabulation of QTs according to phrasing and position

Phrasing U-final S-final XP-final XP-medial

separate (see (4.110)a) 129 102 24 8


integrated (see (4.110)b) 38 20 2 6
ratio 3.4 5.1 12 1.3

Table 4.15  Cross-tabulation of QTs according to phrasing and


verb type

Phrasing be do have modal

separate (see (4.110)a) 124 84 22 33


integrated (see (4.110)b) 46 8 3 9
ratio 2.7 10.5 7.3 3.7

z = 5.2, p <  0.0001). Furthermore, within the group of reversed polarity


tags, pos-neg QTs were significantly more frequently phrased separately
than QTs with neg-pos (beta = 1.3, z = 3.4, p < 0.0005). Distribution num-
bers and ratios are given in Table 4.13. They confirm that hypothesis (4.108)
c, claiming obligatory separation of reversed polarity tags, is too strong, but
that there is a clear tendency in this direction. What is more, the reversed
polarity tags do not behave uniformly in this respect.
Position also had a significant effect on the prosodic phrasing of QTs
(see Table  4.14). There were more separate QTs in utterance-final, sen-
tence-final and XP-final position than in XP-medial position (U-final vs.
XP-medial: beta = 1.3, z = 2.0, p < 0.05; S-final vs. XP-medial: beta = 1.8,
z = 2.6, p < 0.01; XP-final vs. XP-medial: beta = 2.0, z = 2.1, p < 0.05).
There was no difference in prosodic phrasing for utterance-final, sentence-
final and XP-final QTs (all p-values > 0.15).
Finally, a significant effect of verb type on the number of separate QTs
was found (see Table  4.15 for figures and ratios): QTs containing the
verb do were significantly more often separate than QTs with the verb be
(beta = 1.5, z = 3.5, p < 0.001) and QTs with modal verbs (beta = 1.2,
z = 2.0, p < 0.05); there was no difference between QTs containing do and
have (p > 0.3). No other differences were significant (all p-values > 0.2).
These results support Bolinger’s (1989) hunch about verb type generally
affecting phrasing of the QTs, but do not confirm his observation about a
predominant integration of modals.
In order to test the hypothesis that the prosodic structure of the QT
(bisyllabic, e.g. don’t you, or trisyllabic, e.g. haven’t you) might influence
phrasing more than verb type (raised by an anonymous reviewer for Dehé
and Braun 2013), an alternative model was fitted with the factors mentioned
222  Data analysis, results and discussion

above (position and polarity) but replacing verb type by the number of syl-
lables in the QT (two or three). Results showed no effect of syllable number
of the QT (p > 0.15). The factor syllable number is also not significant if
this factor is added to the model mentioned above (p > 0.2). To exclude the
possibility that the effect of syllable number is hidden by the factor polar-
ity (all neg-pos and pos-pos QTs are disyllabic, while pos-neg QTs had
both bisyllabic and trisyllabic forms), the data set was reduced to pos-neg
cases (N = 231) and the model was run again (excluding the factor polar-
ity). In this model also, syllable number did not show an effect (p > 0.15).
Therefore, syllable number does not seem to be the primary factor guiding
prosodic phrasing.
Taken together, these results confirm some of the claims previously made
for QTs, while others are modified and new findings are added. Specifically,
there is a high correlation between prosodic phrasing and intonational real-
ization of QTs, such that rising tags integrate significantly more often than
falling ones. Moreover, the factors polarity, position of the QT and verb
type all have an effect on prosodic phrasing. The results will be discussed in
more detail immediately below. See Dehé and Braun (2013) for effects of the
same factors on the intonational realization of QTs.

4.3.6.5  Remarks and discussion


This section will further discuss the results obtained for QTs in relation
to the hypotheses developed above. First, prosodic separation of QTs is
not obligatory, but it is frequent (73% in the current data set), confirming
hypothesis (4.108)a. However, more than a fifth of the QTs were annotated as
prosodically integrated, mostly as unstressed material following the nuclear
accent in their domain. It follows from this result that there is no strict cor-
respondence between being a QT on the one hand and prosodic separation
on the other, an assumption put forward, for example, in Potts’ (2002b) the-
oretical account. It seems that, like CCs and RVs, QTs provide additional
evidence for the assumption that parentheticals, and in particular short and
light types, are not necessarily prosodically phrased separately. The phono-
logical form may be adjusted according to the given formal and discourse
context. However, prosodic separation is indeed the most frequent pattern
with QTs in English. This may be due to their syntactic status, but also to
the fact that, pragmatically, they make a contribution separate from that of
the host, or a combination of the two. It is striking that despite the similar-
ities between CCs, RVs and QTs discussed in Chapter 2 above (e.g. length,
prosodic weight, apparent syntactic incompleteness, semantic-pragmatic
contribution), a general difference with respect to prosodic phrasing can be
observed between CCs and RVs on the one hand and QTs on the other.
Specifically, CCs and RVs are more often prosodically integrated compared
to QTs. QTs, unlike CCs and RVs, are predominantly phrased separately.
We will return to this matter in Chapter 5.
Analysis and results  223

Overall, prosodic separation on the one hand and integration as postnu-


clear unstressed material on the other account for 94 per cent of all QTs in
the present data set (see (4.110)). This supports Ladd’s (1981) observation
that nuclear QTs are directly preceded by a prosodic boundary (see hypoth-
esis (4.108)b) and cannot be preceded by host material in their IP. QTs are
either nuclear themselves or they are (part of) postnuclear material. This is
shown in (4.127): the QT with nuclear prominence (indicated by underlin-
ing in (4.127)) is preceded by an IP boundary. It may optionally be followed
by host material in the same IP (indicated by (…) in (4.127)), but it may
not be preceded by host material (see (4.127)b; # indicates ‘not predicted’).
Ladd’s observation bears up against the results of the present corpus study:
in accordance with (4.127), the present results show that QTs with a nuclear
accent are either prosodically separate (e.g. (4.111) through (4.114)), or they
are followed by unstressed material such as adverbs, vocatives or hesitant
stretches (e.g. (4.118) through (4.120)), both corresponding to (4.127)a; a
nuclear QT preceded by host material in its IP does not occur, thus is cor-
rectly excluded by (4.127)b.
(4.127) QT: phrasing predicted by hypothesis (2.129)a (Ladd 1981):
underlining indicates nuclear prominence
a. … IP[…]IP IP[QT (…)]IP …
b. # … IP[…]IP IP[… QT (…)]IP …
The statistical analysis strongly supports the assumption that the prosodic
phrasing of QTs is affected by a number of other factors, alongside their
syntactic status as parentheticals. Specifically, the factors polarity, position
and verb type were tested here and they all have an effect on the prosodic
phrasing of QTs to the extent outlined above. First, hypothesis (4.108)c is
based on Gussenhoven’s (1984, 2004) strong claim that reversed polarity
tags are obligatorily phrased separately, while constant polarity tags must
incorporate. The statistical analysis confirms that polarity does indeed affect
phrasing in the predicted direction. However, since the prosodic phrasing
of QTs is not affected by polarity alone, it is not surprising that hypothesis
(4.108)c in its absolute sense is too strong. It turns out that separate phras-
ing of reversed polarity tags is frequent but not obligatory, and that constant
polarity tags do not obligatorily incorporate (see Table 4.13). An example of
a constant polarity tag phrased separately is given in (4.128).
(4.128) QT with constant polarity (pos-pos), phrased separately
a. if they’re in the end called up if not enough of them volunteer
they’ll have to go will they (ICE-GB: s2b-012 #053)
b. [if not enough of them   volunteer] [they’ll have to go]
H* L-H% H* L-H%
[ will they]
L* H-H%
224  Data analysis, results and discussion

The results also show that within the group of reversed polarity tags, pos-
neg QTs are more often separate than neg-pos QTs. One reason for this
novel finding (first reported in Dehé and Braun 2013) may lie in the dis-
course function of the respective QTs, a factor that was not included in
the current analysis. Separately phrased QTs have been argued to consti-
tute a separate speech act. If this holds true, the present results suggest
that positive QTs (neg-pos and more so pos-pos) are more often part of the
interpretation of the host, while negative QTs (pos-neg) are more likely to
form an interrogative speech act that seeks the interlocutor’s confirmation.
Interestingly, the two excluded neg-neg instances were also phrased separ-
ately. In future research, the replies following neg-pos and pos-neg QTs will
have to be analysed to corroborate this interpretation.22
Second, the factor verb type also affected phrasing (cf. hypothesis (2.129)
d), in line with Bolinger’s (1989: 164) assumption that the choice of the tag
may affect its prosodic realization. However, while Bolinger (1989) argues
that it is common for “the central auxiliaries will and can” to be unaccented
and integrated and he remains relatively vague otherwise, the present results
allow us to be more specific. In particular, the verbs do and have seem to be
special such that they phrase separately more often than be and modals. As
yet, there is no satisfactory explanation for this finding. The frequent pros-
odic separation of QTs containing the verb do may be explained by its func-
tion as ‘dummy do’; possibly, the need to replace a verb from the host clause
by do is effortful and results in more prosodic separation. This explanation
cannot be extended to QTs with have, though. An alternative interpretation,
that the word prosodic structure of QTs (whether they are bisyllabic or tri-
syllabic) affects prosodic phrasing was not supported by the statistical ana-
lysis. Note that this effect is not merely a reflex of one or other of the factors
tested as their effect was factored out of the statistical model.
Third, the prosodic phrasing of QTs is affected by their position: QTs
are phrased separately more frequently when located at a phrasal or clausal
syntactic boundary or at the end of an utterance than when positioned
phrase-medially. This result is compatible with the Align/Wrap and
Match theories of the syntax–prosody interface (see Chapter 3 above) and

