0521761921
0521761921
0521761921
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.
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
NICOLE DEHÉ
University of Konstanz
University Printing House, Cambridge CB2 8BS, United Kingdom
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
v
vi Contents
References 234
Index 247
Figures
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
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
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
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)
(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
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.
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
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
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
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
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
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
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
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
(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
(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) (…)
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
(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
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
…ParP…
Anchor Par'
* *
Par Apposition
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
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
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
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
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
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
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
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
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) (…)
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
87
88 Intonational phrasing and prosodic theory
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
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
B1: Sue!?
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
(a)
(b)
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
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
300
200
100
Mary will win, won’t she
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
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
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
UP XP2 XP2 UP
ZP X' X' ZP
X YP X YP
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
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]
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.
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.
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
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.
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
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
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
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
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
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)
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)
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.
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)
(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).
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)
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
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)
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)
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
(4.55) 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
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
c. phrasing 3 1 8 0 12 (3%)
unclear
107 81 267 11 466
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)
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)
200
Pitch (Hz)
150
100
60
N-APP
last month The Defence Secretary Mr Cheney said it was conceivable <SIL> that
0 3.795
Time (s)
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
250
Pitch (Hz)
200
150
100
50
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
(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.
separate 93 75 196 9
integrated 5 5 54 1
ratio 18.6 15 3.6 9
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
separate 300 52 21
integrated 64 1 0
ratio 4.7 52 n/a
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
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.
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
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
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
200
100
20
CC
and where we are still effectively ? I believe <SIL> underestimating the scale <SIL>
0 5.016
Time (s)
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)
150
Pitch (Hz)
100
50
!H* L-L% L+H* !H* L-L%
CC
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 …
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
250
Pitch (Hz)
200
150
100
H* L-L% H* !H* L-L%
CC
<noise>
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>
Figure 4.14 Example (4.89): CC phrased together with preceding and following host material, unstressed
186 Data analysis, results and discussion
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)
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
250
200
Pitch (Hz)
150
100
50
CC
0 3.4
Time (s)
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
separation: 2 5 7
[…][RV]([…]) / […[RV]…] ((4.98)a)
integration: [… RV] 7 15 22
unstressed / (incorporated or encliticized; (4.98)b1/b2)
integration: [… RV …] 0 2 2
unstressed or prenuclear prominence ((4.98)b3)
overall 9 23 32
separation: 4 3 7
[…][RV]([…]) / […[RV]…] ((4.98)a)
integration: [… RV] 20 2 22
unstressed / (incorporated or encliticized; (4.98)b1/b2)
integration: [… RV …] 2 0 2
unstressed or prenuclear prominence ((4.98)b3)
overall 27 5 32
160
Pitch (Hz)
140
120
100
80
H* !H* L-L% H* L-L%
RV
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
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)
150
Pitch (Hz)
100
50
L+H* L-L% L L%
RV
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.
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
300
Pitch (Hz)
200
100
50
H* !H* L+H* L-L% H* L-L%
QT
0 2.278
Time (s)
200
Pitch (Hz)
150
100
50
QT
0.318 4.321
Time (s)
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
H* L-L%
Analysis and results 213
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
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
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
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
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
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.
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
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
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
227
228 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
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
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
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
247
248 Index