French Dislocation: Interpretation, Syntax, Acquisition (Oxford Studies in Theoretical Linguistics)

French Dislocation Interpretation, Syntax, Acquisition

OXFORD STUDIES IN THEORETICAL LINGUISTICS general editors: David Adger, Queen Mary College London; Hagit Borer, University of Southern California. advisory editors: Stephen Anderson, Yale University; Daniel Büring, University of California, Los Angeles; Nomi Erteschik-Shir, Ben-Gurion University; Donka Farkas, University of California, Santa Cruz; Angelika Kratzer, University of Massachusetts, Amherst; Andrew Nevins, Harvard University; Christopher Potts, University of Massachusetts, Amherst; Barry Schein, University of Southern California; Peter Svenonius, University of Tromsø; Moira Yip, University College London. recent titles 10 The Syntax of Aspect: Deriving Thematic and Aspectual Interpretation edited by Nomi Erteschik-Shir and Tova Rapoport 11 Aspects of the Theory of Clitics by Stephen Anderson 12 Canonical Forms in Prosodic Morphology by Laura J. Downing 13 Aspect and Reference Time by Olga Borik 14 Direct Compositionality edited by Chris Barker and Pauline Jacobson 15 A Natural History of Infixation by Alan C. L. Yu 16 Phi Theory: Phi-Features Across Interfaces and Modules edited by Daniel Harbour, David Adger, and Susana Béjar 17 French Dislocation: Interpretation, Syntax, Acquisition by Cécile De Cat published in association with the series The Oxford Handbook of Linguistic Interfaces edited by Gillian Ramchand and Charles Reiss For a complete list of titles published and in preparation for the series, see pp 296.

French Dislocation Interpretation, Syntax, Acquisition

CÉCILE DE CAT

1

3

Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York © Cécile de Cat 2007 The moral rights of the authors have been asserted Database right Oxford University Press (maker) First published 2007 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available Typeset by SPI Publisher Services, Pondicherry, India Printed in Great Britain on acid-free paper by Biddles Ltd. www.biddles.co.uk _________________ ISBN 978–0–19–923047–1 1 3 5 7 9 10 8 6 4 2

Contents List of Figures List of Tables Abbreviations Acknowledgements

x xii xiv xvi

1 Introduction

1

2 Diagnostics for dislocated elements 2.1 Defining the language under investigation: unmarked spoken French 2.1.1 Advanced French: a questionable notion 2.2 French subject clitics are not agreement morphemes 2.2.1 Introduction and background 2.2.2 Testing the predictions of the morphological analysis 2.2.3 Implications for the system of agreement morphology in spoken French 2.2.4 Subject clitics are available for syntactic movement 2.2.5 Getting in the way: ne, en, y, and object clitics 2.2.5.1 Ne is more than an affix 2.2.5.2 Object clitics as affixes? 2.2.5.3 En and y as affixes? 2.2.5.4 Concluding remarks 2.2.6 Spoken French does not allow subject doubling 2.2.6.1 Distributional restrictions 2.2.6.2 The presence of a subject clitic forces the topic interpretation of a coindexed XP 2.2.7 Conclusion 2.2.8 French subject clitics: grammatical or anaphoric ‘agreement’? 2.2.8.1 Locality 2.2.8.2 Questioning of the related argument 2.2.8.3 Topicalization of parts of idioms 2.2.8.4 Peripheral vs. core status of the related argument 2.2.8.5 Conclusion 2.2.9 Information structure and syntactic structure

4 4 7 9 10 11 11 13 15 15 18 18 19 19 21 22 26 26 27 28 28 29 29 29

vi

Contents 2.2.10 The morpheme-like properties of French subject clitics are accidental 2.2.11 Conclusion 2.3 The prosodic characteristics of French dislocation 2.3.1 Right-dislocation prosody in spoken French 2.3.2 Prosodic differences between left-dislocated and heavy subjects in spoken French: a review of the literature 2.3.2.1 The acoustic characteristics of LD 2.3.2.2 The acoustic characteristics of heavy subjects 2.3.2.3 Summary 2.3.3 Diagnostics for LD? A preliminary acoustic analysis 2.3.3.1 Clear cases of LD prosody 2.3.3.2 Comparison with heavy subjects 2.3.3.3 Interfering factors 2.3.4 Summary 2.4 Conclusion

43 47 50 50 51 51 53 57 60 62

3 Interpretation 3.1 Topics 3.1.1 General definition 3.1.2 The information structure partitioning of the sentence 3.1.3 Topics do not have to correspond to old information 3.1.4 The relevance condition 3.1.5 Stage topics and aboutness topics 3.1.6 The role of topics 3.1.7 Summary 3.2 Topics in spoken French 3.2.1 A test case for topichood 3.2.2 Indefinite topics 3.2.2.1 Take 1: generic indefinites 3.2.2.2 Take 2: specific and d-linked indefinites 3.2.3 Topics in specificational pseudo-clefts 3.2.4 Topics take wide scope 3.2.5 Spoken French as a discourse-configurational language 3.3 Conclusion

63 63 64 65 67 70 71 74 76 77 77 81 81 86 91 92 94 96

4 Syntax 4.1 A brief overview of the literature 4.1.1 TopicP or no TopicP? 4.1.1.1 Functional heads to derive peripheral syntax

98 98 99 99

32 33 34 34

Contents

vii

4.1.1.2 Alternatives to the functional projection 100 approach 4.1.2 What moves in narrow syntax (if anything)? 101 4.1.2.1 The topic moves 102 4.1.2.2 The resumptive moves 103 4.1.2.3 Nothing moves in narrow syntax 103 4.1.2.4 Some move, some don’t 105 4.2 Dislocated topics in spoken French: an overview 108 4.2.1 Clause-peripheral topics 108 4.2.2 Conclusion 111 4.2.3 Caveat 111 4.3 French dislocation is not generated by movement 118 4.3.1 French LD does not yield Weak Crossover effects 118 4.3.2 French LD does not license parasitic gaps 119 4.3.3 No Relativized Minimality effects 120 4.3.4 No reconstruction effects in the interpretation of French LD 121 4.3.4.1 A variable in a left-dislocated XP cannot be bound by a clause-mate QP 121 4.3.4.2 Absence of Principle C effects 122 4.3.4.3 Wide scope with respect to negation 123 4.3.4.4 Interpretation of variables 124 4.3.5 French LD is not sensitive to islands 124 4.3.5.1 Native speakers’ judgements 125 4.3.5.2 To what extent are islands a diagnostic 129 for movement? 4.3.5.3 On the status of the ‘resumptive’ pronoun 131 4.3.6 CLLD or Hanging Topic? 134 4.3.7 Which analysis for French RD? 139 4.3.7.1 French RD is not an LF/PF phenomenon 140 4.3.7.2 French RD is not LD lower in the tree 144 4.3.7.3 French RD is not LD+IP-inversion 146 4.3.7.4 Differences are unexpected if RD = LD 147 4.3.7.5 French RD is not subject to the Right-Roof 148 constraint 4.3.8 Summary 149 4.4 A first-merge adjunction analysis of French dislocation 149 4.4.1 The analysis 149 4.4.1.1 Discourse Projections 150 4.4.1.2 D-subarrays 153

viii

Contents 4.4.1.3 Last-resort adjunction 4.4.1.4 Topic interpretation 4.4.1.5 On the relation between the dislocated element and its resumptive 4.4.2 Predictions of the adjunction analysis 4.4.2.1 Problematic predictions of the template approach 4.4.3 French embedded Discourse Projections 4.4.4 Deriving the differences between LD and RD from the properties of the peripheries 4.4.4.1 Prosodic properties and their consequences 4.4.4.2 General salience and its consequences 4.4.4.3 Linear order and its consequences 4.4.5 Theoretical consequences 4.5 Reconciling syntax and information structure 4.6 Conclusion

154 154 155 155 155 157 160 161 163 164 165 166 169 171 171 172 172 173 173 175 176 179 180 181 194 196 204 205 205 205 207

5 Acquisition 5.1 Introduction 5.2 Identifying early dislocated elements 5.2.1 Omissibility 5.2.2 Resumption 5.2.3 Word order and intervening material 5.2.4 Context 5.2.5 Prosody 5.3 Dislocations emerge early 5.4 Early dislocations and the CP projection 5.4.1 Eight diagnostics for the implementation of CP 5.4.2 Discussion 5.5 Sentence fragments: mini root projections 5.6 Primitives, learnability, and early discourse competence 5.6.1 Early discourse competence 5.6.2 Absence of violations of the relevant discourse rules 5.6.2.1 ILP subjects 5.6.2.2 Dislocated indefinites 5.6.3 Positive evidence for the relevant pragmatic competence 5.6.4 Learnability considerations 5.7 Conclusion

209 212 213

6 Concluding remarks

215

Contents

ix

Appendix A Adult data

217

Appendix B Child data

243

Appendix C Judgement elicitation

267

References Index

278 293

List of Figures 2.1 Prosody of a declarative sentence with a right-dislocated element (spontaneous, Geneviève)

37

2.2 Prosody of a yes/no question with a right-dislocated element 2.57b (spontaneous, Geneviève)

38

2.3 Prosody of a declarative sentence with a right-dislocated element (spontaneous, Chloé)

39

2.4 Prosody of a wh-question with a right-dislocated element (spontaneous, Chloé)

39

2.5 Prosody of a wh-question with a right-dislocated element (spontaneous, Geneviève)

40

2.6 Prosody of a yes/no question with a right-dislocated element (spontaneous, Geneviève)

41

2.7 Prosody of a left-dislocated element followed by a heavy subject (spontaneous, Catherine)

48

2.8 Prosody of a left-dislocated object (elicited, Pascale)

52

2.9 Prosody of a left-dislocated subject (elicited, Pascale)

52

2.10 Prosody of a left-dislocated subject (spontaneous, Catherine)

54

2.11 Prosody of a heavy subject (elicited, Pascale)

55

2.12 The ‘medium-high-low’ contour of left-dislocation prosody

55

2.13 Prosody of a heavy subject (spontaneous, Catherine)

57

2.14 Prosody of a monosyllabic left-dislocated subject (spontaneous, A.-Gaël)

58

2.15 Prosody of a left-dislocated subject in a ‘monotonous’ sentence (elicited, Isabelle)

59

2.16 Prosody of a contrastive heavy subject (spontaneous, Catherine)

60

5.1 The prosody of left-dislocated moi ‘me’ (spontaneous, Lisette)

177

5.2 The prosody of left-dislocated moi ‘me’ (spontaneous, Audrey)

177

5.3 Prosody of a declarative sentence with a right-dislocated element (spontaneous, Chloé)

178

5.4 Prosody of a declarative sentence with a right-dislocated element (spontaneous, Geneviève)

179

List of Figures

xi

5.5 Functions associated with dislocated elements in the child data

210

5.6 Functions associated with dislocated elements in the adult data

211

B.1 Distribution of root verbs in Max’s data

250

B.2 Distribution of root verbs in Anne’s data

250

B.3 Distribution of root verbs Tom’s data

251

B.4 Distribution of root verbs and their subjects in Max’s data

251

B.5 Distribution of root verbs and their subjects in Anne’s data

252

B.6 Distribution of root root verbs and their subjects in Tom’s data

252

B.7 Nature of the dislocated elements in the child data

257

B.8 Nature of the dislocated elements in the adult data

258

B.9 Functions associated with dislocated elements in the child data

260

B.10 Proportion of sentences containing a dislocated element in the child data

264

List of Tables 2.1 Verbal agreement morphology in spoken French

12

2.2 Subject (clitic)-verb inversion in root clauses in adult spoken French

14

2.3 Contrast between the melodic contour of a left-dislocated and a heavy subject

56

5.1 First signs of CP implementation

181

5.2 The age at which overtly moved wh-elements and fragments are first attested

200

5.3 Fragment utterances with satellite in the child data

202

5.4 Frequency of wh-questions involving a subject dislocation in child French (Anne, Max, and Tom)

213

A.1 Distribution of dislocated elements expressing selected constituents across dialects of spoken French

223

A.2 Number of dislocated elements associated with a single clause in the adult data

225

A.3 The functions associated with dislocated DPs and strong pronouns in spoken French

229

A.4 The functions associated with dislocated PPs in spoken French

230

A.5 The nature and function of dislocated elements in spoken French

234

A.6 Locality of dislocated elements associated with an argument of the verb

236

A.7 Direction of dislocation in root wh-questions across dialects of spoken French

240

A.8 Direction of dislocation in root yes/no questions across dialects of spoken French

240

A.9 Direction of dislocation in root declarative clauses across dialects of spoken French

241

A.10 Proportion of root clauses involving dislocation across utterance types in spoken French

241

B.1 Max and Anne’s ages at each recording session

245

B.2 Tom and Léa’s ages at each recording session

246

B.3 MLUw and age at each recording of Léa

246

List of Tables

xiii

B.4 MLUw and age at each recording of Max

247

B.5 MLUw and age at each recording of Anne

248

B.6 MLUw and age at each recording of Tom

249

B.7 Direction of dislocation in the children’s declarative sentences

253

B.8 Direction of dislocation in the children’s yes/no questions

253

B.9 Direction of dislocation in the children’s wh-questions

254

B.10 Nature of the dislocated elements in the child data

254

B.11 Nature of the dislocated elements in the adult data

254

B.12 Presence of a dislocated subject and realization of the subject according to the finiteness of the verb

265

C.1 Results of the first Internet judgement elicitation task

272

C.2 Results of the second Internet judgement elicitation task

275

Abbreviations 1 p. 2 p. 3 p. ACC agr. non-agr. C CP DP EXPL f. −FIN +FIN Fr. INDEF. INF INDIC IP IS m. NEG NOM NP OM p.p. PART PASS pl. PP PRES Q REFL sg. SM

first person morphology second person morphology third person morphology accusative agreement non-agreement complementizer Complementizer Phrase Determiner Phrase expletive feminine non-finite verb finite verb French indefinite infinitival indicative Inflection Phrase Information Structure masculine negation nominative noun phrase object marker past participle partitive passive plural prepositional phrase present tense question particle reflexive singular subject marker

Abbreviations SUBJ −T +T UG VP

subject untensed (verb) tensed (verb) Universal Grammar Verb Phrase

xv

Acknowledgements This book is a distant relative of my doctoral dissertation. Sections of Chapters 2 and 4 have appeared in an earlier and less extensive form respectively in Lingua (De Cat 2005)and Natural Language and Linguistic Theory (De Cat in press). The list of people I would like to thank for their help along the way is now very long, too long for me to mention them all here. A special mention is due to David Adger, Joe Emonds, Bernadette Plunkett, and George Tsoulas for having guided me through the initial stages of this research, to Claire Papageorgiadis and Jeannine Monnoye for having first opened my eyes to the structure of language, and to Anthony Pugh for plunging me into it all. Warm thanks also to Aafke Hulk for her unfaltering support and generous interest. For feedback and discussion over the years, I would like to thank Cedric Boeckx, Ute Bohnaker, Alastair Butler, Jo˜ao Costa, Jenny Doetjes, Nigel Fabb, Astrid Ferdinand, Hans-Martin Gärtner, Liliane Haegeman, Caroline Heycock, Paul Hirschbühler, Aafke Hulk, Kyle Johnson, Ruth Kempson, Marie Labelle, Knud Lambrecht, Luis López, Eric Mathieu, Piet Mertens, Fritz Newmeyer, Richard Ogden, Teresa Parodi, Susan Pintzuk, Paul Postal, Dan Robertson, Alain Rouveret, Bonnie Schwartz, Peter Sells, Ben Shaer, Nicolas Sobin, Mark Steedman, Peter Svenonius, Kriszta Szendr˝oi, Satoshi Tomioka, Sharon Unsworth, Anne Vainikka, Anthony Warner, and Roberto Zamparelli. Warm thanks also to the following people: Sally Johnson for being such a caring Head of Department; my informants for their help, their patience and their interest; the families of the children recorded for the York and the Cat corpora—and especially Catherine, Nelly, and Geneviève; my family for they-know-what; Marie-Hélène Crémer pour son amitié; certain inhabitants of Belle Vue for making my life what it is; Jaco for her eagle eye; the York Cemetery Trust for having given me the best of views to think to. And above all, Kester, thanks for the answer to life, the universe, and everything.

1 Introduction In spite of decades of research on French, certain aspects of spoken French are still, surprisingly, poorly understood. Dislocations fall into this category. Their pervasive use is one of the distinctive traits of spoken French. In these constructions, an XP appears in the left- or the right-periphery of the sentence and is usually resumed by an element inside it. In (1.1a), for instance, the left-dislocated DP is coindexed with an object clitic. In (1.1b), there is a right-dislocated DP coindexed with a subject clitic and a second rightdislocated element (a PP) coindexed with the ‘possessor’ in the first dislocated element. 1 (1.1)

a. Ces animaux-lài , on lesi voit partout. those animals-there one them sees everywhere ‘You see these animals everywhere.’ b. Ellei était vraiment bien, son j expositioni , à Julia j . she was really good her exhibition to Julia ‘Julia’s exhibition was really good.’

There are several reasons why the phenomenon of French dislocation is not yet fully understood. The absence of a consensus in the literature regarding the status of spoken French’s subject clitics is not the least of them. Indeed, the claim that such clitics are agreement markers without argument status has led a number of researchers to assume that the DP adjacent to the subject clitic in a sentence like (1.2) occupies the canonical subject position rather than a sentence-peripheral position (see e.g. Jaeggli 1982; Hulk 1986; Roberge 1986; Auger 1994; Miller and Monachesi 2003). (1.2) Les clitiquesi ilsi comptent pour du beurre. they count for some butter the clitics ‘Clitics don’t count.’ 1 These two examples far from exhaust the possibilities, as will shortly be shown. They are only intended to give a pre-theoretical ‘feel’ for what dislocations are like. Dislocated elements will be given in bold font throughout this book.

2

French Dislocation

I will argue in Section 2.2 that the information structural properties of sentences such as (1.2) are unambiguously those of dislocated constructions, irrespective of the status of French subject clitics. This leads me to the second source of confusion in the study of French dislocation, i.e. the general lack of understanding of the information structural properties of dislocated elements. While it has long been uncontroversial that dislocated elements encode the topic of the sentence, the very notion of topic has tended to be oversimplified. From Reinhart’s (1981) seminal paper, usually only the idea that the topic is what the sentence is about is retained. While this definition adequately captures a great many cases, it is not sufficient to account for all the dislocation data, nor is it entirely accurate. I therefore return to Reinhart’s original definition of topic (based on Strawson 1964) as that with respect to which the truth value of predication is computed. This allows a unified analysis of dislocated topics encompassing both aboutness and stage topics— as explained in Chapter 3—which is shown to have welcome consequences for the theory. Also problematic have been the prosodic diagnostics relied on by most syntacticians and acquisitionists to identify dislocated elements. In Section 2.3, I describe the main characteristics of dislocation prosody in French and discuss the limitations of diagnostics aimed at distinguishing dislocated subjects from their non-dislocated counterparts, thus throwing (some) light on examples like (1.2). I have chosen not to use the term topicalization to try and avoid confusion. So-called topicalized elements have been shown by Cinque (1990) to be focusfronted elements in Italian. The same applies to French. Focus fronting has to comply with strict contextual restrictions. The focused element is the only essential element of the utterance; all the other elements can be omitted (provided they are easily recoverable from the context). Focus-fronting is sensitive to islands and to Weak Crossover configurations but dislocation is not (cf. Section 4.3.1). Focus-fronted elements can be existential indefinites; dislocated elements cannot (see Section 3.2). Examples of focus-fronted elements are given in (1.3). (1.3)

a. Des légumes, on a dit. part. vegetables one has said ‘Vegetables, we said.’

(A.-Gaël, F) 2

2 Information in parentheses associated with an example indicates that it was uttered spontaneously and identifies the speaker, their dialectal area (for adults), or their age (for children). Capital letters stand for Belgium, Canada, and France. The age of the children is indicated in years;months.days.

Introduction b. Rien d’ autre, je n’ aime pas. nothing of other I neg like not ‘There’s nothing else I don’t like.’

3

(Léa 2;9.5)

The empirical basis of the present study consists essentially of attested, spontaneous utterances from two corpora 3 and of acceptability judgements from a large group of non-specialist informants. Judgements were elicited experimentally when the issues they are meant to inform are controversial in the literature (such as sensitivity of dislocated elements to islands). Context is shown to be indispensable in the evaluation of dislocated structures. Indeed, a better understanding of the felicity conditions of dislocated topics is argued to contribute significantly to the understanding of the formal mechanisms underlying their structure. The book is organized as follows. It opens with a definition of the field of enquiry (i.e. spoken French) and the careful identification of diagnostics for dislocated structures (including a discussion of the status of (subject) clitics and considerations regarding the prosody of dislocated structures). The following chapter looks at the interpretation of dislocated elements and discusses what it means for a referent to be a topic. Building on this, Chapter 4 provides detailed arguments in favour of an adjunction analysis of French dislocated elements. Further evidence comes from acquisition data (introduced in Chapter 5) and from learnability considerations. In Chapter 5, I also show how the proposed analysis accounts for complex fragment (i.e. non-sentential predications displaying an information structure partition), which are found both in adult and child speech. The last part of this book consists of three appendices which cover in more detail the empirical basis of the research.

3

See Appendix A for information regarding these corpora.

2 Diagnostics for dislocated elements This chapter establishes the empirical basis of my investigation into the interpretive and syntactic properties of dislocated structures in spoken French. It opens with a discussion of what the term ‘spoken French’ covers. The following two sections assess the extent to which prosody and the presence of a resumptive element can be used as diagnostics for dislocated structures in this language.

2.1 Defining the language under investigation: unmarked spoken French It is undisputed today (e.g. Blanche-Benveniste 1997) that spoken French differs quite substantially from what is commonly referred to as standard French (or written, prescriptive French, such as it is taught at school). A few traits characteristic of the differences between these two types of French are illustrated in (2.1)–(2.4). For instance first person plural subject is expressed using the clitic nous and its corresponding morphology in standard French, but in spoken French the clitic on tends to be used instead with third person singular morphology on the verb (2.1) (Blanche-Benveniste 1997; Côté 1999). Spoken French allows the omission of the negative particle ne and in many instances of the impersonal pronoun il (2.2), but standard French does not (Blanche-Benveniste 1997). Standard French uses subordination more than spoken French (2.3) (Frei 1929). Dislocations are rare in standard French, but very frequent in spoken French (2.3). 1 Finally, the impersonal pronoun il tends to be replaced by ce in spoken French (2.4). (2.1) a. Nous sommes allés à la mer. we are gone to the sea ‘We went to the seaside.’

(Standard Fr.)

1 Campion (1984) has shown that the presence of dislocations varies proportionally with the degree of formality of the speech and that this has been confirmed throughout the history of vernacular French.

Diagnostics for dislocated elements

5

b. On est allés à la mer. (Spoken Fr.) one is gone to the sea ‘We went to the seaside.’ (2.2) a. Il n’y a pas d’amour heureux. (Standard Fr.) it neg-there has not of-love happy ‘There’s no happy love.’ b. Y a pas d’amour heureux. (Spoken Fr.) there has not of-love happy ‘There’s no happy love.’ (2.3) a. J’ai vu le chien mordre le petit à la jambe. (Standard Fr.) I-have seen the dog bite the little at the leg ‘I saw the dog bite the little one’s leg.’ b. Le chien, je l’ai vu, il le mordait à la (Spoken Fr.) the dog I it-have seen he him bit at the jambe, le petit.2 leg the little ‘I saw the dog bite the little one’s leg.’ (2.4) a. Il est dommage qu’ il soit si tard.3 (Standard Fr.) it is sad that it be so late ‘It’s sad that it’s so late.’ b. C’est dommage, qu’ il soit si tard. (Spoken Fr.) it-is sad that it be so late ‘It’s sad that it’s so late.’ Contrary to what is sometimes claimed (e.g. Zribi-Hertz 1994), subject-verb inversion is not exclusively a characteristic of standard French: it is obligatory in certain cases (2.5), and it is still very productive in certain contexts in Canadian French and in some varieties of Belgian French (2.6). (2.5) a. Où sont les femmes? where are the women b. ∗ Où les femmes sont? where the women are ‘Where are the women?’

This example comes from Frei (1929: 272). The resumption by il of a postverbal clause expressing the subject only takes place in extraposition constructions (Moreau 1976). According to Moreau, in such sentences the extraposed clause is situated inside the sentence nucleus (i.e. not in a peripheral position). In the spoken French counterpart, only resumption by ce is possible, and the postverbal clause expressing the subject is uttered with rightdislocation intonation. 2 3

6

French Dislocation

(2.6) a. Que fait-on, ma puce? what does-one my flee ‘What shall we do, love?’ b. Es-tu venu toute la bouche rouge rouge are-you come all the mouth red red rouge? red ‘Did you come back with your mouth all red?’

(Dominique, B)

(Catherine, C)

In fact, most of the traits allegedly characteristic of standard French are still productively used across dialects. For instance, the clitic nous ‘we’ is still used by certain (especially older) speakers of Belgian French, and the negative particle ne appears frequently in the speech of all members of this group regardless of age dialect. The nature of standard French is essentially prescriptive, and this has led some to question its psychological reality. ‘En réalité, le vrai français, c’est le français populaire. Et le français littéraire ne serait plus aujourd’hui, à ce point de vue, qu’une langue artificielle, une langue de (Bauche 1926) mandarins—une sorte d’argot’. 4

Similarly, Côté (1999) suggests that standard varieties of French are not totally coherent from a grammatical point of view, and may consequently not exist as such in the mind of any speaker. In spite of this, many linguists (including Blanche-Benveniste 1997) still argue that characteristics such as those illustrated by the (b) examples of (2.1)–(2.4) are errors (even though, as they admit, they are made by the vast majority of speakers). Such linguists continue to entertain the myth of a ‘pure’ French, without really questioning the extent to which this ‘pure’ variety has ever existed outside of the prescriptive norms laid down by scholars influenced by the written standard. Various dichotomies have been proposed in an attempt to isolate ‘real’ French as it ‘exists’ in the minds of speakers. Standard French is usually taken to represent the norm, from which usage can differ to a variable extent. Whatever system lives in the minds of speakers is defined in contrast to this norm. A clear definition of standard French has thus to be provided first. Some equate standard French with the written language, in opposition to which they define spoken standard French (e.g. Barnes 1985). Others equate 4 ‘In reality, it is “popular” French which is the real French. And from this point of view, literary French today is no more than an artificial language, a language of the elite—a kind of slang’.

Diagnostics for dislocated elements

7

the norm with formality, in opposition to which they define Colloquial French (e.g. Auger 1993; Zribi-Hertz 1994). Informal French is often equated with spoken French (e.g. Sportiche 1995). If a social dimension is included in the notion of the norm, sub-groups are defined within non-standard French; the opposition then shifts to one between français standard and français populaire (e.g. Carroll 1982). Further distinctions on the basis of regional differences are also established within spoken varieties (one much studied variety is that of Quebec French—see e.g. Auger 1994). It seems reasonable to assume that whenever a group of speakers can be identified as using the variety in question, the identification of this variety is methodologically justified. One of the guises under which spoken French has been studied does not, however, comply with this minimal requirement. It deserves special mention here, for two reasons: (i) it has been adopted by a number of acquisition researchers as corresponding to the input available to children acquiring French today (e.g. Hulk 1995; Ferdinand 1996; Müller et al. 1996; Jakubowicz et al. 1998), and (ii) the essential distinctive trait of this variety is central to the study of dislocations in French as it rests on the assumption that subject clitics are not arguments in this variety of French and hence that the preverbal DP in a sentence like (2.7) does not contain a left-dislocated DP coindexed with a cliticized subject (as I will argue), but a DP in the canonical subject position (with the subject clitic marking agreement on the verb). (2.7) Les mûres elles sont mûres. the blackberries they are ripe ‘The blackberries are ripe.’ The name by which this variety is designated in such work is Advanced French. Section 2.1.1 looks more closely at what this term is supposed to cover. 2.1.1 Advanced French: a questionable notion The notion of Advanced French (AF) was (re)introduced by Zribi-Hertz (1994). The term originates in Frei (1929), who defines le français avancé as le français qui s’écarte de la norme traditionnelle (‘the (variety of) French that deviates from the traditional norm’). It includes mistakes, innovations, popular language, slang, unusual or controversial uses, grammatical puzzles, etc. (Frei 1929: 32). In this sense, Zribi-Hertz’s AF refers to the same concept: it is ‘a group of dialects which, at various degrees, disregard the norms taken as characterizing “good usage”, and which thus reveal ongoing grammatical

8

French Dislocation

mutations’ (Zribi-Hertz 1994: 460). 5 She further specifies two subtypes within AF: Colloquial French (‘the unmarked informal style used by those speakers who also master Modern Standard French’) and Very Advanced French (which includes ‘all forms regarded as substandard or dialectally marked by Modern Standard French and Colloquial French speakers: so-called “popular French” and “child” French’ (Zribi-Hertz 1994: 461)). The presumed reality of AF has been questioned by a number of researchers (see e.g. Gadet 1998; Côté 1999, 2001). I summarize here the arguments of Côté (1999, 2001). Neither Colloquial French nor Very Advanced French can be identified with a particular group of speakers. Contrary to Zribi-Hertz’s (1994) claim, the mastery of Modern Standard French is not a sufficient criterion to identify speakers of Colloquial French: the language used informally by speakers of Colloquial French can vary quite dramatically (as say between a speaker from Montreal and one from Toulouse) and yet they may master Modern Standard French to the same degree. As for Very Advanced French, Côté argues that it is a mere collection of features unified only by their absence from the speech of educated Parisians (Côté 1999: 26). Only three examples of Very Advanced French are given by Zribi-Hertz (1994), each illustrating one of the features in 2.8. (2.8) a. The third person feminine subject clitic elle is pronounced [a] b. The complementizer que can co-occur with an overtly filled embedded [spec,CP] (i) L’homme à qui que j’en ai parlé. the-man to whom that I-of-it have talked ‘The man to whom I talked about it.’ (Gadet 1989, cited by Zribi-Hertz 1994) c. Nominative proclitics can co-occur with a quantified subject. Côté (1999) denounces the heterogeneity of this amalgam of traits, and points to the fact that (2.8a) and (2.8b) are both characteristic of her own dialect (Montreal French), but that the co-occurrence of subject clitics with a quantified subject is ungrammatical to her and her informants (from Montreal and elsewhere). She concludes that Advanced French is far too heterogeneous to be considered a coherent linguistic system. A distinction between Standard 5 The claim that deviance from the norm is a sign of grammatical change is however a controversial one. Some of the features claimed to be the manifestation of language change have been attested (and criticized by prescriptive grammarians) for centuries (e.g. Campion 1984). For instance, the frequent use of left-dislocation of the subject, as in (2.7), was already condemned by grammarians in the 17th century (Blanche-Benveniste 1997: 37).

Diagnostics for dislocated elements

9

French and Advanced French does not bring ‘significant progress in the description of French, as [it] does not reflect the reality of dialectal variation’ (Côté 1999: 27). It seems to me that the motivation behind the definition of Advanced French stems essentially from the need to identify the variety/varieties of French for which a morphological analysis of subject clitics can be argued to hold. 6 The same motivation led Lambrecht (1981) to label the variety he studied non-standard French, the defining traits of which are, he proposes, (i) the loss of inversion, (ii) the development of an ‘agglutinative’ verb complex (consisting of the verb and the clitics), and (iii) the ‘desyntacticization’ of NPs, whereby the ‘nucleus’ of the sentence is essentially grammatical with most of the lexical elements in the periphery. Remember however that loss of inversion has only (partially) taken place in some dialects of French which nonetheless feature a high frequency of dislocation (the latter encompassing properties (ii) and (iii)). The subject of the present study is spoken French as opposed to written French as it is used by speakers in informal situations. The dialectal regions represented in the corpora of spontaneous production are Liège and Wallonia (Belgium); the Pyrenees, Paris, and Brittany (France); and Montreal and New Brunswick (Canada). Judgements were elicited from a wider range of speakers from Belgium, Canada, France, and (to a smaller extent) Switzerland, randomly spread across regions.

2.2 French subject clitics are not agreement morphemes The status of French subject clitics is a controversial issue within morphosyntactic theory. Two main analyses have been proposed and defended over the past three decades, one advocating the view that French subject clitics are syntactic arguments bearing a theta-role and the other that such clitics are inflectional morphemes on the verb. This section demonstrates that the empirical basis motivating the morphological analysis of French subject clitics is much narrower than has been assumed in the literature, and shows that the implementation of such an analysis faces numerous theoretical and empirical difficulties. It is concluded that the limited similarities between the behaviour of French subject clitics and that of morphemes should be treated as accidental rather than as decisive evidence in favour of a morphological analysis. 7 6 Indeed, this is the main subject of Zribi-Hertz (1994). Note that no such claim is made in Frei (1929). Frei treated sentences like (2.7) involving left-dislocation of the subject, which he considered had an expressive function. 7 Large sections of this chapter were published as De Cat (2005).