22
That pragmatic effects also play a role in the choice of polarity can be seen when looking at
the distribution of polarity across discourse types. Counting only types with more than 10
instances (N) in the corpus, pos-pos QTs were most frequent in legal cross-examinations
(31%, N = 29), business transactions (27%, N = 11), broadcast interviews (20%, N = 25)
and private telephone calls (17%, N  =  18); they are rather infrequent in private direct
conversations (7%, N = 163), in classroom lessons (3%, N = 31) and broadcast discus-
sions (0%, N = 35). One tentative generalization may be that in situations in which the
QT functions as a leading question (rather than a confirmation-seeking or information-
seeking question) pos-pos QTs are more frequent. This complies with Gussenhoven’s
(1984) observation that constant polarity tags are statements rather than questions (see
Section 2.3.6).
Analysis and results  225

it includes syntactic phrasing as one factor shaping prosodic constituent


structure. An IP boundary preceding an utterance-final or sentence-final
QT follows from Selkirk’s (2005) Align R (CommaP, IP) constraint and
Match clause. What is surprising, perhaps, is that XP-final position did
not differ from sentence-final and utterance-final positions, because pros-
odic boundaries at clause edges are predicted to be stronger than those at
XP edges. However, first, it is necessary to keep in mind that “separate”
in the present study means phrased in an ip, IP or British-style TG. It is
therefore impossible to rule out that some QTs phrased in a separate IP
(as well as some CCs and RVs) might have to be reanalysed as ips (in the
terminology of Beckman and Pierrehumbert 1986) in a more fine-grained
analysis. How, for QTs, this distinction relates to position and whether
it would result in a statistically significant difference between QTs at
XP-boundaries and QTs at utterance or sentence boundaries is a topic for
future research. Second, while, according to Align/Wrap and Match, an
ip boundary is predicted at the right edge of an XP, no such boundary is
predicted in XP-medial position. It is conceivable, therefore, that while ip
boundaries may be strengthened to IP boundaries, no such strengthening
would occur XP-medially, and that the statistical effects result from this
difference. However, it appears from Table 4.14 that despite the statistic-
ally significant difference between XP-medial QTs one the one hand and
all three other syntactic types on the other, even within XP-medial QTs
more tokens were phrase separately (N = 8 in Table 4.14) than were inte-
grated (N = 6). Note in this context that the eight XP-medial QTs phrased
separately include the two QTs in the data set phrased in a CPD, i.e. in a
recursive prosodic structure (see (4.117) above for an example). This is in
line with the predictions made by prosodic theory. What is not predicted
is for XP-medial QTs to be phrased separately in a non-recursive prosodic
structure. The number of items phrased in this way is correspondingly
small (N = 6).
Finally, note that the syntactic distribution of QTs in the present corpus
data shows that Ziv’s (1985: 189) strong claim that QTs are “restricted to
sentence-final position” is untenable. The spoken corpus investigated here
includes numerous counterexamples.
In summary, the systematic analysis of the corpus data reveals that several
factors interact in the prosodic phrasing of QTs in English, among them
polarity, position and verb type. Pragmatic factors and discourse effects
have not been included in the analysis but must nevertheless be considered
important factors in the prosodic realization of QTs, as well as in the pros-
odic realization of CCs (see Dehé and Wichmann 2010a) and RVs. A full
analysis of the present data taking all discourse contexts into account and
including them in a (statistical) analysis with the other factors has to await
future research.
226  Data analysis, results and discussion

Table 4.16  Prosodic phrasing: summary of results for six types of parentheticals

Phrasing FPC NRRC N-APP CC RV QT

separation IP[…]IP IP[par]IP 93% 92% 80% 18% 19% 73%


IP[…]IP
[… IP[par]IP
IP 6% n/a <1% 5% 3% <1%
…]IP

integration […]IP IP[…


IP 1% n/a 14% 31% 72%b 21%
par]IP IP[…]IP   
[…]IP IP[par
IP n/a n/a <1% 6% n/a <2%
…]IP IP[…]IP

[…]IP IP[… par


IP n/a n/a 2% 36% 6% <2%
…]IP IP[…]IP

par][par [… part of
IP n/a 7%a n/a n/a n/a n/a
par]IP IP[part of
par]IP IP[…]IP

unclear n/a 1% 3% 4% n/a 3%

a Note that this is a special case: it is always the relative pronoun (and a pied-piped pre-
position where appropriate) that is phrased with preceding host material. See Section 4.3.2.
b This figure includes both incorporated and encliticized cases, like for CCs; the two were
separately discussed in Section 3.5 above.

4.4  Summary
Overall, the prosodic phrasing of the six types of parentheticals analysed
above has been found to be according to the predictions formulated in
Chapter 2. In particular, FPCs, NRRCs and, to a lesser extent, N-APPs are
phrased separately by default, while with CCs and RVs integration is the
default and overall more variation can be observed for CCs and RVs than for
FPCs and N-APPs (see Table 4.16).
QTs are more prosodically separate than integrated despite being grouped
with CCs and RVs at the end of Chapter 2. This, however, makes sense in
light of the predictions for QTs and the statistical effects. For example,
polarity affects phrasing such that reversed polarity tags prefer separation,
while constant polarity tags prefer integration. The data set contained 339
(89%) reversed polarity tags but only 44 (11%) constant polarity tags. It is
thus not surprising that separation is the most frequent pattern for QTs,
even if it is not obligatory even for reversed polarity tags.
The results of this chapter will further be discussed in the next chapter.
5 Final discussion

In Chapters 2 and 3, predictions were formulated as to the prosodic phras-


ing of parentheticals. In Chapter 2, these were formulated for each of the six
types of parentheticals under investigation. They were based on a lot of pre-
vious research specifically on parentheticals, including experimental work,
work based on corpus data, and work based on introspective data, as well as
on more general work in syntax and prosody and the respective interfaces.
In Chapter 3, predictions were formulated based on current syntax–­prosody
interface theories, specifically Align/Wrap and Match theories. These pre-
dictions are repeated for convenience in (5.1) and (5.2) below. This chapter
will further discuss the results reported on in Chapter  4 in light of these
predictions and implications for a theory of prosodic phrasing and the
­syntax–prosody interface.
(5.1) Predictions of the Align/Wrap and Match constraints for
the ­prosodic phrasing of parentheticals (par) (repeated from
(3.35))
a. IP[…]IP IP[par]IP IP[…]IP √Align R, √Wrap, √Match
b. IP[… IP[par]IP …]IP √Align R, √Wrap, √Match
c. IP[…]IP IP[… par]IP IP[…]IP √Align R, √Wrap, *Match
d. IP[…]IP IP[(…) par …]IP IP[…]IP *Align R, √Wrap, *Match
e. # IP[… part of par]IP IP[part
of par]IP IP[…]IP √Align R, *Wrap, *Match
(5.2) Prosodic separation: non-recursive and recursive structure
(repeated from (3.36))
a. IP[clause1]IP IP[par]IP IP[clause2]IP √Align R (S/IP), √Wrap-CP,
√Match clause
b. IP[XP]IP IP[par]IP IP[XP]IP *Match phrase
c. IP[clause1 IP[par]IP clause2]IP *Align R (S/IP), *Wrap-CP,
*Match clause
d. IP[XP IP[par]IP XP]IP √Match phrase (if XP
matched by ip)

227
228  Final discussion

First, FPCs and NRRCs are phrased separately in a non-recursive struc-


ture by default (see Table 4.16), following (5.1)a, in accordance with the
predictions that these kinds of parentheticals are phrased separately in
their own IP. These types of parentheticals thus satisfy the syntax–prosody
interface constraints Align R (the right edge of the parenthetical clause
coincides with the right edge of an IP), Wrap (the parenthetical is wrapped
in one IP) and Match (the parenthetical is matched by the IP). However,
while the parenthetical phrasing satisfies the interface constraints, there
may still be violations due to the phrasing of the host material. Nespor
and Vogel (1986: 189f) observe that the strings preceding and following
the parenthetical form IPs on their own, too, even if they would not do so
otherwise. This is – ignoring for the moment the few exceptions – exactly
what we observe with FPCs and NRRCs in the present data. Unless the
parenthetical is preceded by a syntactic clause, however, a separate IP to
the left of the parenthetical is not predicted by Align/Wrap or Match (see
(5.2)); what is predicted is a weaker boundary of type ip or prosodic word,
depending on the constituency of the string to the left of the parenthet-
ical. Here the presence of the parenthetical affects the prosodic phrasing
of its host, resulting in the promotion of weaker prosodic boundaries to
stronger ones. Consider the examples in (5.3) through (5.7). The NRRC
in (5.3) follows a clause and is followed by a clause, thus phrasing into
three subsequent IPs, as attested in the data and as predicted by Align/
Wrap and by Match clause. In (5.4), the NRRC follows the subject noun
phrase, which, according to the Align, Wrap and Match constraints, is
predicted to be followed by an ip boundary rather than an IP boundary.
This boundary, however, has been promoted to an IP boundary due to
the presence of the parenthetical. The same holds for (5.5) (repeated from
(4.34) above; see there for tonal analysis), which has an as-parenthetical
clause between the subject noun phrase of the host and the VP. In (5.6),
the and-parenthetical is inserted between auxiliary and main verb, a pos-
ition in which no IP or ip boundary is predicted. In this particular case,
there is a H* L-H% nuclear tune associated with are preceding the paren-
thetical, which, also upon inspection of the context (i.e. the immediately
preceding corpus items) may be interpreted as a focus accent (interpret-
ation: are you sure they ARE doing what they can, not just talking about
it), which may be responsible for the IP boundary to its right (e.g. Selkirk
2005). This also holds for the example in (5.7), where the word preceding
the parenthetical is explicitly emphasized. The syntax–prosody interface
constraints are overridden here by a focus constraint, which, in interaction
with eurhythmy constraints, is responsible for the boundary to the right of
the focused word (see, for example, Selkirk 2005; Truckenbrodt 2007). It
remains to be tested whether the promotion of a minor boundary to an IP
boundary is only possible in contexts like these, i.e. involving focus. This
is impossible to say on the basis of the current data.
Final discussion  229

(5.3) NRRC phrased according to (5.1)a, preceded and followed by a


clause (pattern (5.2)a)
a. Monday the twenty-fifth of May is a bank holiday which is half
term so we don’t have Monday Tuesday classes practically all the
children that I coach will be off (ICE-GB: s1a-083 #130)
b. IP1[Monday the twenty-fifth of May is a bank holiday]IP1
IP2[which is half term so we don’t have Monday Tuesday classes]IP2
IP3[practically all the children that I coach will be off]IP3

(5.4) NRRC phrased according to (5.1)a, preceded by a phrase (pattern


(5.2)b)
a. The Villa Somalia which was Siad Barre’s official residence in
Mogadishu still lies abandoned … (ICE-GB: s2b-023 #60)
b. IP1[The Villa Somalia]IP1 IP2[which was Siad Barre’s official resi-
dence in Mogadishu]IP2 IP3[still lies abandoned]IP3
(5.5) FPC (as-parenthetical clause) phrased according to (5.1)a, preceded
by a phrase (pattern (5.2)b)
a. Uhm the title of the talk as you’ll see on the handout is Renewing
the Connection a phrase of Firth’s (ICE-GB: s2a-030 #4)
b. IP1[Uhm the title of the talk]IP1 IP2[as you’ll see on the handout]IP2
IP3[is …

(5.6) FPC (and-parenthetical clause) phrased according to (5.1)a, pre-


ceded by a non-constituent syntactic string
a. Do to answer the the the questioner’s main point do you believe
that the British Government and the American Government
are and I’m quoting his quote from Resolution sixty Six Sixty uhm
supporting all efforts in regard to creating intensive negotiations
between Iraq and Kuwait (ICE-GB: s1b-035 #68)
b. IP1[Do to answer the the the questioner’s main point]IP1 IP2[do
you believe that]IP2 IP3[the British Government and the American
Government are]IP3 IP4[and I’m quoting his quote from Resolution
sixty Six Sixty]IP4 IP5[supporting all efforts …
(5.7) FPC (and-parenthetical clause) phrased according to (5.1)a, pre-
ceded by a non-constituent syntactic string
a. Will the Minister confirm that come the single uh Common
Market that three hundred million EEC nationals could and I
emphasise could seek employment in this country without the
need to obtain a work permit (ICE-GB: s1b-059 #40)
b. … IP1[that three hundred million EEC nationals could]IP1 IP2[and
I emphasise could]IP2 IP3[seek employment in this country …
It can be argued that because the presence of an IP boundary always coin-
cides with the presence of an ip boundary, (5.2)b does not violate Match
230  Final discussion

phrase. However, in order to arrive at the stronger IP boundary, the weaker


boundary will have to be promoted to a stronger one.
The example in (5.8), repeated from (2.44)e, illustrates a parenthetical
clause phrased in a recursive structure, i.e. phrased according to (5.1)b. The
phrasing is in line with the predictions in (5.2), since the parenthetical is not
preceded by any syntactic constituent demanding a right edge boundary.
(5.8) FPC phrased according to (5.1)b: in a recursive structure
a. That’s a little bit of uh how shall I put it uh uhm uh arrogance
that has still got to be eliminated uh in my life (ICE-GB: s1b-
041 #13)
b. IP1[That’s a little bit of IP2[how shall I put it]IP2 arrogance]IP1
IP3[that …

Careful experimental work, which places identical FPCs and NRRCs in


varying positions such as utterance-medial following a clause, utterance-
medial following an XP, and utterance-medial following a non-constituent,
and which controls for factors such as focus, length and scope, will be needed
to test whether the predictions with regard to recursive vs. non-recursive
phrasing as a function of syntactic position is borne out. In the present data,
an IP boundary terminating the pre-parenthetical string regularly occurs in
positions, which, given the syntactic structure, suggest the presence of an
ip boundary (or weaker boundary) rather than an IP boundary, i.e. clause-
internally.
The NRRC exception to prosodic separation  – phrasing of the relative
pronoun (sometimes along with a pied-piped preposition) with preceding
material – was motivated in Dehé (2009) by a constraint Shift IP boundary,
which allows the left edge of an IP to shift rightwards beyond an initial elem-
ent of type F, but not beyond that, where F comprises functional elements
such as relative pronouns, coordinating conjunctions and connectives (Dehé
2009: 602). This constraint takes the default structure as input and is thus a
constraint operating not at the syntax–prosody interface but in the phono-
logical component. It is also possible that the position of the IP boundary in
the position following the relative pronoun is a performance phenomenon,
e.g. as a device to hold the floor (Dehé 2009: 602).
Second, N-APPs were also phrased separately by default, but not to the
same extent as FPCs and NRRCs. Factors influencing phrasing of N-APPs
that were identified in Chapter 4 above were the syntactic function of the
N-APP and its syntactic type, specifically the proper-noun effect. Other
than prosodic separation, the only phrasing pattern that reached some fre-
quency was phrasing of the N-APP with preceding host material (i.e. fol-
lowing pattern (4.58)b1/(5.1)c). This was related to the syntactic type such
that proper-noun N-APPs were integrated more frequently. Accepting that
N-APPs are Comma Phrases (Potts 2005) and Comma Phrases are illo-
cutionary clauses (Selkirk 2009, 2011), the predictions in (5.1) and (5.2)
Final discussion  231

apply to N-APPs. Phrasing of the N-APP with preceding material violates


Match, but it satisfies Align and Wrap, in line with Selkirk’s (2005) account.
Sentences containing N-APPs with the N-APP phrased separately in a non-
recursive prosodic structure typically involve promotion of the predicted ip
boundary following the anchor noun phrase to an IP boundary. Sentences
following pattern (4.58)b1/(5.1)c, on the other hand (i.e. the N-APP
phrased together with preceding host material which typically is or includes
the anchor), may involve downgrading of the predicted ip boundary follow-
ing the anchor NP. The factors accounting for these patterns were discussed
in Section 4.3.3 above.
Third, the prosodic phrasing of QTs resembles that of N-APPs in distri-
butional terms. Like N-APPs, QTs are most frequently phrased separately
at a similar rate (N-APPs: 80%, QTs: 73%), and, like for N-APPs, pros-
odic integration with preceding host material is the second most frequent
pattern (N-APPs: 14%, QTs: 21%). However, N-APPs and QTs which are
phrased together with preceding host material differ with respect to prom-
inence. While QTs integrated in this way are unstressed and follow the
nuclear syllable in their domain, N-APPs phrased in this way typically bear
nuclear prominence. Crucially, in this context, prosodic integration relates
to phrasing only, i.e. it corresponds to phrasing of the parenthetical together
with host material. The integrated QT is postnuclear, thus corresponding
to Ladd’s (1981) postnuclear tag, which, he argues, is integrated in meaning
and intonation. The integrated N-APP, on the other hand, is nuclear, which
makes sense, given that N-APPs provide discourse-new or otherwise salient
information (Meyer 1992; Potts 2005). N-APPs and QTs also differ with
respect to the statistically significant factors affecting prosodic phrasing. For
example, while the syntactic position of the parenthetical does not reach sig-
nificance for N-APPs, it does for QTs, such that QTs which are phrase- or
clause-final are separate more often than QTs in clause-medial position; see
Sections 4.3.3.4 and 4.3.6.4.
Given the similarities between QTs on the one hand and CCs and RVs
on the other in terms of their length and prosodic weight (QTs typically
consist of not more than three syllables), as well as in terms of their syn-
tactic deficiency and the variety of discourse functions they may serve, the
possibility was suggested in Section 2.4 that QTs, CCs and RVs may group
together in terms of prosodic realization. This was not borne out; instead,
QTs are phrased separately, following the default pattern of FPCs, NRRCs
and N-APPs, much more often (i.e. in 73% of cases) than CCs and RVs. The
high proportion of prosodically separate QTs is perhaps surprising, espe-
cially when compared with the results for CCs and RVs. Following Ladd’s
(1981) account, several aspects of which have been corroborated by Dehé
and Braun (2013), this result implies that QTs typically state or assert the
speaker’s assumption or signal a hedge to the proposition expressed by the
host, rather than being truly interrogative. A careful analysis of the relation
232  Final discussion

between pragmatic factors and discourse factors on the one hand and pros-
odic realization on the other awaits further research.1 CCs and RVs, on the
other hand, do behave differently from the other four types such that they
do not typically follow the pattern of prosodic separation; prosodic integra-
tion along with lack of prominence is instead the rule rather than the excep-
tion. Moreover, there is a lot more variation in the prosodic phrasing of CCs
and RVs. In particular, when integrated, CCs and, to a lesser extent, RVs
may also be followed by host material in their domain, following (4.128)d, a
pattern which is non-existent with FPCs and NRRCs and rare with N-APPs
and QTs (see Table 4.16).
So what is the reason for CCs and RVs to behave differently from the
other four types? To answer this question it makes sense to first take a look
at the predictions again. Variation in prosodic phrasing was predicted for
CCs in Section 2.3.4 on the basis of previous literature. Specifically, inte-
grated prosodic phrasing patterns have previously been observed with CCs
(at least since Armstrong and Ward 1926) and these patterns have typically
been accounted for in terms of factors related to meaning and interpretation
in context (e.g. Dehé and Wichmann 2010a) and prosodic weight (e.g. Peters
2006; Dehé 2009). While, on the basis of the data investigated here, I would
underline that these factors play a role, a systematic analysis of the rela-
tion between meaning and prosody, as well as a controlled study of length
effects awaits further research. Given that CCs and RVs are full clauses (cf.
Kluck and de Vries’ 2012 analysis in Sections 2.3.4 and 2.3.5 above), they
are subject to the same predictions regarding the syntax–­prosody interface
constraints as the other four types of parentheticals. In particular, according
to Match Theory, CCs and RVs should match an IP in the prosodic struc-
ture. However, remember that all CCs occurring in a position preceded
by a syntactic constituent such as a subject, full clause or relative pronoun
were removed from the analysis (see Section 2.3.4 above). In the current
data, CCs are thus typically preceded by strings which are not constituents
in the syntax, and if they are, they are constituents of type XP, not clause.
Therefore, phrasing of the string preceding the CC into a separate IP is
not predicted by either Match or Align/Wrap. This excludes non-recursive
prosodic phrasing of type (5.1)a and predicts recursive prosodic structure as
in (5.1)b. Seen from this perspective, it is surprising that 18 per cent of CCs
were in fact phrased according to (5.1)a, and that only 5 per cent of CCs
were phrased in a CPD. RVs did not undergo any similar selection process.
Accordingly, there are RVs in the data which follow clausal constituents and

1
Notice that, in the data reported on above, there is at least one exception to the assumed
correspondence between prosodic integration and true interrogative meaning. In example
(4.113), the QT is phrased separately with a nuclear accent, but there can be no doubt that it
is truly interrogative, also perceived as interrogative by the listener, who steps in and offers a
reply immediately. The results call for a careful analysis of the relation between prosody and
meaning in context, ideally done in joint work by prosodists and pragmatists.
Final discussion  233

which are phrased separately in a non-recursive prosodic structure accord-


ing to the predictions based on the interface constraints; see (4.99) for an
example.
The reason for the frequent integration of CCs and RVs can perhaps
best be found in the semantics. Syntactically, they are clauses and should
therefore match an IP in the prosody. However, semantically, they have been
argued to resemble evidentials (e.g. Rooryck 2001; Brinton 2008; Scheffler
2009; among many others) or to be mixed expressions between an eviden-
tial and a CI meaning (Scheffler 2009). Rooryck (2001) has also argued that
RVs have evidential meaning. If CCs/RVs are evidentials, they are not CIs
in Potts’ (2005, 2007) sense; unlike CIs, they have an impact on the at-issue
entailment such that they indicate that the proposition expressed by the
host is to be added to the common ground but that the speaker is not abso-
lutely certain. Recall (2.90): the speaker is not absolutely certain that ‘Peter
will come today’. Thus the proposition is weakened by the evidential CC.
CCs may thus not be Comma Phrases in Potts’ semantic sense, even though
they are clauses in the syntax. Potts (2005) relates the CI semantics to pros-
odic separation. If, on the other hand, CCs/RVs function as side comments
in the domain of CIs, they are Comma Phrases semantically and syntactic-
ally. Remember that Scheffler (2009) argues that the effect on the epistemic
threshold is in fact the primary function of elements such as I think. The
function served by a particular CC or RV may or may not require or result
in prominence. In particular, expressions of conventional implicatures may
require prominence, while those that function as evidentials may not, and
mixed expressions may go in one or the other direction, depending on which
function prevails. In the case of lack of prominence, the unaccented CC or
RV will be prosodically integrated or encliticized. The details of prosodic
integration (whether the target element phrases with preceding or follow-
ing material or both) may well depend on prosodic factors such as length
and weight, or on factors such as semantic scope. Accordingly, the syntax–
prosody interface constraints Align/Wrap and Match may be overridden by
other constraints, for example phonological constraints or constraints which
relate to the prosody–meaning interface. Like the results for QTs (see foot-
note 1), this issue will best be addressed in joint work by prosodists and
pragmatists.
References