10

French Dislocation

2.2.1 Introduction and background Two opposing analyses of French subject clitics have been proposed in the literature. According to the ‘syntactic’ analysis, French subject clitics are syntactic entities (i.e. elements available for syntactic operations) which bear a θ-role and which transit via the canonical [spec,TP] subject position, from where they cliticize phonologically onto the verb. This is broadly the position held by Kayne (1975); Rizzi (1986); Laenzlinger (1998); Belletti (1999); among others. According to the ‘morphological’ analysis, French subject clitics are agreement morphemes generated directly on the finite verb in the lexicon (or at a lexicon-syntax interface). This analysis is generally argued to apply to spoken French only, and not to standard/formal/written French. Proponents of the ‘morphological’ analysis include Jaeggli (1982); Hulk (1986); Roberge (1986); Auger (1994); Miller and Monachesi (2003). The mechanisms implementing the morphological analysis have been described in e.g. Cummins and Roberge (1994); Auger (1994); and related work. Cummins and Roberge (1994) propose that Romance clitics are associated with lexical roots prior to syntax (more precisely at the posited LexiconSyntax Interface). At that point, the clitics are said to be mere bundles of features. Ungrammatical representations are later filtered out at LF (for semantic reasons, which are left vague) and PF (where language-particular constraints are defined in the form of templates). An additional filter mechanism is argued to operate in the Computational Component of grammar in the form of agreement between the clitic and the empty category in argument position. 8 Auger (1994: 196) proposes that ‘verbal forms are directly generated by morphological processes’ (though these processes are not explicitly defined) and argues that no template need be postulated to derive the French facts. She assumes that French object clitics bear a Case feature (to account for the fact that object doubling is ungrammatical in this language) and that French subject clitics do not bear any Case feature. Object clitics are thus claimed to be affixes, while subject clitics are (affixal) agreement morphemes. The remainder of this section will abstract away from the exact mechanisms at work in the morphological analysis and concentrate instead on the broader issues raised by it. 8 It is far from clear how such an empty category is endowed with the relevant features in the first place. Object clitics have argument status in spoken French (Auger 1994), so the verb cannot possibly select another (albeit empty) argument with the relevant features.

Diagnostics for dislocated elements

11

2.2.2 Testing the predictions of the morphological analysis The morphological analysis of French subject clitics has a number of theoretical and empirical consequences. (2.9)

a. Subject-verb agreement can be marked twice morphologically. b. Subject clitics are not available for syntactic operations independent of their host. c. Preverbal clitics appearing between the subject clitic and the verb also have to be analysed as affixes. These elements include en, y, object clitics, and the negative particle ne. d. Subject doubling is predicted (i.e. the co-occurrence of an XP in [spec, TP] and an adjacent subject clitic).

To my knowledge, these consequences have not been fully investigated in the literature. I address them in turn in Sections 2.2.3–2.2.6 and show that they are problematic for the morphological analysis. 2.2.3 Implications for the system of agreement morphology in spoken French The first consequence of the morphological analysis (2.9a) is that it generates a certain amount of random redundancy in the French morphological paradigm of verbal agreement. The features identified by ‘agreement prefixes’ (i.e. subject clitics) are identical to those identified by agreement suffixes whenever the finite verb does not correspond to an elsewhere form. Elsewhere forms are underspecified in the lexicon with respect to particular features such as person and number (following Halle and Marantz (1993)). The full paradigm of verbal agreement morphology in spoken French is given in Table 2.1. Elsewhere forms appear in phonetic transcription (indicating homophony). First person plural nous has been omitted from this table as it is not generally considered to be part of the grammar of spoken French (Blanche-Benveniste 1997; Côté 2001). The only feature that can be marked on agreement prefixes but not on suffixes is [gender], although this feature cannot be marked on first and second person clitics, and it is not always marked on third person plural clitics (feminine is only marked if the individuals concerned are all female). (2.10) je (1 p.sg.), tu (2 p.sg.), on, 9 vous (2 p.pl.): il (3 p.sg.):

not marked for gender masculine

9 On can denote third person referents (usually under an arbitrary interpretation), first person plural referents, and second person referents.

12

French Dislocation

Table 2.1. Verbal agreement morphology in spoken French

1 p.sg. 2 p.sg. 3 p.sg. 2 p.pl. 3 p.pl.

present

imperfect

future (synthetic)

subjunctive (present)

1 p.sg. 2 p.sg. 3 p.sg. 2 p.pl. 3 p.pl. 1 p.sg. 2 p.sg. 3 p.sg. 2 p.pl. 3 p.pl. 1 p.sg. 2 p.sg. 3 p.sg. 2 p.pl. 3 p.pl.

-er, -oir

-ir, -re, vouloir

être, avoir, aller

⎫ ⎪ ⎬

⎫ ⎪ ⎬

suis 

⎪ ⎭

[plœ:r]

⎪ ⎭

[vφ]

[ε]

pleurez [plœ:r] ⎫ ⎪ ⎬ [plœrε] ⎪ ⎭

voulez veulent ⎫ ⎪ ⎬ [vulε] ⎪ ⎭

êtes sont ⎫ ⎪ ⎬ [etε] ⎪ ⎭

pleuriez [plœrε] pleurerai 

vouliez [vulε] voudrai 

êtiez [etε] serai 

[plœr:a]

[vudra]

pleurerez pleureront ⎫ ⎪ ⎬ [plœ:r] ⎪ ⎭

voudrez voudront ⎫ ⎪ ⎬ [vœj] ⎪ ⎭

serez seront ⎫ ⎪ ⎬ [swa] ⎪ ⎭

pleuriez [plœ:r]

veuillez [vœj]

soyiez [swa]

elle (3 p.sg.), elles (3 p.pl.): ils (3 p.pl.):

[sra]

feminine masculine or mixed

An important question that should be addressed by the proponents of the morphological analysis is the following: Why should subject-verb agreement ever be allowed to be marked twice in an entirely redundant fashion, and yet not systematically across verbal paradigms or tense paradigms? In other words: why is it not possible to omit the clitic or the suffix in cases like (2.11)? (2.11) a. je v-ais [1 p.sg.] go-[1 p.sg.]

Diagnostics for dislocated elements

13

b. vous pleur-ez [2 p.pl.] cry-[2 p.pl.] Another oddity is that subject clitics would be the only prefixal agreement markers in spoken French: agreement is otherwise exclusively marked by suffixes on affected lexical classes such as participials and adjectives. (2.12) a. elle-est parti-e. [3 p.sg.f.]-is-[3 p.sg] gone-[sg.f.] b. elle-est joli-e. [3 p.sg.f.]-is-[3p.sg.] pretty-[sg.f.] 2.2.4 Subject clitics are available for syntactic movement The second consequence of the morphological analysis (2.9b) is that subject clitics, by virtue of being affixes inserted pre-syntactically, should not be available for syntactic operations independent of their host. Auger (1994: 67–68) claims that only second person subject clitics ever appear postverbally in Quebec Colloquial French, and that subject clitics never appear postverbally in European Colloquial French. I dispute both these claims. In the corpora investigated here (the York and Cat corpora—see Appendix A) inversion of verbs and subject clitics is clearly productive. Table 2.2 gives an overview of the distribution of subject clitic-verb inversion in matrix clauses in yes/no and wh-questions across dialects for adult speakers in the York and Cat corpora. In this table, questions introduced by est-ce que (i.e. so-called locutionary questions) are treated separately, as their analysis in terms of inversion structure is controversial. Non-root clauses are excluded from the counts as they do not allow inversion. The following whquestions were also excluded as they do not allow subject-verb inversion: questions with a wh-element in situ and structures with a doubly-filled Complementizer. Questions involving the question particle ti (sometimes pronounced tu) were excluded as the status of this particle is unclear (see e.g. Vecchiato 2000). 10 This table shows that inversion is attested at least in Belgian and Canadian French, and suggests that it is used differently across dialects. It is used 10 Extracts read from books were excluded as they are in most cases representative of written French. There are no instances of complex inversion, illustrated in (i), in the York and the Cat corpora.

(i) Comment les voleurs sont-ils entrés? how the thieves are-they entered ‘How did the thieves come in?’

14

French Dislocation Table 2.2. Subject (clitic)-verb inversion in root clauses in adult spoken French No inversion

Inversion

Locutionary

Belgium Canada France

wh-questions (with wh-movement) 8% (59) 40% (308) 53% (408) 27% (200) 1% (4) 72% (536) 25% (229) 2% (18) 73% (656)

Belgium Canada France

96% (1900) 61% (1616) 98% (2133)

Yes/no questions 2% (45) 21% (567) 0% (0)

2% (38) 17% (462) 2% (45)

Total

775 740 903 1983 2645 2178

productively in wh-questions only in Belgian French (forty per cent of cases), and in yes/no-questions only in Canadian French (twenty-one per cent of cases). Each type of question (inverted, non-inverted, locutionary) appears to have become pragmatically specialized in a different way across dialects, a point I cannot explore here in detail. Contrary to what is claimed in Auger (1994), inversion of the subject clitic and the verb is productive in Canadian French as it affects not only second person clitics (as claimed by Auger) but also third person clitics as in (2.13) (see De Cat 2002 for details). (2.13) Peut-il avoir une petite bouchée? may-he have a little mouthful ‘Can he have a bite?’ Auger (1994) derives the orders ‘second person subject clitic–verb’ and ‘verbsecond person subject clitic’ by postulating the existence in the lexicon of two types of second person subject clitics: one prefixal, marked [−interrogative], and the other suffixal, marked [+interrogative]. This is insufficient to account for the fact that in Canadian French tu (2 p.sg.) can appear preverbally or postverbally in yes/no questions and that it never appears postverbally in wh-questions in that dialect. To account for the inversion facts reported above while maintaining the morphological analysis of subject clitics, one would need to postulate the existence in the lexicon of a wide range of homomorphous clitics with different feature specifications and different affixation requirements. In order to account for the distribution of second

Diagnostics for dislocated elements

15

person singular subject clitics, the lexical entries given in (2.14), for example, would be required. The [?] stands for an extra feature that would differentiate the exact discourse conditions determining when tu appears postverbally. (2.14) tu [− (yes/no) interrogative] prefix tu [+ (yes/no) interrogative] [?] suffix tu [+ (yes/no) interrogative] [?] prefix While something like (2.14) is theoretically possible, it is highly uneconomical, and therefore unlikely to provide an adequate model of how language works following minimalist assumptions. 2.2.5 Getting in the way: ne, en, y, and object clitics The third consequence of the morphological analysis (2.9c) concerns preverbal clitics appearing between the subject clitic and the verb. Such elements include the clitics en, y, object clitics, and the ‘negative’ particle ne (given in bold in (2.15)). (2.15) a. Je la lui donnerai. [1 p.sg.] it to-him will-give ‘I’ll give it to him.’ b. Je ne t’ en veux pas. [1 p.sg.] ne to-you of-it want not ‘I don’t begrudge you.’ c. On y va? [3 p.sg.] there goes ‘Shall we go?’ If subject clitics are verbal agreement morphemes, they must by definition be attached to a (possibly morphologically complex) agreeing verb. One of the criteria proposed by Zwicky and Pullum to distinguish affixes from clitics is that ‘clitics can attach to material already containing clitics, but affixes cannot’ (Zwicky and Pullum 1983: 503). This means that if French subject ‘clitics’ are in fact agreement morphemes (i.e. affixes), any element intervening between the subject ‘clitic’ and the verb must also be an affix inserted pre-syntactically. However, as we will see, analysing such elements as affixes poses numerous problems. 2.2.5.1 Ne is more than an affix Auger (1994) avoids having to deal with ne by claiming that it is not part of the grammar of spoken French anymore. This claim is however unwarranted. The York and Cat corpora provide clear

16

French Dislocation

evidence that the particle ne is used in spontaneous speech in Belgium, Canada, and France. While this particle is used more frequently by Belgian speakers, it is attested in the speech of all the speakers in both corpora (a total of twelve adults and seven children) and in a variety of contexts (i.e. productively; see also Gadet (2000)). Ne can be (and is) omitted most of the time, but in some cases its presence is obligatory (2.16) or strongly preferred (2.17). (2.16) a. Personne ∗ (n’)a chanté ni crié. nobody ne-has sung nor shouted ‘Nobody sang nor shouted.’ (Rowlett 1998: 169) b. Rien du tout ∗ (n’)a été fait. nothing of all ne-has been done ‘Nothing at all has been done.’ (Rowlett 1998: 169) c. Personne d’intéressant ∗ (n’)a été invité. nobody of-interesting ne-has been invited ‘Nobody of interest has been invited.’ (2.17) a. Aucun des animaux ∗ ?(n’)est parti. gone none of-the animals ne-is ‘None of the animals is gone.’ aidée. b. Pas un employé ∗ ?(ne) m’a me-has helped not one employee ne ‘Not a single employee helped me.’ c. Les profs d’aucune école ∗ ?(n’)ont fait grève. the teachers of-no school ne-have done strike ‘The teachers at none of the schools have gone on strike.’ A sentence containing ne does not have to be interpreted as negative (Godard 2004: fn.9): (2.18) Paul n’aime que Mozart, et Marie [aussi/∗ non plus]. neither Paul ne-likes only Mozart and Marie too/ ‘Paul likes only Mozart and so/∗ neither does Marie.’ On a morphological analysis, this would require listing two ne entries in the lexicon (one negative, one restrictive). In addition to these two verbal affixes, the lexicon would also need to contain a non-affixal ne to account for the fact that ne can directly precede pas or the presentative voilà (Rowlett 1999): (2.19) a. C’est pour ne pas tomber. it-is to ne not fall ‘It’s so as not to fall.’

Diagnostics for dislocated elements

17

b. Ne voilà-t-i(l) pas qu’il m’invite chez lui. ne pres-t-it not that-he me-invites at him ‘And what do you know, he invites me to his place.’ However, the most intractable problem for a morphological analysis relating to ne is that its distribution is structurally determined. The role of ne is to mark the scope of negation (Kayne 1984): negation cannot take scope higher than the clause containing ne. In (2.20a), only the infinitival clause is interpreted negatively; in (2.20b), the whole sentence is interpreted negatively. (2.20) a. Paul accepte de ne renvoyer personne. Paul agrees to ne dismiss anybody ‘Paul agrees not to dismiss anybody.’ b. Paul n’accepte de renvoyer personne. Paul ne-agrees to dismiss anybody ‘Paul doesn’t agree to dismiss anybody.’ Examples like (2.20b) show that ne can appear in a different clause to the negative expression (personne ‘nobody’ in (2.20b)). This long-distance relation is only possible under certain structurally defined circumstances (as shown by Milner (1979); Rowlett (1998); and Godard (2004); among others). First, the embedded clause must be non-finite. (2.21) a. Paul ∗ n’accepte qu’on renvoie personne. Paul ne-agrees that-one dismiss nobody b. Paul n’accepte de renvoyer personne. Paul ne-agrees to dismiss nobody ‘Paul doesn’t agree to dismiss anybody.’ Second, the relation between ne and the negative expression is sensitive to syntactic islands. For instance, such a relation cannot hold across the boundaries of a complex DP (2.22) or those of a sentential subject (2.23). ∗

Il ne reste [de potager [avec aucun arbre fruitier]]. it ne remains some allotments with no tree fruiting b. Il ne reste [aucun potager [avec des arbres fruitiers]]. it ne remains no allotment with some trees fruiting ‘There aren’t any allotments with fruit trees left.’ (2.23) a. ∗ [Engager personne] n’est permis. to-appoint nobody ne-is allowed (2.22) a.

18

French Dislocation b. Il n’est permis [d’engager personne]. it ne-is allowed to-appoint nobody ‘No one is allowed to be appointed.’

If ne is a morpheme affixed to the verb pre-syntactically, its distribution should not be syntactically constrained. 2.2.5.2 Object clitics as affixes? If subject clitics are attached to the verb stem pre-syntactically, any clitic intervening between the subject clitic and the verb should also undergo affixation pre-syntactically. Given the absence of lookahead, no information regarding the syntactic structure to be derived is available at that point, yet the choice of which verb stem an object clitic is affixed to depends in some cases entirely on the (future) structural configuration. This is illustrated in (2.24). In these sentences, the object clitic les ‘them’ is an argument of the verb lire ‘to read’. The clitic can appear either (i) on the infinitival verb of which it is an argument (2.24a)), (ii) on a higher infinitival (2.24b)), or (iii) on the finite verb (2.24c)). The choice is however not free: in each case, only one position is licit (indicated by the absence of∗ ). (2.24) a. Il [∗ lesi ] va [lesi ] lire 0i . he (them) will (them) read ‘He’ll read them.’ b. Il [∗ lesi ] va [lesi ] faire [∗ lesi ] lire 0i . he (them) will (them) have (them) read ‘He’ll have them read.’ c. Il [lesi ] a [∗ lesi ] fait [∗ lesi ] lire 0i . he (them) has (them) had (them) read ‘He’s had them read.’ If object clitics are affixed to verb stems in the lexicon, it is not clear why they should be allowed to appear on a different verb to the one of which they are an argument, nor how the appropriate target verb stem is identified in the absence of structural information. 11 2.2.5.3 En and y as affixes? The clitics en and y are generally associated with non-human referents. However, in certain cases, they can take a human antecedent, as illustrated in (2.25). 11 On the syntactic analysis, the object clitic is predicted to surface on the highest verb in the clause containing the verb which selects that object. In (2.24a) it appears in the infinitival clause PRO les lire. In (2.24b) it appears in the infinitival clause PRO les faire lire, in which the causative verb and its infinitival complement form a verbal complex (Labelle and Hirschbühler in preparation). In (2.24c) the relevant clause corresponds to the whole sentence, with fait lire as a verbal complex.

Diagnostics for dislocated elements

19

est (2.25) a. Jean-Jacques a présenté Émilei à Sophie. Elle eni Jean-Jacques has introduced Émile to Sophie she of-him is tout de suite tombée amoureuse. immediately fallen in-love ‘Jean-Jacques introduced Émilei to Sophie. She fell in love with himi immediately.’ b. Elle pensait à Émilei . Elle yi pensait tous les jours. she thought of Émile she of-him thought all the days ‘She thought about Émilei . She thought about himi every day.’ (Ruwet 1990: 52) The relevance of this observation to the present discussion lies in the fact that the conditions licensing a human antecedent for en and y are structurally determined. Observe the difference between (2.25) and (2.26). (2.26) a. Émilei pense que Sophie [∗ eni ] est amoureuse [de lui]. Émile thinks that Sophie (of-him) is in-love (of him) ‘Émile thinks that Sophie is in love with him.’ pense [à lui]. b. Émilei croit que Sophie [∗ yi ] Émile believes that Sophie (to-him) thinks (to him) ‘Émile believes that Sophie thinks about him.’ (Ruwet 1990: 53) Lamiroy (1991) argues that en and y can take a human antecedent as long as they are not bound by a c-commanding DP. This correctly predicts that coreference should be allowed in (2.27), where Émile does not c-command en, given that the si- clause is attached at least as high as TP. très malheureux si Sophie eni disait du mal. (2.27) Émilei serait Émile would-be very unhappy if Sophie of-him said some evil ‘Émile would be very unhappy if Sophie spoke ill of him.’ These facts are unexplained under an affixal analysis of en and y. 2.2.5.4 Concluding remarks The evidence presented in this subsection raises serious problems for a morphological analysis of French subject clitics. Elements intervening between such subject clitics and the verb stem clearly behave like syntactic entities rather than affixes, contrary to expectations. 2.2.6 Spoken French does not allow subject doubling The fourth consequence of the morphological analysis (2.9d) is that subject doubling (i.e. the co-occurrence of an XP in [spec, TP] and of an adjacent

20

French Dislocation

subject clitic) is predicted to be possible, given that the subject clitic does not have argument status. 12 The string ‘XPi –subject clitici ’ is amenable to two analyses. Either the XP is the subject of the sentence and surfaces in the canonical [spec,TP] position as in (2.28a), or it is the topic of the sentence and surfaces in a clause-peripheral position as in (2.28b). 13 (2.28) a. [ CP [TP XP [ T . . . ] ] ] [T  . . . ] ] ] b. [ CP XP . . . [TP

Canonical Subject position Topic position

The latter possibility must be acknowledged irrespective of which analysis of subject clitics one adopts, given that in certain cases an element intervenes between the XP and the clitic. The XP cannot be in [spec,TP] in examples like (2.29a), where the XP expressing the subject precedes a fronted wh-word, and (2.29b), where the XP expressing the subject of the embedded clause appears in the left periphery of the matrix clause. ellei est? (2.29) a. Et la cléi , où and the key where she is ‘And where’s the key?’ b. La cléi , je pense qu’ ellei est restée dehors. the key I think that she is stayed outside ‘I think the key was left outside.’ I will argue on interpretive grounds that, contrary to the standard assumption made by proponents of the morphological analysis of subject clitics, structures like (2.30) are never found in spoken French (or at least not in its most widely spoken varieties across Belgium, France, and Quebec). (2.30) [ CP [ TP XPi [ T’ clitici +T . . . ] ] ]

12 This is one of the diagnostics most commonly used to support the morphological analysis of (French) subject clitics. 13 I am abstracting away from the question of whether CP consists of several layers of projections. As explained in Chapter 4, the XP in (2.28b) is taken to be adjoined to either TP, possibly as in (i), or CP, as in (2.29).

(i) Tu crois que les autresi , ilsi aimeraient ça? you think that the others they would-like that ‘Do you think the others would like that?’

Diagnostics for dislocated elements

21

2.2.6.1 Distributional restrictions It is a well established fact that peripheral XPs resumed by an element within a sentence are interpreted as the topic of that sentence (see e.g. Gundel 1975; Larsson 1979; Reinhart 1981; Lambrecht 1981, 1994). In Chapter 3, I demonstrate that: (2.31) Only possible topics can be dislocated in spoken French. A number of researchers (e.g. Roberge 1990; Auger 1994; Zribi-Hertz 1994) have taken examples like (2.32) to indicate that, in dialects allowing such a sentence, subject clitics cannot be syntactic entities bearing a θ-role, but rather are a kind of agreement morpheme on the verb. This analysis relies on the assumption that indefinites like un enfant ‘a child’ in (2.32) cannot receive a topic interpretation. pose une (2.32) Un enfanti ili arrive pi il te a child he arrives then he to-you asks a question. question ‘A child arrives and he asks you a question.’

(Auger 1994)

Their reasoning proceeds as follows. If subject clitics are syntactic entities, any XP coindexed with such a clitic must appear outside the canonical subject position by virtue of the Subcategorization Principle (Chomsky 1965). Dislocated XPs are obligatorily interpreted as the topic of the sentence. Therefore, if the subject clitic in (2.32) is a syntactic entity, the coindexed un enfant ‘a child’ must be dislocated. However, indefinites cannot be topics, so a dislocation analysis of un enfant ‘a child’ is impossible. Therefore a sentence such as (2.32) is representative of a (dialectal) variety of French in which subject clitics are not arguments. The only alternative is that such elements are morphemes in T in that (dialectal) variety. This reasoning is based on Rizzi (1986). However, what is often overlooked by Rizzi’s followers is that indefinites per se are not banned as topics: it is only under their existential reading that they are incompatible with a topic interpretation; under a generic reading, indefinites can be topics (Côté 2001). The sentence in (2.32) is precisely one that receives a generic interpretation: this sentence is not about a particular child, but about a behaviour that is typical of children in general. If a specific reading is forced (by using a past tense (2.33)), this sentence is no longer acceptable for speakers of the main varieties of spoken French, including speakers of Canadian French, to one of whom sentence (2.32) is attributed.

22 (2.33)

French Dislocation ∗

a posé une question. Un enfanti ili est arrivé pi il t’ a child he is arrived then he to-you has asked a question

I will demonstrate in Section 2.2.6.2. that in spoken French, a heavy (i.e. nonweak) element expressing the subject is interpreted as a topic only if it is resumed by a subject clitic. The presence of a subject clitic in (2.32) is therefore not only possible but obligatory, which turns the argument of Auger, amongst others, on its head. 2.2.6.2 The presence of a subject clitic forces the topic interpretation of a coindexed XP In this section, three arguments are presented in support of the claim that in spoken French the presence of a subject clitic has a direct impact on the information structure of a sentence. Specifically, it is argued that an XP can only receive a focus reading in the absence of a coindexed clitic, and that an XP can only be interpreted as a topic in the presence of a coindexed clitic. The first argument relates to the availability of a focus reading of an XP. The focus is traditionally understood to be the most informative part of a sentence (Rochemont 1986). It can be restricted to the subject, as in (2.34) (a marked option, cf. Cinque 1993 and Reinhart 1996). The focus in the answer is in capitals, indicating stress prominence. (2.34) Q: Who’s eaten my porridge? A: GOLDILOCKS has. In (2.34), only the subject conveys new information, an analysis which is confirmed by the fact that the VP has been elided. Dislocated DPs cannot be focused. Hence they cannot convey the new information in the answer to a wh-question. This is illustrated for a dislocated object in (2.35). (2.35) Q: Qu’est-ce qu’il a senti? ‘What did he smell?’ A: [LA CHAIR FRAICHE]i , il (∗ l’i ) a senti(e). he it has smelled the flesh fresh ‘He has smelled THE FRESH FLESH.’ If an XP coindexed with a(n adjacent) subject clitic allows for a focus interpretation of that XP, it implies that the XP in question is not dislocated and hence that it occupies the canonical subject position, yielding a structure like (2.36b) rather than (2.36a). (2.36) a. [ CP XPi . . . [TP clitici [ T . . . ] ] ] b. [ CP . . . [TP XPi [T  clitici +T . . . ] ] ]

Diagnostics for dislocated elements

23

To test whether such a structure is allowed in (certain varieties of) spoken French, a judgement elicitation task was carried out involving fourteen native speakers from Belgium, Canada, and France. The informants were presented with eighteen contexts (including nine distractors), each with three possible follow-ups pre-recorded on CD (no transcription was provided). The set of possible follow-ups contained one sentence with an XP subject and no coindexed clitic, one sentence with an XP subject coindexed with an adjacent subject clitic, and one clearly unacceptable distractor (either completely inappropriate in the context in question, or clearly ungrammatical in any variety of spoken French). The prosody of the recorded sequence ‘XPi –subject clitici ’ was intended to be as close as possible to that of the corresponding construction without the clitic so as to avoid prompting a dislocation analysis of that XP. The contexts all forced a focus interpretation of the subject. In example (2.37), C stands for context and F for follow-up. (The follow-up distractor is omitted.) (2.37) C: La voiture bleue est foutue. ‘The blue car’s knackered.’ F: (i) Non, la voiture ROUGE elle est foutue. no the car red she is knackered ‘No, the red car’s knackered.’ (ii) Non, la voiture ROUGE est foutue. no the car red is knackered ‘No, the red car’s knackered.’ The option in which the XP expressing the subject is resumed by a clitic was accepted randomly across speakers and dialects only 4.7 per cent of the time (corresponding to six out of 126 answers, distractors excluded). Each speaker accepted at most one instance of ‘XPi -subject clitici ’ over the whole test (i.e. out of the nine test conditions). Most speakers rejected all configurations, which forced a focus interpretation of the XP, in the context provided. 14 The six answers mentioned previously can thus be treated as noise in the data. These results are consistent with a dislocation analysis of the XP coindexed with an adjacent subject clitic: dislocated XPs are topics, and topics by definition cannot be focused (see e.g. Erteschik-Shir 1997). The second argument relates to variable binding. Zubizarreta (1998: 11) argues that in several languages, including French, English, and Spanish, a 14 The informants were allowed to choose more than one option as long as they indicated which option they preferred. In almost all cases though, they only selected one option.

24

French Dislocation

QP object each/every N may bind a variable contained within the subject if and only if the subject is focused. In spoken French, the binding of a quantifier in subject position is only possible in the absence of a resumptive clitic (2.38). (2.38) Sai mère (∗ elle) accompagnera chaque enfanti . his mother she will-accompany each child Not all of my fourteen informants allowed the variable in subject position to be bound by the distributive object QP; some rejected entirely a wide-scope interpretation of the object. Those who did allow such binding almost categorically rejected sentences where the DP containing the variable was resumed by a(n adjacent) subject clitic. Of twenty-eight expected responses (i.e. testing two such sentences), nine were blank (indicating the impossibility of a widescope reading of the object) and only one out of nineteen corresponded to the string ‘DPi –subject clitici , as in (39a)’. The same speaker abstained from providing a judgement for the other test sentence, which suggests that she only marginally allowed for a distributive object QP to bind a variable in subject position. (2.39) Il faut encore décider qui rentrera chaque cheval au box. ‘We still need to decide who will take each horse to its box.’ a. Son jockey il ramènera chaque cheval its jockey he will-take-back each horse b. Son jockey ramènera chaque cheval. its jockey will-take-back each horse Again, these results are consistent with an analysis according to which a dislocated XP is resumed by a subject clitic: such XPs cannot be focused because they are topics. The third argument relates to the availability of a topic interpretation of the XP. Not all sentences can take an aboutness topic (see Chapter 3); whether this is possible depends on the information structure of the sentence. One of the key relevant factors is the span of the focus, i.e. how much of the sentence is new information. In certain contexts, all the information conveyed by the sentence is new. Such sentences are thetic as opposed to categorical. Thetic sentences describe a state of affairs and are typically uttered in answer to a question like What happened? In contrast to categorical sentences, they do not predicate something about a referent whose existence is presupposed: they cannot have an

Diagnostics for dislocated elements

25

aboutness topic. Imagine a situation in which person A sees person B in tears, prompting the exchange in (2.40). (2.40) A: Qu’est-ce qui s’est passé? ‘What happened?’ B: Les voisinsi (#, ilsi ) ont mangé mon lapin. the neighbours they have eaten my rabbit ‘The neighbours have eaten my rabbit.’ In the context given in (2.40), the referent of les voisins ‘the neighbours’ is not established, so it is not available to be the sentence’s topic. B’s response has to describe a state of affairs; it consists entirely of new information. What is interesting for our present purpose is that certain predicates can never appear in thetic sentences. Such predicates belong to a relatively uniform class defined as Individual Level Predicates (Milsark 1974). Individual Level Predicates (henceforth ILPs) generally express permanent properties. 15 Their subject is obligatorily interpreted as the topic of the sentence (Erteschik-Shir 1997), except when there is a narrow focus on that constituent (as I argue in Chapter 3). This is illustrated in (2.41), where both A’s statement and B’s amendment contain an ILP. In B’s utterance, there is narrow (contrastive) focus on the DP expressing the subject. This DP is thus forced to appear in the canonical subject position where a contrastive focus reading is allowed. However, a property reading obtains in both utterances. In A’s statement, the topic is ta sœur ‘your sister’, while in B’s it is a covert aboutness topic corresponding to something like ‘musical people in the speaker’s family’. (2.41) A: Ta sœur, elle est musicienne. your sister she is musician ‘Your sister’s a musician.’ B: [MON FRÈRE AUSSI]FOCUS est musicien. is musician my brother too ‘My brother’s a musician too.’ If a topic interpretation is only allowed in the presence of a resumptive clitic in spoken French, one can therefore expect that ILPs will always take a subject clitic except when there is narrow focus on the subject. This prediction was confirmed by analysis of the York and Cat corpora: of a random sample of 4,030 clauses from these corpora, ILPs did not appear without a subject clitic 15 This is a simplification. For an in-depth discussion of the properties of ILPs, see Jäger (2001). I return to ILPs in Chapter 3, where the distinction between ILPs and SLPs is introduced.

26

French Dislocation

except in those rare instances in which there was a narrow focus reading of the subject. Examples are given in (2.42). Maman fait à manger. (2.42) a. La cuisine, c’ est le lieu où the kitchen it is the place where Mum makes to eat ‘The kitchen is the place where Mum cooks.’ b. [Luc aussi]FOCUS a les yeux de son père? has the eyes of his dad Luc too ‘Luc too has his dad’s eyes?’ The quasi-obligatory presence of a subject clitic with an ILP is a direct consequence of the fact that the subject of an ILP is interpreted as a topic except when it is in narrow focus. 2.2.7 Conclusion The evidence discussed so far contradicts claims made by the proponents of the morphological analysis of French subject clitics and raises a number of problems for its implementation. We have seen that subject (clitic)-verb inversion is productive, and that XP subjects in [spec,TP] are never doubled by a clitic in the most widely spoken varieties of French. As a result, the empirical basis motivating the morphological analysis of French subject clitics is considerably reduced. This takes us to the question of why such an analysis should have been proposed in the first place. 2.2.8 French subject clitics: grammatical or anaphoric ‘agreement’? The idea that French subject clitics might be agreement markers was inspired originally by comparisons with polysynthetic languages such as Amerindian ones, in which verb forms are long and morphologically complex, and include morphemes whose features correspond to the argument(s) of the verb (Vendryes 1920). Givón (1976) proposes that this state of affairs is the result of historical changes which begin with the incorporation of pronouns into verbs and lead to their reanalysis as grammatical agreement markers. In their seminal paper, Bresnan and Mchombo (1987) discuss the reasons for the close relation between incorporated pronouns and grammatical agreement markers, and propose a series of well-motivated diagnostics to distinguish them. In their terminology, which I adopt in the discussion below, grammatical agreement corresponds to agreement markers or morphemes and anaphoric agreement corresponds to incorporated pronouns or clitics.

Diagnostics for dislocated elements

27

On the basis of such diagnostics, Bresnan and Mchombo (1987) demonstrate that in Chichewa, ˆ a Bantu language spoken in East Central Africa, subject markers are ambiguous between grammatical and anaphoric agreement markers, whereas object markers are unambiguously anaphoric agreement markers. I propose to apply these diagnostics to French subject clitics in order to determine whether a morphological analysis in terms of agreement markers (henceforth grammatical agreement) is plausible. Following Bresnan and Mchombo (1987), if the morphological analysis is correct, French subject clitics are expected to behave like grammatical agreement markers. 2.2.8.1 Locality Only grammatical agreement markers require the DP with which they are associated to be local, i.e. to appear in the same clause. Hence in spoken French, at least those cases where the DP is not adjacent to the subject clitic are instances of anaphoric agreement (as already pointed out with respect to (2.29)). This also means that in examples like (2.43) the subject clitic elle ‘she’ cannot be an example of grammatical agreement. (2.43) Salma Hayeki , ça faisait huit ans [qu’ellei travaillait sur ce film]. Salma Hayek it made eight years that-she worked on that film ‘It had been eight years that Salma Hayek had been working on that film.’ This suggests that the possibility of a non-local relation between a DP and a clitic could be used as a diagnostic for anaphoric agreement. If it is possible to separate the DP and the clitic in a sentence like (2.44a) by placing the DP in a higher clause and leaving the clitic in what then becomes the lower clause (as in (2.44b)) without altering the information structure interpretation of the DP (i.e. whether is it a topic or not), the clitic in question should be treated as an anaphoric agreement marker (incorporated pronoun) rather than as a grammatical agreement marker (morpheme) not only in sentences like (2.44b) but also in sentences like (2.44a). (2.44) a. Les autres, ils sont là. the others they are there ‘The other ones are there.’ b. Les autres, je pense qu’ils sont là. the others I think that-they are there ‘I think that the other ones are there.’