Ackema, Peter and Ad Neeleman. 2004. Beyond Morphology: Interface Conditions on


Word Formation. Oxford University Press.
Adli, Aria. 2005. Die Extraktions- vs. Parenthesedebatte: eine Diskussion anhand
französischer Daten. In Georg Kaiser (ed.), Deutsche Romanistik  – generativ,
13–30. Tübingen: Narr.
Aijmer, Karin. 1997. Think  – an English modal particle. In Toril Swan and
Olaf Jansen Westvik (eds.), Modality in Germanic Languages: Historical and
Comparative Perspective, 1–47. Berlin/New York: Mouton de Gruyter.
Altmann, Hans. 1981. Formen der ‘Herausstellung’ im Deutschen: Rechtsversetzung,
Linksversetzung, Freies Thema und verwandte Konstruktionen (Linguistische
Arbeiten). Tübingen: Niemeyer.
Armstrong, Lilias E. and Ida C. Ward. 1926. A Handbook of English Intonation.
Cambridge: Heffer & Sons.
Arnold, Doug. 2007. Non-restrictive relatives are not orphans. Journal of Linguistics
43(2), 271–309.
Asher, Nicholas. 2000. Truth conditional discourse semantics for parentheticals.
Journal of Semantics 17(1), 31–50.
Astruc-Aguilera, Lluïsa. 2005. The Intonation of Extra-Sentential Elements in
Catalan and English. Doctoral dissertation, University of Cambridge.
Astruc-Aguilera, Lluïsa and Francis Nolan. 2007a. Variation in the intonation
of sentential adverbs in English and Catalan. In Carlos Gussenhoven and
Tomas Riad (eds.), Tones and Tunes, Volume i: Typological and Comparative
Studies in Word and Sentence Prosody, 235–64. Berlin: Mouton de
Gruyter.
  2007b. Variation in the intonation of extra-sentential elements. In Pilar Prieto,
Joan Mascaró and Maria-Josep Solé (eds.), Segmental and Prosodic Issues in
Romance Phonology, 85–107. Amsterdam: John Benjamins.
Auran, Cyril and Rudy Loock. 2011. The prosody of discourse functions: the case
of appositive relative clauses in spoken British English. Corpus Linguistics and
Linguistic Theory 7(2), 181–201.
Baayen, Harald R. 2008. Analyzing Linguistic Data: A Practical Introduction to
Statistics Using R. Cambridge University Press.
Bald, Wolf-Dietrich. 1980. English tag-questions and intonation. In Kuno
Schuhmann (ed.), Anglistentag 1979: Vorträge und Protokolle, 263–91. Berlin:
Technische Universität Berlin.

234
References  235

Beckman, Mary E. and Gayle Ayers Elam. 1993. Guidelines for ToBI labelling (ver-
sion 3.0). www.ling.ohio-state.edu//~tobi/ame_tobi/labelling_guide_v3.pdf
(27 March 2013): The Ohio State University Research Foundation.
Beckman, Mary E., Julia Hirschberg and Stefanie Shattuck-Hufnagel. 2005. The
original ToBI system and the evolution of the ToBI framework. In Sun-Ah Jun
(ed.), Prosodic Typology, 9–54. Oxford University Press.
Beckman, Mary E. and Janet B. Pierrehumbert. 1986. Intonational structure in
Japanese and English. Phonology Yearbook 3, 255–309.
Behaghel, Otto. 1909. Beziehungen zwischen Umfang und Reihenfolge von
Satzgliedern. Indogermanische Forschungen 25, 110–42.
  1930. Von deutscher Wortstellung. Zeitschrift für Deutschkunde 44, 81–9.
Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward
Finegan. 1999. Longman Grammar of Spoken and Written English. Harlow:
Pearson Education.
Bing, Janet M. 1985. Aspects of English Prosody. New York/London: Garland.
Blaauw, Eleonora. 1994. The contribution of prosodic boundary markers to the per-
ceptual difference between read and spontaneous speech. Speech Communication
14, 359–75.
Blakemore, Diane. 2005. and-parentheticals. Journal of Pragmatics 37(8), 1165–81.
  2006. Divisions of labour: the analysis of parentheticals. Lingua 116(10),
1670–87.
  2007. ‘Or’-parentheticals, ‘that is’-parentheticals and the pragmatics of reformu-
lation. Journal of Linguistics 43(2), 311–39.
Boersma, Paul. 2001. Praat, a system for doing phonetics by computer. GLOT
International 5(9–10), 341–5.
Boersma, Paul and David Weenink. 2012. Praat: doing phonetics by computer
[computer program]. Version 5.3.04. www.praat.org.
Bolinger, Dwight. 1986. Intonation and Its Parts: Melody in Spoken English. London:
Arnold.
  1989. Intonation and Its Uses: Melody in Grammar and Discourse.
London/Melbourne: Edward Arnold.
Brinton, Laurel J. 2008. The Comment Clause in English: Syntactic Origins and Pragmatic
Developments (Studies in English Language). Cambridge University Press.
Burton-Roberts, Noel. 1999. Language, linear precedence and parentheticals. In
Peter Collins and David Lee (eds.), The Clause in English, 33–52. Amsterdam/
Philadelphia: John Benjamins.
  2006. Parentheticals. In E. K. Brown (ed.), Encyclopaedia of Language and
Linguistics, 179–82. Elsevier Science.
Bussmann, Hadumod. 1996. Routledge Dictionary of Language and Linguistics.
London/New York: Routledge.
Cattell, Ray. 1973. Negative transportation and tag questions. Language 49(3),
612–39.
Chomsky, Noam. 1981. Lectures on Government and Binding: The Pisa Lectures.
Dordrecht: Foris.
  1995. The Minimalist Program. Cambridge, MA: MIT Press.
Cole, Jennifer, Yoonsook Mo and Soondo Baek. 2010. The role of syntactic structure
in guiding prosody perception with ordinary listeners and everyday speech.
Language and Cognitive Processes 25(7/8/9), 1141–77.
236  References

Cooper, William E. and Jeanne Paccia-Cooper. 1980. Syntax and Speech. Cambridge,
MA: Harvard University Press.
Cooper, William E. and John M. Sorensen. 1981. Fundamental Frequency in Sentence
Production. New York/Heidelberg/Berlin: Springer-Verlag.
Corver, Norbert and Craig Thiersch. 2002. Remarks on parentheticals. In Marc van
Oostendorp and Elena Anagnostopoulou (eds.), Progress in Grammar: Articles at
the 20th Anniversary of the Comparison of Grammatical Models Group in Tilburg.
www.meertens.knaw.nl/books/progressingrammar/ (30 March 2013).
Cruttenden, Alan. 1997. Intonation (2nd edn). Cambridge University Press.
Crystal, David. 1969. Prosodic Systems and Intonation in English. Cambridge
University Press.
  1972. The intonation system of English. In Dwight Bolinger (ed.), Intonation:
Selected Readings, 110–36. Harmondsworth: Penguin.
Culicover, Peter. 1992. English tag questions in Universal Grammar. Lingua 88(3–
4), 193–226.
D’Avis, Franz Josef. 2005. Über Parenthesen. In Franz Josef D’Avis (ed.),
Deutsche Syntax: Empirie und Theorie, 259–79. Göteborg: Acta Universitatis
Gothoburgensis.
D’Imperio, Mariapaola, Gorka Elordieta, Sónia Frota, Pilar Prieto and Marina
Vigário. 2005. Intonational phrasing in Romance: the role of syntactic and
prosodic structure. In Sónia Frota, Marina Vigário and Maria J. Freitas (eds.),
Prosodies: With Special Reference to Iberian Languages, 59–97. Berlin/New
York: Mouton de Gruyter.
Dehé, Nicole. 2007. The relation between syntactic and prosodic parenthesis.
In Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals (Linguistik
Aktuell/Linguistics Today 106), 261–84. Amsterdam/Philadelphia: John
Benjamins.
  2009. Clausal parentheticals, intonational phrasing, and prosodic theory. Journal
of Linguistics 45(3), 569–615.
Dehé, Nicole and Bettina Braun. 2013. The prosody of question tags in English.
English Language and Linguistics 17(1), 129–56.
Dehé, Nicole and Yordanka Kavalova. 2006. The syntax, pragmatics and prosody of
parenthetical what. English Language and Linguistics 10(2), 289–320.
Dehé, Nicole and Yordanka Kavalova (eds.). 2007. Parentheticals. Amsterdam/
Philadelphia: John Benjamins.
Dehé, Nicole and Anne Wichmann. 2010a. The multifunctionality of epistemic par-
entheticals in discourse: prosodic cues to the semantic–pragmatic boundary.
Functions of Language 17(1), 1–28.
  2010b. Sentence-initial I think (that) and I believe (that): prosodic evidence for
use as main clause, comment clause and discourse marker. Studies in Language
34(1), 36–74.
Del Gobbo, Francesca. 2007. On the syntax and semantics of appositive relative
clauses. In Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals, 173–201.
Amsterdam/Philadelphia: John Benjamins.
Dixon, Robert M.W. 1991. A New Approach to English Grammar, on Semantic
Principles. Oxford: Clarendon.
Downing, Bruce T. 1970. Syntactic Structure and Phonological Phrasing in English.
Doctoral dissertation, University of Texas at Austin.
References  237

  1973. Parenthesization rules and obligatory phrasing. Papers in Linguistics 6,


108–28.
Elordieta, Gorka, Sónia Frota and Marina Vigário. 2005. Subjects, objects and
intonational phrasing in Spanish and Portuguese. Studia Linguistica 59(2/3),
110–43.
Emonds, Joseph. 1973. Parenthetical clauses. In C. Corum, T. C. Smith-Stark and
A. Weiser (eds.), You’ll Take the High Node and I’ll Take the Low Node, 333–47.
Chicago Linguistic Society.
  1976. A Transformational Approach to English Syntax. New York: Academic
Press.
  1979. Appositive relatives have no properties. Linguistic Inquiry 10(2), 211–43.
Espinal, M. Teresa. 1991. The representation of disjunct constituents. Language
67(4), 726–62.
Fabb, Nigel. 1990. The difference between English restrictive and nonrestrictive
relative clauses. Journal of Linguistics 26(1), 57–77.
Fagyal, Zsuzsanna. 2002. Prosodic boundaries in the vicinity of utterance-medial
parentheticals in French. Probus 14(1), 93–111.
Ferreira, Fernanda. 1991. Effects of length and syntactic complexity on initi-
ation times for prepared utterances. Journal of Memory and Language 30(2),
210–33.
  1993. Creation of prosody during sentence production. Psychological Review
100(2), 233–53.
Féry, Caroline. 2010. Indian Languages as intonational ‘phrase languages’. In Imtiaz
Hasnain and Shreesh Chaudhury (eds.), Problematizing Language Studies.
Festschrift for Rama Agnihotri, 288–312. Delhi: Aakar Books.
Fortmann, Christian. 2007. The complement of reduced parentheticals. In Nicole
Dehé and Yordanka Kavalova (eds.), Parentheticals, 89–119. Amsterdam/
Philadelphia: John Benjamins.
Fraser, Bruce. 1980. Conversational mitigation. Journal of Pragmatics 4(4), 341–50.
Frazier, Lyn, Charles Clifton and Katy Carlson. 2004. Don’t break, or do: prosodic
boundary preferences. Lingua 114(1), 3–27.
Frota, Sónia. 2000. Prosody and Focus in European Portuguese: Phonological Phrasing
and Intonation. New York/London: Garland.
  2012. Prosodic structure, constituents and their implementation. In Abigail C.
Cohn, Cécile Fougeron and Marie K. Huffman (eds.), The Oxford Handbook of
Laboratory Phonology, 255–65. Oxford University Press.
Gee, James Paul and Francois Grosjean. 1983. Performance structures: a psycholin-
guistic and linguistic appraisal. Cognitive Psychology 15(4), 411–58.
Ghini, Mirco. 1993. Phi-formation in Italian: a new proposal. In Carrie Dyck (ed.),
Toronto Working Papers in Linguistics 12, 41–78. Toronto.
Grenoble, Lenore A. 2004. Parentheticals in Russian. Journal of Pragmatics 36(11),
1953–74.
Grewendorf, Günther. 1988. Aspekte der deutschen Syntax: Eine Rektions–Bindungs–
Analyse. Tübingen: Gunter Narr.
Grice, H. P. 1975. Logic and conversation. In P. Cole and J. Morgan (eds.), Syntax
and Semantics 3: Speech Acts, 41–58. Academic Press.
Grice, Martine, D. Robert Ladd and Amalia Arvaniti. 2000. On the place of phrase
accents in intonational phonology. Phonology 17(2), 143–85.
238  References

Griffiths, James. 2012. Remarks on alleged subject-oriented parenthetical construc-


tions. Presented at the workshop Parenthetical verbs: hypotaxis, parataxis or
parenthesis? Les verbes parenthétiques: hypotaxe, parataxe ou parenthèse?
24–26 May 2012, Université Paris Ouest, Nanterre La Dé fense, France.
  In press. Parenthetical verb constructions, fragment answers, and constituent
modification. Natural Language and Linguistic Theory.
Griffiths, James and Mark de Vries. 2013. The syntactic integration of appositives:
evidence from fragments and ellipsis. Linguistic Inquiry 44(2), 332–44.
Gundel, Jeanette K. 1988. Universals of topic-comment structure. In Michael
Hammond, Edith Moravcsik and Jessica Wirth (eds.), Studies in Syntactic
Typology, 209–39. Amsterdam/Philadelphia: John Benjamins.
Günes, Güliz. 2012. The prosody of parenthesis in a phrase language: a study in
Turkish. Presented at the workshop Parenthetical verbs: hypotaxis, parataxis
or parenthesis? Les verbes parenthétiques: hypotaxe, parataxe ou parenthèse?
24–26 May 2012, Université Paris Ouest, Nanterre La Dé fense, France
Gussenhoven, Carlos. 1984. On the Grammar and Semantics of Sentence Accents.
Dordrecht: Foris.
  1990. Tonal association domains and the prosodic hierarchy in English. In Susan
M. Ramsaran (ed.), Studies in the Pronunciation of English. a Commemorative
Volume in Honour of A. C. Gimson, 27–37. London: Routledge.
  2004. The Phonology of Tone and Intonation. Cambridge University Press.
Gussenhoven, Carlos and Toni Rietveld. 1992. Intonation contours, prosodic struc-
ture and preboundary lengthening. Journal of Phonetics 20, 283–303.
Haegeman, Liliane. 1991. Parenthetical adverbials: the radical orphanage approach.
In S. Chiba, A. Ogawa, Y. Fuiwara, N. Yamada, O. Koma and T. Yagi (eds.),
Aspects of Modern English Linguistics: Papers presented to Masatomo Ukaji on his
60th Birthday, 232–54, Tokyo: Kaitakusha.
Haider, Hubert. 1993. Deutsche Syntax: Generativ: (Vorstudien zur Theorie einer pro-
jektiven Grammatik). Tübingen: Gunter Narr.
  2005. Parenthesen – Evidenz aus Bindungstheorie. In Franz Josef D’Avis (ed.),
Deutsche Syntax: Empirie und Theorie, 281–93. Göteborg: Acta Universitatis
Gothoburgensis.
Hand, Michael. 1993. Parataxis and parentheticals. Linguistics and Philosophy 16(5),
495–507.
Hanote, Sylvie. 2004. Des introducteurs de discours aux indices de frayage. In Juan
Manuel Lopez Muñoz, Sophie Marnette and Laurence Rosier (eds.), Le dis-
cours rapporté dans tous ses états: Actes du Colloque International Bruxelles – 8–11
novembre 2001, 538–48. Paris: L’Harmattan.
Hellmuth, Sam. 2008. Prosodic weight and phonological phrasing in Cairene Arabic.
Chicago Linguistic Society (CLS) 40, 97–111.
Heringa, Herman. 2011. Appositional Constructions. Groningen: LOT: Netherlands
Graduate School of Linguistics.
  2012. A multidominance approach to appositional constructions. Lingua 122(6),
554–81.
Hirschberg, Julia. 2000. A corpus-based approach to the study of speaking style.
In Merle Horne (ed.), Prosody: Theory and Experiment, 335–50. Dordrecht:
Kluwer.
  2004. Pragmatics and intonation. In Laurence R. Horn and Gregory Ward (eds.),
The Handbook of Pragmatics, 515–37. Oxford: Blackwell.
References  239

Hock, Hans Henrich and Indranil Dutta. 2010. Prosody vs. syntax: prosodic
rebracketing of final vocatives in English. In Proceedings of Speech Prosody, 5th
International Conference, Chicago, IL, 11–14 May 2010.
Hoffmann, Ludger. 1998. Parenthesen. Linguistische Berichte 175, 299–328.
Hooper, Joan B. 1975. On assertive predicates. In John P. Kimball (ed.), Syntax and
Semantics iv, 91–124. New York: Academic Press.
Howell, Peter and Karima Kadi-Hanifi. 1991. Comparison of prosodic properties
between read and spontaneous speech material. Speech Communication 10(2),
163–9.
Huddleston, Rodney. 1970. Two approaches to the analysis of tags. Journal of
Linguistics 6(2), 215–22.
Huddleston, Rodney and Geoffrey K. Pullum. 2002. The Cambridge Grammar of the
English Language. Cambridge University Press.
Ifantidou, Elly. 2001. Evidentials and Relevance (Pragmatics and Beyond).
Amsterdam/Philadelphia: John Benjamins.
Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge,
MA: MIT Press.
Jespersen, Otto. 1937. Analytic Syntax. London: Allen & Unwin.
Jun, Sun-Ah. 2003. The effect of phrase length and speech rate on prosodic phras-
ing. Presented at the 15th International Congress of Phonetic Sciences (ICPhS)
Barcelona, Spain.
Jun, Sun-Ah (ed.) 2005. Prosodic Typology: The Phonology of Intonation and Phrasing.
Oxford University Press.
Kaltenböck, Gunther. 2007. Position, prosody, and scope: the case of English com-
ment clauses. Vienna English Working Papers 16(1), 3–38.
  2008. Prosody and function of English comment clauses. Folia Linguistica 42(1),
83–134.
  2009a. Is that a filler? On complementizer use in spoken object clauses. Vienna
English Working Papers 18(1), 28–63.
  2009b. Initial I think: main or comment clause? Discourse and Interaction 2(1),
49–70.
  2010. Pragmatic functions of parenthetical I think. In Gunther Kaltenböck,
Wiltrud Mihatsch and Stefan Schneider (eds.), New Approaches to Hedging,
243–70. Bingley: Emerald.
Kaltenböck, Gunther, Bernd Heine and Tania Kuteva. 2011. On thetical grammar.
Studies in Language 35(4), 852–97.
Kärkkäinen, Elise. 2003. Epistemic Stance in English Conversation: A Description of
Its Interactional Functions, with a Focus on I think. Amsterdam/Philadelphia:
John Benjamins.
Kavalova, Yordanka. 2007. And-parenthetical clauses. In Nicole Dehé and Yordanka
Kavalova (eds.), Parentheticals, 145–72. Amsterdam/Philadelphia: John
Benjamins.
Kawahara, Shigeto. 2012. The intonation of nominal parentheticals in Japanese.
In Toni Borowsky, Shigeto Kawahara, Takahito Shinya and Mariko Sugahara
(eds.), Prosody Matters: Essays in Honor of Elisabeth Selkirk, 304–40. Sheffield:
Equinox.
Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT
Press.
240  References

Kearns, Kate. 2007. Epistemic verbs and zero complementizer. English Language
and Linguistics 11(3), 475–505.
Kingdon, Roger. 1958. The Groundwork of English Intonation. London: Longman.
Kiziak, Tanja. 2007. Long extraction or parenthetical insertion? Evidence from
judgement studies. In Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals,
121–44. Amsterdam/Philadelphia: John Benjamins.
Kluck, Marlies. 2011. Sentence Amalgamation. Groningen: LOT: Netherlands
Graduate School of Linguistics.
Kluck, Marlies and Mark de Vries. 2012. On V2, gaps and operators in comment and
reporting parentheticals. Presented at the workshop Parenthetical verbs: hypo-
taxis, parataxis or parenthesis? 24–26 May 2012 (Les verbes parenthétiques:
hypotaxe, parataxe ou parenthèse? 24–26 May 2012) Université Paris Ouest,
Nanterre La Défense, France.
Knowles, Gerry. 1991. Prosodic labelling: the problem of tone group boundaries.
In Stig Johansson and Anna-Brita Stenström (eds.), English Computer Corpora:
Selected Papers and Research Guide, iii, 149–61. Berlin/New York: Mouton de
Gruyter.
Knowles, John. 1980. The tag as a parenthetical. Studies in Language 4(3), 379–409.
König, Ekkehard. 1977. Form und Funktion: Eine funktionale Betrachtung ausgewähl-
ter Bereiche des Englischen. Tübingen: Niemeyer.
Kratzer, Angelika and Elisabeth Selkirk. 2007. Phase theory and prosodic spellout:
the case of verbs. The Linguistic Review 24(2/3), 93–135.
Kreiman, J. 1982. Perception of sentence and paragraph boundaries in natural con-
versation. Journal of Phonetics 10, 163–75.
Krivokapić, Jelena. 2007. Prosodic planning: effects of phrasal length and complex-
ity on pause duration. Journal of Phonetics 35(2), 162–79.
Kutik, Elanah J., William E. Cooper and Suzanne Boyce. 1983. Declination of fun-
damental frequency in speakers’ production of parenthetical and main clauses.
Journal of the Acoustical Society of America 73(5), 1731–8.
Ladd, D. Robert. 1978. Stylized intonation. Language 54(3), 517–40.
  1981. A first look at the semantics and pragmatics of negative questions and tag
questions. In Chicago Linguistic Society (ed.), Papers from the 17th regional
meeting (CLS 17), 164–71. Chicago Linguistic Society.
  1986. Intonational phrasing: the case for recursive prosodic structure. Phonology
Yearbook 3, 311–40.
  1996. Intonational Phonology. Cambridge University Press.
  2008. Intonational Phonology (2nd edn). Cambridge University Press.
Lahiri, Aditi and Frans Plank. 2010. Phonological phrasing in Germanic: the judge-
ment of history, confirmed through experiment. Transactions of the Philological
Society 108(3), 370–98.
Lakoff, George. 1974. Syntactic amalgams. In Papers from the 10th Regional Meeting:
Chicago Linguistic Society, 321–44. Chicago Linguistic Society.
Lakoff, Robin. 1969. A syntactic argument for negative transportations. In Robert I.
Binnick, Alice Davison, Georgia M. Green and Jerry C. Morgan (eds.), Papers
from the 5th regional meeting of the Chicago Linguistics Society, 140–7. University
of Chicago, Dept of Linguistics.
Lambrecht, Knud. 1994. Information Structure and Sentence Form: Topic, Focus, and
the Mental Representations of Discourse Referents. Cambridge University Press.
References  241