28

French Dislocation

2.2.8.2 Questioning of the related argument A grammatical agreement marker should be present even when the argument to which it is related is questioned. This is true of Chichewa’s ˆ subject markers (SM) but not its object markers (OM). (2.45) a. (Kodí) chíyâni chi-ná-ónek-a?16 what SM-past-happen-indic Q ‘What happened?’ b. ?? (Kodí) mu-ku-chí-fún-á chíyâni?17 you-pres-OM-want-indic what Q The presence of a subject clitic in subject questions is impossible in spoken French. (2.46) a. Quii (∗ ili ) veut du gouda ? who he wants some gouda-cheese ‘Who wants some Gouda cheese?’ b. Quels soldatsi (∗ ilsi ) sont partis ? which soldiers they have left ‘Which soldiers have left?’ 2.2.8.3 Topicalization of parts of idioms DPs associated with an anaphoric agreement marker are interpreted as topics, which implies that their reference should be recoverable from the context. This observation leads Bresnan and Mchombo (1987) to predict that it should not be possible to associate parts of idioms with anaphoric agreement markers, given that the meaning of idioms is not established compositionally and that consequently DPs used in idioms tend not to correspond to discourse referents so they should not be interpretable as topics. In contrast, grammatical agreement markers do not entail a topic interpretation of the DP associated with them and so are predicted to be acceptable in idioms. This is verified for Chichewa’s ˆ subject marker. (2.47) Bôndo li-ná-nóng’onez-ˇedw-a. knee SM-past-whisper.to-pass-indic ‘The knee was whispered to.’ (i.e. remorse was felt)

16

ˇ indicates a rising tone, ´ indicates a high tone, and ˆ indicates a falling tone. The relative acceptability of the object question in (2.45b) is due to the fact that it can be interpreted as a reduced cleft. In Chichewa, ˆ object markers are allowed to appear as resumptive elements in such constructions (Sam Mchombo, p.c.). 17

Diagnostics for dislocated elements

29

The same is not possible in spoken French. The idiomatic interpretation—the game is up is not retained in (2.48). (2.48) Les carottesi ellesi sont cuites. the carrots they are cooked ‘The carrots are cooked.’ (Rather than ‘The game is up’.) 2.2.8.4 Peripheral vs. core status of the related argument A DP associated with a grammatical agreement marker does not have to be peripheral: it should be allowed to occupy its canonical (argument) position in the presence of such a marker. This is not possible with anaphoric agreement markers, given that these are incorporated pronouns. If an anaphoric agreement marker co-occurs with a coreferential argument in the core of the sentence, this yields a violation of the Subcategorization Principle (Chomsky 1965). Prosodic evidence confirms that an XP is indeed dislocated whenever it is resumed by a subject clitic (see Deshaies et al. 1993 and Guilbault 1993 for Quebec French, and also Section 2.3). 2.2.8.5 Conclusion In Chichewa, ˆ subject agreement markers are always present, even when the subject is not interpreted as the topic of the sentence, when it is questioned, or when it is in focus (Sam Mchombo, p.c.). Bresnan and Mchombo (1987) conclude that subject markers in this language have a dual status of anaphoric agreement marker when a topic interpretation of the subject obtains, and grammatical agreement marker in all other cases. They argue that anaphoric agreement markers are incorporated pronouns, i.e. that they have argument status in spite of the fact that they are affixed to the verb. The diagnostics applied in this section reveal that French subject clitics do not behave as predicted under a morphological analysis: they do not behave like grammatical agreement markers. They are more akin to Chichewa’s ˆ object markers, i.e. to anaphoric agreement markers (incorporated pronouns). In the next section, I investigate how the anaphoric agreement relation between a subject clitic and an XP can be captured under the syntactic and the morphological analyses of subject clitics. 2.2.9 Information structure and syntactic structure We have seen that in spoken French, an XP coindexed with an adjacent subject clitic is obligatorily interpreted as a topic. For a focus interpretation of an XP to obtain, a subject clitic cannot be present. Under a syntactic analysis of subject clitics, the topic interpretation of the XP follows directly from the syntactic structure of the sentence. That XP is

30

French Dislocation

obligatorily dislocated whenever there is a subject clitic because it cannot occupy the canonical subject position since the latter is occupied by the clitic. The resulting dislocated structure (2.36a) is syntactically transparent to what Erteschik-Shir (1997) calls f(ocus)-structure, the grammatical level at which the scope of topic and focus is defined and which mediates between syntax and PF/LF. At f-structure, the topic has to take scope over the rest of the sentence. Under the syntactic analysis, no ad-hoc mechanism is required to account for the information structure contrast between sentences with a subject clitic and those without one. Under a morphological analysis of subject clitics, the topic interpretation of the XP expressing the subject is purely accidental, from a syntactic point of view. Some authors even assume that the XP cannot possibly be a topic in all cases because the string ‘XPi –subject clitici ’ occurs in such a high proportion of the sentences in corpora of spontaneous production (e.g. Auger 1994: 116). However, a high proportion of subject topics is exactly what is to be expected given that crosslinguistically the topic has a strong tendency to coincide with the grammatical subject of the sentence (see e.g. Gundel 1975; Givón 1976; Li and Thompson 1976; Reinhart 1981; Lambrecht 1994). Proponents of the morphological analysis consider that, by default, the XP occupies the subject position whenever it is adjacent to the subject clitic, as in (2.30). Under this analysis, the absence of a subject clitic in certain sentences/clauses is not therefore predicted. This claim is clearly stated in Auger (1994: 93): (2.49) ‘Subject markers are true agreement markers and are thus expected to show up on every finite verb.’ Given (2.49), the absence of a subject clitic is predicted to be random or speaker-dependent (as suggested by Auger 1994: 13). To account for the facts discussed in Section 2.2.6.2, it would therefore be necessary to stipulate that the realization of the subject clitic is blocked whenever the subject is focused (i.e. when it is in narrow focus or when the sentence is thetic). In addition to this, the realization of the clitic must be made obligatory in cases where there is no coindexed XP adjacent to (and preceding) the finite verb, so as to rule out sentences like (2.50) but not (2.51). (2.50) is correctly ruled out by virtue of (2.49) but (2.51) is wrongly predicted to be impossible (or marked) in spoken French due to the absence of a subject clitic. (2.50)

∗

A pas de sujet, cette phrase. has no subject this sentence

Diagnostics for dislocated elements

31

(2.51) Où sont les crayons? where are the pencils ‘where are the pencils?’ One solution to this problem is proposed by Roberge (1986), who stipulates that the clitic can only absorb Case in the absence of material in the canonical subject position. However, following Baker (1996), the absorption of case features is triggered by some requirement of the agreement morpheme itself, making overt NPs in A-positions impossible. The presence of an NP thus cannot block case absorption: it is itself barred as a result of case absorption. This brings us back to the problem of how to account for the distribution of the clitic in the first place. Another possibility would be to consider subject clitics to be topic markers, e.g. by endowing them with a topic feature. This would account for the fact that such clitics are only realized when the subject is interpreted as topic. An interesting consequence would be the blurring of the distinction between topics and pronoun-like elements, at least in the case of those associated with the subject: if subject clitics bear a topic feature, any sentence with such a clitic would force a topic interpretation of the subject, i.e. whether or not Tim is uttered in (2.52). This idea is compatible with the claim that topics can be covert (cf. e.g. Gundel 1975). (2.52) (Timi ) ili a retrouvé ses framboisiers. Tim he has retrieved his raspberry-canes ‘Tim has got his raspberry-canes back.’ It would take me beyond the scope of this section to discuss whether it is theoretically justified to consider pronominal (subject) clitics always to be topics. Note that analysing subject clitics as topic markers would leave certain issues unresolved if the morphological analysis is adopted. One such issue relates to the fact that whenever a subject clitic is omitted, the subject position has to be filled by overt lexical material, as indicated by the contrast between (2.53a) and (2.53b). However, under the morphological analysis when a subject clitic is present, the subject position can be left empty, as in (2.53c). ∗

[T M’ aidera]]. me will-help b. [ TP Kester [ T m’ aidera]]. Kester me will-help

(2.53) a.

[TP 0

32

French Dislocation c. [TP 0 [T Il m’ aidera]]. he me will-help ‘He/Kester will help me.’

Another issue that would not be resolved by endowing subject clitics with a topic feature is that in some cases there is feature mismatch between the putative subject DP and the finite verb, as in (2.54) where the DP is plural and the verb agreement morphology singular. (2.54) Les banques pl . c’ ests g . les banques. it is the banks the banks ‘Banks will be banks.’ 2.2.10 The morpheme-like properties of French subject clitics are accidental By their nature, clitics are hybrid elements with a status somewhere between pronouns and affixes, so they can be expected to share properties of both. The analysis of clitics (i.e. syntactic or morphological) therefore depends ultimately on the relative weight attributed to these sets of properties. There is presently no consensus on the issue of what should be the decisive criteria for a morphological rather than a syntactic analysis of clitics. In (2.55), I list the properties of French subject clitics that are expected under a morphological analysis but not under a syntactic one and that have not yet been addressed here. 18 (2.55) Subject clitics in spoken French a. cannot be conjoined: ∗ Il et elle viendront (‘He and she will come’) b. cannot take wide scope: ∗ Je pleure et dors (‘I cry and sleep’)

18 I do not include complex inversion in this list because this construction, illustrated in (i), is not productively used in spoken French and therefore is unattested in the York and Cat corpora.

(i) Les carottes sont-elles cuites ? the carrots are-they cooked ‘Are the carrots cooked?’/‘Is the game up?’ (literal vs. idiomatic meaning) Marie Labelle (p.c.) points out that complex inversion follows straightforwardly from a morphological analysis of subject clitics. However, the morphological analysis was proposed to account for spoken French and it is explicitly stated that it does not apply to Standard French, so the fact that complex inversion, a construction exclusively attested in Standard French, can be accounted for neatly under a morphological analysis of subject clitics cannot be taken as an argument supporting such an analysis. For an account of complex inversion under a syntactic analysis of French subject clitics, the reader is referred to Laenzlinger (1998).

Diagnostics for dislocated elements

33

c. display (very rare  instances of) idiosyncrasy: e.g. je suis (‘I am’) is pronounced [ y] only when it is a form of the copula, not when it is suivre ‘to follow’ The first two properties are ascribable to the phonologically weak nature of subject clitics, as suggested by e.g. Labelle (1985). Whether cases of alleged idiosyncrasy such as the example in (2.55c), are due to the morphological status of subject clitics is controversial. Marie Labelle (p.c.) notes that the pronunciation of je ‘I’ as [ ] results from a phonological rule of devoicing in front of an unvoiced consonant (as in je fais ‘I make’, je te dis ‘I tell you’), and that the omission of [i] in the pronunciation of suis is observed independent of whether the first person singular subject clitic je is attached to it (e.g. the auxiliary suis in je me suis fait mal ‘I hurt myself ’ is pronounced [sy] in Quebec French). 2.2.11 Conclusion Given the empirical facts, the morphological analysis of French subject clitics reveals itself to be more problematic than has previously been acknowledged in the literature. Some of the empirical assumptions on which that analysis is founded are not verified by relevant data. I have demonstrated that in the most widely spoken varieties of French, subject clitic–verb inversion and the scope marker ne are used productively, and that the appearance of subject clitics is prohibited whenever the subject is not the topic of the sentence. This considerably reduces the empirical basis of the morphological analysis. Whenever an XP is associated with a subject clitic in spoken French, that XP is dislocated and interpreted as the topic. This means that French subject clitics are more akin to anaphoric agreement markers (incorporated pronouns) than to grammatical agreement markers (morphemes). It might well be possible to account for these facts under a morphological analysis of French subject clitics, but only at a cost. Ad hoc rules or mechanisms need to be postulated to derive the following: (2.56) a. Subject clitics are only realized when the subject is the topic. b. Subject clitics are obligatory when no XP expressing the subject immediately precedes or follows the verb. c. There can be feature mismatch between the subject clitic and the coindexed XP. d. Subject clitics can appear preverbally or postverbally under certain discourse conditions.

34

French Dislocation

Implementing a morphological analysis of French subject clitics leads to redundancies in the verb agreement paradigm and places a heavy burden on the lexicon (e.g. requiring the existence of numerous variants for each clitic). In addition, it has been shown that the distribution of elements intervening between a subject clitic and the verb stem is syntactically constrained. This is incompatible with the hypothesis that such elements are affixed to the verb stem in the lexicon. I conclude that the morphological analysis of French subject clitics is untenable, at least in its present form. The limited number of similarities between the behaviour of French subject clitics and morphemes or affixes listed in (2.55) should thus be treated as ‘noise’ rather than as decisive evidence in favour of a morphological analysis of the former.

2.3 The prosodic characteristics of French dislocation All studies of dislocation in spoken French mention prosody as one of the diagnostics for this type of construction. However, while the characteristics of right-dislocation prosody are clearly defined (see e.g. Rossi 1999), issues regarding the prosody of left-dislocation are not yet fully resolved. I start by presenting briefly the established prosodic diagnostics for French rightdislocation (Section 2.3.1), before turning to a more detailed investigation of left-dislocation prosody (Section 2.3.2). 2.3.1 Right-dislocation prosody in spoken French Nowadays, it is widely acknowledged that right-dislocated elements in spoken French have a distinctive prosody. Three main characteristics have been proposed. 19 The first characteristic is that Right-dislocated XPs are destressed, especially in contrast to the immediately preceding intonation peak on the last clauseinternal element (Lambrecht 1981: 84–85). They are also of a smaller amplitude (i.e. loudness) than the nucleus of the sentence (Ashby 1994: 39). Ashby (1994: 37) notes that the intonation peak, which marks ‘the most salient piece of information in the sentence’, does not always immediately precede the right-dislocated element, but that there is nonetheless a sharp contrast between the ‘intonation’ of the sentence nucleus and that of the right-dislocated element. By ‘intonation’, Ashby means ‘pitch’, i.e. variations in melody (his observations are based on graphs representing variations in 19 The findings of Labelle and Valois (1996) suggest that these three characteristics are also valid for right-dislocation in child French.

Diagnostics for dislocated elements

35

fundamental frequency). Ashby’s (1994) observation corresponds to the second characteristic of right-dislocation prosody: right-dislocated XPs are lower in pitch than the nucleus of the sentence. In the twenty-eight spontaneous examples he studied, Ashby (1994: 39) reports that the drop in pitch from the sentence nucleus to the right-dislocated element measured nearly five semitones on average. This prosodic contour was first defined by Delattre (1966) as the parenthesis intonation, which manifests itself in declarative sentences as a flat pitch, typically on a low level (a plateau). In the terminology of Mertens et al. (2001), it corresponds to the appendix contour: the pitch of the right-dislocated element reaches the lowest level of the speaker’s range. 20 Finally, contrary to what is often claimed (e.g. Ronat 1979; Pierce 1992, 1994; Kaiser 1994), there is no obligatory pause between the sentence nucleus and a right-dislocated element (e.g. Ashby 1994; Rossi 1999). More thorough investigation of the phenomenon in question (Rossi 1973, 1981, 1999) reveals that these three characteristics are not valid for all cases of right-dislocation in spoken French. 21 In order to account for the prosody of right-dislocation in all utterance types and not just declaratives, Rossi (1999: 83) characterizes right-dislocation prosody as ‘lacking a discrete intonation’. The right-dislocated element is analysed as being prosodically dependent on the rest of the sentence, and in particular on the final contour of the sentence nucleus: it copies the preceding marker, but it has decreased intensity, a lower pitch, and a flat contour which ends with a further pitch drop. Hence in a sentence which has ‘conclusive’ intonation (i.e. descending pitch) the rightdislocated element will be still lower in pitch than the sentence nucleus and will have a slight terminal drop of its own. By contrast, in a sentence which ends with a pitch rise on the sentence nucleus (as in a yes/no question, for instance) the right-dislocated element will itself display a pitch rise, but one that is lower and of a smaller size than that of the preceding element. Rossi further argues that sentences stripped of a right-dislocated element are prosodically well-formed (Rossi 1973, 1999). In other words, the non-dislocated part of the sentence is prosodically self-sufficient. I will now briefly illustrate each of the three characteristics of rightdislocated elements with examples from the Cat corpus. These examples have been extracted from the same recording session to control for variations in 20

The range of a speaker corresponds to the total melodic span covered by his/her voice in spoken language (e.g. Rossi 1999). 21 Various approaches to the prosody of dislocation in spoken French including that of Rossi (1999) and Mertens et al. (2001) are discussed in more detail in Section 2.3.2. I have tried to keep the section about right-dislocation prosody as concise as possible, given that there is little controversy (if any) regarding what the essential prosodic characteristics of this construction are.

36

French Dislocation

recording conditions. This session involved a child (Chloé) and her mother (Geneviève), both from Brussels, Belgium. 22 Eighteen utterances containing a right-dislocated element were extracted from the recording. The sentences in (2.57) exhibit the typical characteristics of right-dislocation prosody, as defined previously. In each figure, the thick line represents pitch, i.e. variations in fundamental frequency, and the thin line intensity. 23 Each syllable (indicated by vertical lines) is identified separately on the basis of the way the string was pronounced rather than its orthography. In the corresponding text in each figure, the unpronounced material is given in parentheses. (Geneviève, B) (2.57) a. Ah ben non ili a pas de chapeau, celui-cii . this-one ah well no he has not part hat ‘Ah this one doesn’t have a hat.’ (surprise intonation) b. Tu vas parfois leuri donner à manger, aux canardsi ? you go sometimes to-them give to eat to-the ducks ‘Do you sometimes go and feed the ducks?’ (Geneviève, B) The right-dislocated element in (2.57a), as represented in Figure 2.1, appears after the main intonation peak which falls on the last syllable of chapeau ‘hat’, i.e. outside the sentence nucleus. The pitch of the dislocated element is quite high, consistent with the hypothesis that right-dislocation prosody is a copy of the preceding intonation marker (Rossi 1999). 24 This sentence is uttered with a tone of surprise, with a rise on the last word of the sentence nucleus (chapeau ‘hat’). The right-dislocated element is only slightly lower in pitch and intensity than the last word of the sentence nucleus. This first example illustrates that even in declarative sentences right-dislocation prosody is not always characterized by a sharp lowering in pitch or a flat pitch lower than that of the sentence nucleus; the prosodic contour of a right-dislocated element has to be captured in terms of being a copy of the preceding intonation marker. 22 These data were collected in Belgium in 1999 using a Sony MZ-R91 minidisc recorder and a Sony ECM-F8 boundary microphone. The sound files were subsequently converted into wave files at a rate of 22,050 Hz. 23 The level 0 of the speaker’s range (represented by the y-axis in the graph) corresponds to the lowest pitch reached by the speaker in spontaneous speech. The range of a speaker is the total melodic span covered by his/her voice in spoken language (e.g. Rossi 1999). The value of 0 was calculated on the basis of five minutes’ worth of extracts from the speaker in question, including very quiet moments as well as lively, loud, and more high-pitched ones. All calculations were performed using the sound analysis programme Praat written by Paul Boersma. 24 The continuation of the pitch trace of the last syllable of chapeau ‘hat’ into the following one is due to echo in the room. The [s] of celui ‘this’ starts at the point indicated by the vertical line between these two syllables.

37 85

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

74

63

52

INTENSITY (dB)

PITCH(semitones)

Diagnostics for dislocated elements

41 ah 0

ben

non i(l) n’a pas d(e) cha-

-peau 1

0.5

c(el)ui-

-ci 30 1.5

TIME(s)

Figure 2.1. Prosody of a declarative sentence with a right-dislocated element (spontaneous, Geneviève)

The pitch of the right-dislocated element in (2.57b), as represented in Figure 2.2, is higher than most of the sentence nucleus, but is not quite as high as that of the preceding intonation marker on manger ‘eat’. The process of copying in this example is limited to the last syllable of the sentence nucleus and the last syllable of the dislocated element: the latter copies the former in the sense that its pitch is higher than that of the rest of the sentence, but to a lesser degree than the last syllable of the nucleus. There is however no rise in pitch inside the dislocated element copying the rise in pitch on the last part of the sentence nucleus. In Section 2.3.2, which contains a more in-depth analysis of dislocation prosody, I propose that French prosody is best captured using the notion of Intonation Group. Only the accented (i.e. stressed) syllables of a given Intonation Group are relevant to determining the prosodic contour of an utterance. Hence the fact that the first syllable of the right-dislocated element (which is destressed) in Figure 2.2 starts on a fractionally higher pitch than the preceding element is of no relevance. The sentences in (2.58), which were uttered by a child, also display typical right-dislocation prosody. (2.58) a. Je la connais pas, hein, moi. I her know not eh me ‘I don’t know her, do I?’

(Chloé, B)

38 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

85

74

63

52

INTENSITY (dB)

PITCH(semitones)

French Dislocation

41 et tu vas par- -fois leur do(n)- -ner à man- -ger aux ca- -nards? 30 0

0.5

1

1.5

TIME(s)

Figure 2.2. Prosody of a yes/no question with a right-dislocated element (spontaneous, Geneviève)

b. Qui c’ est, ça? who it is that ‘Who’s that?’

(Chloé, B)

The prosody of the right-dislocated element in (2.58a), represented in Figure 2.3, exhibits the classic low, flat pitch reported in the literature (e.g. Delattre 1966) accompanied by a decrease in intensity. The dislocated element is preceded by a tag hein ‘eh’, which is typically found between dislocated elements and the sentence nucleus in spontaneous speech. Figure 2.4 shows that the right-dislocated element in (2.58b) is not set off from the sentence nucleus as clearly as in previous examples. Sentence-final ça ‘that’ in (2.58b) is unambiguously right dislocated though. Note that ça ‘that’ is monosyllabic, which gives less room for variations in pitch and intensity. In spite of this, the pitch of right-dislocated ça is lower than that of the copula at the end of the sentence nucleus. This is significant given that the overall pitch variation is very small in this sentence. 25 In the corpus I have examined, there are a few right-dislocated utterances that do not display exactly the characteristics defined as typical. I will briefly describe two such examples, given in (2.59). Despite their apparently a typical prosody, it is clear to native speakers that the sentence-final constituent is right 25

This is discussed further in Section 2.3.2.

39 85

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

74

63

52

INTENSITY (dB)

PITCH(semitones)

Diagnostics for dislocated elements

41 j(e) (l)a

co(n)-

-nais

pas

hein

moi 30

0.5

0

1 TIME(s)

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

85

74

63

52

INTENSITY (dB)

PITCH(semitones)

Figure 2.3. Prosody of a declarative sentence with a right-dislocated element (spontaneous, Chloé)

41 qui

c’est

ça 30 0.5

0 TIME(s)

Figure 2.4. Prosody of a wh-question with a right-dislocated element (spontaneous, Chloé)

French Dislocation 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

85

74

63

52

INTENSITY (dB)

PITCH(semitones)

40

41 et qu’est-ce que tu leur donnes à man-

-ger

aux

ca-

-nards? 30

887.5

888

888.5

TIME(s)

Figure 2.5. Prosody of a wh-question with a right-dislocated element (spontaneous, Geneviève)

dislocated even in these cases. Such examples illustrate how much spontaneous speech can vary from those patterns defined on the basis of controlled sentences elicited out of context on which most prosodic analyses are based. One has to keep this variation in mind when applying strict prosodic diagnostics for dislocated constructions to examples gathered from spontaneous speech. (2.59) a. Et qu’ est-ce que tu leuri donnes à manger, aux and what is-it that you to-them give to eat to-the (Geneviève, B) canardsi ? ducks ‘And what do you feed the ducks?’ b. Ellei est pas belle, hein, celle-lài ? (Geneviève, B) she is not pretty eh that-one ‘That one isn’t pretty, is she?’ In sentence (2.59a), 26 the prosody of the right-dislocated element is not a lowered copy of that of the last element in the sentence nucleus. Instead, the pitch of the dislocated element is a higher copy of that of the sentence 26 There is a sudden noise when Geneviève utters que. This explains the peak of intensity on that element.

41

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

85

74

63

52

INTENSITY (dB)

PITCH(semitones)

Diagnostics for dislocated elements

41 e(l)- -l(e) est

pas

be(l)-

-l(e) hein

celle-

-là? 30

0

0.5 TIME(s)

1

Figure 2.6. Prosody of a yes/no question with a right-dislocated element (spontaneous, Geneviève)

nucleus’s final part. This leads me to conclude, contra Rossi (1999), that having a lower pitch than the final part of the sentence nucleus is not the essential characteristic of right-dislocation prosody, at least not in questions. 27 The sentence in (2.59b) is not strictly speaking a yes/no question, but the presence of the tag hein together with the rising final contour (both of which are typically used to seek confirmation from the addressee) has the effect of making the sentence more like a yes/no question than a declarative. (This is reflected by the inclusion of a tag question in the English translation.) In this case, as in (2.59a), the prosody of the right-dislocated element does not copy that of the end of the sentence nucleus in the expected way. Figure 2.6 shows a clear rise in pitch on the dislocated element. The end point of this rise has the highest pitch in the whole sentence. In addition, the right-dislocated element is only marginally lower in intensity than the rest of the sentence and the last syllable of the sentence nucleus (i.e. belle ‘beautiful’). Geneviève reaches the lowest possible level of her range on belle ‘beautiful’ and the following syllable. 28 The pitch of belle is so low in fact that it is uttered with a creak. 29 The prosody copied by the right-dislocated element might be 27 Rossi’s (1999: 88) original formulation is: ‘L’intonème de thème est une copie abaissée de l’intonème qui le précède.’ (‘The theme intoneme is a lowered copy of the preceding intoneme’). 28 See n. 22 for a definition of the notion of range. 29 Creak is the ‘broken voice’ effect observed when a speaker reaches the bottom of his/her range.

42

French Dislocation

that of the tag hein rather than that of the last part of the sentence nucleus, but even if that is the case the copy is higher than the copied element and not lower as claimed by Rossi (1999). I will leave for further research the task of integrating these observations into a principled account of right-dislocation prosody. I have used right-dislocation prosody as a diagnostic for cases like (2.60), where there is no other indication that the sentence contains a dislocated element. (2.60) a. Je sais, tout ça, tu sais? I know all that you know ‘I know all that, you know.’ b. Tu aimes bien, les colliers? you like well the necklaces ‘Do you like necklaces?’

(A.-Gaël, F)

(Nelly, B)

It will be confirmed in Chapter 3 that a right-dislocation analysis of sentences such as those in (2.60) is well-founded. Briefly, the justification will be that in both cases the element uttered with right-dislocation prosody is the topic of the sentence. In (2.60a), the referent of the dislocated element is salient in the context provided; (it corresponds to a d-linked set; see Chapter 3). 30 In (2.60b) the interpretation is generic and favours the omission of the resumptive. Additional justification for the use of prosody as a diagnostic for syntactic right-dislocation is provided by the examples in (2.61). (2.61) a. #Télesphore est heureux, avec une femme. Télesphore is happy with a woman b. *Qu’i est-ce qu’elle a dit, qu’elle avait fait ti ? what is-it that-she has said that-she had done Sentence (2.61a) is anomalous if the PP is uttered with right-dislocation prosody under an existential interpretation, i.e. this sentence cannot mean that Télesphore is happy with a particular woman. (2.61a) becomes felicitous under a generic reading, where no reference is made to any woman in particular. These facts are straightforwardly derived from a right-dislocation analysis of avec une femme ‘with a woman’ (see Section 3.2.2). 30

D-linked elements are given in or are recoverable from the preceding discourse (Pesetsky 1986).

Diagnostics for dislocated elements

43

Sentence (2.61b) is ungrammatical if the sentence-final clause is uttered with right-dislocation intonation. As shown in Section 5.2.1, a sentence must be syntactically well-formed when it is stripped of its dislocated elements. This particular sentence is ungrammatical because the final clause containing the trace of the wh-element cannot be omitted, which is incompatible with it being right-dislocated. 2.3.2 Prosodic differences between left-dislocated and heavy subjects in spoken French: a review of the literature There are two main reasons for wanting to establish definite prosodic diagnostics for left-dislocated elements in spoken French. First, they could be used to test the claim that in certain varieties of French an element in the subject position can be doubled by a subject clitic affixed to the verb, as detailed in Section 2.2. For this claim to be justified, the prosody of doubled subjects would have to differ from that of left-dislocated subjects in the dialects in question. Second, clear prosodic diagnostics for left-dislocated subjects would be of great significance to the study of subjects in child French. As is well known, children go through a null subject stage (Bloom 1970; Hyams 1986), during which they frequently omit the subject of the verb. It has been shown that even when the subject clitic is missing, a right-dislocated element coreferential with the subject may appear in child French (Ferdinand 1993, 1996; Labelle and Valois 1996), as exemplified by (2.62). gagné, moi. (2.62) (J’) ai (I) have1 p.s g . won me ‘I’ve won.’

(Tom 2;1.11)

In De Cat (2004a), I show that left-dislocated elements coreferential with the subject are also possible in child French when the subject clitic is missing, as in (2.63). (Max 2;3.20) (2.63) Moi, 0 mettre ça comme Pol.31 Pol me put−FIN that like ‘I (want to) put it like Pol.’ (meaning derived from context) In cases like (2.63), the non-nominative case of the preverbal element is sufficient to identify it as a left-dislocated subject. In cases like (2.64), however, no such indicator is available, and the only way to determine whether the 31

See Appendix B.1 for transcription conventions.

44

French Dislocation

preverbal element is a left-dislocated subject or a heavy subject 32 is to examine its prosody and the context in which the sentence was uttered. (2.64) Là, tout (le) monde fait (l)a fête. there all the people makes the party ‘There everybody’s partying.’

(Anne 2;4.20)

To establish clear prosodic diagnostics for left-dislocated elements, as in (2.65a), a comparison must be made with the prosody of heavy subjects, as in (2.65b). (2.65) a. Les Belgesi , ilsi sont les plus braves. the Belgians they are the most brave b. Les Belges sont les plus braves. the Belgians are the most brave ‘Belgians are the bravest.’ Diagnostics are therefore needed to distinguish left-dislocated subjects from heavy subjects. Various attempts have been made to describe the prosody of left-dislocated elements in spoken French. These can be found in studies whose main focus is the syntax or the pragmatics of dislocated elements (e.g. Larsson 1979; Carroll 1982; Barnes 1985; Cadiot 1992; Kaiser 1994; Rizzi 1997), as well as in studies specifically dedicated to the prosody of dislocated elements (Deshaies et al. 1993; Rossi 1999; Mertens et al. 2001). In both types of studies, it is taken for granted that left-dislocation prosody is easily identified by native speakers. In the syntax/pragmatics studies, the basic characteristics attributed to leftdislocated elements are a pause separating the dislocated element from the sentence nucleus, and the greater relative prominence of the dislocated element in comparison with the rest of the sentence. This prominence manifests itself in terms of pitch (i.e. melody) and intensity. This type of prosody is sometimes referred to as comma intonation (e.g. Emonds 1979; Rizzi 1997). Heavy subjects are claimed not to display these characteristics. Below I review three different approaches to the acoustic analysis of leftdislocated elements in spoken French. The approaches considered are (i) that of Deshaies et al. (1993) and Guilbault (1993), (ii) that developed in the work of Mario Rossi (Rossi 1971, 1972, 1981, 1999), and (iii) that developed by Piet Mertens (Mertens 1987, 1993, 1997; Mertens et al. 2001). The first 32 I use the term heavy subject to designate a non-clitic element appearing in subject position. Such an element is not coindexed with a resumptive element.

Diagnostics for dislocated elements

45

of these approaches investigates the relevance of various arbitrarily chosen traits for left-dislocation prosody. The other two integrate the prosody of leftdislocated elements into a coherent system aimed at describing and explaining the prosody of spoken French. A short introduction to these two approaches is provided before I turn to the actual acoustic characteristics of left-dislocation prosody. Mario Rossi adopts the morphological approach to intonation, in the tradition of the Prague school. Under this approach, left-dislocation prosody is defined as and identified by a bundle of traits or intonemes which relate to various parameters (intensity, time, melody). Such traits should not therefore be considered in isolation. Rossi’s research is based on a fine-grained acoustic analysis of the data, whereby variations in each parameter are measured in terms of Perception Units (PU). One PU corresponds to the minimum difference in a given parameter that can be perceived by the human ear in naturalistic speech. 33 In a nutshell, left-dislocated topics are identified by the following combination of intonemes.

r Their fundamental frequency (F0 ) 34 dominates the utterance and is characterized by a rise of at least three PUs on the accented (i.e. stressed) syllable of the topic. The peak of the rise reaches the infra-high or high level of the speaker’s range. 35 r Lengthening of approximately five PUs on the accented syllable. r Peak of intensity on the vowel of the accented syllable. Piet Mertens’ theory (Mertens 1987, 1993, 1997; Mertens et al. 2001) is situated at the interface between syntax and prosody. It follows the distributional approach to intonation. The core idea is that prosody is determined by the presence of intonation markers. These markers correspond to combinations of tones (or height morphemes) and syllables (i.e. localization points). The whole model is based on a multi-layered representation of prosodic structure, where each layer results from the combination of units at the preceding level. The levels of direct relevance to the present study result from the combination into syllables (which can be stressed or not), the combination of syllables into Intonation Groups (IGs), and the combination of IGs into packages. 33 ‘Une unité de perception est égale à une fois le seuil de perceptibilité (ou seuil différentiel) du paramètre considéré’ (Rossi 1999: 212). 34 The F0 or fundamental frequency corresponds to the main frequency of a periodic sound and determines that sound’s pitch. 35 See n. 22 for a definition of the notion of range. In the three approaches to the prosody of French left-dislocation discussed here, the notion of dominance has to be understood in terms of ‘quantity’, not structural superiority, e.g. an element X dominates Y in pitch if X is higher than Y.