Lee-Goldman, Russell. 2012. Supplemental relative clauses: internal and external


syntax. Journal of Linguistics 48(3), 573–608.
Lehiste, Ilse. 1973. Phonetic disambiguation of syntactic ambiguity. Glossa 7(2),
107–22.
Local, John. 1992. Continuing and restarting. In Peter Auer and Aldo Di Luzio
(eds.), The Contextualization of Language, 273–96. Amsterdam/Philadelphia:
John Benjamins.
Loock, Rudy. 2007. Appositive relative clauses and their functions in discourse.
Journal of Pragmatics 39(2), 336–62.
  2010. Appositive Relative Clauses in English: Discourse Functions and Competing
Structures (Studies in Discourse and Grammar). Amsterdam/Philadelphia:
John Benjamins.
McCawley, James D. 1982. Parentheticals and discontinuous constituent structure.
Linguistic Inquiry 13(1), 91–106.
  1998. The Syntactic Phenomena of English (2nd edn). The University of Chicago
Press.
Meyer, Charles F. 1992. Apposition in Contemporary English (Studies in English
Language). Cambridge University Press.
Mittwoch, Anita. 1979. Final parentheticals with English questions – their illocut-
ionary function and grammar. Journal of Pragmatics 3(5), 401–12.
Mo, Yoonsook. 2008. Duration and intensity as perceptual cues for naïve listen-
ers’ prominence and boundary perception. Proceedings of the 4th International
Conference on Speech Prosody, Campinas, Brazil, 6–9 May 2008.
Moya Guijarro, Arsenio Jesús. 2005. The introduction of topics in two discourse
genres. Word 56(2), 201–22.
Murphy, M. Lynne. 1993. Discourse markers and sentential syntax. Studies in the
Linguistic Sciences 23(1), 163–7.
Neeleman, Ad and Hans van de Koot. 2002. The configurational matrix. Linguistic
Inquiry 33(4), 529–74.
Nelson, Gerald, Sean Wallis and Bas Aarts. 2002. Exploring Natural Language:
Working with the British Component of the International Corpus of English.
Amsterdam: John Benjamins.
Nespor, Marina and Irene Vogel. 1986. Prosodic Phonology. Dordrecht: Foris.
Newmeyer, Frederick J. 2012. Parentheticals, ‘fragments’, and the grammar of com-
plementation. Presented at the workshop Parenthetical verbs: hypotaxis, para-
taxis or parenthesis? Parenthetical verbs: hypotaxis, parataxis or parenthesis?
Les verbes parenthétiques: hypotaxe, parataxe ou parenthèse? 24–26 May
2012, Université Paris Ouest, Nanterre La Défense, France.
Nolan, Francis. 2006. Intonation. In Bas Aarts and April McMahon (eds.), The
Handbook of English Linguistics, 433–57. Oxford/Malden, MA: Blackwell.
Nosek, Jirí. 1973. Parenthesis in Modern Colloquial English. Prague Studies in
English (15), 99–116.
O’Connor, Joseph D. and Gordon F. Arnold. 1971. Intonation of Colloquial English:
A Practical Handbook. London: Longman.
O’Connor, Kathleen M. 2008. Aspects de la syntaxe et de l’interprétation de l’apposition
à antécédent nominal. Doctoral dissertation, Université Charles de Gaulle  –
Lille 3, UFR Angellier.
242  References

Palmer, Harold E. 1922. English Intonation with Systematic Exercises. Cambridge:


Heffer & Sons.
Payà, Marta. 2003. Prosody and pragmatics in parenthetical insertions in Catalan.
Catalan Journal of Linguistics 2, 207–27.
Peters, Jörg. 2006. Syntactic and prosodic parenthesis. Proceedings of the 3rd
International Conference on Speech Prosody. Dresden, 2–5 May 2006.
Peterson, Peter. 1999. On the boundaries of syntax. In Peter Collins and David
Lee (eds.), The Clause in English, 229–50. Amsterdam/Philadelphia: John
Benjamins.
  2004. Non-restrictive relatives and other non-syntagmatic relations in an LF
framework. Presented at LFG 2004. Christchurch, New Zealand.
Pierrehumbert, Janet. 1980. The Phonology and Phonetics of English Intonation.
Doctoral dissertation, MIT.
Pierrehumbert, Janet and Julia Hirschberg. 1990. The meaning of intonational
contours in the interpretation of discourse. In Ph R. Cohen, Jerry C. Morgan
and M. Pollack (eds.), Intentions in Communication, 271–311. Cambridge, MA:
MIT Press.
Pitrelli, John F., Mary E. Beckman and Julia Hirschberg. 1994. Evaluation of pros-
odic transcription labeling reliability in the ToBI framework. In 3rd International
Conference on Spoken Language Processing (ICSLP 94). Yokohama, Japan.
Pitt, M. A., Laura Dilley, K. Johnson, S. Kiesling, W. Raymond, E. Hume and E.
Fosler-Lussier. 2007. Buckeye corpus of conversational speech (2nd release).
Columbus, OH Department of Psychology, Ohio State University.
Potts, Christopher. 2002a. The lexical semantics of parenthetical-as and appositive-
which. Syntax 5(1), 55–88.
  2002b. The syntax and semantics of as-parentheticals. Natural Language and
Linguistic Theory 20(3), 623–89.
  2005. The Logic of Conventional Implicatures (Oxford Studies in Theoretical
Linguistics). Oxford University Press.
  2007. Into the conventional-implicature dimension. Philosophy Compass 2(4),
665–79.
Price, Patti, Mari Ostendorf, Stefani Shattuck-Hufnagel and Cynthia Fong. 1991.
The use of prosody in syntactic disambiguation. Journal of the Acoustical
Society of America 90(6), 2956–70.
Prieto, Pilar, Mariapaola D’Imperio and Barbara Gili Fivela. 2005. Pitch accent
alignment in Romance: primary and secondary associations with metrical
structure. Language and Speech 48(4), 359–96.
Quirk, Randolph, Anne P. Duckworth, Jan Svartvik, J. P. L. Rusiecki and A. J. T.
Colin. 1964. Studies in the correspondence of prosodic to grammatical features
in English. In Horace G. Lunt (ed.), Ninth International Congress of Linguists,
679–91. Berlin: Mouton.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik (eds.). 1985.
A Comprehensive Grammar of the English Language. New York: Longman.
Radford, Andrew. 2004. English Syntax: An Introduction. Cambridge University
Press.
Rando, Emily. 1980. Intonation in discourse. In Linda R. Waugh and C. H. van
Schooneveld (eds.), The Melody of Language, 243–78. Baltimore, MD:
University Park Press.
References  243

Rappaport, Gilbert. 1983. On the linguistic function of detachment (comma inton-


ation). In John F. Richardson, Mitchell Marks and Amy Chukerman (eds.),
Papers from the Parasession on the Interplay of Phonology, Morphology and
Syntax, 185–97. Chicago Linguistic Society.
Reinhart, Tanya. 1983. Point of view in language: the use of parentheticals. In Gisa
Rauh (ed.), Essays on Deixis, 169–94. Tübingen: Gunter Narr.
Reis, Marga. 1995. Wer glaubst du hat recht? On so-called extractions from verb-
second clauses and verb-first parenthetical constructions in German. Sprache
und Pragmatik 36, 27–83.
  1996. Extractions from verb-second clauses in German? In Uli Lutz and
Jürgen Pafel (eds.), On Extraction and Extraposition in German, vol. xi, 45–88.
Amsterdam: John Benjamins.
  2002. Wh-movement and integrated parenthetical constructions. In C. Jan-
Wouter Zwart and Werner Abraham (eds.), Studies in Comparative Germanic
Syntax, 3–40. Amsterdam/Philadelphia: John Benjamins.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman
(ed.), Handbook in Generative Syntax, 281–337. Dordrecht: Kluwer.
Roach, Peter. 1994. Conversion between prosodic transcription system: “Standard
British” and ToBI. Speech Communication 15(1–2), 91–9.
Rooryck, Johan. 2001. Evidentiality, Part i. GLOT international 5(4), 125–33.
Ross, John R. 1973. Slifting. In Maurice Gross, Morris Halle and Marcel-Paul
Schützenberger (eds.), The Formal Analysis of Natural Languages, 133–69. The
Hague/Paris: Mouton.
Sadock, Jerrold. 1974. Toward a Linguistic Theory of Speech Acts. New York:
Academic Press.
Safir, Ken. 1986. Relative clauses in a theory of binding and levels. Linguistic Inquiry
17(4), 663–89.
Sandalo, Filomena and Hubert Truckenbrodt. 2002. Some notes on phonological
phrasing in Brazilian Portuguese. MIT Working Papers in Linguistics 42, 285–310.
Savino, Michelina. 2012. The intonation of polar questions in Italian: where is the
rise? Journal of the International Phonetic Association 42(1), 23–48.
Schafer, Amy J., Shari R. Speer, Paul Warren and David S. White. 2000. Intonational
disambiguation in sentence production and comprehension. Journal of
Psycholinguistic Research 29(2), 169–82.
Scheer, Tobias. 2011. Guide to Morphosyntax–Phonology Interface Theories. How
Extra-Phonological Information is Treated in Phonology since Trubetzkoy’s
Grenzsignale. Berlin: Mouton de Gruyter.
Scheffler, Tatjana. 2009. Evidentiality and German attitude verbs. University of
Pennsylvania Working Papers in Linguistics 15(1), 183–92.
Schelfhout, Carla, Peter-Arno Coppen and Nelleke Oostdijk. 2004. Finite comment
clauses in Dutch: a corpus-based approach. Journal of Germanic Linguistics
16(4), 331–50.
Schneider, Stefan. 2007a. Reduced parenthetical clauses as mitigators: a corpus
study of spoken French, Italian and Spanish (Studies in Corpus Linguistics).
Amsterdam/Philadelphia: John Benjamins.
  2007b. Reduced parenthetical clauses in Romance languages: a pragmatic typ-
ology. In Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals, 237–58.
Amsterdam/Philadelphia: John Benjamins.
244  References

Schubiger, Maria. 1958. English Intonation: Its Form and Function. Tübingen: Max
Niemeyer.
Schwyzer, Eduard. 1939. Die Parenthese im engern und im weitern Sinne. Berlin:
Verlag der Akademie der Wissenschaften, Walter de Gruyter.
Selkirk, Elisabeth. 1981. On prosodic structure and its relation to syntactic struc-
ture. In Thorstein Fretheim (ed.), Nordic Prosody 11, 111–40. Trondheim:
TAPIR.
  1984. Phonology and Syntax: The Relation between Sound and Structure.
Cambridge, MA: MIT Press.
  1986. On derived domains in sentence phonology. Phonology Yearbook 3,
371–405.
  1995. Sentence prosody: intonation, stress, and phrasing. In John A. Goldsmith
(ed.), The Handbook of Phonological Theory, 550–69. Oxford: Blackwell.
  2000. The interaction of constraints on prosodic phrasing. In Merle Horne (ed.),
Prosody: Theory and Experiment, 231–61. Dordrecht: Kluwer.
  2005. Comments on intonational phrasing in English. In Sónia Frota, Marina
Vigário and Maria J. Freitas (eds.), Prosodies: With Special Reference to Iberian
Languages, 11–58. Berlin/New York: Mouton de Gruyter.
  2009. On clause and intonational phrase in Japanese: the syntactic grounding of
prosodic constituent structure. Gengo Kenkyu 136, 35–73.
  2011. The syntax–phonology interface. In John Goldsmith, Jason Riggle and
Alan C.L. Yu (eds.), The Handbook of Phonological Theory (2nd edn), 435–84.
Oxford/Malden, MA: Wiley-Blackwell.
Shattuck-Hufnagel, Stefanie and Alice E. Turk. 1996. A prosody tutorial for inves-
tigators of auditory sentence processing. Journal of Psycholinguistic Research
25(2), 193–247.
Silverman, Kim E. A., Mary E. Beckman, John F. Pitrelli, Mari Ostendorf, Colin
Wrightman and Patti J. Price. 1992. ToBI: a standard for labeling English
prosody. In The 2nd International Conference on Spoken Language Processing
(ICSLP). Banff, Canada.
Speer, Shari R., Paul Warren and Amy J. Schafer. 2011. Situationally independent
prosodic phrasing. Journal of Laboratory Phonology 2(1), 35–98.
Sperber, Dan and Deirdre Wilson. 1986/1995. Relevance: Communication and
Cognition. Oxford: Blackwell.
Steinbach, Markus. 2007. Integrated parentheticals and assertional complements. In
Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals, 53–87. Amsterdam/
Philadelphia: John Benjamins.
Steinhauer, Karsten, Kai Alter and Angela D. Friederici. 1999. Brain potentials
indicate immediate use of prosodic cues in natural speech processing. Nature
Neuroscience 2(2), 191–6.
Stenström, Anna-Brita. 1990. Pauses in monologue and dialogue. In Jan Svartvik
(ed.), The London-Lund Corpus of Spoken English: Description and Research,
211–52. Lund University Press.
Swerts, Marc. 1997. Prosodic features at discourse boundaries of different strength.
Journal of the Acoustical Society of America 101(1), 514–21.
Syrdal, Ann K. and Julia McGory. 2000. Inter-transcriber reliability of ToBI pros-
odic labeling. Paper presented at the 6th International Conference on Spoken
Language Processing (ICSLP) Beijing, China.
References  245

Taglicht, Josef. 1998. Constraints on intonational phrasing in English. Journal of


Linguistics 34(1), 181–211.
Tappe, Hans Thilo. 1981. Wer glaubst du hat recht? Einige Bemerkungen zur
COMP-COMP-Bewegung im Deutschen. In Manfred Kohrt and Jürgen
Lenerz (eds.), Sprache: Formen und Strukturen, vol. iic, 203–12. Tübingen:
Niemeyer.
Thompson, Sandra A. and Anthony Mulac. 1991. A quantitative perspective on
the grammaticization of epistemic parentheticals in English. In Elizabeth C.
Traugott and Bernd Heine (eds.), Approaches to Grammaticalization, 313–27.
Amsterdam/Philadelphia: John Benjamins.
Tottie, Gunnel and Sebastian Hoffmann. 2006. Tag questions in British and
American English. Journal of English Linguistics 34(4), 283–311.
Trim, J. L. M. 1959. Major and minor tone groups in English. Le Maitre Phonetique
74(111–12), 26–9.
Truckenbrodt, Hubert. 1995. Phonological Phrases: Their Relation to Syntax, Focus,
and Prominence. (Slightly revised version of) Ph.D. thesis, MIT.
  1999. On the relation between syntactic phrases and phonological phrases.
Linguistic Inquiry 30(2), 219–55.
  2005. A short report on intonation phrase boundaries in German. Linguistische
Berichte 203, 273–96.
  2007. The syntax-phonology interface. In Paul de Lacy (ed.), The Cambridge
Handbook of Phonology, 435–56. Cambridge University Press.
Turk, Alice E. 2012. The temporal implementation of prosodic structure. In Abigail
C. Cohn, Cécile Fougeron and Marie K. Huffman (eds.), The Oxford Handbook
of Laboratory Phonology, 242–53. Oxford University Press.
Turk, Alice E. and Stefanie Shattuck-Hufnagel. 2007. Multiple targets of phrase-
final lengthening in American English words. Journal of Phonetics 35(4),
445–72.
Urmson, James O. 1952. Parenthetical verbs. Mind 61(244), 480–96.
Vaissière, Jacqueline. 1983. Language-independent prosodic features. In Anne
Cutler and Robert Ladd (eds.), Prosody: Models and Measurements, 53–66.
Berlin: Springer.
Vallduví, Enric. 1992. The Informational Component. New York/London: Garland.
Viesel, Yvonne. 2011. glaubt er, glaub ich, glaub. Integrierte V1-Parenthesen,
Extraktion aus V2-Komplementen, Grammatikalisierung. Linguistische Berichte
226, 129–69.
Vries, Mark de. 2003. Three-dimensional grammar. Linguistics in the Netherlands,
201–12.
  2005. Coordination and syntactic hierarchy. Studia Linguistica 59(1), 83–105.
  2006. The syntax of appositive relativization: on specifying coordination, false
free relatives, and promotion. Linguistic Inquiry 37(2), 229–70.
  2007. Invisible constituents? Parentheses as B-merged adverbial phrases. In
Nicole Dehé and Yordanka Kavalova (eds.), Parentheticals, 203–34. Amsterdam/
Philadelphia: John Benjamins.
  2009. Specifying coordination: an investigation into the syntax of dislocation,
extraposition and parenthesis. In Cynthia R. Dreyer (ed.), Language and
Linguistics: Emerging Trends, 37–98. New York: Nova.
246  References

  2012a. Unconventional mergers. In Myriam Uribe-Etxebarria and Vidal Valmala


(eds.), Ways of Structure Building, 143–66. Oxford University Press.
  2012b. Parenthetical main clauses  – or not? On appositives and quasi-relatives.
In Lobke Aelbrecht, Liliane Haegeman and Rachel Nye (eds.), Main Clause
Phenomena: New Horizons, 177–202. Amsterdam: John Benjamins.
Wagner, Michael. 2004. Prosody as a diagonalization of syntax: evidence from com-
plex predicates. In Keir Moulton and Matthew Wolf (eds.), Proceedings of
NELS 34.
Ward, Gregory and Betty J. Birner. 1996. On the discourse function of right-
ward movement in English. In Adele E. Goldberg (ed.), Conceptual Structure,
Discourse and Language, 463–79. Stanford, CA: Center for the Study of
Language and Information Publications.
  2001. Discourse and information structure. In Deborah Schiffrin, Deborah
Tannen and Heidi Hamilton (eds.), Handbook of Discourse Analysis, 119–37.
Oxford: Basil Blackwell.
Watson, Duane and Edward Gibson. 2004. The relationship between intonational
phrasing and syntactic structure in language production. Language and
Cognitive Processes 19(6), 713–55.
Wells, John C. 2006. English Intonation: An Introduction. Cambridge University
Press.
Wichmann, Anne. 2000. Intonation in Text and Discourse: Beginnings, Middles and
Ends. London: Longman.
  2001. Spoken parentheticals. In Karin Aijmer (ed.), A Wealth of English, 177–93.
Göteborg: Acta Universitatis Gothoburgensis.
  2007. Can English tag questions grammaticalise? Nouveaux cahiers de linguistique
française 28, 349–57.
Wichmann, Anne and Nicole Dehé. 2011. Inconsistencies in the prosodic analysis of
corpus data. Presented at the 2nd Conference of the International Society for
the Linguistics of English (ISLE 2), Boston, MA.
Wightman, Colin W., Stefanie Shattuck-Hufnagel, Mari Ostendorf and Patti J.
Price. 1992. Segmental durations in the vicinity of prosodic phrase boundaries.
Journal of the Acoustical Society of America 91(3), 1707–17.
Xu, Yi. 2010. In defense of lab speech. Journal of Phonetics 38(3), 329–36.
Yoon, Tae-Jin, Sandra Chavarría, Jennifer Cole and Mark Hasegawa-Johnson. 2004.
Intertranscriber reliability of prosodic labeling on telephone conversation
using ToBI. Proceedings of ICSA International Conference on Spoken Language
Processing, (Interspeech 2004) Jeju, Korea.
Yoon, Tae-Jin, Jennifer Cole and Mark Hasegawa-Johnson. 2007. On the edge:
acoustic cues to layered prosodic domains. Presented at the International
Congress of Phonetic Sciences Saarbrücken, Germany.
Ziv, Yael. 1985. Parentheticals and functional grammar. In A. M. Bolkestein,
C. Groot and J. L. Mackenzie (eds.), Syntax and Pragmatics in Functional
Grammar, vol. i, 181–99. Dordrecht: Foris.
Index

adjunct, 31, 105, see also parenthetical, as Autosegmental-Metrical Framework,


adjunct 43–44, 90–91, 97–103, see also prosodic
adverb, 26, 65, see also sentence adverb framework
adverbial clause, 3
Align constraints, 104–5, 190, 225, backtracking, 7–8, 42, 117, 144, 152–53
see also Align/Wrap theory Beckman, Mary E., 98–102
Align/Wrap theory, 104–8, 114–15, 145, 154, binding, 19, 22–24, 50
169, 189, 224–25, 227–33 A-binding, 19
amalgamation, see syntactic amalgamation Condition C effect, 19, 46, 50, 61, 69, 76
ambiguity, 62–63, see also disambiguation Q-binding, 19
semantic scope, 39–40 Bolinger, Dwight, 34, 35–36, 40–42, 52, 82,
structural, 38–39, 49 221, 224
anchor, 8, 78, 119–21, 160, 161, 163, 164, 175 boundary, see prosodic boundary
nominal, 54, 119–21 boundary tone, 35, 36, 91, 138, 155
pronominal, 60, 120–21, 156, 161, 166 Brinton, Laurel J., 64, 66, 175
sentential, 119–20 British tradition of intonation analysis, 43–44,
and-parenthetical, 10, 26, 46–47, 138, 140, 228 88–90, 97–98, 99, see also prosodic
apposition framework
close, see apposition, restrictive
full, 55–57 Catalan, 32, 121
loose, see apposition, non-restrictive CC, see comment clause
nominal, 3, 10, 26, 54–63, 84–85, 120–22, Class 0 expressions, 34, 72, 77
154–75, 230–31 classifications of parentheticals,
non-restrictive, 51, 54–55, 62, 63 see parenthetical, classification of
partial, 55–57 cleft construction, 18, 19, 25, 61, 165, 167, 171
pragmatic relation, 58 comma feature, 13, 27–28, 47, 51, 104, 108
addition, 58 comma intonation, 28, 46, 47, 52, 55, 81, 82
restrictive, 54–55, 62, 63, 170–71 Comma Phrase, 104–5, 108, 110–12, 113
semantic relation, 57–58, 167–68, 171 comment clause, 3, 34, 64–74, 84–86, 122–23,
attribution, 57–58 175–93, 231–33
coreference, 57 main clause analysis, 66–68
equivalence, 57–58, 175 parenthetical analysis, 68–71
identification, 57 syntax of, 66–71
inclusion, 57–58 complete tonal contour (CTC), 38, 91, 128,
strict, 55 129, 131, 138, 141, 142, 147, 151, 155,
weak, 55, 121 156, 177, 207, 211
as it were, 65, 123, 186–88 Compound Prosodic Domain (CPD), 32–34,
as-parenthetical, 26, 28, 46, 138–40, 140–42, 177–80, 197, 204–205, 211–212
142–44, 228 compound tone, 99
Astruc-Aguilera, Lluïsa, 12, 43, 51–52, 63, 77, Condition C effect, see binding
121, 153, 169, 202 continuation rise, 91, 129, 131, 138, 183