46

French Dislocation

An IG corresponds to a sequence of syllables bearing a unique stress or final accent. In French, the syllable bearing this accent is the only one that can be lengthened. It also contains two tones, one on each mora. 36 The composition of the French IG is given in (2.66), where sequences in square brackets indicate optionality. NA stands for non-accented syllable, IA for syllable bearing the initial accent, and FA for syllable bearing the final accent. (2.66) IG −→ [[NA] IA] [NA] FA [NA] A package is the result of the combination of two or more IGs following the rule in (2.67). (2.67) ‘For any two successive IGs: if the tone in the FA position of the last IG dominates that of the first IG, then there is an embedding effect of the first IG with the second. Otherwise, the two IGs are independent (juxtaposition).’ (Mertens 1993: 3) Dominance is defined according to the tone of the syllable bearing the FA (Final Accent). The tone of a syllable corresponds to the level associated with it. Four levels, defined relative to each other, are distinguished: infralow, low, high, and ultra-high. The passage from one level to the next is marked by a major interval (typically five semitones). Within a given level, minor intervals (typically three semitones) can create a heightening or lowering. The delimitation of IGs and packages is largely dependent on speaker choice. An example of grouping into packages (indicated by square brackets) is given in (2.68). Each sequence inside a set of brackets corresponds to an IG. In this example each IG ends with a FA, indicated by two capital letters. 37

36 A mora is a subword prosodic constituent smaller than a syllable which is used to analyse the metrical structure of speech. 37 In Mertens’ notation, l stands for ‘low’ and h for ‘high’. Capital letters indicate that a syllable is stressed. A sequence l....l involves only low tones. The minus sign indicates that the pitch reached represents the lowest level in the speaker’s range. The / and \ signs indicate heightening and lowering within a given level respectively. See Mertens (1987) for further details.

(i) La lecture n’était pas un niveau auquel on s’intéressait quand on faisait une théorie the reading NEG-was not a level to-which one REFL-interested when one made a theory de la littérature. of the literature ‘Reading wasn’t a level relevant to literary theory.’

Diagnostics for dislocated elements

47

(2.68) La lecture n’était pas un niveau auquel on s’intéressait [l . . . l HH] [l . . . l HH] [l . . . l HH ll/LL] [\l . . . l HH] quand on faisait une théorie de la littérature. [[l . . . l/ LL \l . . . l HH] l . . . \l L−L−] The assumption underlying this model is that there is a certain level of congruence between prosody and syntax. Intonation boundaries define a structure that cannot be mapped one-to-one onto the syntactic structure, but one which nonetheless respects syntactic structure (Mertens et al. 2001). I now turn to the characterization of left-dislocation prosody, in Deshaies et al. (1993); Guilbault (1993); Rossi (1999) and Mertens et al. (2001). 2.3.2.1 The acoustic characteristics of LD On the whole, the defining traits attributed to left-dislocation prosody reflect how native speakers report that they perceive this construction. Left-dislocated elements are perceived as being separate from the rest of the sentence, and as being somehow prominent, both in terms of intensity and melody. It is important to stress that this is how leftdislocation prosody is reported to be perceived by native speakers. As Rossi (1999) argues, the role of fine-grained acoustic analyses should be to define the threshold of perception with respect to which an element is analysed as being left-dislocated rather than as occupying subject position. Each of the posited defining traits is now presented and briefly discussed in order to lay the foundations for the preliminary analysis presented in Section 2.3.3. Pause or no pause? The presence of a pause after the left-dislocated element, although very often mentioned in syntactic or pragmatic analyses as a diagnostic for this construction (e.g. Cadiot 1992; Pierce 1992; Kaiser 1994; Auger 1994), is not in fact confirmed by fine-grained acoustic analyses (e.g. Deshaies et al. 1993). The impression of a pause results from the lengthening of the last syllable of the dislocated constituent (Rossi 1999). Crucially, the possibility of inserting a pause after a potentially left-dislocated element cannot be taken as a diagnostic confirming that this element is not in subject position because in some rare instances in spontaneous production, speakers insert such a pause after an element that is clearly a heavy subject (given the absence of a coindexed subject clitic). Intensity prominence: accent and volume The prominence of leftdislocated elements is perceived in terms of both intensity (Deshaies et al. 1993) and melody (see below). Intensity prominence manifests itself in two ways: the last syllable of the left-dislocated element is stressed, and the peak of intensity on that syllable is perceived as particularly prominent.

French Dislocation 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

85

74

63

52

INTENSITY (dB)

PITCH(semitones)

48

41 comm(e) ça le cra-

-yon

va

res-

-ter a- -vec

le

ca-

-hier 30

0

0.5

1

1.5

TIME(s)

Figure 2.7. Prosody of a left-dislocated element followed by a heavy subject (spontaneous, Catherine)

The presence of stress (or accent) on the last syllable of the left-dislocated element is a necessary condition for this element to be an Intonation Group (IG) in Mertens’ model. Spoken French is a language which displays group stress rather than word stress (as in English), and thus the intonation boundaries of each IG are easy to perceive (Simon 2001). By definition, each IG contains one and only one final accent (see (2.66)). Intensity prominence arises essentially from the presence of a peak on the last syllable of the left-dislocated element: this syllable is supposed to dominate (most of) the rest of the sentence in terms of loudness. When a sentence contains both a left-dislocated element and a heavy subject, the highest peak of intensity is on the left-dislocated element (Rossi 1999), as illustrated in Figure 2.7 for sentence (2.69). The intensity criterion is not however sufficient to distinguish a left-dislocated element from a heavy subject in sentences containing only one of these elements (Rossi 1999: 68), a point which is illustrated in Section 2.3.3. (2.69) Comme ça, le crayon va rester avec le cahier. like that the pencil will stay with the notebook ‘Like that, the pencil will stay with the notebook.’

Diagnostics for dislocated elements

49

Melody: pitch prominence The most striking perceptual characteristic of left-dislocated elements is their melody, i.e. variations in their fundamental frequency (F0 ). The pitch of a left-dislocated element’s last syllable is perceived to be high. In most cases there is a rise from the beginning of the left-dislocated element to its last syllable; if this element is monosyllabic, a rise can sometimes be observed within it. According to both Rossi’s (1999) and in Mertens’ (2001) theories, there must be a rise in F0 culminating on the last syllable of the left-dislocated element. Rossi claims this rise should be of at least three PUs (i.e. approximately three semitones) and that, as a result, the pitch of that syllable dominates that of the rest of the sentence. Mertens claims the rise in question should be of five semitones on average (see below). Both authors consider that the level reached should be high, although this is defined in different ways in the two approaches. For Rossi, high level is defined with respect to the speaker’s range; for Mertens, each level is defined in relation to the preceding one, and as a consequence the relative height of e.g. two HH tones is irrelevant per se. Researchers diverge on the issue of to what the pitch of the dislocated element’s final syllable should be compared. Some argue it should be compared with the pitch of the first syllable of the following constituent (Deshaies et al. 1993; Guilbault 1993). Others claim that it is not methodologically sound to compare isolated points in an utterance and that the relative pitch levels of all the syllables which form an utterance are relevant (Mertens 1987, 1997). Deshaies et al. (1993) and Guilbault (1993) show that there is a drop in F0 between the last syllable of a left-dislocated element and the last syllable of the following element. In the corpus of spontaneous production they analyse, which contains sixty-five left-dislocated utterances in total, Deshaies et al. (1993) observe a drop of 1.96 tones on average. They interpret this drop as a manifestation of the prosodic rupture between the left-dislocated element and the sentence nucleus. Under Mertens’ approach, by contrast, the prosody of the left-dislocated element is analysed in relation to that of the rest of the sentence. The presence of a drop in pitch between the last syllable of the left-dislocated element and the first one of the following element is regarded as an artifact of other factors rather than an intrinsic characteristic of left-dislocation prosody. Mertens argues that French left-dislocated elements are characterized by the presence of a major boundary tone on the last syllable of their IG: the pitch rises by five semitones on average, which triggers a change of level (e.g. from low to high). Left-dislocated elements thus have to comply with three prosodic requirements: they must constitute an IG, they must receive a major boundary

50

French Dislocation

tone, and they must not be combined with the following IG. This is illustrated in (2.70) for a left-dislocated object. (2.70) Les livres, elles les avaient envoyés. (Mertens et al. 2001: 18) ll HH l . . . l HH l . . . l LL− ‘They had sent the books.’ (Literally: The books, they had sent them.) Left-dislocated elements are defined as prosodic islands: they are IGs that cannot be combined with the following IG (Mertens et al. 2001). In (2.70), the IG following the left-dislocated element does not end on a higher level than the latter, hence grouping into a package does not take place. 2.3.2.2 The acoustic characteristics of heavy subjects Of those studies mentioned previously, the only one that explicitly compares the prosody of left-peripheral topics with that of heavy subjects is Rossi (1999: 66–7). 38 Rossi argues that the heavy subject intoneme is defined by the following traits: accent, high, long. This contrasts with the traits of the topic intoneme: accent, high, rising, long, pause. 39 The two essential differences between these two intonemes are thus the optionality of the pitch rise on heavy subjects, and the absence of a subjective pause on heavy subjects (which is due to the fact that the lengthening of the last syllable of a heavy subject is not great enough to give the impression of a pause). The highest pitch reached on a heavy subject is also claimed to be lower on average than that reached on a left-dislocated topic: the peak in F0 is situated between the medium and infra-high levels of the speaker’s range in the former. By contrast, the two intonemes are not expected to differ in terms of intensity according to Rossi (1999: 69). In other words, the topic intoneme typically dominates the subject intoneme in all respects apart from intensity. 2.3.2.3 Summary The proposed prosodic characteristics of left-dislocated elements are summarized in (2.71). (2.71) a. Presence of a stress on the last syllable of the left-dislocated element. b. Rise in F0 (by three semitones at least) culminating on or occurring within the last syllable of the left-dislocated element. The pitch of this final syllable dominates that of the following IG. It is perceived as being dominant in the sentence. 38 A full definition of the notion of topic is given in Section 3.1. At this point, a topic can be characterized as what the sentence is about. 39 Recall that pause does not correspond to a rupture of the sound wave, but to the impression of a pause resulting from a significant lengthening of the left-dislocated element’s last syllable.

Diagnostics for dislocated elements

51

c. If there is a heavy subject, the left-dislocated element dominates the subject in all its prosodic characteristics, especially pitch. 2.3.3 Diagnostics for LD? A preliminary acoustic analysis For this preliminary acoustic analysis of left dislocation, I have used two types of data: elicited data produced in almost ideal recording conditions, and spontaneous data produced in relatively poor recording conditions. I briefly describe the elicited data in this section. For a description of the spontaneous data, see Appendix A. The elicited data come from studio recordings of two native speakers of French, both female (one from Belgium, Isabelle, and one from France, Pascale). Data were elicited in two different ways. In the first session, in which both informants took part though at different times, the informant was given a series of sentences which she was told were part of a dialogue. She was asked to act the part of a character in an attempt to maximize her involvement in the elicitation and enhance the chances of obtaining a natural reading, i.e. as close as possible to spontaneous speech despite the highly controlled conditions. Before each sentence, I gave a context (a situation with which the character was confronted in her ‘daily life’) to put the informant in the right frame of mind necessary for her reply to have a naturalistic intonation. In the second session (in which only Pascale participated), the informant was given a series of mini dialogues in which she had to play the role of both speakers. These dialogues contained near minimal pairs of sentences, each including either a left-dislocated subject or a heavy subject. A DAT recorder was used at both sessions. Subsequent transformation into wave files (using a digital sound card) was at a sample rate of 22MHz, with a mono resolution of 16 bit. The speech analysis programme Praat was used to create all graphs and perform all calculations. 2.3.3.1 Clear cases of LD prosody The data studied contain many clear examples of left-dislocation prosody, as defined in the summary above. Consider first sentence (2.72), whose prosody is represented in Figure 2.8. (2.72) La portei , elle l’i a fermée. h LL] [ll HH] [l . . . the door she it-has closed ‘The door, she’s closed it.’

(Pascale, F)

The left-dislocated object meets the requirements proposed by Mertens et al. (2001): it constitutes an IG with a major boundary tone on the FA syllable and this IG cannot be combined with the following one into a package (because

French Dislocation 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

95

82

69

56

INTENSITY (dB)

PITCH(semitones)

52

43 la

port(e)

ell(e)

l’a

fer-

-mée 30

0.5

1

TIME(s)

Figure 2.8. Prosody of a left-dislocated object (elicited, Pascale)

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

95

82

69

56

43 la

port(e)

e(l)- -l(e) est

fer-

-mée 30

0.5

1 TIME(s)

Figure 2.9. Prosody of a left-dislocated subject (elicited, Pascale)

INTENSITY (dB)

PITCH(semitones)

the tone of the second IG’s FA does not dominate that of the dislocated element’s FA). Furthermore, as predicted by Rossi (1999), the last syllable of the dislocated element dominates the rest of the sentence in pitch and intensity.

Diagnostics for dislocated elements

53

Sentence (2.73) is a near-minimal pair with (2.72). In this sentence, the dislocated element is coreferential with the subject rather than the object. The prosody is represented in Figure 2.9. (2.73) La portei , ellei est fermée. [ll HH] [l . . . h LL] the door it is shut ‘The door, it is shut.’

(Pascale, F)

The characteristics of the two IGs in this sentence are exactly identical to those in sentence (2.72), as indicated by the notation accompanying (2.73). Furthermore, as in (2.72), the dislocated element dominates the rest of the sentence in pitch and intensity. This is another clear case of ‘classical’ leftdislocation prosody, as it is known in the literature. Consider now the spontaneously uttered sentence (2.74), whose prosody is shown in Figure 2.10. (2.74) Et des papillons, ça se trouve où, ça, Max?40 [l . . . l HH] [l . . . l HH h] [hh HH] and the butterflies it finds where that Max ‘And butterflies, where are they to be found, Max?’

(Catherine, C)

Again in this example, the left-dislocated element meets the requirements that it be an IG with a major boundary tone and that it not be combined with the following IG. The last syllable of the dislocated element further dominates the rest of the sentence in pitch (which is not seen as relevant under Mertens’ approach). Note however that the prediction that the last syllable of the dislocated element papillons ‘butterflies’ will be higher in intensity than the rest of the sentence is not verified. (Again, this is not strictly relevant under Mertens’ approach.) 2.3.3.2 Comparison with heavy subjects In contrast with these clear instances of left-dislocation prosody, consider now two cases of typical heavy subject prosody. Sentence (2.75) is a near-minimal pair with (2.73). (2.75) La porte est fermée. [ll /LL] [l . . . l L−L−] the door is shut ‘The door is shut.’

(Pascale, F)

40 Max is the pseudonym of the child. His real name is bisyllabic and is replaced by xx-xx in Figure 2.10.

French Dislocation 90

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

80

70

60

50

INTENSITY (dB)

PITCH(semitones)

54

40 et des pa-

pillons

ça

s(e) trouv(e)

où

ça

xx-

-xx? 30

0

0.5

1 TIME (s)

1.5

2

Figure 2.10. Prosody of a left-dislocated subject (spontaneous, Catherine)

The heavy subject forms an IG in this sentence and this IG does not get combined with the next one into a package, given that the latter does not dominate the former. Mertens’ system thus cannot be used as a diagnostic to distinguish left-dislocated subjects from heavy subjects because heavy subjects are not obligatorily combined into a package with the following IG either. Let us turn now to the traits defined by Rossi. There is no notable peak of intensity on the heavy subject. Aside from the unavoidable troughs at the points where consonants are pronounced and the decrease on the sentence-final syllable, the intensity curve is fairly flat. There is however a rise of approximately four semitones on the heavy subject. So far the overall picture is rather inconclusive, and thus it is necessary to add an extra variable. An obvious candidate is duration, following Rossi (1999). Left-dislocated elements show a lengthening on their last syllable; heavy subjects are predicted not to do the same. In this case, given that we have near minimal pairs elicited in conditions which are otherwise as identical as possible (i.e. same recording conditions, same speaker), a comparison is possible. Even without going into details, it is very easy to see in relation to the x-axis in Figures 2.9 and 2.11 that the DP la porte ‘the door’ is uttered faster when it is a heavy subject. The exact duration of the last syllable of that DP (i.e. the most relevant one) is 0.2856 sec when the DP is left-dislocated, and 0.1865 sec when the DP is a heavy subject. The difference is thus one of

55 90

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

80

70

60

50

INTENSITY (dB)

PITCH(semitones)

Diagnostics for dislocated elements

40 la

por-

-t(e) est

fer-

-mée 30

0

0.5 TIME(s)

Figure 2.11. Prosody of a heavy subject (elicited, Pascale)

0.099 sec. I must leave for further research the task of determining whether this is a significant difference in terms of perception or not. Another difference between the prosody of (2.73) and (2.75) is that there is a drop in F0 after la porte ‘the door’ in Figure 2.9 but not in Figure 2.11. Deshaies et al. (1993) and Guilbault (1993) have shown that left-dislocated elements are followed by a drop in F0 of nearly four semitones on average. Such a drop can be observed in Figure 2.9. In Figure 2.11, the pitch gets lower after the heavy subject, but it does not reach a level lower than that of the first syllable of the heavy subject. Only the left-dislocated subject manifests a ‘medium-highlow’ contour, as illustrated schematically in Figure 2.12 and shown in detail in Table 2.3.

pitch

high

medium low time

Figure 2.12. The ‘medium-high-low’ contour of left-dislocation prosody

56

French Dislocation

Table 2.3. Contrast between the melodic contour of a left-dislocated and a heavy subject Sentence-initial element Left-dislocated

Heavy subject

Melody (in semitones), syllable by syllable la 12.43

por17.46

t(e) e(l)12.44

-l(e) est 9.25

fer15.3

mée 11.97

medium

high

low

la

por-

t(e) est

fer-

mée

14.01

18.93

13.9

16.67

11.67

medium

high

medium

To conclude, the heavy subject in Figure 2.11 can be distinguished prosodically from its left-dislocated counterpart. The characteristics which differentiate it have been found to be the absence of a ‘medium-high-low’ contour and possibly duration. Whether such indicators are reliable will be discussed later in the text. Consider now in Figure 2.13 a heavy subject uttered spontaneously by the same speaker as (2.74) (which contained a left-dislocated subject). (2.76) Le canard fait quoi? [l . . . /l HH hh H+H+] the duck does what What does the duck do?

(Catherine, C)

In this case, the IG formed by the heavy subject is combined with the following IG into a package. As a result, the heavy subject cannot be considered to form a prosodic island and cannot be analysed as an example of left-dislocation. This is further confirmed by the fact that the pitch of the last syllable of the heavy subject does not dominate any part of the rest of the sentence (though it does dominate it in intensity). To summarize the conclusions so far, if the IG of a left-dislocated or heavy subject is combined into a package with the following IG, the element in question cannot be analysed as left-dislocated; it must be a heavy subject. If such a combination does not obtain, it is necessary to rely on other diagnostics. The examples considered in this section suggest that the element in question is leftdislocated if (i) there is a rise in pitch of more than three semitones culminating on the last syllable of that element, (ii) there is a ‘significant’ lengthening

57 90

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

80

70

60

50

INTENSITY (dB)

PITCH(semitones)

Diagnostics for dislocated elements

40 le

ca-

-nard

fait

quoi? 30

0

0.5

1

TIME(s)

Figure 2.13. Prosody of a heavy subject (spontaneous, Catherine)

of that last syllable, or (iii) a ‘medium-high-low’ contour is observed on that syllable and the following two. Before reaching a final conclusion, more factors must be considered.

2.3.3.3 Interfering factors Many factors can blur the distinction between the prosody of left-dislocated elements and that of heavy subjects. Two such factors are monosyllabicity and the general melody of the sentence. Some left-dislocated subjects are monosyllabic. If one of the key traits is a rise from the penultimate to the last syllable of a left-dislocated element, it is to be expected that monosyllabic left-dislocated subjects will tend not to display such a rise. Rossi (1999: 79) notes that when the left-dislocated subject is a strong pronoun, its prosody is less clearly distinct from that of true heavy subjects, especially when the strong pronoun is coreferential with the subject and appears in a matrix clause of the type je pense ‘I think’. Figure 2.14 illustrates this point. The monosyllabic nature of the left-dislocated element is probably the reason why the pitch rise is only of 1.1 semitone, i.e. only just enough to be perceptible (Rossi 1999). In spite of this, the pitch of the leftdislocated element clearly dominates that of the rest of the sentence, which is a typical characteristic of left-dislocated topics.

French Dislocation 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

90

80

70

60

50

INTENSITY (dB)

PITCH(semitones)

58

40 moi

je

veux

euh

pa-

-to 30

0

0.5

1 TIME(s)

Figure 2.14. Prosody of a monosyllabic left-dislocated subject (spontaneous, A.-Gaël)

(2.77) Moi, je veux euh Pato. me I want er Pato ‘I want Pato.’

(A.-Gaël, F)

It has also been observed by Deshaies et al. (1993) and Guilbault (1993) that the drop in pitch after a left-dislocated element is smallest between left-dislocated pronouns and a coindexed resumptive clitic, measuring on average 1.24 tone (Deshaies et al. 1993). This trait is not included in the defining ones considered by Rossi (1999) and Mertens et al. (2001), but I believe that the size of this drop in pitch may reflect the extent to which the pitch on the last syllable of the leftdislocated element dominates the pitch of the rest of the sentence, and/or be a reflection of the general pitch contour of the sentence. The general melody of the sentence may also be an issue. When the prosody of the whole sentence is fairly flat, the pitch trace of the dislocated element or heavy subject is fairly flat too. In the elicited sentence in Figure 2.15, the variations in pitch are very small, and consequently the rise on the left-dislocated element is very small too (0.76 semitone), if not inexistent. This lack of salience in terms of pitch might be compensated for by a clear intensity peak on the last syllable of the dislocated element and possibly by the lengthening of that syllable. The significance of these two factors with respect to left-dislocation prosody in an example like that in Figure 2.15 remains a matter for further research.

59 90

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

80

70

60

50

INTENSITY (dB)

PITCH(semitones)

Diagnostics for dislocated elements

40 le

mi-

-roi(r)

-r i(l)

-(i)l est

où? 30

0

0.5 TIME(s)

Figure 2.15. Prosody of a left-dislocated subject in a ‘monotonous’ sentence (elicited, Isabelle)

(2.78) Le miroir, il est où? [l . . . l /LL] [l . . . l /LL] the mirror it is where Where is the mirror?

(Isabelle, B)

Similarly, in sentences expressing contrast, emphasis, surprise, etc., variations in pitch seem to be greater throughout the sentence, including on the left-dislocated element. The pitch contour of the heavy subject in a sentence of this type is thus likely to resemble that of a left-dislocated element, especially when the subject itself is contrastive. Such a contour is exemplified by the spontaneously uttered sentence (2.77) illustrated in Figure 2.16. (2.79) Mais Hercule peut pas rentrer dedans. [l . . . l HH] [[ll/ LL] [ll/LL] [ll HH]] but Hercules can not enter inside ‘But Hercules cannot go in there.’

(Catherine, C)

The heavy subject is contrastive in this context and is uttered with a very marked, emphatic intonation. The whole sentence is affected by this, and it ends on a rising tone as if it were a question (which the context indicates it

French Dislocation 90

25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

80 70 60 50

INTENSITY (dB)

PITCH(semitones)

60

40 mais

Her-

-cul(e)

peut

pas

ren-

-trer

de-

dans 30

0

0.5

1

1.5

TIME(s)

Figure 2.16. Prosody of a contrastive heavy subject (spontaneous, Catherine)

is not). The prosody of the heavy subject cannot be distinguished from that of a left-dislocated element in this sentence. 41 Note that cases like this are very rare in the data studied here. In the overwhelming majority of cases, the prosody of left-dislocated subjects was clearly distinct from that of heavy subjects. 2.3.4 Summary One of the main conclusions of this preliminary analysis is that the prosody of left-dislocated elements cannot be narrowed down to a single defining characteristic (e.g. the presence of a pause or a pitch rise), and that consequently the absence of a given characteristic cannot in isolation be taken to indicate that the element in question is not left dislocated. In particular, the absence of a pause after a sentence-initial DP expressing the subject does not mean that such an element occupies the subject position, contrary to what has been claimed in a number of studies (e.g. Roberge 1990; Auger 1994; Kaiser 1994). Clearly, more work needs to be carried out to establish a set of diagnostics which reliably differentiate the prosody of left-dislocated elements from that of heavy subjects. The various phonetic traits this preliminary analysis has concentrated on need to be incorporated into a more general picture, one that would reveal and explain the relevance of each such characteristic. For 41 It is telling that in the initial transcription this sentence was erroneously taken to involve leftdislocation of the subject. It was only on closer examination that I noticed the absence of a subject clitic.

Diagnostics for dislocated elements

61

instance, variations in the duration of the last syllable of a left-dislocated element should be studied in relation to the general tempo of the utterance. 42 It is likely that the lengthening observed is due to a general slowing down on left-dislocated topics but not on heavy subjects. Similarly, pitch and intensity should be studied in relation to the expressivity contour of the utterance as variations in these two parameters are more acute in sentences expressed with emphasis. (Hence a heavy subject in an emphatic sentence is likely to be similar in pitch and intensity to a left-dislocated subject.) The tentative incorporation into the present analysis of the melodic contour of the constituent following a left-dislocated or heavy subject is a step in this direction. This preliminary investigation has shown that prosody alone cannot distinguish left-dislocated subjects from heavy subjects when they occur in apparently identical structural positions. However, prosody can be one of the determining factors in accepting or rejecting a dislocation analysis of a particular example, especially if discourse factors are also taken into account. For instance, left-dislocation prosody might not indicate syntactic left-dislocation in contexts favouring a contrastive interpretation of a potential heavy subject. In the present study, I retain the diagnostics in (2.78) to distinguish leftdislocated subjects from heavy subjects in the absence of structural cues. (2.80) a. If the IG of the element in question is combined with the following IG into a packet, that element must be a heavy subject. b. If such combination does not take place, the element in question is taken to be a left-dislocated subject if: (i) there is a pitch rise of more than three semitones (culminating) on the last syllable of that element; (ii) the last syllable of that element is strongly prominent, either in intensity or duration; or (iii) a ‘medium-high-low’ contour is observed on the last syllable of that element and the following two syllables, except (i) when the context favours a contrastive or emphatic reading of the element in question; or (ii) when the whole sentence is uttered with great variations in pitch and intensity (i.e. usually in emphatic sentences), in which case no definite choice can be made between the two analyses. 42

I owe this suggestion to Richard Ogden.

62

French Dislocation

Before closing, I would like to outline some directions for future research in this area. First, a comparison of spontaneous and elicited speech may reveal that the latter is more monotonous due to the artificial character of the situation, a factor which could have contributed to blurring the distinctions between left-dislocated subjects and heavy subjects in this study. Second, perception experiments involving native speakers could be used to determine the importance of the various characteristics of left-dislocation prosody. For instance, it would be interesting to test whether left-dislocated subjects are still perceived as such in sentences where the resumptive element has been removed. Another possibility is to remove the actual words from a set of sentences involving either a left-dislocated subject or a heavy subject (by replacing all the articulated sounds by [a], for instance) in order to test whether native speakers perceive differences between the two types of prosody.

2.4 Conclusion Prosody and the presence of a resumptive element are reliable diagnostics for dislocated structures in spoken French, but with certain limitations. Firstly, in sentences uttered with monotone intonation, variations in pitch and intensity may not be sufficiently marked to identify dislocation prosody. Secondly, when the context favours a contrastive or emphatic reading of the subject, it may not be possible to distinguish a dislocated from a non-dislocated subject prosodically. The presence of a resumptive element is a sufficient diagnostic for dislocated elements in spoken French, but it is only a necessary diagnostic with respect to subjects (and, as argued in De Cat 2004a, only in adult French). Indeed, where null objects are allowed, a dislocated object may appear without a resumptive element (see Chapter 4). In a nutshell, and as a direct corollary of the properties discussed previously, dislocated elements are syntactically and prosodically non-essential. A sentence stripped of its dislocated element(s) should be well-formed syntactically (Chapter 4) and prosodically (as argued by e.g. Rossi 1999).

3 Interpretation It is uncontroversial that the essential function of conversation is to exchange information and that speaker and hearer cooperate to that effect (see e.g. Roberts 1998). Given this, the way we say things depends on what we think our addressee already knows or believes. Put more formally, the information structure of an utterance is ‘the tailoring of [that] utterance by a sender [ . . . ] that reflects the sender’s hypotheses about the receiver’s assumptions and beliefs and strategies’ (Prince 1981: 224). The module of linguistic competence which determines that sort of encoding and decoding of information is referred to in various ways in the literature, e.g. information structure (Lambrecht 1994), information packaging (Vallduví 1992), f-structure (Erteschik-Shir 1997). The place this component is claimed to occupy in the grammar varies depending on the model. For Vallduví (1992), information packaging does not incorporate strictly syntactic or semantic requirements and therefore falls within the realm of pragmatics. For Erteschik-Shir (1997), f-structure is an integral part of grammar and may ultimately prove the level of LF redundant. However, as Kadmon (2001: 4) points out, ‘there is no a priori reason to assume that semantics is in the grammar while pragmatics is not’. The view defended here is that information structure is part of grammar in the sense that it is inherently linguistic, has a direct impact on syntax and prosody, and contributes to determining the truth value of propositions. I will remain agnostic at this point as to whether information structure (IS) phenomena can be accounted for entirely in terms of semantics (in which case IS should be recoverable at LF) or whether it is necessary to postulate the existence of an independent module within UG.

3.1 Topics Many languages feature dislocated elements, and it is standardly assumed that dislocated elements are interpreted as topics. This has been claimed to be the case in many languages, across language families, e.g. Italian (Cinque 1990;

64

French Dislocation

Rizzi 1997), Greek (Iatridou 1995; Anagnostopoulou 1997), Lebanese Arabic (Aoun and Benmamoun 1998). French is no exception to this, as argued by e.g. Larsson (1979); Lambrecht (1981); Campion (1984); Barnes (1985); Ashby (1988); Lambrecht (1994); Delais-Roussarie et al. (2004). 3.1.1 General definition The notion of topic was fully formalized for the first time by Reinhart (1981). Until then, Reinhart points out, several attempts had been made to establish criteria for the identification of topics, but none were entirely satisfactory: either the definition was too broad, or it was too vague. In the first case, the definition did not provide criteria precise enough to distinguish the topic from the non-topic part of the utterance. Either the topic was defined in terms not easily amenable to formalization (like ‘speaker’s intentions’, e.g. Schachter 1973) or, when stricter defining criteria were proposed, they erroneously overextended a possible characteristic of topics to the whole class, thus missing potential candidates and wrongly selecting others. The topic was variously conceived as the subject of the sentence (e.g. Gundel 1974), the first expression in the sentence (e.g. Halliday 1967), or the part of the sentence conveying old information (e.g. Chafe 1976). In the second case, topic was viewed as a gradient notion. Firbas (1975), for instance, defines topics as having the lowest communicative dynamism. This led to the investigation of interesting parameters (like the human/non-human and definite/indefinite scales defined by Givón 1976) but did not provide a working definition of the concept. The cornerstone of Reinhart’s proposal is the notion of aboutness. 1 Topichood is defined as the relation between an argument and a proposition relative to a context. The topic of a sentence is what the sentence is about. Reinhart’s definition of topic is directly inspired by the work of Strawson (1964), who argues that the topic of a sentence is whatever in the sentence is already part of the presumed knowledge of the hearer if the assertion expands his/her knowledge of that topic. What is known and what the sentence is about often overlap, but this is not always the case. To formalize the notion of topichood, Reinhart (1981) summons the file card metaphor, which models the speaker and hearer’s knowledge store and how its state is changed by the addition of information gleaned from the discourse. Reinhart and her followers (e.g. Erteschik-Shir 1997; Portner and Yabushita 1998) define sentence topics as entries under which the hearer classifies the propositions that carry information gathered from the discourse. The 1 The notion of aboutness originates in the work of Vilém Mathesius, the founder of the Prague School. Mathesius examined how the information status of utterance chunks is encoded by word order, intonation, or other means. See Firbas (1964) for details.

Interpretation

65

set of propositions accepted as true at a point in a discourse form the context set of the discourse at that point, i.e. the set of possible worlds compatible with the presuppositions which speaker and hearer both hold (Stalnaker 1974). In order to build the context set, the hearer assesses what he or she already knows about the topic in question, and subsequently stores under the relevant entry (or topic) whatever has been learned about it. In some cases, this involves creating a new entry in the context set. Discourse referents are thus conceived as mental constructs. Within this approach, the predication is evaluated with respect to its topic (e.g. Li and Thompson 1976; Reinhart 1981; Erteschik-Shir 1997). This idea can also be traced back to the work of Strawson (1964), who argues that the truth value of a sentence is assessed with respect to the topic. Strawson (1964) claims that topics carry existential presuppositions. A sentence containing a topic that fails to refer will yield a truth value gap: it will be neither true nor false. Hence sentence (3.1) fails to have a truth value, at least under a reading where it refers to the ‘real’ world as we know it, in which Father Christmas does not exist. (3.1)

Le père Noël, c’est une ordure. the father Christmas it-is a rubbish ‘Father Christmas is a bad guy.’

By contrast, Strawson (1964) argues sentence (3.2) is false (assuming that exhibition refers successfully), as the King of France cannot be among the set of visitors to the exhibition. Whether the King of France really exists is immaterial in this case. (3.2)

The exhibition was visited yesterday by the King of France.

3.1.2 The information structure partitioning of the sentence The notions of topic and focus are information structure primitives. 2 Their identification is a prerequisite for determining the IS of a given sentence. The focus of a sentence corresponds to the most informative part of that sentence (Lambrecht 1994: 55) or the part to which the speaker wishes to direct the attention of the hearer (Erteschik-Shir 1986). 3 All sentences have a topic (as I will argue below) and a focus (Vallduví 1992: 43, Lambrecht 1994: 206, 2 For some, the first part of this claim (i.e. that the notion of topic is a primitive) is controversial (see e.g. McNally 1998; Roberts 1998). It has been argued that topics are in fact a special case of focus (von Fintel 1994). 3 This is not incompatible with approaches that treat focus as a set of alternatives (see e.g. Rooth 1992).