247
248  Index

conventional implicature (CI), 1, 13, 26–28, I assume, 65, 72


38, 46, 47, 51, 62, 65, 66, I believe, 65, 177–80, 190
189–90, 233 I suppose, 186
conversational implicature, 26 I think, 65, 175, 182, 183, 188–89, 191–93
coordination, 47 ICE-GB, 13–15, 117, 126–28
creaky voice, 129, 138, 200, 207 Icelandic, 69–70
Cruttenden, Alan, 89–90 if you will, 72, 123
Crystal, David, 88–89, 99, 107–8 illocutionary clause, 113–14
CTC, see complete tonal contour illocutionary force, 9, 18, 20, 50, 61, 113
imperative tag, see tag
declination, see downtrend Inclusiveness, 25
disambiguation, see prosodic disambiguation, information structure, 12, 161, 171–72,
see also ambiguity 173–75
discourse, 12, 38, 53, 138, 189–90, 202 new information, 58, 164
discourse marker, 65 Insertion, 25
distribution, see parenthetical, distribution of integrated approach (to parentheticals), 25, 50,
domain-final lengthening, see final see also parentheticals, as adjuncts
lengthening interjection, 3
downstep, see downtrend intermediate phrase (ip), 92, 97–103
downtrend, 32–33, 34, 94–95, 131, 135, 147, interrogative parenthetical, see parenthetical,
207, 211, see also reset interrogative
Dutch, 18, 20, 21, 24, 46, 50, 67, 69–71, intonation domain, 88
75–76, 129 Intonational Phrase (IP), 31, 36, 90–95, 97–103
and relation to syntax, 108–15
edge tone, 91, see also boundary tone intonational separation, see prosodic separation
edge-alignment, 104–5, 110–12, see also Align/ invisibility, 19, 28–30, 61
Wrap theory
encliticization, see prosodic encliticization Japanese, 16, 32, 104
end-weight, 12, 58, 173–75
epistemic modality, 21, 64, 65–66, 72, Kaltenböck, Gunther, 22, 127, 190–93
189, 233 Kluck, Marlies, 8, 21, 70–71, 75–76
eurhythmy, 42, 107, 113, 228
evidential meaning, 64, 65, 205, 233 Ladd, D. Robert, 32, 82–83, 92, 94–95, 97,
experimental work on parentheticals, 213, 223
15–16, 32–33, 39, 51–53, 63, 67, 77, length, 43, 73, 190
153, 202 prosodic, 42, 154, 161, 213
expressive, 26 syntactic, 42, 175
extraposition, 59 loudness, 35, 36

final lengthening, 95, 129–30, 135, 138, 182, manner adverbial, 28


195, 200, 207 Match clause, 225, 228
floor-holding device, 73, 154, 230 Match constraints, 112–15, 190, see also Match
focus, 42, 228 theory
Force Phrase, 113 Match clause, 112–15
French, 16, 32, 67 Match theory, 112–15, 145, 154, 169, 189, 224,
full parenthetical clause (FPC), 44–47, 84–86, 227–33
118–19, 137–45, 228–30 matrix clause hypothesis, 66
merge, 28–30
German, 18, 66, 69–71, 75–76 parenthetical merge, 28–30, 61–62, 71
grammaticalization, 66 regular merge, 28–30
Gussenhoven, Carlos, 34–35, 77, 82–83, 182, sup-merge, see parenthetical merge
200–202, 223 Meyer, Charles F., 58, 62–63, 172–74
mitigation, 65, 73, 85, 189, 190
head, see Tone Group movement, see syntactic movement
Heringa, Herman, 54–55, 59, 60–62, 169
hesitation, 16, 73, 117, 129, 177, 183, 186, 188, N-APP, see apposition, nominal
212, see also pause, hesitant negative polarity item, 19, 21, 50, 82
Index  249
new information, see information structure, phonological phrase, see intermediate phrase
new information phrase accent, 91
nominal apposition (N-APP), see apposition, Pierrehumbert, Janet, 98–102
nominal pitch, 35–36
non-restrictive relative clause (NRRC), compression, 34, 36, 94, 138, 155, 160
see relative clause, non-restrictive expansion, 34, 36, 94
NRRC, see relative clause, non-restrictive pitch accent, 90–91
nucleus, 88–89, see also Tone Group pitch reset, 129, 131, 135, 142, 147, 155, 207
polarity, see question tag, polarity
omission of parenthetical, see parenthetical, Portuguese, 32, 33–34, 94
omission of position, see parenthetical, position of
orientation, 26, 75, 77 Potts, Christopher, 26–28, 46, 62, 169
addressee-oriented, 12 pre-boundary lengthening, see final
speaker-oriented, 1, 2, 21, 24, 26, 46, 75, 77 lengthening
subject-oriented, 75, 77 prehead, see Tone Group
orphanage approach (to parentheticals), 22, proper noun effect, 166, 169–71, 175, 230
24, 25, 49, 106, see also unintegrated prosodic boundary cues, 89–90, 93–96,
approach (to parentheticals) 128–35
non-radical, 24 prosodic disambiguation, 38–40, 53, 109,
radical, 24, 49 172–73
prosodic encliticization, 34–35, 77, 78, 151,
parasitic gap, 23, 46 182, 193, 198, 202
parenthetical prosodic framework, 43–44, 87, 97–103,
as adjunct, 24, 25–30, 106 see also Autosegmental-Metrical
anchored, 8–9, 10, 21, 46 Framework; British tradition of
classification of, 8–9 intonation analysis
complex, 6 prosodic hierarchy, 44, 92–93, 96–97
definition of, 1–2, 13 prosodic incorporation, 34, 77, 78, 202,
detached, 9, 46, 138, 140 see also prosodic integration
distribution of, 10–12, 225, prosodic integration, 43, 128, 135
see also parenthetical, position of of appositions, 156–64
elliptical clause, 3 of comment clauses, 73–74, 180–86, 189,
floating, 8–9, 46, 140 232–33
free, 8–9, 140 of full parenthetical clauses, 142–44
interrogative, 2, 3, 9, 46, 140 of question tags, 83–84, 212–16
main clause, 2, see also full parenthetical of reporting verbs, 77–78, 197–200,
clause 232–33
multiple, 6, 61 prosodic parenthesis, 40–42
non-finite, 3 prosodic separation, 27, 31, 32–38, 53, 62, 97–
omission of, 1, 18, 75 103, 114, 128, 130–35, 200, 202, 228
position of, 10–12, 24, 47, 58–60, 75, 81, of appositions, 63, 155–56, 168–69
165, 169, 216, 231, see also question of comment clauses, 73–74, 177–80, 189
tag, position of; parenthetical, definition of, 103
distribution of of full parenthetical clauses, 47, 138–42,
what, see what 144–45
parenthetical merge, see merge, of non-restrictive relative clauses, 51–52,
parenthetical merge 53–54, 146–49, 153
parenthetical phrase (ParP), 30, 47, 61 of question tags, 83–84, 206–212, 222–24
parenthetical verb, 3, 77, see also comment of reporting verbs, 77–78, 194–97
clause, reporting verb prosodic weight, see weight
ParP, see parenthetical phrase (ParP) psycholinguistics, 43, 52, see also experimental
pause, 31, 35, 36, 38, 77, 93, 96–97, 117, 129, work on parentheticals
135, 138–42, 155, 156, 161, 188, 191, punctuation, 2, 16, 51, 77
195, 207
filled, 3, 93 QT, see question tag
hesitant, 93, 186, 188, 191 question tag, 3, 78–84, 84–86, 98–101, 125–26,
silent, 93 205–25, 231–32
250  Index
QT (cont.) Structure-Preserving Constraint, 69
information-seeking, 210 subject orientation, see orientation
intonational realization, 83 subject-verb inversion, 70, 74
polarity, 79–80, 82, 83, 216, 220, sup-merge, see merge, parenthetical merge
223–24, 226 syntactic amalgamation, 3, 21–22
constant, 79–80, 82 syntactic deficiency, 64, 66, 68, 70–71, 74,
reduplicative, 79, see also question tag, 81, 85
polarity, constant syntactic independence (of parentheticals),
reversed, 79–80, 82, 224 18–19, 22, see also orphanage approach
position of, 81, 221, 224–25 (to parentheticals)
syntactic movement, 18, 46
read speech, 16, 93
recursivity (in prosodic structure), 32, 95, 115, tag, see also question tag
135, 164, 230 imperative tag, 3
relative clause statement tag, 3
appositive, see relative clause tempo, see speech rate
non-restrictive (NRRC), 3, 8, 26, 32, 43, ToBI, 90–91, 92, 95
47–54, 84–86, 119–20, 145–54, 228–30 tonal harmony, see tonal parallelism
restrictive, 6, 40, 49–50, 51–52 tonal parallelism, 63, 151
relevance theory, 47 tonal reduplication, see tonal parallelism
reporting verb, 3, 34, 74–78, 84–86, 123–25, tone concord, see tonal parallelism
193–205, 231–33 tone copy, 35, 151, 182, 193, see also prosodic
restarts, 117 encliticization
right-alignment, see edge-alignment Tone Group, 31, 88–90, 96–103
right-dislocation, 121 major, 96–97
right-node raising, 3, 24 minor, 96–97
Romance, 67 and relation to syntax, 107–8
root clause, 46, 108, 113 topicalization, 24
RV, see reporting verb Truckenbrodt, Hubert, 105–6

sandhi, 31, 33, 93–94, 130, 135 unintegrated approach (to parentheticals),


Scheffler, Tatjana, 65–66, 86, 189, 205, 233 24–25, see also orphanage approach (to
secondary predicate, 23, 24 parentheticals)
segmental rule, see sandhi
Selkirk, Elisabeth, 32, 53, 104–15 V1, see verb-first
Sense Unit Condition, 108–10 V2, see verb-second
sentence adverb, 3, 68 verb-first, 69, 70
Shift IP boundary, 154, 230 verb-second, 18, 46, 50, 69, 70
size, see also length vocative, 3, 34, 43, 77, 102, 212, 223
prosodic constituent, 42, 145, 160, Vries, Mark de, 28–30, 46, 51, 70–71, 75–76
163, 173
syntactic constituent, 43 weight, 43, see also end-weight
slifting, 24 prosodic, 42, 190, 232
speaker attitude, 73, 92 syntactic, 58
speaker orientation, see orientation what, 33, 43
speech act, 50, 65, 82, 100, 108, 224 Wichmann, Anne, 13, 16, 34, 42, 188, 213
speech rate, 16, 35, 36, 42, 145, 160, 163, 213 Wrap constraints, 105, 114–15, 145, 154, 169,
statement tag, see tag 189, 190, 228, 231, see also Align/Wrap
structural ambiguity, see ambiguity theory

You might also like