66

French Dislocation

Reinhart 1996, Erteschik-Shir 1997, McNally 1998, Zubizarreta 1998). 4 In other words, any proposition must be restricted to a given domain (defined by the topic) and contributes information. The information structure of sentences is however not bipartite: topic and focus do not always span over the whole sentence. Topic and focus, though they interact, belong to separate dichotomies: the topic/comment dichotomy (or theme/rheme in Halliday’s (1967) terminology) and the focus/presupposition dichotomy (e.g. Chomsky 1971; Zubizarreta 1998) (the latter also corresponds to the focus/(back)ground dichotomy, as in Chafe (1976) and Vallduví (1992)). The comment is understood to be what is predicated about the topic in a given sentence; the presupposition covers whatever is not ‘informative’ in the sentence. Vallduví (1992) argues that the identification of information packaging units start with identifying what is in focus, as the rest of the sentence (i.e. the ground) is defined in relation to the focus. The focus is the only informative part of the sentence; the ground acts as a vehicular frame for the informative focus. The focus is identified by the context; it is the only unelidable part of the sentence, and it invariably bears the main stress of the sentence. Whatever is not in the focus is in the ground. The part of the ground that precedes the focus is the link (which corresponds to the a sentence-initial topic), 5 and the part that follows the focus is the tail (which Vallduví claims corresponds to the antitopic in Lambrecht’s terminology, but see Section 4.4.4 for discussion). Links are pointers to address cards in the knowledge store (which correspond to file cards in the context set in Reinhart’s terminology, adopted here). Tails are said to convey instructions as to how the information has to be entered under a specific address. One aspect of Vallduví’s analysis is problematic. It predicts that in a sentence like (3.3), the whole VP el va ficar al calaix is in focus because all non-link, non-tail (i.e. not right-detached) material is predicted to be focused. Vallduví (1992) does not allow for the possibility of only focusing the PP in such a sentence, as I endeavour to show below. 6

4 Reinhart (1996) claims that each sentence is associated with a set of possible foci rather than an actual focus. The context determines which focus is to be selected. 5 The idea of a fixed order of topic and focus originates in the work of the Prague School scholars, in particular Mathesius, Daneš, and Firbas, who postulated that topics were always sentence-initial. This idea was further developed in Halliday (1967), which analyses the theme as always being the first expression in a sentence. For Vallduví too, the link always precedes the focus. The link is said to take scope over the information provided by the focus: the operator involved is a quantifier-like element. 6 Note that for Vallduví (1992), unaccented pronominals are never links. Hence the resumptive of a dislocated element can be in the focus, as el ‘it’ is in (3.3).

Interpretation (3.3)

67

ficar ti al El Iu [eli va calaix]focus , el ganiveti . the Iu obj 3sg.-past put in-the drawer the knife Approx.:‘Iu put it in the drawer, the knife.’

In a sentence where it is obvious from the context that only the object is in focus (like B’s sentence in (3.4)), both the subject and the verb should in principle be analysed as links under Vallduví’s approach because they precede the focus. The problem is that only elements that have a corresponding address in the knowledge store can be links. Analysing the verb as (part of) a link is not possible in Vallduví’s system, hence sentences like B’s in (3.4) are predicted not to exist. 7 (3.4) A: Qu’est-ce qu’elle veut? what is-it that-she wants ‘What does she want?’ B: Elle veut [des brocolis]focus . she wants indef broccoli ‘She wants broccoli.’ However, it is desirable for a theory of focus to be able to derive the fact that the preferred answer to a question such as that in (3.4) is des brocolis ‘broccoli’, i.e. the object DP alone. Therefore, I will adopt the view that not all the elements of a given sentence are necessarily part of either the ground or the focus, contra Vallduví (1992). This implies that some elements can be neither topic nor focus in a sentence. I will also assume, following Lambrecht (1994) and ErteschikShir (1997) among others, that the focus of a given sentence is not necessarily coextensive with the predicate. 3.1.3 Topics do not have to correspond to old information There is little consensus in the literature as to the conditions under which a referent can be endowed with the topic function (and hence be dislocated in those languages that permit/require it). In particular, it is frequently assumed that topics must correspond to old information (e.g. Chafe 1976; Iatridou 1995; Rizzi 1997). While it is often the case that topics convey old information, this cannot be a necessary condition for topichood. Indeed, as pointed out by Reinhart (1981), imposing such a requirement on topics would rule out indefinites (which she argues are by definition new information) when in fact certain types of indefinites can be topics, as we will see in Section 3.2.2. 7 Note that verbs are not excluded from the ground by definition as Vallduví (1992) claims they can be right-dislocated and hence part of the tail. See Section 4.4.4 for discussion.

68

French Dislocation

In spite of this, it is desirable to derive the fact that old information is a common property of topics from the definition of topics itself. This is achieved by the salience condition. Lambrecht (1994) convincingly argues that topics can encode new (or old) referents to the extent that they are active. The more active a referent is, the more it will be perceived as old information. Only referents that are active (or salient) enough may be topic. The search space for topic referents is not limited to what has been said in the discourse: what is relevant for identification of the topic referent is the context as a whole. This includes not only what has been said, but also what is physically present in the surroundings as well as what the speaker and hearer share (memories, relations, etc.). Barnes (1985) provides a salience scale along these lines. 8 To give an example, in the situation in (3.5), the reference to a particular film (which both speaker and hearer have seen) renders the referent ‘the woman’ salient enough for it to be encoded as a topic and even as a right-peripheral topic, contrary to the claim that such elements have to be more salient than their left-peripheral counterparts; see e.g. Lambrecht (1981). (3.5)

A: Je pense souvent au film L’enfance d’un chef. I think often to-the film the childhood of-a chief ‘I often think of the film A Chief ’s Childhood.’ B: Oui. C’était un beau personnage, hein, la femme. yes it-was a beautiful character eh the woman ‘The woman was a beautiful character, wasn’t she?’

The salience requirement is met by default if the referent is part of the permanently available knowledge store, also called the permanent registry in Kuno’s (1972) terminology. A sentence like (3.6) can be uttered completely out-of-theblue, and yet be acceptable. 9 (3.6) Le mauvais temps, c’est déprimant. the bad weather it-is depressing ‘Bad weather is depressing.’

8 Authors diverge a lot on the exact conditions topics have to meet to be felicitous. However, the bottom line is always that the hearer has to be able to identify the topic referent. 9 Non-salient referents used as topics in out-of-the-blue contexts need not appear as leftdislocations rather than right-dislocations. One can easily imagine a context where someone who has been day-dreaming about rain and grey skies suddenly utters the sentence below.

(i) C’est déprimant, le mauvais temps. it-is depressing the bad weather ‘Bad weather is depressing.’

Interpretation

69

Lambrecht (1994) argues that the degree of identifiability of a referent can be measured by the amount of discourse encoding necessary for the hearer to retrieve it correctly from his/her knowledge store. Definiteness is typically associated with identifiability/salience, but there is no obligatory correlation between the two. I will come back to this issue in Section 3.2.2. Prince (1981) and Ward and Prince (1991) derive the salience condition from the requirement that the referent of the preposed topic must be in a poset relation (i.e. a partially ordered set relation) 10 to a referent in the previous discourse. This is also the view adopted by Vallduví (1992). Some of the theories based on File Change Semantics (Heim 1982, 1983) account for the salience requirement by ordering (at least part of) the common ground file: cards corresponding to recently mentioned referents are placed on top of the file and are therefore more readily available (see e.g. ErteschikShir 1997). 11 An additional ‘re-ranking’ mechanism should be incorporated to account for the intrinsic salience of referents available in the physical/situational context and in the ‘permanent registry’. Referents belonging to the permanent registry are typically generic: they denote prototypical individuals which form part of our knowledge of the world (see e.g. Kuno 1972). Such referents are sufficiently salient by definition, and can be encoded directly as topics in spoken French (via dislocation, as will be argued in Section 3.2), even when uttered ‘out of the blue’, as shown in the following exchange. (3.7)

A: Les vieux films, c’est quand même vachement bien. the old films it-is nonetheless really good ‘Old films are great.’ B: Pourquoi tu dis ça? why you say that ‘Why do you say that?’

In this exchange, the topic is felicitous: the hearer does not question its reference nor its relation with the predication. What is judged ‘illicit’ by the 10 A partially ordered set is a set containing elements that are ordered according to a relation that is either reflexive, transitive, and antisymmetric, or irreflexive, transitive and asymmetric. Without going into detail, this partial ordering is what defines relations like e.g. is taller than, is part of, is a subtype of. A referent will thus be a possible topic if it enters into such a relation with a referent mentioned previously in the discourse. In Section 4.4.4, I will argue that the same poset relation holds in some cases between the referent of a dislocated topic and that of its resumptive clitic. 11 An alternative has been proposed by Portner and Yabushita (1998), who do away with the notion of file and define the discourse context in terms of Domain (the set of discourse referents currently in use) and Common Ground (the propositional information supplied by the context). What is said about re-ranking in the discussion above can be seen to apply equally to what should be included in the Domain.

70

French Dislocation

speaker, as shown by their request for further information, is the appropriateness of this move in the conversation (see Roberts 1998). On first mention of a referent that is present in the situational context, the trigger for re-ranking tends to be physical deixis (such as gaze or pointing), which could be seen as a licensing mechanism. Imagine, for instance, a situation in which speaker and hearer, who have not been discussing light at all, suddenly become aware of a very strange light. In such a situation, merely looking at the light would be sufficient to license an utterance in which the light is directly encoded as a topic (hence dislocated in spoken French), in spite of the fact that it does not constitute old information. (3.8) C’i est quoi, cette lumièrei ? it is what this light ‘What’s this light?’ ‘Old information’ should therefore not be seen as a defining property of topics (i.e. a necessary requirement) but rather as a property derived from the salience requirement and which is neither necessary nor sufficient for a referent to be a topic.

3.1.4 The relevance condition What is often ignored in the literature is the fact that topics have to be relevant to the predication they are associated with (Strawson 1964): the assertion has to expand the hearer’s knowledge of the topic (which triggers the updating of the topic’s file card in Reinhart’s and Erteschik-Shir’s systems). I think this explains why sentences like (3.9) (originally mentioned in Reinhart 1981 as a puzzle) seem odd: losing a book about Marilyn Monroe does not expand the knowledge one might have of her. (3.9)

# Marilyn Monroei , I’ve lost a book about heri .

The relevance requirement may appear to be reminiscent of the definition of topic as Question Under Discussion (von Fintel 1994; Roberts 1998; Büring 1999). Under that approach, which also builds on the notion of common ground, a topic is identified with a question which the speaker chooses to answer as part of a strategy to ultimately satisfy the goal of the conversation (Roberts 1998). This question (which does not need to be overtly formulated) is one of a salient set of presupposed alternatives. In a sentence like (3.10), the topic would be Zebedee if the sentence were understood to be the answer to the implicit question What about Zebedee?

Interpretation

71

(3.10) Zebedee sold his old bike. The Question-Under-Discussion approach does not, however, capture the relevance requirement on topics: while the choice of Marilyn Monroe as a topic may be adequate in a particular discourse context, it does not automatically imply that what is then predicated of Marilyn Monroe will be relevant to that particular discourse referent. 12 What needs to be captured here is the (quasi-?) impossibility of imagining a context that would render (3.11Bi) appropriate. This can only be achieved with the notion of sentence topic, and only if a condition of relevance to the predication is built into its definition. (3.11) A: She’s collecting photographs of Marilyn Monroe and Benoît Poelvoorde. B: Really? B: (i) #Marilyn Monroe, I’ve lost a book about her. (ii) Marilyn Monroe, I’ve got a book about her. 3.1.5 Stage topics and aboutness topics Under the Reinhart-type approach, topics are often assumed to correspond by definition to an entity in the discourse context. This is however not always the case. Consider the sentence in (3.12). (3.12) Quand j’étais petite, c’était dur. when I-was little it-was hard ‘Life was hard when I was little.’/‘That was hard when I was little.’ This sentence can be interpreted in two ways: (i) one in which the dislocated clause is resumed by the subject of the predication, in which case the sentence is equivalent to something like My childhood was difficult, or (ii) one in which the dislocated clause is the (spatio)-temporal frame within which the predication is to be evaluated (in which case the reference of the subject is not defined sentence-internally). Under the former reading, the information would have to be entered on a card headed with something like ‘X’s (i.e. the speaker’s) childhood’. Under the latter reading, it would be entered under a heading corresponding to whatever the speaker is talking about at 12 The QUD approach is however unique in its ability to predict which focus is felicitous in a given context (Roberts 1998). Such an approach can predict which question the speaker may choose to address next (i.e. the discourse topic) and which information contribution would be appropriate in the context (i.e. the predication associated with the topic), but as far as I understand it does not explain why certain combinations of topic and focus are infelicitous.

72

French Dislocation

the time (for instance, the conditions of work in a mustard factory). In both cases, the predication is evaluated with respect to the referent of dislocated clause. The idea that sentence modifiers such as locative or temporal clauses (as in (3.12)) can be topics, although already present in the work of Gundel (1974, 1975), has seldom been implemented in the literature. Lambrecht (1994: 125) claims that adverbs can sometimes be topics, to the extent that they express ‘pragmatically presupposed propositions’, e.g. they provide temporal background for the proposition expressed by the main clause. Barbosa (2000: 61) defines the role that peripheral adverbs play in the information structure of a sentence as that of ‘introduc[ing] a point of reference with respect to which the whole clause is predicated’. Von Fintel (1994) argues that conditional ifclauses can be topics in the sense that they are domain restrictors. I will adopt Erteschik-Shir’s (1997: 32) denomination stage topic for such clausal modifiers. Stage topics act as restrictors for the interpretation of the sentence. Such topics define the world or situation within which the truth of the predication is evaluated. They are not presupposed in the way that other topics are, but they are nonetheless discourse specified. Their role is best understood in relation to the distinction between Individual Level Predicates and Stage Level Predicates. This distinction was originally made by Milsark (1974) and more extensively defined in Carlson (1977). It is based on the availability of two types of predicate reading: the property reading and the existential reading. Roughly, Stage Level Predicates express temporary or accidental properties (3.13a); Individual Level Predicates express more permanent properties (3.13b). 13 (3.13) a. The reader is tired. b. Literature is a window on the world. Stage Level Predicates are known to be less restrictive than Individual Level Predicates in that they allow for both a property reading and an existential reading in the appropriate context, while Individual Level Predicates only allow the property reading (Diesing 1989, 1992; Kratzer 1989; Erteschik-Shir 1997). In (3.14), the sentence contains a Stage Level Predicate and can consequently be interpreted in two ways: either agony aunts are understood to be available by definition (property reading), or agony aunts are understood to be available right now to do what is expected of them (existential reading). 13

A more accurate definition, along with diagnostics, will be provided in Section 3.2.1.

Interpretation

73

(3.14) Agony aunts are available. In (3.15), by contrast, only the property reading is possible: agony aunts are said to have the quality of being fine psychologists by definition or by nature, and not just in the current situation. This is an instance of Individual Level Predicate. (3.15) Agony aunts are fine psychologists. Erteschik-Shir (1997) claims that the existential reading of sentences like (3.14) is due to the presence of a (covert) stage topic, which refers to the time and place of discourse. By default, the topic of Stage Level Predicates is a stage topic. The rest of the sentence (essentially the subject and the predicate) is in focus. 14 In English, any XP appearing as the subject of an Individual Level Predicate is argued by Erteschik-Shir to be the topic of the sentence (unless another constituent is expressly marked as the topic). 15 She justifies this claim by citing (i) the restriction imposed by Individual Level Predicates on the f-structure of the sentence containing them which prevents them from taking a stage topic, and (ii) the Topic Constraint, which states that the unmarked option in languages like English is for the subject to be the topic of the sentence—a welldocumented cross-linguistic fact (Gundel 1975; Givón 1976; Li and Thompson 1976; Reinhart 1981; Lambrecht 1994). In Section 3.2.1, I will show how these two constraints can be used to investigate the ontology of possible topics in spoken French, and propose a refinement of restriction (i) on the basis of the spoken French data. Erteschik-Shir (1997: 27) argues that Individual Level Predicates cannot take a Stage Level topic as their sole topic, although she acknowledges that Individual Level Predicates can be interpreted with an indexed stage topic. The examples below illustrate that Individual Level Predicates tend not to be compatible with stage topics (3.16a) except in particular cases like (3.16b), where the restriction imposed by the stage topic does not impinge on the defining property of the predicate. (3.16) a. b.

# These days, lemurs are mammals. In our classification, lemurs are mammals.

14 The focus is the most informative part of the sentence or what the speaker wishes to attract the attention of the hearer to (Erteschik-Shir 1986). 15 This was first pointed out by Kuroda (1972) with respect to Japanese.

74

French Dislocation

In spite of the differences, stage topics and aboutness topics play the same role in terms of information structure: they provide the frame within which the predication is to be evaluated. This is particularly clear in (3.17), where the stage topic defines the (hypothetical) world within which the question applies. (3.17) Où irais-tu avec une mobylette, si tu en avais where would-go-you with a moped if you of-them had une? (Nelly, B) one ‘Where would you go with a moped if you had one?’ We are now in a position to see that every sentence has a topic. This was originally proposed in Gundel’s seminal work (1974, 1975). Gundel argued that topics always correspond to a dislocated element, and that a topic can be identified by means of a question like What about X? (where X corresponds to the topic referent) or What’s wrong? What’s new?, etc. The topic corresponds in many cases to time t, place p (Gundel 1975: 110), by default the time and place of the utterance. In other words, it can be a (covert) stage topic. 3.1.6 The role of topics The question any theory needs to address is that of the function of topics in the information structure of sentences, i.e. what it is that topics uniquely contribute to the encoding and retrieval of information. Within the Reinhart tradition, a topic determines where in the file new information can be entered (Section 3.1.1). Portner and Yabushita (1998) adopt the radical view that information gleaned from a predication can only be entered on the card associated with the link/topic, and not on a card associated with any other referent. 16 However, this cannot be the whole story: thetic sentences, for instance, do not have an aboutness topic, but the information they contribute becomes part of the common ground and hence must somehow be entered in the file. Also, if information can only be entered on the card corresponding to a topic T, that information should not be available to summon a referent other than T in subsequent utterances (as claimed by Portner and Yabushita 1998). This prediction is not however borne out. Consider the following piece of discourse. The six sentences below are uttered by the same speaker, with either (i) or (ii) as a follow-up to sentence (3.18e). 16 The same is predicted under Erteschik-Shir’s (1997) approach, except that such information is also entered on the card(s) corresponding to any referent in focus. This would have no impact on the example discussed below in the text.

Interpretation

75

(3.18) a.

Il y a deux femmes qui sont entrées à cinq minutes d’intervalle. ‘Two women came in within five minutes’. b. La première, elle a commandé des calamars et la the first-one she has ordered some squid and the deuxième, rien du tout. second-one nothing at all ‘The first one ordered squid and the other one nothing at all.’ c. Puis il y a un homme qui est entré avec une caisse de figues. ‘Then a man came in with a crate of figs.’ d. Il l’a donnée à celle qui était entrée la première. ‘He gave it to the one who had come in first.’ e. Après ça, je me souviens pas bien. ‘After that, I don’t really remember.’ (i) Je pense que celle qui avait reçu les figues, elle est I think that the-one who had received the figs she is par derrière. sortie gone-out by the-back ‘I think the one who had received the figs left through the back.’ (ii) Je pense que celle qui avait commandé des I think that the-one who had ordered some calamars, elle est sortie par derrière. squid she is gone-out by the-back ‘I think the one who had ordered squid left through the back.’

The sentence in (3.18a) introduces two referents into the discourse (the women), which implies that a card is created for each of them. On the card for woman X, the information entered is (i) that she entered before woman Y 17 and (ii) that she ordered squid. Information regarding woman Y is entered on the card for woman Y, not on the one for woman X. The sentence in (3.18c) triggers the creation of an additional card, for man Z, upon which the information entered is that (i) he entered and (ii) that he was carrying a box of figs. The link/topic for sentence (3.18d) is the man, hence the information that he gave the figs to woman X is entered on the man’s card and not on the card of woman X. Portner and Yabushita (1998) predict that for this reason a follow-up like (3.18ei) should be anomalous because the guise under which the referent of woman X is summoned as a topic is illicit: it uses information 17

I leave aside here any specifics regarding place and time.

76

French Dislocation

that was entered not on the card for woman X but on the man’s card. Only (3.18eii) is predicted to be licit (i.e. not to require accommodation). However, this is not verified by spoken French: either way of encoding woman X as a topic is accepted without hesitation by my informants; neither is reported to be marked. It is not possible to capture the facts discussed above if one assumes that the information contributed can only be entered on the card corresponding to the topic which the predication is about. The essential information structural role of topics therefore cannot be to restrict where information can be entered in the Common Ground file. Instead, the view adopted here is that the main function of the topic is to delimit the domain of applicability of the predication. From this perspective, the properties of aboutness and relevance can be seen to follow from the ‘domain restrictor’ function of topics. When the topic corresponds to a referent identifiable in the context, the impression of aboutness and relevance derives directly from the fact that the predication is interpreted as true with respect to that topic. Defining topics as domain restrictors predicts their behaviour better than an approach which defines them as pointers in the Common Ground file. 3.1.7 Summary The picture drawn so far is the following. The topic of a sentence defines the domain of applicability of the predication (i.e. the frame within which the predication is assumed to hold true). When the topic corresponds to a discourse referent in the context set (or common ground between speaker and hearer), the predication is interpreted as being about that topic. Each predication is assessed with respect to a topic. Even in cases where a topic is not overtly expressed, the predication is assessed with respect to a (covert) stage topic. By default, the stage topic is coindexed with the time and place of discourse. If there is more than one topic, the frame defined is restricted cumulatively by the topics. Topics do not have to convey old information. However, aboutness topics have to correspond to a referent which is sufficiently salient in the context. This referent has to be identifiable by the hearer 18 and the predication has to contribute information that is relevant to the topic. Topic is an information structure primitive in the sense that (i) it cannot be derived from other notions such as focus or discourse topics, and (ii) it has a central role to play in the evaluation of the predication, which is the prerequisite for information gathering. 18

The ease with which a referent can be identified by the hearer is a factor which the speaker must evaluate.

Interpretation

77

3.2 Topics in spoken French I demonstrate below that in spoken French phonologically non-weak elements are obligatorily dislocated if they are topics, and that dislocated XPs which do have a non-dislocated counterpart are obligatorily interpreted as topics. Given these properties, dislocation can be as a diagnostic for topichood. Using this tool, I will investigate the extent to which indefinites can be topics in spoken French, a point which has led to some confusion in the literature. 3.2.1 A test case for topichood I contend that in spoken French, non-pronominal elements are obligatorily dislocated if they are topics. This can be tested using Individual Level Predicates, which we have seen force a topic interpretation of their subject. Individual Level Predicates provide the only reliable test for topichood given that their subject is argued to be a topic irrespective of the context. Consider the contrasts in (3.19) and (3.20). (3.19) a.

Le malais, c’est difficile. the Malay it-is difficult b. #Le malais est difficile. the Malay is difficult ‘Malay is difficult.’ (3.20) a. Le bacille, c’est pas laid. the bacillus it-is not ugly b. #Le bacille est pas laid. the bacillus is not ugly ‘Bacillus isn’t ugly.’ The predicates in (3.19) and (3.20) could be viewed as prototypical Individual Level Predicates in that the copula provides mere ‘introductory support’ for the property that is predicated of the subject. As illustrated by the contrast between the (a) and (b) sentences in these examples, the subject of such predicates is obligatorily dislocated in spoken French. This is not due to a constraint that would prevent the copula from taking a heavy subject, as the examples below illustrate. (3.21) a.

Le directeur est là. the director is there ‘The director’s there.’

78

French Dislocation b. Le fromage est moisi. the cheese is mouldy ‘The cheese’s gone mouldy.’

The difference between the sentences in (3.21) and those in (3.19) and (3.20) in terms of information structure is that the sentences in (3.21) are only felicitous in an out-of-the-blue context, i.e. if all of the sentence is in focus. According to Erteschik-Shir (1997), the topic of such sentences is a stage topic referring to the here-and-now of the discourse. The subject cannot be interpreted as the topic in such sentences. In (3.19) and (3.20), by contrast, the predication cannot be restricted in time or space; it expresses a defining property of what is being talked about. The subject must be interpreted as the topic and the sentence receives a generic reading. 19 The contrast between (3.21) and (3.19)–(3.20) thus confirms the hypothesis that heavy subjects must be part of the focus in spoken French. The obligatory dislocation of the subject with c’est ‘it is’ is thus not the result of a ‘grammaticalization’ of certain dislocations (a position advocated by Barnes 1985 who went on to claim that such dislocations were not pragmatically motivated, as they were obligatory). Instead, the obligatory nature of such dislocations is due to the fact that c’est x (x = property) is an Individual Level Predicate whose subject must be the topic of the sentence, and that in spoken French an XP can only be a topic if it is dislocated. In what follows, I propose to test the predictions arising from this double claim using the York and Cat corpora. This requires first a refinement of the definitions of Individual Level Predicate and Stage Level Predicate. Jäger (2001) shows that no uniform contrast differentiates ILPs from SLPs, and that permanence is neither necessary nor sufficient for a property to be Individual Level (a point already made by Carlson 1977). Jäger argues in favour of a continuum between ILPs and SLPs, and distinguishes eight cases. Each case is defined by the combined results of three tests: (i) their acceptability in perception reports, (ii) the possibility of a weak reading of the subject, and (iii) the transitory character of the property denoted by the predicate. Each case is briefly covered below. Perception verbs can only take in their embedded clause predicates which denote transitory or accidental properties. This is illustrated for English in (3.22). In (3.22a) the predicate of the embedded clause is a SLP. In (3.22b) that predicate is an ILP and as a result the sentence does not make sense: knowing 19 Characterizing or generic sentences express regularities which transcend particular facts (Krifka et al. 1995). I will come back to this point in Section 3.2.2.1.

Interpretation

79

English is not a transitory or accidental property, so you cannot have seen someone or something know English. (3.22) a. I saw the Muppets sing. b. #I saw the Muppets know English. SLPs but not ILPs allow for a weak reading of indefinite subjects. Under the weak reading, the indefinite is interpreted existentially, as in (3.23a): this sentence is about a particular group of chimps. Under the strong reading, the indefinite is interpreted generically, as in (3.23b): this sentence is about chips in general. 20 The predicate in (3.23a) is therefore a SLP while that in (3.23b) is an ILP. (3.23) a. Chimps are dancing in the yard. b. Chips are crisp little things. To help evaluate the transitory character of the property denoted by the predicate, I will rely on Milsark’s (1974) proposal that SLPs express qualities that can be removed without causing any changes in the essential qualities of the individual. 21 For instance, in (3.24a) being naked or not does not affect the identity of the king, and hence this is clearly one of his transitory properties. This predicate is thus clearly an SLP. Being the head of the nation is a defining property of the king, which makes the predicate in (3.24b) an ILP. (3.24) a. The king is naked. b. The king is the head of the nation. Another indicator of the permanence of a property is the ‘lifetime effect’ induced when an ILP appears in the past tense. In (3.25a), the past tense implies that the individual in question does not exist anymore. In (3.25b), this individual may or may not still exist at the time of speech. The predicate in (3.25b) is thus an SLP, while that in (3.25a) is an ILP.

20

Sentence (i) clearly shows that ILPs are not necessarily incompatible with a stage topic.

(i) In Belgium, chips are crisp little things. In England, this is rarely true. Here the stage topic in Belgium specifies the world in which the statement expressed is true. This does not change the fact that the predicate is an ILP. Erteschik-Shir’s (1997) claim that ILPs cannot take a stage topic is thus too strong, a point she herself acknowledges. 21 Even so, the transitory character of a property is still difficult to judge in some cases, as for instance in (i). (i) The king is the proud owner of a motorbike.

80

French Dislocation

(3.25) a. Claude Deschamps was a wonderful woman. b. Claude Deschamps was in her garden. The claim that the subject of an ILP is obligatorily dislocated in spoken French can be tested as follows. 22 The data to consider consist of all the phonologically non-weak elements expressing the subject in a given sample of spontaneous production (here, the adult sample from the York and Cat corpora). Phonologically weak elements (i.e. subject clitics and weak ça ‘it’) need not concern us here as they can never by definition be dislocated. 23 The claim above will be verified if none of the non-weak elements expressing the subject appears in the canonical subject position of an ILP. A total of 292 non-weak elements appear in the canonical subject position in the sample under investigation. These can be identified on the basis of the absence of a resumptive clitic, as argued in Section 2.2. In all but two cases, the predicate of these sentences is eventive, as in (3.26), or it denotes a transitory property, as in (3.27). (3.26) a.

Tout le monde est rentré? (Sarah, F) all the people is re-entered ‘Everybody’s come back in?’ b. Heureusement Bruno est venu. (Jaco, B) happily Bruno is come ‘Luckily Bruno came.’ (3.27) a. Les dauphins sont verts. (Catherine, C) the dolphins are green ‘The dolphins are green.’ (talking about a particular picture) b. Catherine a un gros nez. (Catherine, C) Catherine has a big nose ‘Catherine’s got a big nose.’ (she’s got a clown nose on) c. Est-ce que le jus de fruits est meilleur dans la is-it that the juice of fruit is best in the gourde? (Dominique, B) bottle ‘Is the fruit juice better from the bottle?’

22 Whether the subject of a SLP is a topic or not depends on the context and is therefore much more difficult to evaluate. 23 Note that there are two pronominals ça in French: a weak one and a strong one. The strong one can be dislocated; it cannot appear in the canonical subject position.

Interpretation d. Toute la partie du haut est faite. all the part of-the top is done ‘All the top part is done.’ (while doing a jigsaw)

81

(Dominique, B)

I have found only two exceptions, in which an ILP occurs with a heavy subject. They are given in (3.28). (3.28) a.

Elise est une petite fille. Elise is a little girl ‘Elise is a little girl.’ b. Si, Isabelle est une fille. yes Isabelle is a girl ‘Yes, Isabelle is a girl.’

(Jaco, B)

(Jaco, B)

Both sentences are part of an exchange between Jaco and her grandson Tom, in which Tom is asked to name all the people he knows who are girls. The context clearly indicates that the subject is in fact in narrow focus in both cases: Elise and Isabelle are contrasted with other ‘candidates for girlhood’. In other words, these two referents belong to a set previously introduced in the discourse, i.e. the set of people known by Tom to be girls. In neither case is the sentence thetic: the main topic is not a stage topic but a covert aboutness topic. This calls for a refinement of Erteschik-Shir’s (1997) proposal: the information structure restriction on ILPs is not that their subject must be a topic, but that their main topic cannot be a stage topic. The subject of an ILP must be an aboutness topic corresponding either to the subject or to a set including the subject referent. I conclude that phonologically non-weak topics are obligatorily dislocated in spoken French. 3.2.2 Indefinite topics 3.2.2.1 Take 1: generic indefinites The topic status of indefinites has been discussed at length in the literature (see e.g. Gundel 1975; Larsson 1979; Rizzi 1986; Ward and Prince 1991). Various contradictory claims have been made with respect to the ability of such elements to be topics. Some argue that indefinites can be topics only if they are specific (e.g. Gundel 1975), others claim that specific indefinites can never be topics (e.g. Larsson 1979), and others still propose that quantified expressions in general are banned from topic positions (e.g. Zribi-Hertz 1994). Consider the sentences in (3.29).

82

French Dislocation

(3.29) a.

∗

Quelqu’un, il est venu. somebody he is come b. ∗ Un enfant, il est arrivé et il t’a posé une question. a child he is arrived and he to-you-has asked a question

An initial, intuitive explanation of such data is that the peripheral elements in (3.29) are non-felicitous as topics because it is impossible to identify to who quelqu’un ‘someone’ or un enfant ‘a child’ refers, and that this is due to the nature of quantifiers and indefinites. However, quantifiers and indefinites as such are not excluded from dislocated (topic) position, as shown in (3.30). Tous les copains ‘all the friends’ in (3.30a) refers to the entire set of characters with which the child is playing. It has an identifiable referent, as there is a limited number of such characters, and the child knows about all of them. In (3.30b), what is described is not the behaviour of a particular child, but behaviour that is typical of children in general. (3.30) a.

Et après tous les copainsi , ilsi viendront en train pour and then all the friends they will-come by train to (A.-Gaël, F) visiter le zoo. visit the zoo ‘And then all the friends will come by train to visit the zoo.’ b. Tsé, un enfanti , ili arrive pi ili te pose une you-know a child he arrives and he to-you asks a question. (Auger 1994) question ‘A child comes and asks you something . . . ’

Un enfant ‘a child’ in (3.30b) is interpreted generically; it cannot refer to a particular child. Note that the locus of genericity is not the NP itself; this is illustrated by the contrast in (3.31). 24 Krifka et al. (1995) argue that NPs are interpreted generically because they are bound by a generic operator in the sentence. (3.31) a.

∗

On leuri coupera le cou, à des we to-them will-cut the neck to indef aristocratesi . aristocrats

(Cadiot 1992)

24 Example (3.31a) originally comes from Cadiot (1992), who uses it as evidence in favour of the hypothesis that indefinites cannot be (right) dislocated. The present discussion of this example shows that a total ban on indefinites being topics is too strict, though.

Interpretation

83

couperait le cou, à des aristocratesi . b. On leuri we to-them would-cut the neck to indef aristocrats ‘We’d chop off the heads of aristocrats.’ The sentence in (3.31a) is ungrammatical because the specific temporal reference requires a specific reference for the indefinite, which is not possible in this case. The sentence in (3.31b), on the other hand, is grammatical because in this case the interpretation is generic: it transcends any particular situation. The indefinite is thus acceptable as a topic, and it can be dislocated. DPs when interpreted generically do not receive an existential reading (Diesing 1992): they do not refer to a particular individual or set of individuals. 25 In (3.32), for instance, no particular pig is being referred to: the sentence expresses a general truth about pigs, and un cochon ‘a pig’ is obligatorily interpreted as the topic. (3.32) Un cochon, ça s’échappe de n’importe où. a pig it refl-escapes of no-matter where ‘A pig can escape from anywhere.’ XPs expressing the subject of a generic sentence are obligatorily dislocated in spoken French. In this language, the dislocation construction is one possible means of triggering a generic reading of the object. Both claims are substantiated below. The following sentences initially appear to undermine the claim that subject XPs are obligatorily dislocated in generic sentences. (3.33) a.

Un maçon ne trahit pas. a mason neg betrays not ‘A mason doesn’t betray.’ b. Le paresseux passe près de quatre-vingts pourcents de son the sloth spends near to eighty per cent of his temps à dormir. time to sleep ‘Sloths sleep eighty per cent of the time.’

25 In French, plural DPs that appear to be definite can be interpreted generically, in which case the definite article can be viewed as a dummy definite article (i). As pointed out in Krifka et al. (1995: 68), the definite article used with a plural which is interpreted generically in French is a reflex of the requirement that every NP must have a determiner in this language.

(i) Les cobras, c’est long. the cobras it-is long ‘Cobras are long.’

84

French Dislocation c.

Lire le Braille est un atout. to-read the Braille is an asset ‘To read Braille is an asset.’

However, native speakers report an impression of formality when presented with such examples. My informants report that (3.33a), for instance, sounds like an extract from a code of conduct, (3.33b) like an extract from a nature programme, and (3.33c) like a specification from a job recruitment advert. These sentences are thus not representative of spoken French, but of written/formal French. 26 We saw in Section 3.2.1 that the subject of an ILP is obligatorily dislocated in spoken French. As such predicates express inherent, permanent properties, they are by their nature generic (a point argued by Chierchia 1995). The obligatoriness of the dislocation in such cases in spoken French thus comes as no surprise. The availability of a generic reading is not limited to cases where the dislocated XP expresses the subject. With the verb aimer ‘to like/love’ in particular, the object tends to be dislocated when a generic interpretation is required. In such sentences in Canadian French, the resumptive element is always ça ‘that’, even when the dislocated object is a clause (3.35). (3.34) MAX:

CAT:

MAX:

CAT:

MAX:

CAT:

Ili aime ça j , les poissons j , luii ? he likes that the fish him ‘Does he like fish?’ Mais oui il aime çai , les poissonsi . but yes he likes that the fish ‘Yes, he does like fish.’ Parce que moi, j’ aime pas çai , les poissonsi . because me I like not that the fish ‘Cos me, I don’t like them.’ T’ aimes pas çai , les poissonsi ? you like not that the fish ‘You don’t like fish?’ Non. no ‘No.’ T’ aimes pas çai , [les manger]i ou tu les aimes pas? You like not that them to-eat or you them like not ‘You don’t like to eat them or you don’t like them?’

26 Information structure is encoded differently in written French. In particular, postverbal subjects are used to encode thetic structures, and elements in the canonical subject position allow a topic interpretation (see e.g. Lahousse 2003).

Interpretation MAX:

(3.35) a.

b.

Je les aime pas. I them like not ‘I don’t like them.’ Moi, j’aimerais çai , [être un homard]i . me I-would-like that to-be a lobster ‘I would enjoy being a lobster.’ Il aime çai , Crocro, [qu’on le caresse]i . he loves that Crocro that-one him stroke ‘Crocro loves to be stroked.’

85

(Pol, C)

(Catherine, C)

In European dialects of spoken French, the equivalent construction does not involve resumption (although resumption with ça is also possible). 27 For a discussion of the omissibility of object clitics, see e.g. Fónagy (1985); Cummins and Roberge (2004, 2005); Prévost (2006). (3.36) Les gâteaux, j’adore. the cakes I-adore ‘I love cakes.’ The generic reading of a plural NP can be blocked when a certain type of resumptive element is associated with it, as shown in (3.37). 28 (3.37) a.

b.

Les gâteauxi , j’adore çai . the cakes I-adore that ‘I love cakes.’ Les gâteauxi , je lesi adore. I them adore the cakes ‘I love the cakes.’

When the resumptive element is ça ‘that’, the sentence retains its generic reading. When the resumptive element is les ‘them’, a specific reading is obligatory in a sentence like (3.37b). The resumptive pronoun ça ‘that’, when bound by

27

The right-dislocated equivalent is also attested in the corpora under investigation. This is a simplification. The possibility of a generic reading does not depend solely on the presence or the nature of the resumptive element, as illustrated by the contrast in (i). The sentence in (ia) is ambiguous between a generic and a specific reading in spite of the presence of the specific resumptive les ‘them’. The use of ça ‘that’ to force a generic reading is only marginally acceptable (ib). 28

(i) a.

[Les lapins malades]i , il faut lesi tuer. the rabbits sick it must them kill ‘Sick rabbits must be killed.’/ ‘The sick rabbits must be killed.’ b. ?[Les lapins malades]i , il faut tuer çai . the rabbits sick it must kill that

86

French Dislocation

a generic operator, seems to be stripped of any ϕ-feature it might otherwise have. In such cases ça, can refer to a [+ human] nominal (something it cannot do under a specific reading). (3.38) a.

b.

Les enfantsi , çai fait du bruit. it makes some noise the kids ‘Kids are noisy.’ ∗ Les enfantsi , çai vient demain. the kids that comes tomorrow

Similarly, the resumption of an indefinite DP by ça ‘that’ is only possible in a sentence with a generic reading. (3.39) a.

b.

jamais vu [Un robinet à piles]i , je n’ai a tap with batteries I neg-have never seen çai . that ‘I’ve never seen a battery-powered tap.’ Ben parce qu’une camérai , çai parle pas. well because that-a camera it talks not ‘Because cameras don’t talk.’

(Nelly, B)

(Denis, F)

Note however that resumption by ça does not automatically entail that a sentence has a generic interpretation. (3.40) a. Çai raconte quoi, ce livrei ? it tells what that book ‘What’s that book about?’ b. Çai marche pas, luii ? it works not him ‘Is this one not working?’

(A.-Gaël, F)

(Martine, C)

3.2.2.2 Take 2: specific and d-linked indefinites English and spoken French differ significantly with respect to specific and d-linked indefinites. Apparently, English allows specific indefinite topics just as it does generic indefinites (Gundel 1975). The same is not true of spoken French, as has been widely reported in the literature (e.g. Larsson 1979; Lambrecht 1981; Barnes 1985; Lambrecht 2001). 29 Consider the following contrast.

29 Barnes (1985) argues, following Larsson (1979), that there is a ‘grey area’ between specific indefinites and generic indefinites in which examples like the following are found:

Interpretation

87

(3.41) a. A friend of mine is intelligent. (Erteschik-Shir 1997: 39) b. #Une de mes copines, elle est géniale. one of my friends she is genius-like Modulo the requirement that topics be left-dislocated rather than appear in the subject position of ILPs in spoken French, the two sentences in (3.41) are equivalent from an information structure point of view. However, the sentence in (3.41b) is odd in spoken French. A felicitous equivalent is given in (3.42). (3.42) J’ai une copine, elle est géniale. I-have a friend she is genius-like ‘A friend of mine is a genius.’ Crucially, the utterance in (3.42) is a single sentence, with the first clause bearing left-dislocation prosody. If each clause were uttered as a separate sentence (i.e. with the typical declarative final contour (Delattre 1966) on the first clause instead of dislocation prosody), the meaning would be different, just as uttering I have a friend serves a different purpose to uttering a friend of mine. This use of a full clause with left-dislocation intonation for the purpose of referent identification has been reported before in the literature on spoken French (e.g. Carroll 1982; Blasco 1993). Other variants are also found, as illustrated in (3.43). a un garsi , ili était entré Moi, en mécanique, il y me in mechanics it there has a guy he was entered à l’école des métiers. to the-school of-the trades ‘When I was in mechanics, there was this guy who went to engineering college.’ (Carroll 1982: 233) b. Vous avez des garsi , ilsi ont des diplômes, ils n’ont you have some guys they have some diplomas they neg-have pas de boulot. not of job ‘Some guys have diplomas and yet they don’t have a job.’ (Blasco 1993: 277)

(3.43) a.

(i) Mais une soupe, une soupe à la betterave, j’en ai déjà mangé, moi, au xx. but a soup a soup to the beet I-of-it have already eaten me at-the xx ‘But I’ve already eaten beet soup at the xx.’ (Barnes 1985: 93) I treat such examples as generic, given that what they refer to (here a type of soup) transcends particular individuals. The dislocated indefinite could be resumed by ça ‘that’ in this sentence without yielding a different interpretation.

88

French Dislocation

I have come across two types of constructions in spoken French that can be used to introduce specific indefinites as topics. These constructions are both ‘presentational’ in the sense that they introduce a referent into the discourse context. One of them could be classed as simply ‘existential’: it is the dislocated il y a x ‘there is x’ construction, as in (3.43a). Uttering il y a x introduces the referent of x into the discourse by asserting its existence. When this clause is dislocated, I propose that it makes the referent x available as topic of the main clause (as I explain below). The second presentational construction could be classed as ‘speaker-related existential’. It is introduced by a dislocated clause of the type j’ai x ‘I have x’ or je connais x ‘I know x’. Alternatively, the reference can be made to an ‘abstract’ addressee, 30 as in (3.43b) ‘you have x’. These constructions bring to the hearer’s attention the existence of an individual or group of individuals with special reference to the speaker or hearer, and announce that the sentence is going to be about that/these individual(s). 31 As pointed out by Kuno (1972: 297), ‘the most natural way to introduce an entirely new event in conversations seems to be to talk about the existence, or coming into existence toward the place of the speaker, of something’. I propose that the left-dislocated clauses in (3.42) and (3.43) serve the same purpose as the modifier of the indefinite in (3.41a): they trigger what Erteschik-Shir (1997) calls the subordinate update rule. This rule creates a new card in the hearer’s file (i.e. it introduces a new discourse referent). The label (or heading) of the card corresponds to the referent of the indefinite, and the modifier of the indefinite is added as an entry on that card. ErteschikShir (1997: 41) argues that the reference of this card is speaker-oriented, as only the speaker is assumed to have a full card for that item. The reference attributed by the hearer to this sort of card is something like ‘x which is known of the speaker’. This process is similar to the licensing of a referent as topic by virtue of it being in a poset (partially ordered set) relation with a referent in the context set (Ward and Prince 1991), but it is wider-ranging because it does not rely in all cases on a referent from the context (as in (3.43b) for instance).

30 The addressee is thought of as abstract or arbitrary in the sense that there need not be any connection between the actual addressee and the referent introduced. The effect of this construction is to implicate the addressee, to call them as ‘witness’ to the fact that the referent of x exists, even though the addressee might not have been aware of its existence before. 31 Linearity plays a role in this construction, as expected given its introductory function: such dislocations are impossible in the right-periphery of the clause. See Section 4.4.4.

Interpretation

89

In spoken French, indefinites with an existential reading cannot be introduced as topics by a plain dislocation in the way that new but salient referents can: they need to be embedded in a presentational construction, which is itself dislocated. In other words, subordinate f-structures involving indefinites need to be expressed as a secondary predication in spoken French: they cannot be encoded by a DP alone (as they can in English). I believe this is what explains the contrast in acceptability between sentences like (3.44) and (3.45). diable. (3.44) Tous tes cousinsi , tu peux lesi envoyer au all your cousins you can the to-send to-the devil ‘You can tell all your cousins to go to hell.’ (3.45) ∗ Quelques-unsi , ilsi sont saltimbanques. they are acrobats some ones In (3.44), the referent of the dislocated DP corresponds to a finite set of people who can be identified from the context (i.e. they are d-linked in the sense of Pesetsky (1986)) without the need for subordinate update. In this case, the referent of the dislocated element is part of the speaker and hearer’s Common Ground which means it qualifies as a topic. In (3.45), by contrast, the reference of the quantified DP is not available directly from the context: it corresponds to a subset that can only be identified via subordinate update. In English, though, a DP suffices to trigger subordinate update (as argued by ErteschikShir 1997), and a sentence like (3.46) is acceptable. (3.46) Some are acrobats. In spoken French, this type of referent can only be identified via a presentational construction of the type il y a ‘there is’. (3.47) Il y en a, ils sont saltimbanques. it there of-them has they are acrobats ‘Some of them are acrobats.’ Subordinate update normally points to a subset of a previously established topic. Given that indefinites cannot by themselves trigger this mechanism in spoken French, it is to be expected that those plural indefinites that are successfully dislocated in this language will point to an entire set rather than a subset in the context. In other words, a dislocated quantified expression is only predicted to be possible in spoken French when it is interpreted as a sum individual (Erteschik-Shir 1997), i.e. a finite set defined in the discourse context and treated as if it were an individual (represented by a single

90

French Dislocation

card). This is verified by the rare cases I have come across in the literature, as well as in the York and Cat corpora. Representative examples are given below. Si tous mes soldatsi , ilsi étaient curés . . . if all my soldiers they were priests ‘If all my soldiers were priests . . . ’ (Vanelli 1987, from Sanfeld 1928) b. Et après, tous les copainsi , ilsi viendront en train pour and then all the friends they will-come in train to visiter le zoo. visit the zoo ‘And then all the friends will come by train to visit the zoo.’ (A.-Gaël, F)

(3.48) a.

The present analysis also predicts that non-dislocated presentational constructions will not create a topic, i.e. they will not allow subordinate update. Indefinites that always require an existential reading (like quelqu’un ‘somebody’) cannot be interpreted as topics and hence are not acceptable in subordinate f-structures. My analysis of the il y a construction thus predicts that such indefinites should be acceptable in non-dislocated il y a but not in left-dislocated il y a. This is precisely the contrast between sentences (3.49) and (3.50). (3.49)

#Y a quelqu’un, il sonne à la porte. there has someone he rings at the door (3.50) Y a quelqu’un qui sonne à la porte. there has someone who rings at the door ‘Someone’s ringing the doorbell.’ When the presentative construction is not dislocated, as in (3.50), it does not create a topic; there is no subordinate f-structure in that case. The sentence in (3.50) is a thetic sentence. Its topic can only be a (covert) stage topic. To conclude, indefinites can be topics in spoken French under a specific reading (contra e.g. Larsson 1979; Lambrecht 1981, 2001), but this is only possible via subordinate update, which in spoken French is achieved via the dislocation of a presentative clause when it involves an indefinite. The situation is slightly different when definites are involved, as discussed in Section 4.5.

Interpretation

91

3.2.3 Topics in specificational pseudo-clefts Consider the specificational pseudo-clefts in (3.51): (3.51) a. Ce qu’ il veut, c’ est rester près de toi. (Nelly, B) that that he wants it is to-stay next to you ‘What he wants is to stay near you.’ b. Ce qui est amusant, c’ est d’ effacer ce qu’ on fait, that that is fun it is to wipe-off that that one did hein? (Nelly, B) eh ‘What’s fun is to erase what you’ve done, isn’t it?’ By definition, specificational pseudo-clefts involve a predicate in the subject position (Heycock and Kroch 1999). (3.52) What they wanted to buy was a picture of the sea. In spoken French, however, this predicate must be dislocated. (3.53) Ce qu’ils voulaient acheter*(, c’) était une photo de la that which-they wanted to-buy it was a picture of the mer. sea ‘What they wanted to buy was a picture of the sea.’ This suggests that such predicates might be topics. Indeed, I argue below that a topic analysis is desirable. In specificational sentences that do not involve a pseudo-cleft, one NP is interpreted predicatively, e.g. mon seul ami ‘my only friend’ in (3.54), and the other one identifies (i.e. specifies) an individual with that property, e.g. mon chien ‘my dog’ in (3.54). (3.54) a. Mon chien est mon seul ami.32 my dog is my only friend ‘My dog is my only friend.’ b. Mon seul ami ∗ (, c’) est mon chien. my only friend it is my dog ‘My only friend is my dog.’ Comparison of the French examples above with their English translations shows that the predicative NP is obligatorily dislocated if it is (associated with) the subject in French but not in English. 32

The subject can also be dislocated in this sentence.

92

French Dislocation

The parallelism of (3.54b) and the specificational pseudo-clefts in (3.51) is clear: the dislocated clause identifies a property, and the post-copular clause identifies an instantiation of that property. By identifying a property, the dislocated elements in (3.51) and (3.54b) summon a set of possible topics from which the relevant but not fully specified one is picked (this corresponds to the head of the free relative in (3.51)). This is a special case of subordinate update, in which the exact referent of the topic remains unspecified: it is a variable, whose value is provided by the post-copular focus. This analysis predicts that the dislocated relative clause will be presupposed, and that the head of the relative will find its reference among a set defined by the context. In other words, this construction is obligatorily d-linked: the reference for the variable contained in the topic must be part of a set available in the discourse. This is verified by the examples above. For instance, before (3.51a) was uttered, the speaker and the hearer had been discussing what a particular doll would like best. The options were either to be left in the room upstairs or to stay in the hearer’s arms: a two-member set from which only one member is picked as matching the property identified by the topic. The context preceding (3.51b) is one in which the child has been drawing and then erasing things: again, a two-member set from which one member is selected. 3.2.4 Topics take wide scope As a direct consequence of the requirement that the referent of a topic be identifiable in the context, anaphoric expressions can only be dislocated in spoken French if they draw their reference from the context rather than from an element inside the sentence. This is illustrated by the contrast in (3.55). emmenée, sa j/#i fille. (3.55) Je connais l’hommei qui l’a I know the-man who her-has taken-away his/her daughter ‘I know the man who took his/her daughter away.’ For the majority (91 per cent) of my informants, the daughter in question cannot be the child of the man in the sentence. 33 This is because the reference of the man in (3.55) is defined by the restrictive relative clause. It thus cannot have been introduced previously in the discourse

33 The same judgement obtains if sa fille ‘his/her daughter’ is left-dislocated. The sentence in (3.55) and its left-dislocation equivalent were presented to thirty-one native speakers as part of a judgement elicitation task. I come back to this in Chapter 4. For details, see Appendix C.

Interpretation

93

(at least not under this guise). As a consequence, the reference of the daughter, being a topic, cannot be dependent on that of the man; it has to be related to some other person identified in the context. As their reference is by definition independent of the rest of the sentence, dislocated topics cannot receive a bound interpretation. They obligatorily take wide scope. (3.56) illustrates this with respect to negation. (3.56) Il parlait tout le temps des tas d’amis qu’il avait. Ben he talked all the time of-the loads of-friends that-he had well tous ses amisi , sa mère a pas voulu lesi inviter à son all his friends his mother has not wanted them to-invite to his enterrement. burial ‘He was always going on about the many friends he had. Well his mother refused to invite any of them/#did not want to invite all of them to his burial.’ Topics defined by subordinate update are no exception to the requirement that their reference be defined in the context rather than in the sentence. The set in question does not have to have been entirely defined in the preceding discourse, though. It can correspond to a set in the mind of the speaker, as we saw is possible for certain indefinites in Section 3.2.2. In the sentences in (3.57), the hearer interprets the topic as x: a member of a set y, with y defined in the speaker’s file. (3.57) Une que j’aimerais bien rencontrer, c’est Marie Yacoub. one who I-would-like well to-meet it-is Marie Yacoub ‘Someone I’d love to meet is Marie Yacoub.’ Note that in (3.57), the indefinite does not receive a specific reading (in contrast to (3.41b)). Rather, it receives a variable interpretation: the indefinite is the variable and the relative clause identifies the set from which the referent introduced in the focus of the main clause will be picked (as happens in specificational pseudo-clefts). The one apparent exception to the requirement that the referent of the topic be defined in the context or via subordinate update is illustrated in (3.58). (3.58) [Le premier qui bouge]i , je lei mate. the first who moves I him subdue ‘If anyone moves, I’ll subdue them.’

94

French Dislocation

In this case, the predication cannot be evaluated as true or false. Rather, the sentence defines a possible situation in which something would be true. As indicated by the translation, the dislocated element is informationally equivalent to an if-clause introducing a variable but, because this clause acts as a topic, the referent of the variable must be picked from a d-linked set. Indeed, it is difficult to imagine how a sentence like (3.58) could be uttered if the speaker were not in the presence of all the people concerned. A typical situation in which this sentence could be uttered is that of a robber threatening people in a bank. The d-linked set would correspond to the entire group or to a subset of the people in the bank who are clearly being targeted by the gaze of the robber. I conclude that the reference of a dislocated topic must be defined in the context and not (exclusively) in the sentence. The apparent exception to this rule is that the reference of a topic introduced via subordinate update can be identified by the main predication. However, in such a case the choice is obligatorily constrained from a discourse point of view: the referent must be picked from a d-linked set. 3.2.5 Spoken French as a discourse-configurational language The distinctive trait of discourse-configurational languages is that they feature a ‘one-to-one correspondence between the syntactic and the notional predication structures’ (É. Kiss 1995: 14). The obligatory mapping of (phonologically non-weak) topics onto dislocated structures in spoken French strongly suggests that it belongs to the discourse-configurational category. Indeed, categorical and thetic judgements are systematically realized in different syntactic structures in spoken French, and it is possible to have multiple, freely ordered dislocated topics. Both these characteristics are typical of discourseconfigurational languages (É. Kiss 1995: 8-9). I present below a set of informal observations about the encoding of focus in spoken French which demonstrate further the high level of correspondence between syntax and discourse structure in this language. In informal speech, there is a strong tendency for a new referent in the discourse to be introduced via a presentational construction. 34 Such constructions usually consist of a biclausal structure with an existential predication in the main clause, as in (3.59). For instance, should a speaker wish to report the arrival of a policeman who has not yet been mentioned in an informal context, he or she would say (3.59) rather than (3.60)—a fact also reported in Côté (2001). 34 The use of a presentational construction might even be obligatory, but this is very hard to test on native speakers of French because in testing situations it is almost impossible to elicit judgements that would not be influenced by the prescriptive norm, which is based on written French.

Interpretation

95

(3.59) Il y a un policier qui arrive. it there has a policeman who arrives ‘A policeman’s coming.’ (3.60) Un policier arrive. a policeman arrives ‘A policeman’s coming.’ The definiteness of the DP introduced by il y a ‘there is’ determines whether the reading is existential, as in (3.59), or presentational, as in (3.61). An existential reading introduces an entirely new referent into the discourse by stating its existence (Kuno 1972). By contrast, a presentational reading introduces a new referent into the discourse, the existence of which is known by the hearer. Note that the introduction of a referent associated with the grammatical function of subject (as in the examples discussed here) is only possible in an out-of-theblue context, i.e. one in which all the information conveyed by the sentence is new (a thetic sentence). 35 Typically, such a sentence will be the answer to a question like What’s the matter? a Gladys qui dort. (3.61) a. Il y it there has Gladys who sleeps ‘Gladys is sleeping.’ b. Il y a ton chien qui m’embête. it there has your dog who me-bothers ‘Your dog’s bothering me.’ When the complement of il y a ‘there is’ is a small clause, the reading is obligatorily existential: definite DPs cannot appear as the subject of the small clause, as shown by the following contrast. (3.62) a. Il y a [une souris dans la cuisine]. it there has a mouse in the kitchen ‘There’s a mouse in the kitchen.’ a [la souris dans la cuisine]. b. ∗ Il y it there has the mouse in the kitchen Exhaustive/contrastive focus is achieved via the cleft construction in spoken French (É. Kiss 1998). Clefts have the form c’est X qui/que Y ‘it is X that Y ’. The copular clause in a cleft performs an identificational function, not a 35 The only exception to this is when the new referent is salient in the context (for instance because it can be seen by both speaker and hearer). In such a case, the new referent can be directly encoded as a topic in spite of the fact that it has not yet been mentioned in the discourse (De Cat 2004b).

96

French Dislocation

predicative one: it identifies an individual X as displaying a certain property defined by the dependent clause Y . X could be viewed as the answer to an implicit question like What is Y? or What does Y do? In (3.63), the implicit question would be What smells so strong? 36 (3.63) C’est le houblon qui sent si fort. it-is the hops that smells so strong ‘It’s the hops that smell so strong.’ The evidence discussed above indicates that spoken French is a discourseconfigurational language. The following correspondences between syntactic structure and discourse (or information) structure have been shown to hold in this language: (i) categorical judgements are encoded using a dislocated topic (unless the topic is expressed by a phonologically weak element only), while thetic judgements do not involve dislocated elements; 37 (ii) thetic sentences introducing a new referent which expresses the subject are conveyed via the il y a ‘there is’ construction (triggering a presentational or an existential reading); and (iii) contrastive focus is expressed using a cleft construction.

3.3 Conclusion The aim of this monograph is not to propose a new or revised framework within which to capture topics. However, this detailed empirical investigation of clause-peripheral topics in spoken French contributes to the evaluation of existing frameworks and will provide grounds for further research. I have argued that the notion of topic is linguistic in nature, and that it is a primitive. It cannot be reduced to a notion such as ‘old information’, nor can it be defined as a derivative of discourse topic (as Question Under Discussion) or focus. Rather, the topic of a sentence delimits the domain of application of the predication with which it is associated: it defines the world or referent with respect to which the predication should be assumed (by the hearer) to be true. This has been argued to hold even in cases where the topic is not overtly expressed. Two kinds of topic have been distinguished: aboutness topic and stage topic. Both provide the domain of application of the predication. By default, the stage topic is coindexed with the time and place of discourse. Aboutness topics need to be salient (so that their reference is recoverable by the 36 See De Cat (2002) for a detailed study of clefts, pseudo-clefts, and pseudo-pseudo-clefts in spoken French. 37 This is true except for dislocated elements which correspond to a stage topic that defines the world within which the judgement holds true.

Interpretation

97

hearer) and in a relation of relevance with the predication that accompanies them. Sufficient salience can be achieved in three ways:

r by virtue of being part of the permanently available registry (e.g. the kinds and referents that are unique in the world relevant to the conversation);

r by virtue of being salient in the discourse or situational context (in the latter case, the referent enters the discourse context at the same time as it is encoded as a topic); or r by being introduced via subordinate update, which in most cases requires a new topic to be in a poset relation with another salient discourse referent. The cross-linguistic association between topic and subject manifests itself in the dislocated position in spoken French. In this language, non-pronominals have to be dislocated in order to be interpreted as topics.

4 Syntax In this chapter, I will argue that French Dislocation is a unified syntactic phenomenon including both left and right dislocation, regardless of whether it is resumed by a clitic or a non-clitic element. Narrow syntax will be shown to play a minimal role in its derivation: all that is required is the merging of the dislocated element by adjunction to a Discourse Projection (generally a finite TP with root properties). No agreement or feature checking is necessary, hence no syntactic movement of any sort need be postulated. The so-called resumptive element is argued to be a full-fledged pronoun rather than a true syntactic resumptive. The syntactic and interpretive mechanisms underlying left and right dislocation will be shown to be identical; differences between the two constructions can be derived straightforwardly from the properties of the two sides of the clause periphery.

4.1 A brief overview of the literature Dislocated elements have been attested in a variety of languages. The most widely studied type of left dislocation is so-called Clitic Left Dislocation (CLLD). In CLLD, a left-peripheral XP (which appears in bold in (4.1)) is coindexed with a resumptive clitic within the clause. This construction has been attested in many languages (e.g. Italian: Cinque 1990, Rizzi 1997; Greek: Iatridou 1995, Anagnostopoulou 1997; Lebanese Arabic: Aoun and Benmamoun 1998). (4.1) Les malotrusi , on ne lesi invite pas. one neg them invites not the louts ‘We don’t invite louts.’ Traditionally, CLLD has essentially been exemplified with dislocated objects. This appears to be due to the fact that most of the languages discussed in this respect are PRO-drop. As French is not a PRO-drop language (at least in its most widely spoken varieties, as I demonstrated in Section 2.2), it is a prime source of examples of CLLD involving subjects.

Syntax

99

(4.2) Les clitiquesi , ilsi comptent pas pour du beurre. they count not for some butter the clitics ‘Clitics do count.’ Right dislocation has been studied less, but it is also attested in a large number of languages (e.g. Italian: Frascarelli 2002; Catalan: Vallduví 1992; Villalba 2000; López 2003; French: Kayne 1975; Larsson 1979; Ashby 1994; Cantonese: Cheung 1997; Japanese: Tanaka 2001). (4.3) On ne lesi invite pas, les malotrusi . one neg them invites not the louts ‘We don’t invite louts.’ The literature on dislocation is so vast that an entire volume would be necessary for a comprehensive review. I will limit myself here to presenting the central questions addressed in generative analyses, and to identifying the motivations that underly the solutions proposed. An in-depth discussion of the main issues will follow in subsequent sections. 4.1.1 TopicP or no TopicP? 4.1.1.1 Functional heads to derive peripheral syntax From an interpretive point of view, dislocated elements are widely acknowledged to be topics. Under the assumption that the discourse notions of topic and focus should be handled by the syntactic component of grammar, it has been assumed in many generative analyses over the past decade that the dislocated element occupies the specifier of a TopicP at the left edge of the clause (e.g. Rizzi 1997; Frascarelli 2002; Grohmann 2003; among others). Given the relatively free distribution of left-dislocated topics, the Topic projection is assumed to be recursive (Frascarelli 2000; Benincà 2001) and, some authors argue, can appear as multiple separate projections. Rizzi (1997) for instance proposes two (recursive) TopicPs, one on either side of the position which hosts focus-fronted elements. Some authors postulate the existence of an additional projection to host so-called Hanging Topics, which are assumed to be derived in a different way and to have a different distribution (see Section 4.1.2). Benincà (2001) for instance proposes a DiscourseP at the top of the clause, above Rizzi’s ForceP. 1 (Hanging Topics will be discussed with respect to spoken French in Section 4.3.6.) 1 Benincà (2001) claims that the embedded complementizer can occupy either the head of one of the projections hosting topics in their specifier (DiscP or TopP) or the head of ForceP. An important question Benincà does not address is why dedicated topic projections should be necessary at all if topics can appear in the specifier position of a projection hosting a complementizer. Such an analysis requires the complementizer itself to bear some kind of [Topic] feature.

100

French Dislocation

Not all researchers provide an account of right-dislocated topics even when they are attested in the language they investigate. Under a strict functional projection approach, right-dislocated elements cannot be hosted by the right periphery of the clause because functional projections are generally assumed to be absent from that side of the periphery. 2 Thus under this approach, rightdislocated elements can only be hosted either by a left-peripheral TopicP in the CP field (as proposed by Frascarelli (2002), who argues that in this case the remnant of IP is fronted to [spec,GroundP]) or by a TopicP lower in the clause. Cecchetto (1999); Villalba (2000); Belletti (2001); and others have argued that right-dislocated topics sit in the specifier of a TopicP situated just above vP (see Section 4.3.7). The essential motivation for postulating the existence of TopicP in the syntax is that it can contribute to accounting for the distribution of dislocated constructions in two ways. First, the conditions for the projection of a TopicP can be constrained ‘cartographically’. It has been argued, for instance, that the ordering of projections in the CP field is responsible for ordering restrictions between topic and fronted focus or wh-phrases (see e.g. Rizzi 1997; Benincà 2001) and with respect to finite vs. non-finite complementizers (e.g. Rizzi 1997). The same claim has been made regarding the different distribution of dislocated topics in root vs. embedded clauses (see e.g. Frascarelli 2002), and to explain why Hanging Topics allegedly have to precede other topics (e.g. Benincà 2001; Delais-Roussarie et al. 2004). Second, the presence of TopicP implies that a functional [Topic] feature is involved in the derivation of dislocated structures, which in turn triggers feature-driven movement (in overt or covert syntax) or Agree relations. Independently motivated constraints on these operations can then be invoked to (further) restrict the distribution of dislocated elements. This approach predicts (i) a strict ordering of elements at the left edge of the clause and (ii) clear interpretive and structural effects on both left- and right-peripheral elements. These predictions will be investigated with respect to spoken French in Section 4.4.4. 4.1.1.2 Alternatives to the functional projection approach The assumption that (narrow) syntax is responsible for word-order phenomena associated with a particular discourse function is far from being uncontroversial (see e.g. Newmeyer 2003; Gill and Tsoulas 2004; Culicover and Jackendoff 2005). A number of recent generative analyses of such clause-peripheral phenomena 2 This can partly be explained partly by the fact that rightward movement is ruled out in most recent versions of the generative framework. Note however that even the advocates of rightward movement do not postulate the existence of right-peripheral functional projections. See e.g. the contributions in Beerman et al. (1997).

Syntax

101

have been proposed that do not rely on purpose-made functional projections, whether to account for peripheral topics (e.g. Aoun and Benmamoun 1998; Barbosa 2000; López 2003; Contreras 2004; Gill and Tsoulas 2004) or focus fronting (e.g. Erteschik-Shir 1997; Szendr˝oi 2003). Alternatives to the functional projection approach to peripheral topics are of two kinds. The first one posits that topics are base-generated in adjunction structures (e.g. Iatridou 1995; Barbosa 2000). The second postulates that topics reach their peripheral position by syntactic movement, but that the features driving this movement are not syntactic in nature (López 2003). Many have argued for the adjunction analysis over the years (e.g. Hirschbühler 1975; Cinque 1990; Iatridou 1995; Barbosa 2000). According to this analysis, dislocated elements are generated directly at the edge of the clause 3 but not in (the specifier of) a projection endowed with dedicated discourse properties. The topic is generally assumed to be licensed in this adjoined position by a rule of predication (Iatridou 1995; Barbosa 2000). Adjunction per se does not automatically imply absence of syntactic movement. Iatridou (1995), for instance, argues that a dislocated element is directly generated as an adjunct to the CP containing its resumptive element, and that it can move to a higher clause provided no island boundary is crossed. 4.1.2 What moves in narrow syntax (if anything)? Two types observations have been taken to indicate a close relationship between a dislocated element and either its resumptive or the empty argument position with which it is associated. Firstly, in languages with overt case marking such as Greek, dislocated elements have been shown to agree with their resumptive element not only in gender and number but also case (see e.g. Anagnostopoulou 1997). Languages like Italian and French do not feature overt morphological case unambiguously. However, on the assumption that case manifests itself via prepositions (on oblique complements), agreements effects on dislocated elements (traditionally labelled ‘connectivity’ or ‘connectedness’, see Vat 1981; van Haaften et al. 1983; Cinque 1983) have also been argued to arise in these languages. According to this criterion, the dislocated elements in (4.4a) display connectivity, but those in (4.4b) do not. (The issue of connectedness will be addressed in Section 4.1.2.4.)

3

Note that this is possible under a functional approach too, e.g. Frascarelli (2002).

102

French Dislocation

(4.4) a. A Mariei , de ce crime j , je crois que je ne luii en j to Marie of this crime I think that I neg to-her of-it parlerai pas. will-speak not ‘I don’t think I’ll tell Marie about this crime.’ (Delais-Roussarie et al. 2004: 505) b. Mariei , ce crime j , je crois que je ne luii en j parlerai pas. Marie this crime I think that I neg to-her of-it will-speak not ‘I don’t think I’ll tell Marie about this crime.’ (Delais-Roussarie et al. 2004: 505) Secondly, the relationship between a peripheral topic and its resumptive has been argued not to hold across strong islands in a number of languages such as Italian (Cinque 1977, 1990), Greek (Iatridou 1995; Anagnostopoulou 1997), Spanish (Escobar 1997), and Catalan (Vallduví 1992; Villalba 2000). Some other languages show a mixed picture in that respect (e.g. Lebanese Arabic: Aoun and Benmamoun 1998). It has been argued that these two facts indicate the involvement of syntactic movement in the derivation of dislocated structures. Various analyses have been proposed, as explained below. 4.1.2.1 The topic moves Few analyses argue that dislocated topics move from an argument position to a clause-peripheral one. The classic objection to such an approach is that it postulates an initial argument doubling structure, with both the resumptive and the topic in the core of the sentence. This is problematic given that CLLD is attested in languages that do not allow doubling (e.g. Italian (Cinque 1990) and French, see Sections 2.2 and 4.3.7) and that even in languages which do allow argument doubling, the types of NPs allowed in doubling structures differ from those allowed in cases of CLLD (Iatridou 1995; Anagnostopoulou 1997). 4 To avoid this problem, it has been proposed that the resumptive element is inserted late in the derivation, as a result of movement. Grohmann (2003) for instance argues that the resumptive element is a reflex of the movement (by Copy) of the dislocated element, which he argues is triggered to avoid anti-locality effects. Another solution is proposed by López (2003), who argues that the resumptive clitic is a reflex of the pragmatic feature that triggers movement of a topic (see below). The implication of these approaches is that some kind of (possibly

4

The doubling analysis proposed by Kayne (1994) will be discussed in Section 4.3.7.

Syntax

103

covert) movement must have taken place whenever a clitic is present. 5 In the case of López (2003), the prediction is that clitics are always interpreted as topics. 6 Alternatively, the dislocated element has been argued to move from a nonargumental position, such as the specifier of a CliticP (Zubizarreta 1998) or the specifier of a big DP headed by the resumptive clitic and generated in an argument position of the verb (Cecchetto 1999, 2000). This big DP then raises to [spec,AgroP] and then to the specifier of a Topic projection in the extended VP field. The topic is argued to remain in this position while the clitic climbs further up the structure. Note however that, as Cecchetto (1999) himself points out, the latter option is only viable for dislocated DPs. Cecchetto (1999 fn. 22) assumes that dislocated PPs are generated directly in an IP-peripheral topic position. A third option is to limit the movement of dislocated elements to cases where they need to reach the periphery of a clause higher than the one containing their resumptive element (Iatridou 1995). Frascarelli (2002) argues that this is triggered by ECM requirements in the lower clause. 4.1.2.2 The resumptive moves Rizzi’s (1997) solution to account for the relevant distribution and connectivity facts is to generate dislocated topics in the specifier of a TopicP and to assume that the resumptive element moves to that projection at LF. He argues that the resumptive is an anaphoric operator which binds a null epithet or null constant in the argument position and thereby establishes the connection between the topic and the open position in the comment. Island sensitivity is argued to follow from restrictions on the LFmovement of the operator/resumptive. LF-movement of the resumptive element has also been argued for by Demirdache (1991) and Anagnostopoulou (1997), albeit within an adjunction approach. For Anagnostopoulou (1997), the resumptive mediates the relation between the base-generated dislocated element and the gap in argument position. The resumptive is the head of a predicate variable chain, while the fronted XP is the subject of the predication. 4.1.2.3 Nothing moves in narrow syntax Cinque (1990) takes a different approach to the problem of island sensitivity and connectedness and argues 5 I assume that these authors do not postulate a different type of licensing for clitics in the absence of a coreferential element at the periphery of the clause. 6 This prediction is possibly controversial as it entails that thetic (i.e. all-focus) sentences cannot contain clitics. A possible solution is proposed by Erteschik-Shir (1997), who argues that clitics are subordinate topics, and that this allows them to appear in thetic structures.

104

French Dislocation

that these two properties are not indicators of movement (at that time, Move Alpha). Rather, he claims, they are the property of chains, whether derived by movement or not. He argues that CLLD involves a representational binding chain, not a derivational one. The property that CLLD lacks, and which distinguishes it from its movement counterpart (English-style Topicalization, 7 illustrated in (4.5)), is the ability to enter into government chains. At that point in the theory’s development, this required a (possibly null) operator binding a gap. Cinque argues that because it features such an operator Topicalization has quantificational force, but CLLD does not (a fact also pointed out by Rizzi (1997)). cercato, non Piero. (4.5) GIANNI, (∗ l’)ho Gianni him-I-have looked-for not Piero ‘I have looked for Gianni, not Piero.’

(Cinque 1990: 63)

This proposal explains why a topicalized constituent can license a parasitic gap, but a CLLD constituent cannot (4.6) (4.6)

∗

mesi, senza trovare Giannii , l’i ho cercato per Gianni I him have looked-for for months without ei . (Cinque 1990: 62) finding

More recent versions of the no-movement-in-syntax approach have been proposed for CLLD (Aoun and Benmamoun 1998) and for CLRD (Kayne 1994). Both these analyses posit that dislocated elements do not move in narrow syntax but rather that movement can take place at PF or at LF. Aoun and Benmamoun (1998) argue that Lebanese Arabic has two types of CLLD: one is base-generated in its clause-peripheral position, and one is moved there at PF. As reconstruction is only possible when movement has taken place, Aoun and Benmamoun argue that the dislocated element in CLLD can only be interpreted in its reconstructed position when it has been derived by (PF) movement. Hence whenever a dislocated element is separated from its resumptive by an island (indicating absence of movement, as in (4.7a)), reconstruction will not be possible. By contrast, in the absence of islands (4.7b), Aoun and Benmamoun (1998) argue that movement has taken place and that the dislocated element is interpreted via reconstruction. 7 Cinque points out a core interpretive difference between Topicalization and dislocation: the former is a focus construction in Italian (see Cinque 1990: 62, n. 11) and the latter a topic construction.

Syntax

105

∗

[T@lmiiz-ai SSitaan] j , fallayto Pablma k@ll mQallmei student-her the-naughty MS you-left before every teacher f Paas.as.@t-o j . (Aoun and Benmamoun 1998) she-punished-him ‘Her naughty student, you left before every teacher punished him.’ b. [T@lmiiz-ai SSitaan] j , btaQrfo P@nno k@ll mQallmei student-her the-naughty MS you-know that every teacher f Paas.as.@t-o j . (Aoun and Benmamoun 1998) she-punished-him ‘Her naughty student, you know that every teacher punished him.’

(4.7) a.

Aoun and Benmamoun (1998) propose that PF-moved CLLD (4.7b) does not involve a pro (coindexed with the dislocated NP) in its numeration. The NP is said to originate in the specifier position of its resumptive’s CliticP. It is forced to leave this position by the Doubly-Filled Specifier/Head Filter (which Aoun and Benmamoun suggest should therefore be understood as a condition on PF well-formedness). Base-generated CLLD (4.7a), on the other hand, is argued to be merged directly in its clause-peripheral position and has a numeration which features a pro. In Section 4.3.4, I will show that French CLLD never displays reconstruction, which rules out a PF-movement analysis à la Aoun and Benmamoun. Kayne (1994) proposes a different post-Spell-Out-movement analysis to account for right-dislocated structures. One of his main claims is that rightadjunction is prohibited by UG. Given this, Kayne proposes that rightdislocated elements in Romance languages in fact occupy object position, with clitic doubling and dislocation intonation. Kayne argues that left-dislocated phrases are situated in the specifier of a left-peripheral functional projection to which they are moved from an argument position. He proposes that right-dislocation is an instance of CLLD at LF. Right-dislocation intonation is argued to be triggered by an optional feature in the ‘overt syntax’ that would feed both LF (triggering movement) and PF (triggering dislocation intonation). In Section 4.3.7, I will show that the predictions of Kayne’s (1994) analysis are not verified by the French data, ruling out an LF analysis along the lines he proposes. 4.1.2.4 Some move, some don’t In the generative literature of the 1970s, the debate on dislocated structures revolved essentially around the question of whether the dislocated element reached its peripheral position via movement

106

French Dislocation

or base-generation (the only two options available in the theory at the time). Ross’s (1967) original analysis of Left Dislocation (LD) as a movement rule had been challenged by demonstrations of the fact that left-dislocated structures were not subject to movement constraints in a range of languages including French and Italian (Postal 1970; Gundel 1974, 1975; Hirschbühler 1974, 1975; Rodman 1974; van Riemsdijk and Zwarts 1974). In answer, Cinque (1977) argued that two types of left-dislocated construction needed to be distinguished: a base-generated one and one derived by movement. He proposed to label the base-generated version of LD hanging topic (a term apparently coined by A. Grosu), which he defined as ‘a construction that mainly serves to promote an NP to topic status at a point in the discourse when it was not a topic’ (Cinque 1977: 406). The essential difference between the two constructions was shown to be that LD displayed signs of connectedness only when it was derived by movement. Connectedness is the overt manifestation of a close relation between a dislocated element and the empty argument position with which it is associated. It has been argued to manifest itself in the following ways (Cinque 1977, 1983, 1990; Vat 1981; Iatridou 1995). 8 (4.8) The dislocated element: a. contains a preposition ‘governed’ by an element in the core of the clause; hanno scritto, una ragione (i) Se [a Giorgio]i loro non glii if to Giorgio they not to-him have written a reason c’è. there-is ‘If they haven’t written to Giorgio, there is a reason.’ (Cinque 1983:(22)) b. bears a case assigned by the verb; (i) Ton Kosta, i Maria ton the Kosta-ACC the Mary-NOM him-ACC idhe. (Iatridou 1995:(1)) saw ‘Maria saw Kosta.’ c. is part of an idiom (the other part not being dislocated); (i) Le cuoia, le tirerai prima tu, bello mio! the leathers them you-will-draw earlier you beautiful mine ‘YOU will die earlier, my dear!’ (Cinque 1977:(18)) 8 The term syntactic connectedness was coined by Higgins (1973), and applied to left dislocation for the first time by Vat (1981).

Syntax

107

d. gets reconstructed at LF. (i) A suoi figlio, credo che [ognunoi di loro] finirà of them will-end-up to his son I-think that each per lasciargli un appartamento. by leaving an apartment ‘I think that each of them will end up leaving his son an apartment.’ (Cinque 1983:(33)) The distinctive formal properties of Hanging Topic Left Dislocation (henceforth HTLD) have been argued to be the following. (4.9) In HTLD, the dislocated element: a. must be an NP (or a DP) 9 (Cinque 1983), b. has to precede CLLD (Cinque 1977; Benincà 2001; Delais-Roussarie et al. 2004), c. can be separated from its resumptive by an island boundary (Cinque 1977), d. can have a non-clitic resumptive (Cinque 1983), e. cannot occur in embedded clauses (Cinque 1983), f. is unique in the sentence (HTLD cannot be iterated) (Cinque 1983), g. is separated from the clause by a longer pause than is the case in CLLD. By contrast, movement-derived left-dislocation has been defined by the formal properties in (4.10). The Romance instantiation of this construction is usually labelled Clitic Left Dislocation (CLLD), following Vat (1981), because the dislocated element is resumed by a clitic. (4.10) In CLLD, the dislocated element: a. cannot be resumed by a strong pronoun (Cinque 1977, 1983), b. can occur in embedded clauses (Cinque 1983), c. cannot have a clefted resumptive (Cinque 1977). ont parlé. (i) (∗ De) Marie, c’est d’elle qu’ils of Marie it-is of-her that-they have talked ‘It is of Mary that they have talked.’ Interpretive differences have also been claimed to distinguish HTLD from CLLD, as summarized in (4.11) and (4.12).

9

It can be a PP if it is clearly not subcategorized by the predicate (Cinque 1983: n. 5).

108

French Dislocation

(4.11)

a. The left-dislocated phrase in HTLD encodes a new or unexpected topic. b. The topic in HTLD is a discourse topic, not a sentence topic (Cinque 1983). (4.12) a. The left-dislocated phrase in CLLD encodes given information (Cinque 1983). (i) It cannot be contrastive (Cinque 1983). (ii) It can be contrastive. (iii) It is always contrastive (López 2003). b. The topic in CLLD is a sentence topic, not a discourse topic. There is however considerable disagreement in the literature as to the precise characteristics of each construction. Benincà (2001) and Delais-Roussarie et al. (2004), for instance, claim that the pragmatics of HTLD and CLLD is the same (at least in Italian and French respectively). Villalba (2000) claims (with respect to Catalan) that HTLD is used when the topic it introduces is in contrast with the previous topic, and that generic NPs are ‘odd’ in HTLD. This restriction does not apply to spoken French, as illustrated in (4.13), which is clearly acceptable with a habitual interpretation. (4.13)

parlait pas, dans mon Les nouveaux-venusi , on ne leuri one neg to-them would-talk not in my the new-come village. village ‘One wouldn’t talk to newcomers in my village.’

In spite of these alleged differences, Cinque (1977: 43) points out that ‘in many cases, it is not simple to decide what construction we are observing’. In languages like Italian and French, for instance, nominative and accusative case are not overtly marked on DPs. As a consequence, a dislocated phrase associated with a subject or a direct object can never overtly manifest connectedness, and thus HTLD and CLLD are indistinguishable in such cases (Cinque 1977). This issue will be discussed with respect to spoken French in Section 4.3.6.

4.2 Dislocated topics in spoken French: an overview 4.2.1 Clause-peripheral topics In French, left-dislocated elements can be resumed by elements that are not clitics. Such LDs are thus by definition not examples of CLLD. However, there is no principled reason to distinguish CLLD from other types of LD in spoken French: the nature of the resumptive element does not affect the syntactic

Syntax

109

or interpretive properties of LD in that language, as demonstrated below. 10 Examples of non-clitic LDs are given in (4.14). (4.14) a. Claasi , sesi chaussettes ont disparu. Claas his socks have disappeared ‘Claas’ socks have disappeared.’ plus jamais entendu parler de luii . b. Kambii , je n’ai Kambi I neg-have not ever heard to-talk of him ‘I never heard anything about Kambi again.’ c. Le laiti , j’adore çai . the milk I-adore that ‘I love milk.’ The dislocated element expresses the topic of the sentence just as it does in CLLD: the sentence is interpreted as being about the referent of the dislocated element and is evaluated with respect to that referent. Resumption by an epithet is also possible (see Hirschbühler 1975), though rarely used in spontaneous speech. (4.15)

tous les disques de ce farfelui . Plastic Bertrandi , j’ai Plastic Bertrand I-have all the records of this weirdo ‘I have all of Plastic Bertrand’s records.’

In (4.14) and (4.15), the dislocated element expresses the topic of the sentence just as it does in CLLD: the sentence is interpreted as being about the referent of the dislocated element and is evaluated with respect to that referent. The examples below show that non-clitic LDs behave similarly to CLLD (the properties of which were presented in the previous section). First, observe that non-clitic LDs are not sensitive to islands (CLLD’s insensitivity to islands will be demonstrated in Section 4.3.5). pris [une photo [de sesi chaussettes]]. (4.16) a. Claasi , j’ai Claas I-have taken a photo of his socks ‘I’ve taken a picture of Claas’ socks.’ b. Kambii , je me souviens [du banc [où je m’asseyais Kambi I refl remember of-the bench where I refl-sat avec luii ]]. with him ‘I remember the bench where I sat with Kambi.’ 10

There are no intrinsic prosodic differences between a LD that is resumed by a clitic element and one that is not.

110

French Dislocation c. Le laiti , il vaut mieux avoir [un frigo [pour conserver the milk it is-worth better to-have a fridge to conserve çai en été]]. that in summer ‘It’s best to have a fridge to keep milk in summer.’

Second, like CLLD, multiple instances of non-clitic LD are allowed. (4.17)

jamais ça j dans a. Claasi , du contre-plaqué j , tu verrais Claas some plywood you would-see never that in sai maison. his house ‘You’d never see plywood in Claas’ house.’ b. Kambii , l’école j , ça j ne luii allait pas trop. Kambi the-school it neg to-him went not too-much ‘Kambi didn’t like school very much.’ c. Le laiti , Steph j , çai ne va pas avec son j estomac. the milk Steph it neg goes not with his stomach ‘Milk doesn’t agree with Steph.’

Third, LD can always have a right-hand counterpart, whether it is resumed by a clitic or by a non-clitic. 11 Examples of CLRD are given in (4.18), and RD in (4.19). (4.18) a. Tu ne lui parles plus, à ta poupée? you neg to-her talks no-more to your doll ‘You don’t talk to your doll anymore?’ b. Il y en a plein dans le jardin, des guêpes. it there of-it is lots in the garden of wasps ‘There are lots of wasps in the garden.’ à Claasi . (4.19) a. Sesi chaussettes ont disparu, his socks have disappeared to Claas ‘Class’ socks have disappeared.’ b. Je n’ai plus jamais entendu parler de luii , Kambii . I neg-have not ever heard talk of him Kambi ‘I never heard anything about Kambi again.’ c. J’adore çai , le laiti . I-adore that the milk ‘I love milk.’ 11

Right-dislocation will be discussed in detail in Section 4.3.7.

Syntax

111

Fourth, CLLD (4.20) and LD (4.21) can appear in embedded contexts. (4.20) a. Je ne savais pas que les cochonsi , ilsi avaient des they had indef I neg knew not that the pigs salles de bain. bathrooms ‘I didn’t know pigs had bathrooms.’ trouverait à b. Elle pensait que des magasinsi , elle eni she of-them would-find at she thought that indef shops chaque coin de rue. each corner of street ‘She thought she’d find shops everywhere.’ (4.21) a. Il paraît que Claasi , on a caché sesi chaussettes. it appears that Claas one has hidden his socks ‘It appears that Claas’ socks have been hidden.’ b. Tu te souviens que Kambii , tout le monde voulait you refl remember that Kambi all the people wanted toujours danser avec luii ? always to-dance with him ‘Do you remember how everybody always wanted to dance with Kambi?’ c. Je pense pas que la bièrei , çai soit très bon pour le foie. I think not that the beer it be very good for the liver ‘I don’t think beer is very good for the liver.’ 4.2.2 Conclusion French LD is a unified phenomenon in that the nature of the resumptive element does not alter its essential properties. In all cases, the left-dislocated element expresses the topic of the sentence, can be resumed by an element inside an island, can appear in embedded clauses, is recursive, and is preferably stripped of dependency markers (such as prepositions), a feature I come back to in Section 4.4.4. Supporting evidence for these claims is provided in the remainder of this chapter. 4.2.3 Caveat French Dislocation is quite an ambitious title for a monograph, given the magnitude of the field of enquiry. More specifically this work investigates clause-peripheral XPs expressing the topic in spoken French. This excludes

112

French Dislocation

clause-peripheral adverbs that cannot be interpreted as stage topics and certain right-dislocated PPs headed by de ‘of ’. 12 I expand on the latter below. Spoken French features a particular kind of right-dislocation that involves de ‘of ’ + a bare N and disambiguates a DP by naming the set from which one (or a subset of) element(s) is singled out in the predication. Typical examples are given in (4.22). (4.22) a. J’en ai vu un, de cheval noir. (Larsson 1979:(64)) I-of-it have seen one of horse black ‘I have seen a black horse.’ b. On ne trouve rien dans celle-là, de boutique. one neg finds nothing in that-one-there of shop ‘You never find anything in that shop.’ c. Elle a la plus grande, de chambre. she has the most big of room ‘She’s got the biggest room.’ The type of right-dislocation exemplified in (4.22) (and in particular (4.22a)) has been discussed extensively in the literature (e.g. Kayne 1975; Milner 1978; Ruwet 1990; Lamiroy 1991; Emonds 2001; Heyd and Mathieu 2005). I will show below that in this construction the dependency between the dislocated element and the DP it is associated with is stronger than that between a ‘standard’ right-dislocated element and its resumptive. This will be argued to indicate that the dislocated PP is not adjoined to a clausal projection, and that it has to appear in close proximity to its resumptive element (hence the label ‘very local RD’). Furthermore, this type of right-dislocation will be shown not to receive a topic interpretation. As a consequence, it deserves a different account to the one proposed in this book, which only applies to clause-peripheral XPs expressing the topic in spoken French. The characteristics which distinguish ‘very local RD’ from other rightdislocated PPs (such as those in (4.24) below) are listed in (4.23). (4.23) In ‘very local RD’: a. the non-dislocated equivalent of the dislocated PP is part of a DP, b. the dislocated PP contains a bare noun, c. the dislocated PP is resumed by en ‘of ’ only when indefinite or partitive, d. the dislocated PP appears strictly to the right of the element it modifies, 12

head.

It is unclear whether de is really a preposition. Hulk (1996a) argues that it is a quantificational

Syntax

113

e. the dislocated PP has a close syntactic relation with the element it modifies. The above properties do not hold for ‘standard’ dislocated PPs (illustrated in (4.24)), as will be demonstrated below. I will now address each property listed in (4.23) in turn. (4.24) a. On en sort vivant, de la prison. one from-it goes-out alive of the prison ‘One comes out of prison alive’. b. J’en ai parlé avec Admunsen, de son expédition. I-of-it have talked with Admunsen of his expedition ‘I talked with Admunsen about his expedition’. c. Ses cheveux, à ma poupée, ils sont sales. her hair to my doll they are dirty ‘My doll’s hair is dirty’. Firstly, in the non-dislocated equivalent of very local RD, the PP corresponds to part of a DP (4.23a). In some cases, it modifies a noun. 13 This is shown in (4.25), where the material in bold corresponds to the dislocated PP (indicated by the comma). (4.25) a. la première partie du roman → en . . . la première of-it the first the first part of-the novel partie, du roman part of-the novel ‘the first part of the novel’ b. les portes de l’immeuble → en . . . les portes, de of-it the doors of the doors of the building l’immeuble the building ‘the doors of the building’ . . . peu, de clients c. peu de clients → en of-them few of clients few of clients ‘few clients’ d. un cheval noir → en . . . un, de cheval noir of-it one of horse black a horse black ‘a black horse’ 13 The dislocated PP and its resumptive en have been argued to be genitives (see e.g. Larsson 1979; Emonds 2001). Larsson points out that in Swedish the dislocated element bears overt genitive case in this construction.

114

French Dislocation

Alternatively, what is dislocated is the head of the DP, excluding its modifiers. (4.26) a. un cheval noir → en . . . un noir, de cheval of-it a black of horse a horse black ‘a black horse’ b. la plus grande chambre → (∗ en) . . . la plus grande, de the most large room of-it the most large of chambre room ‘the largest room’ c. cette boutique-là → (∗ en) . . . celle-là, de boutique of-it that-one of shop that shop-there ‘that shop’ This first property does not apply to ‘standard’ dislocated PPs (4.24): either the non-dislocated equivalent is selected by the verb, as in (4.27a) and (4.27b), or there is no exact non-dislocated equivalent, as in (4.27c). 14 (4.27) a. On sort vivant de cette grotte. one goes-out alive of that grotto b. J’ai parlé de son expédition avec Admunsen. I-have talked about his expedition with Admunsen c. [Les/∗ Ses] cheveux [de/∗ à] ma poupée sont sales. the/her hair to/of my doll are dirty Secondly, the PP in very local RD contains only a preposition and a bare noun (4.23b), and the preposition is obligatorily de ‘of ’. In ‘standard’ RD, the PP cannot contain a bare noun (4.28) 15 and the preposition does not have to be de ‘of ’ (4.24c).

14 The non-dislocated DP in (4.27c) is not an exact equivalent of the dislocated DP in (4.24c) because the non-dislocated version (i) cannot include a possessive determiner and (ii) requires a different preposition. 15 A reviewer correctly points out that standard RD can contain bare nouns in examples like (i).

(i)

Il en parle souvent, de livres qu’il a lus. he of-them talks often of books that-he has read ‘He often talks about books he’s read.’

Note however that if the noun modifier is removed leaving the noun entirely bare, the resulting sentence is ungrammatical. (ii)

∗ Il

en parle souvent, de livres. he of-them talks often of books

Syntax

115

(4.28) a. On en sort vivant, de ∗ (cette) grotte. one of-it goes-out alive of that grotto ∗ b. J’en ai parlé avec Admunsen, de (son) expédition. I-of-it have talked with Admunsen about his expedition Thirdly, the PP of very local RD can only be resumed by en (4.23c). Such resumption arises when the dislocated element is indefinite (as shown in (4.29)) or when it refers to a set or a whole out of which a subpart is relevant to the predication (as in (4.30)). In both cases, en receives a partitive reading. If a partitive reading is not available, resumption by en is impossible (see (4.26b) and (4.26c)). a peu, de clients. (4.29) Elle en/∗ les she of-them/them has few of clients ‘She has few clients.’ (4.30) J’en ai mangé la moitié, [de la tarte / des pommes]. I-of-it have eaten the half of the pie / of-the apples ‘I’ve eaten half of the pie/apples.’ With respect to ‘standard’ right-dislocated PPs, resumption with en does not force a partitive reading (see (4.28)). Fourthly, in ‘very-local RD’, the dislocated PP appears obligatorily to the right of the element it modifies (4.23d) as shown in (4.31)). 16 By contrast, in standard RD, there are no such ordering restrictions, as shown in (4.32). Removing the preposition renders the LD equivalent of standard RD more acceptable, but it does not improve the LD equivalent of very local RD. (4.31)

16

(i)

a.

∗

(De) cheval noir, j’en ai vu un. of horse black I-of-it have seen one ∗ b. (De) chambre, elle a la plus grande. of room she has the most big

A reviewer suggests that this is not true, on the basis of the following example. D’école, il n’ en existe qu’une en Californie. of-school there neg of-them exists only-one in California ‘There’s only one school in California.’

This is puzzling given the clear ungrammaticality of (4.31), and calls for further investigation, which I will not undertake here. It may be that some of the criteria proposed above to identify very local RD are a bit too strong and should be adjusted so that they exclude only dislocated elements that are not adjoined at clause level and interpreted as topics. The decisive test might be the existence of a LD counterpart. This would automatically indicate that the locality requirement (4.23d) does not apply.

116

French Dislocation

(4.32) (À) ma poupée, ses cheveux, ils sont sales. to my doll her hair they are dirty ‘My doll’s hair is dirty.’ This strict ordering requirement on very local RD is due to its discourse function: it disambiguates the reference of a DP in the sentence, and disambiguation is a strictly linear process. Standard RD can also disambiguate the reference of a DP, but this is a secondary effect of its main function, which is to express the topic of the sentence. This explains why standard RD has a left-dislocated equivalent, as shown in (4.33), but very local RD does not, as shown in (4.34). (4.33) a. (À) ma poupée, ils sont sales, ses cheveux. to my doll they are dirty her hair ‘My doll’s hair is dirty.’ b. De la prison, on en sort vivant. Pas de la guerre. of the prison one of-it comes-out alive not of the war ‘One comes out alive from prison, not from war.’ (Céline, cited in Larsson 1979: 35) ∗ (4.34) (De) stylo, il coule tout le temps, le mien. of pen it runs all the time the mine Fifthly, very local RD has a close syntactic relation with the element it modifies (4.23e). They can be dislocated together and appear in the clause’s left periphery, but cannot be independently dislocated: they cannot be split, one appearing at each edge of the clause, as shown in (4.35). In standard RD, by contrast, the dislocated element can be separated from the element it modifies in such a way (4.36). 17 (4.35) a. Le mien, de stylo, il coule tout le temps. the mine of pen it runs all the time ‘My fountain pen runs all the time’. b. ∗ Le mien, il coule tout le temps, de stylo. the mine it runs all the time of pen 17 DP + PP do not form a single dislocated constituent in either case, as they cannot appear together in an argument position.

(i)

a. b.

∗ [Le

mien de stylo] coule tout le temps. the mine of pen runs all the time ∗ [Ses cheveux à ma poupée] sont sales. her hair to my doll are dirty

Note that there should be no intonation break between the DP and the PP in these examples.

Syntax

117

(4.36) a. Ses cheveux, à ma poupée, ils sont sales. her hair to my doll they are dirty b. Ses cheveux, ils sont sales, à ma poupée. her hair they are dirty to my doll ‘My doll’s hair is dirty.’ I will follow Emonds (2001), who argues that in cases like (4.35a), the PP is attached to a DP sister to V. For Emonds, the structure is the following: (4.37) [ VP eni V [ DP DP Øi ]], [ PPi ] The prediction is that the PP has to be adjacent to the constituent containing the verb and its complements (one of which is the DP with which the PP is associated). This prediction is verified: ∗

Il [ VP en a [ DP trois/ des tas]] dans chacune de ses he of-them has three/ some heaps in each of his maisons, de tableaux. houses of paintings mur], b. Il en a [ VP mis [ DP trois/ des tas] au put three/ some heaps on the wall he of-them has de tableaux. of paintings

(4.38) a.

‘He’s hung lots of paintings.’ The same does not apply to standard RD. parlais] dans la rue chaque matin, au Baron (4.39) Je [ VP luii to-him talked in the street every morning to-the Baron I Rothschildi . Rothschild ‘I talked to Baron Rothschild every morning in the street.’ Heyd and Mathieu (2005) analyse [de N] structures in French as scopeless nominals which are semantically incorporated in that the noun does not denote an individual but a property. The [de N] in ‘very local RD’ cannot therefore encode the topic of the sentence as this would require an individual interpretation (at least in non-generic sentences—see Chapter 3). 18 Owing to this and the fact that it cannot be separated from the element it modifies/disambiguates, very local RD is excluded from the present field of inquiry. 18 A reviewer proposes that very local RD is a subordinate topic in the sense of Erteschik-Shir (1997). What is relevant to the present purpose is that it is not a ‘main’ topic.

118

French Dislocation

4.3 French dislocation is not generated by movement Diagnostics for movement traditionally considered relevant to CLLD include Weak Crossover effects, the licensing of parasitic gaps, Relativized Minimality effects, reconstruction effects, and sensitivity to (strong) islands. In the following sections, these diagnostics are applied in turn to the spoken French data, and shown to indicate consistently that movement is not involved in the derivation of French dislocation. Island sensitivity is discussed in more detail in Section 4.3.5, as it is the most controversial diagnostic. 19 4.3.1 French LD does not yield Weak Crossover effects Examples of the type in (4.40) have been cited in the literature as evidence that CLLD configurations do not induce Weak Crossover (WCO) effects (see e.g. Iatridou 1995). The argument goes that, in such configurations, a pronoun can freely intervene between its A -binder (here Abélard) and the element at the foot of the ‘chain’ (here the resumptive element l’ ‘him’). 20 (4.40) Abélardi , sai mère l’i aimait trop. Abélard his mother him loved too-much ‘Abélard’s mother loved him too much.’ The same is true when the resumptive element is not a clitic. (4.41) Abélardi , on sait que soni élève passait des heures délicieuses Abélard one knows that his pupil spent some hours delicious avec luii . with him ‘It’s well known that Abélard’s pupil spent delightful hours with him.’ 19 All the data in this section is from spoken French, where dislocations are used pervasively. The way information structure is encoded in written French is quite different (see e.g. Lahousse 2003), and I will not have any specific claims to make about it in this book. 20 Contrary to what Zubizarreta (1998) observed in Spanish, even long-distance CLLD does not yield WCO effects in spoken French. A topic in the matrix clause can be associated with a resumptive in an embedded clause and bind a pronominal element in the matrix clause.

(i)

Abélardi , sai femme savait qu’Héloïse passait des heures délicieuses avec luii . Abélard his wife knew that-Héloïse spent some hours delicious with him ‘Abélard’s wife knew that Héloïse spent delightful hours with him.’ (ii) Abélardi , sai mère savait bien qu’Héloïse l’i adorait. Abélard his mother knew well that-Héloïse him adored ‘Abélard’s mother knew full well that Héloïse adored him.’

Syntax

119

Absence of WCO effects is exactly what is expected on a base-generation analysis of dislocation. 21 However, WCO effects are only expected to arise in configurations involving the binding of a variable by a quantifier (Lasnik and Stowell 1991). 22 Given the intrinsically non-quantificational nature of dislocated topics (see e.g. in Rizzi 1997), the absence of WCO in dislocated constructions does not constitute independent evidence for the absence of movement. Indeed, the existence of sentences like (4.42a) indicates that the coindexing of the so-called resumptive and the topic does not entail that the ‘resumptive’ be syntactically bound by its antecedent. Such coindexing is equally possible across sentences, as shown in (4.42b). (4.42) a. Abélardi , sai mère était une sainte. Abélard his mother was a saint ‘Abélard’s mother was a saint.’ b. J’ai bien connu Abélardi . Sai mère était une sainte. I-have well known Abélard his mother was a saint ‘I knew Abélard well. His mother was a saint.’ Rizzi (1997) claims that the absence of WCO effects in CLLD can be accounted for in terms of non-operator A -movement. On the basis of the evidence presented below, which uniformly points out to the absence of movement effects from dislocated configurations in spoken French, I will argue that no movement (even non-operator A -movement) applies in the case of French CLLD. Significantly, I will show in Section 4.3.4 that Strong Crossover effects are not attested either, which rules out A -movement altogether. 4.3.2 French LD does not license parasitic gaps French is more restrictive than English with respect to the configurations in which parasitic gaps can be licensed (Tellier 2001). Examples of the parasitic gap construction in French are given in (4.43) (all these examples are from Tellier 2001). The most deeply embedded verb is obligatorily interpreted as transitive in sentences (4.43a) and (4.43c). Following Tellier, the t is the ‘real’ gap and e is the parasitic gap.

21 Strictly speaking there is no ‘base’ anymore, under current assumptions of the Minimalist Programme. I will however retain the label ‘base-generation’ so as not to confuse the reader with the introduction of non-standard terminology. 22 Thanks to Kyle Johnson for pointing out this issue to me.

120

French Dislocation

(4.43) a. Voilà les livres que tu as déchirés t au lieu de presentative the books that you have torn in place of consulter e. consult ‘These are the books that you tore up instead of consulting.’ b. Un homme dont l’honnêteté t se voit dans les yeux e. a man of-who the-honesty refl sees in the eyes ‘A man whose honesty shows in his eyes.’ c. C’est le genre de plat que tu dois cuire t avant it-is the kind of dish that you must cook before of de consommer e. eat ‘It’s the kind of dish that you have to cook before eating.’ The left-dislocated constructions below are comparable to the examples above, but parasitic gaps are not possible in the former: the presence of a resumptive pronoun in the location of the would-be parasitic gap is obligatory. (4.44) a. Les livresi , tu lesi as déchirés au lieu de in place of the books you them have torn ∗ (lesi ) consulter. them to-consult ‘You have torn the books instead of consulting them.’ b. [Un homme comme ça]i , soni honnêteté se voit dans that his honesty refl sees in a man like ∗ [sesi / les] yeux. his/the eyes ‘You see the honesty of a man like that in his eyes.’ c. [Ce genre de plat]i , tu dois lei cuire avant de ∗ (lei ) this kind of dish you must it cook before to it consommer. eat ‘You have to cook that kind of dish before eating it.’ Again, this is unexpected under a movement analysis of French dislocation. 4.3.3 No Relativized Minimality effects If movement is involved in the derivation of left-dislocated elements, Relativized Minimality effects should arise when a dislocated XP intervenes

Syntax

121

between another dislocated XP and its resumptive element. The examples in (4.45) show that dislocated subjects and objects do not disrupt each other’s chains. du bien. (4.45) a. La pluiei , ta salade j , ellei lui j fera the rain your lettuce it to-it will-do some good ‘The rain will do your lettuce some good.’ b. Cette toile j , je pense que Juliai , ellei ne l’ j a pas vendue. that canvas I think that Julia she neg it has not sold ‘I think Julia didn’t sell that picture.’ It is even possible to ‘intertwine’ two topic chains associated with the same grammatical role. dit qu’il j était charmant. (4.46) Rosii , mon père j , ellei m’a Rosi my father she to-me-has said that-he was charming ‘Rosi told me my father is charming.’ I conclude that Minimality does not constrain topic chains, which corroborates a base-generation analysis. 4.3.4 No reconstruction effects in the interpretation of French LD One of the main arguments for a movement analysis of CLLD in various languages (aside from sensitivity to islands) has been reconstruction effects in the interpretation of dislocated elements (see e.g. Zubizarreta 1998; Cecchetto 1999; Frascarelli 2000; Villalba 2000). Support for a movement analysis is found when dislocated elements are interpreted as if they occupied the argument position with which they are associated. A number of facts suggest that in French a dislocated element is not interpreted in its reconstructed position: (i) a dislocated element cannot be bound by a quantifier in subject position; (ii) no Principle C effects are observed; (iii) dislocated elements obligatorily take wide scope with respect to clausal negation; and (iv) when a dislocated element contains a variable, native speakers by default search for a binder in the context rather than in the sentence. 4.3.4.1 A variable in a left-dislocated XP cannot be bound by a clause-mate QP Consider the sentence in (4.47). The variable which is part of the possessive determiner contained within the object can be bound by the universal quantifier in subject position. A distributive interpretation of this sentence is therefore possible.

122

French Dislocation

(4.47) [Chaque maître]i a renvoyé un de sesi /x disciples. master has dismissed one of his disciples each ‘Each master dismissed one of his (own) disciples.’ If the object is dislocated, as in (4.48), the distributive reading is lost. The only possible interpretation of this sentence is one in which the possessor corresponds to a referent identified in the discourse context (represented below by the index x), and not to the subject of the sentence. (4.48) [Un de sesx/∗i disciples] j , [chaque maître]i l’ j a renvoyé. disciples each master him has dismissed one of his ‘Each master dismissed one of his disciples.’ This contrast indicates that a left-dislocated element is not interpreted in its reconstructed position (and presumably that quantifier raising targets a position that is lower than the dislocated element). 4.3.4.2 Absence of Principle C effects If dislocated elements were interpreted in their reconstructed position, one might expect Principle C effects to arise in cases like (4.49). However, such sentences are clearly acceptable, which shows that reconstruction is at least not obligatory in such contexts in French. (Translations are given without dislocation to highlight the intended meaning while still providing a grammatical English sentence). (4.49) a. Tes sales petites remarques sur Léoni , ili ne les your dirty little remarks on Leon he neg them apprécierait sûrement pas. would-appreciate surely not ‘Leon would surely not appreciate your dirty little remarks about him.’ b. Le dernier livre que j’ai prêté à ma sœuri , ellei l’a the last book that I-have lent to my sister she it-has lu en une nuit. read in one night ‘The last book I lent her, my sister read in one night.’ c. [La meilleure toile d’un artiste j ]i , on ne lui j one neg to-him the best canvas of-an artist, pardonnerait jamais de l’i avoir vendue trop tôt. would-forgive never to it have sold too early ‘One would never forgive an artist for selling his best work too early.’

Syntax

123

Cecchetto (1999, 2000) argues that in Italian, Principle C effects are cancelled in two cases because the potential binder (i.e. the ‘resumptive’ pronoun) does not c-command the reconstruction site. In the first case, a dislocated element is coreferential with a postverbal subject. Cecchetto (2000) explains the lack of Principle C effects by proposing that left-dislocated elements are reconstructed in a position that is c-commanded by [spec,AgrSP] (which hosts null subjects) but not by [spec,VP] (which hosts postverbal subjects). This explanation of anti-reconstruction effects does not hold for French, though: pronominal subjects never occupy a position lower than the canonical subject position in this language, so the alleged reconstruction site for dislocated elements which are coreferential with the subject is always c-commanded by the subject, even under an analysis à la Cecchetto (2000). The second case of anti-reconstruction predicted by Cecchetto occurs when the R-expression is in an adjunct (as opposed to a complement) inside the dislocated element (which he explains by the late insertion of adjuncts). In French, Principle C effects are absent even when the R-expression is not contained inside an adjunct, as shown in the examples above. The absence of Principle C effects in French thus remains unexplained on Cecchetto’s account. Finally, native speakers tend to disallow coreference between the Rexpression Elias and its resumptive element in (4.50b) but not in (4.50a). This is unexpected under a reconstruction analysis. parle plus. (4.50) a. (A) Eliasi , sai sœur ne luii (to) Elias his sister neg to-him speaks no-more à Eliasi . b. Sa j/#i soeur ne parle plus his sister neg speaks no-more to Elias ‘Elias’ sister doesn’t speak to him anymore.’ A movement analysis of French (CL)LD simply makes too many wrong predictions with respect to Principle C effects.

4.3.4.3 Wide scope with respect to negation If dislocated elements were (or could be) interpreted in their reconstructed position, it should be possible for them to get a narrow-scope reading with respect to sentential negation. However, this is not a possiblity for French dislocated elements, as illustrated in (4.51). (4.51)

elle ne lesi a pas vendues. Toutes ces toilesi -là, all those canvases-there she neg them has not sold

124

French Dislocation ‘She didn’t sell any of (all of) those pictures.’ # ‘She didn’t sell some of those pictures.’

This sentence is false in a context where the woman in question sold some but not all of the pictures under discussion. A follow-up saying that she sold half of them would be anomalous. The dislocated quantified phrase thus cannot be interpreted in a reconstructed position within the scope of the sentential negation. 23 4.3.4.4 Interpretation of variables When interpreting a sentence which contains a potential binder for a variable inside a dislocated element, the overwhelming majority of native speakers choose to associate the variable with a referent in the discourse context rather than with the sentence-internal binder. This was tested by presenting sentences like (4.52) to native speakers out of context in order to maximize the chances of sentence-internal binding. je connais l’homme qui l’i a emmenée. (4.52) Sa fillei , his daughter I know the-man who her-has taken-away ‘I know the man who took his daughter away.’ Ninety-four per cent of my informants (i.e. twenty-eight out of thirty-two speakers from Belgium, Canada, France, and Switzerland) interpret sa fille ‘his daughter’ in (4.52) 24 as the daughter of a person other than the man mentioned in the sentence. The same obtains if the dislocated element appears in the right-periphery. This clear preference is unexpected if sa fille is interpreted in object position. If it were, sentence (4.52) would be ambiguous with respect to whether the female in question is the daughter of the man mentioned in the sentence or somebody else. 4.3.5 French LD is not sensitive to islands It is standardly assumed that syntactic movement plays a part in the relationship between two elements if that relationship cannot hold across (strong) 23 A reviewer suggests that the impossibility of dislocating Negative Polarity Items (NPIs) would constitute incontrovertible evidence that French dislocated elements do not reconstruct. However, dislocated NPIs are independently ruled out on interpretive grounds as they do not satisfy the referentiality requirement on topics. The unacceptability of (i) therefore cannot be attributed to basegeneration of the dislocated element alone.

(i)

∗ [Le

livre de personne]i , je ne l’i ai lu. the book of nobody I neg it have read

24 The possessor in sa fille could equally be translated as ‘her’ but I have ignored this in the text in order to highlight the possibility of interpreting sa fille as the daughter of the man who is mentioned in sentence (4.52).

Syntax

125

syntactic islands, which were originally defined by Ross (1967). This diagnostic has been granted precedence over other tests for movement (such as those considered above) in the abundant literature on CLLD (see e.g. DelaisRoussarie et al. 2004 for French). CLLD has been argued to be sensitive to strong islands in e.g. Italian (Rizzi 1997), Greek (Iatridou 1995), and Spanish (Escobar 1997). However, this is not the case in all languages. Lebanese Arabic has been argued to be an exception (Aoun and Benmamoun 1998) and, as I will argue below, the same is true of spoken French. 25 This is not a new claim (see e.g. Hirschbühler 1975; Zribi-Hertz 1984). What is new here though is that the information structure requirements on topics were fully considered, and the key data were evaluated experimentally by a large number of native speakers who took part in two judgement elicitation tasks. 4.3.5.1 Native speakers’ judgements In order to test the sensitivity of (CL)LD to strong islands, two judgement elicitation tasks were designed. Thirty-two native speakers of French from Belgium, Canada, France, and Switzerland took part in the first task, and seventy-five native speakers from Belgium, Canada, and France took part in the second. The second task was designed in the hope of acquiring additional evidence to corroborate the findings of the first one. In both tasks, a short context was provided for each sentence to ensure that the dislocated element was a likely topic for the test sentence. Informants were presented with four levels of acceptability, the English translations of which are given in (4.53). If an informant failed to chose anything from the pull-down menu a ‘no choice’ value was assigned and the token was discarded. (4.53) a. I could say that sentence. b. I could say that sentence but in another context. c. I could never say a sentence like that, but I know that other French speakers could. d. That sentence is too weird. No French speaker talks like that. The context for each test sentence was given in writing prior to the informant clicking on a link to hear the test sentence (which was not transcribed 25 To be more precise, Aoun and Benmamoun (1998) argue that Lebanese Arabic displays two types of CLLD: one insensitive to islands (which they analyse as base-generated) and one sensitive to islands (which they say involves syntactic movement). Alexopoulou et al. (2004) argue that only the latter is genuine CLLD, and that the former is in fact a Broad Subject construction. French CLLD is not amenable to a Broad Subject analysis though, given that Broad Subjects are not obligatorily interpreted as topics.

126

French Dislocation

as the language of enquiry is spoken French). Example test sentences are given in (4.54) for the first test (involving thirty-two informants), and in (4.56) for the second test (involving seventy-five informants). The level of acceptability of each sentence is given in parentheses: the first rating (in bold) indicates ungrammaticality (i.e. the proportion of informants who rejected the sentence as something nobody would say), the second rating indicates markedness (i.e. possible dialectal variation). The remainder corresponds to the percentage of informants who accepted the sentence as something they could say themselves (i.e. straightforward acceptability). Justification for this interpretation of the informants’ judgements is provided in Appendix C, where the empirical methods employed are explained in detail. I start with the results from the first test (thirty-two informants). (4.54) a. Les autresi , je vais attendre [avant de lesi the other-ones I will wait before to them relire]. (0%; 6%) to-reread ‘I will wait before reading the other ones again.’ b. Mais le juge j , çai a surpris tout le monde, [qu’elle l’ j but the judge it has surprised all the people that-she him ait invité]i . (0%; 19%) has invited ‘It surprised everybody that she should have invited the judge.’ c. Aux petitsi , je sais pas [ce [qu’elle leuri to-the little-ones I know not that that-she to-them lit]]. (0%; 25%) reads ‘I don’t know what she reads to the little ones.’ d. Ta mèrei , je ferai tout pour être parti [quand ellei your mother I will-do all to be gone when she viendra]. (3%; 28%) will-come ‘I will do all I can to have left by the time your mother comes.’ on va attendre [avant de leuri e. Aux autresi , before to to-them to-the other-ones one will wait (13%; 31%) parler]. to-speak ‘We’ll wait before speaking to the others.’

Syntax

127

The overall picture reveals that, for this randomly selected group of speakers, the relation between a dislocated element and a coreferential clitic holds across (and in spite of) the following types of islands: adjuncts (4.54a), (4.54d), (4.54e), moved XPs (4.54b), and complex NPs (4.54c). Compare the acceptability ratings of the examples above with those for whextraction across an island: 26 (4.55) À qui est-ce que tu ne sais pas [ce [qu’elle to whom is-it that you neg know not that that-she lit]]? reads

(41%; 19%)

Test sentences from the second elicitation task are given below (seventy-five informants): (4.56) a. Par exemple, elle raconte tout à son mari. Mais à by example she tells everything to her husband but to [plein de trucs [qu’elle ne luii a son psyi , il y a her analyst it is there lots of things that-she neg to-him has (0%; 23%) jamais dits]]. never said ‘For instance, she tells her husband everything. But there are lots of things she’s never told her analyst.’ b. Je n’ai rien lu encore de ton article. Mais de ton I neg-have nothing read yet of your paper but of your [j’avais le temps] et [j’eni ai lu la romani , hier, novel yesterday I-had the time and I-of-it have read the moitié]. (0%; 22%) half ‘I haven’t yet read any of your paper. But yesterday, I had the time and I read half of your novel.’ c. La pluiei , [j’ai espéré toute la semaine] mais [ellei n’est but she neg-is the rain I-have hoped all the week (3%; 23%) jamais venue]. never come ‘I hoped all week but the rain never came.’ 26 Note that this sentence was given in a context favouring a d-linked interpretation, which is supposed to alleviate island effects. In spite of this, the acceptability rating is still fairly low, and in any case it is significantly lower than any of the acceptability ratings for the dislocated constructions tested.

128

French Dislocation d. Ses devoirsi , [elle fait pas exprès] mais [elle lesi her homework-pl. she makes not purpose but she them oublie toujours]. (1%; 35%) forgets always ‘She doesn’t do it on purpose but she always forgets her homework.’ e. Ta voiturei , [j’économiserai] puis [je te your car I-will-save-up then I to-you (0%; 19%) l’i achèterai]. it-will-buy ‘I’ll save up and then I’ll buy your car.’

Contrast this with the clearly ungrammatical (4.57a), and with (4.57b) which is grammatical only in some dialects of French. 27 (4.57) a. Si tu veux, on ira à Rio voir le carnaval. (0%; 0%) if you want we will-go to Rio to-see the carnival ‘If you want, we’ll go to Rio to see the carnival.’ a b. Mais je ne sais pas qu’est-ce qu’elle leur but I neg know not what-is-it that-she to-them has (7%; 61%) dit. said ‘But I don’t know what she said to them.’ Example (4.56a) provides additional evidence that most native speakers (seventy-seven per cent) accept left-dislocated PPs whose resumptive is inside a complex NP (none judged it ungrammatical). Examples (4.56b)–(4.56e) demonstrate that dislocated elements can be associated with a resumptive element inside a coordinate structure (an issue which was not covered in the first task). Seventy-eight per cent of my informants accepted a dislocated PP associated with the object of the second conjunct (4.56b). The same structure but with a dislocated DP was accepted by seventy-five per cent of informants when the DP was associated with the subject of the second clause ((4.56c)), and by sixty-four per cent 28 or eightyone per cent of informants when the DP was associated with the object of the second clause (in (4.56d) and (4.56e) respectively). Some of the sentences above are similar in structure to examples presented as ungrammatical in Cinque (1977): compare (4.58) with (4.56a) (both of 27

See Appendix C for discussion. This lower rating may have been due to the absence of the negative particle ne in the first conjunct, as thirty-five per cent of informants considered this to be marked. 28

Syntax

129

which contain a left-dislocated PP whose resumptive is an indirect object inside a relative clause) 29 and compare (4.59) with (4.56b). penses jamais est (4.58) (∗ À) notre frère, le fait que tu n’y (to) our brother the fact that you neg-of-him think never is absurde. (Cinque 1977) absurd ‘The fact you never think of our brother is absurd.’ j’avais le temps et j’en ai lu la (4.59) (∗ De) ce livre, hier, (of) this book yesterday I-had the time and I-of-it have read the moitié. (Cinque 1977) half ‘Yesterday I had the time and I read half of this book.’ The elicitation tasks presented above demonstrate that, if care is taken to make sentences with these structures plausible in a given context, hardly any native speaker rejects them as ungrammatical, even if the dislocated element is a PP, contrary to what is often claimed (e.g. Cinque 1977; Delais-Roussarie et al. 2004). The vast majority of speakers (seventy-seven per cent) accept such structures as unmarked, and the rest consider them to be possible for speakers other than themselves. I conclude that French (CL)LD is insensitive to strong islands. But is this sufficient to justify a base-generation analysis? 4.3.5.2 To what extent are islands a diagnostic for movement? Ross (1967) originally identified islands as a constraint on Chop, not on Copy. Both were conceived as rules of syntactic movement. What distinguished them was that Chop left a gap in the moved element’s original position, while Copy left a resumptive pronoun behind. Islands were thus originally not a diagnostic for movement per se but a diagnostic for types of movement. 30 29 My informants rejected even the non-dislocated equivalent of (4.58), which they found too formal to be representative of spoken French, so a more plausible alternative was chosen for the test.

(i)

Le fait que tu ne penses pas à notre frère est absurde. the fact that you neg think not to our brother is absurd ‘The fact that you don’t think of our brother is absurd.’

30 I will leave aside Cinque’s (1990) proposal to view islands as a representational constraint on binding chains rather than a derivational constraint on syntactic movement. On that view, the (in)sensitivity of French CLLD to islands would not be an indication of whether movement is involved, and consequently only the diagnostics discussed in Sections 4.3.1–4.3.4 would be relevant in that respect. Delais-Roussarie et al. (2004) analyse French LD along the lines of Cinque (1990), arguing that left-dislocated PPs are the only clear cases of CLLD in that language, all other cases being ambiguous

130

French Dislocation

This idea has recently been revived and reinterpreted by Boeckx (2003), who argues that islands preclude agreement relations but not movement. Building on Cecchetto (2000), Boeckx postulates that resumptive pronouns head a bigDP, and that the moved XP is first-merged as the object of the resumptive element. 31 Boeckx proposes that chains induced by A -movement can be the product of two kinds of operations: either Match+Agree or Match alone. In the former case, illustrated in (4.60), the ϕ requirements of the moved XP have to be satisfied by an agreeing complementizer. In the latter case, illustrated in (4.61), such requirements are fulfilled by the resumptive pronoun. (4.60) An fear aL bhuail tú the man C-agr. struck you ‘The man that you struck’ (4.61) An fear aN bhuail tú é the man C-non-agr. struck you him ‘The man that you struck’

(Irish)

(Irish)

In French subject relatives and clefts (4.62a), no resumptive pronoun is left behind at the extraction site. That the complementizer (qui) should be marked for agreement (as argued by Rizzi 1990) is exactly what is predicted by Boeckx (2003). In (4.62b), by contrast, the presence of a resumptive element (il ‘he’) bleeds the requirement for an agreeing complementizer. Chains like that in (4.62a) are correctly predicted to be sensitive to islands, while chains like that in (4.62b) are correctly predicted not to be. (4.62) a. C’est cet homme-là qui a mangé le raisin vert. it-is that man-there who has eaten the grapes green ‘It’s that man who’s eaten the green grapes.’ between CLLD and HTLD. The extent to which it is possible to distinguish these two configurations in French will be discussed in Section 4.3.6. 31 Boeckx (2003) notes that this big-DP structure is reminiscent of that proposed by Sportiche (1988) for floated quantifiers. Direct evidence in support of a big-DP structure for floated quantifiers can be found in sentences like (i) and (ii), where the quantifier and the DP surface adjacent to each other. However, to my knowledge, there is no such evidence to support the big-DP hypothesis for resumptive elements. See Boeckx (2003: Chapter 2) for discussion.

(i)

[Tous les révolutionnaires] ne sont pas des terroristes. all the revolutionaries neg are not indef terrorists ‘Not all revolutionaries are terrorists.’

(ii) Les révolutionnaires ne sont pas tous des terroristes. the revolutionaries neg are not all indef terrorists ‘Revolutionaries are not all terrorists.’ Note also that according to Cecchetto’s proposal the moved XP is the specifier and not the object of the resumptive element.

Syntax

131

b. Cet homme-là, il a mangé le raisin vert. that man-there he has eaten the grapes green ‘That man has eaten the green grapes.’ What is not predicted is that in French, agreeing complementizers are only possible in subject extraction contexts, not in object extraction contexts (which are equally sensitive to islands). A way out might be to postulate that qui is not an agreeing complementizer after all (contra Rizzi 1990) and that agreement is invisible on French complementizers. (4.63) C’est [le raisin vert]i que ton père a mangé ti . it-is the grape green that your father has eaten ‘Your father has eaten the green grapes.’ Boeckx’ theory predicts that CLLD is insensitive to islands if the resumptive clitic is a true resumptive pronoun, which for him requires that it should head a big-DP structure in which its ‘antecedent’ is the first-merge complement of the resumptive. Evidence for a big-DP is the presence of a resumptive pronoun, and the only evidence that Match alone has applied is that the resulting configuration is insensitive to islands. This renders Boeckx’ proposal unfalsifiable as far as French data is concerned (because of circularity). What it suggests nonetheless is that French CLLD is not derived by Match+Agree because it is insensitive to islands. The other diagnostics for movementinduced configurations, which were examined in Sections 4.3.1–4.3.4, suggest that French CLLD is not derived by Match alone either, but by firstmerge adjunction (i.e. so-called base-generation). One consequence is that the resumptive element cannot be analysed as the Spell-Out of a trace and should have full argument status. This is the subject of the next section. 4.3.5.3 On the status of the ‘resumptive’ pronoun A core property of true resumptive pronouns (RPs), as defined by Sells (1984), is that they are interpreted as bound variables and that this binding is not simply anaphoric. This makes them syntactically dependent in a way that I will show the so-called resumptive of CLLD is not. The bound variable reading of a ‘true’ resumptive pronoun is illustrated below with a Swedish relative clause. (4.64) Det finns mycketi som man önskar att deti skulle vara there is much that one wishes that it should be (Sells 1984: 56) annorlunda. different

132

French Dislocation

According to Sells (1984), English does not have true RPs but what he calls Intrusive Pronouns. Such pronouns appear mainly where they can alleviate island violations, and they are not interpreted as true variables. This is illustrated by the contrast in (4.65). A gap in the extraction position inside the relative clause can be interpreted as a variable (4.65a). In contrast, when a pronoun occupies the extraction position (4.65b), it is interpreted as referring to one particular individual. (4.65) a. [Which of the linguists]i do you think that if Mary marries ti then everyone will be happy? b. [Which of the linguists]i do you think that if Mary marries himi then everyone will be happy? Aside from the availability of true RPs in a given language, Sells argues that a variable interpretation can only be obtained where there is a binder with operator-like properties (such as a quantifier or a wh-element). If the resumptive element of (CL)LD is a true RP, we should expect it to receive a variable interpretation whenever the dislocated element has operator-like properties. Topics per se do not have quantificational properties (Rizzi 1997). There is nonetheless the possibility that a topic might act as an operator due to the inherent properties of the type of XP that instantiates it. However, as is well known, quantifiers, (non-generic) indefinites, and wh-elements (which are standardly regarded as operators) cannot be topics and hence cannot be dislocated: they do not meet the requirement that topic referents be readily identifiable in the context. This is illustrated in (4.66). (4.66) a.

∗

Tout hommei , ili est mortel. he is mortal any/every man ∗ b. Chaque potageri , ili a son robinet. each allotment it has its tap

A possible exception to this rule is what Erteschik-Shir (1997) calls subordinate update, which, as explained in Section 3.2.2, usually consists of identifying the main topic of the sentence from a pre-established set available in the discourse context. In (4.67), for instance, the dislocated element summons the set of exceptionally gifted individuals known to the speaker and identifies one individual in that group. That individual then becomes the topic of the sentence. (4.67) [Un qui est surdoué]i , c’i est le fils Fiorini. one who is over-gifted it is the son Fiorini ‘One who’s gifted is the Fiorini’s son.’

Syntax

133

Crucially, even in instances where the dislocated element has quantifierlike properties, the resumptive pronoun is attributed a fixed reference. The relation between the resumptive and its antecedent is merely anaphoric and is no different to what it would be if the antecedent were omitted from the sentence. Omission of a left-dislocated element does not alter the interpretation of the sentence significantly (provided its referent is salient enough in the context to be the preferred topic), as illustrated in (4.68). (4.68) a. Les Racts, c’étaient les monstres de mon frère. the Racts it-were the monsters of my brother ‘The Racts were my brother’s monsters.’ b. C’étaient les monstres de mon frère. it-were the monsters of my brother ‘They were my brother’s monsters.’ By contrast, omission of a left-peripheral element associated with a true RP (such as Piayya b@nt ‘which girl’ in (4.69)) radically modifies the interpretation of the sentence, as illustrated below for Lebanese Arabic. 32 (4.69) a. Piayya b@nt Seef-a Kariim mbeeriQ which girl saw.3p.sg.m.-her Karim yesterday ‘Which girl did Karim see yesterday?’ b. Seef-a Kariim mbeeriQ saw.3p.sg.m.-her Karim yesterday ‘Karim saw her yesterday.’ There is in fact no syntactic requirement for a dislocated element to be (overtly) present. 33 The ‘resumptive’ element does not need to be licensed syntactically by its ‘antecedent’. Not only can the dislocated element be omitted, in certain cases it is even banned from appearing at the periphery of the clause containing its ‘resumptive’ element. (4.70) a. Les tartesi , elle a oublié d’acheter des œufs pour lesi the pies she has forgotten to-buy some eggs to them faire. make 32

Thanks to Ghada Khattab for providing native-speaker judgements. An anonymous reviewer points out that dislocated elements are obligatory when contrastive. I do not dispute this, but this requirement is not syntactic in nature: it is due to information structure considerations. 33

134

French Dislocation b.

∗

Elle a oublié d’acheter des œufs pour les tartesi , lesi she has forgotten to-buy some eggs to the pies them faire.34 to-make ‘She’s forgotten to buy eggs to make the pies.’

pas envie de lai vendre. (4.71) a. [La vieille MG jaune]i , j’ai the old MG yellow I-have not desire to it sell b. ∗ J’ai pas envie de, [la vieille MG jaune]i , lai vendre. I-have not desire to the old MG yellow it to-sell ‘I don’t feel like selling the old yellow MG.’ The label ‘resumptive’ is therefore misleading in the case of (French) LD. Indeed, the clitic involved in French LD has the same pronominal status as it would have in a sentence without a coreferential dislocated element. In other words, the pronoun il ‘he’ is fundamentally the same in sentences (4.72a) and (4.72b). (4.72) a. Kesteri dit qu’ili aime bien les poissons. Kester says that-he loves well the fish ‘Kester says he loves the fish.’ b. Kesteri , tu sais qu’ili aime bien les poissons. Kester you know that-he loves well the fish ‘You know that Kester loves the fish.’ I conclude that the ‘resumptive’ element in French LD is not a true resumptive but a full-fledged pronoun (with deficient characteristics in the case of clitics). This configuration cannot therefore be derived by movement, whether Chop or Copy. The facts discussed above indicate that French LD cannot be derived by Copy. If it were, reconstruction effects would arise because the dislocated element and its resumptive would in effect be a single constituent with two manifestations in the sentence. However, we have seen that this is not the case. I conclude, together with Hirschbühler (1975), that French LD is not derived by movement, be it (today’s version of) Copy or Chop. 4.3.6 CLLD or Hanging Topic? In the wake of Hirschbühler (1975) and other work arguing for a basegeneration analysis of LD, a distinction was introduced to distinguish between 34 The same judgement would obtain if the left-dislocated element preceded the non-finite complementizer in (4.70b).

Syntax

135

movement-generated LD and base-generated LD (Cinque 1977, 1990; Vat 1981). Instances of the former are standardly considered to include CLLD and (Germanic) Contrastive Left Dislocation (which I will not consider here). Instances of the latter are mainly considered to be Hanging Topic Left Dislocation (HTLD). The question that arises is: is French (CL)LD actually HTLD? The distinction between CLLD and HTLD, though assumed by many to be clear-cut, 35 has, to my knowledge, never been backed up by diagnostics robust enough to distinguish between the two constructions unambiguously in languages that lack morphological case marking on DPs. For instance, Benincà (2001) acknowledges that the dislocated element in a sentence like (4.73) can be analysed as either CLLD or HTLD. (4.73) Marioi , loi rivedrò. Mario him I-will-see-again ‘Mario, I’ll see him again.’

(Benincà 2001: 44)

I show below that there is in fact no reliable empirical base for a distinction between HTLD and CLLD in spoken French. Seven properties traditionally distinguish HTLD from CLLD. (4.74) a. CLLD is syntactically ‘connected’ to the rest of the sentence; HTLD is not (see e.g. Vat 1981; Cinque 1983). A dislocated phrase is ‘connected’ when it bears marks of dependency from a sentenceinternal element (e.g. via case assignment). Connectedness manifests itself essentially in terms of case matching between the dislocated element and its resumptive. b. CLLD gets reconstructed at LF but HTLD does not (Cinque 1983).36 c. HTLD is not recursive but CLLD is, i.e. more than one dislocated element is allowed (Cinque 1990). d. HTLD is a root phenomenon while CLLD can occur in embedded clauses (Cinque 1990). e. HTLD tends not to be resumed by a clitic, though authors diverge as to whether HTLD can be resumed by a clitic at all. 35 36

(i)

See e.g. Cinque 1983, 1990; Villalba 2000; Benincà 2001; Delais-Roussarie et al. 2004. We have seen however that CLLD does not get interpreted via reconstruction in spoken French. luii donnerait sa j clé. A son[x/∗ j ] photographei , aucun de mes amis j ne his key photographer none of my friends would to-him give to his ‘None of my friends j would give his j key to hisx/∗ j photographer.’

136

French Dislocation Delais-Roussarie et al. (2004), for instance, assume that resumption by a clitic is equally possible in CLLD and HTLD. Others argue that HTLD involves by definition a non-clitic resumptive (e.g. Grohmann 2000). f. HTLD is not sensitive to islands but CLLD is (Cinque 1990). g. HTLD precedes CLLD (Cinque 1977; Benincà 2001; DelaisRoussarie et al. 2004).

However, as far as I am aware, no clear interpretive differences have been identified that would distinguish HTLD from CLLD: in both cases, the dislocated element is interpreted as the topic, the level of ‘givenness’ is unaffected by a change in (clitic vs. non-clitic) resumptive, and one construction does not seem to be more marked than the other. 37 I am not aware either of any analysis identifying prosodic differences between these two types of LD. Delais-Roussarie et al. (2004) argue that in French there are no interpretive or prosodic differences between HTLD and CLLD. The presence of dependency markers has been used as the only reliable diagnostic to distinguish between HTLD and CLLD (Cinque 1977; Larsson 1979; Delais-Roussarie et al. 2004). The clearest case of HTLD would therefore be one in which the dislocated element could be introduced by a preposition (determined by the ‘case’ of its resumptive), but that preposition was absent. Hence (4.75) would be an example of HTLD but (4.76) would be an example of CLLD. 38 en a jamais parlé. (4.75) Son photographei , elle luii her photographer she to-him of-it has never talked ‘She never talked about it to her photographer.’ en a jamais parlé. (4.76) A son photographei , elle luii to her photographer she to-him of-it has never talked ‘She never talked about it to her photographer.’ In spoken French, however, such allegedly clear cases of HTLD do not conform to the generalizations in (4.74). First, there can be more than one ‘HTLD’

37

Some have suggested that there are interpretive differences between the two constructions: Cinque (1983: 95) claims HTLD encodes new or unexpected topics, and Villalba (2000) claims that HTLD encodes discourse topics and not sentence topics. However, neither provides diagnostics to identify unexpected topics or discourse topics. 38 French does not have morphological case on DPs. Dislocated pronouns appear in default case irrespective of the grammatical function of their resumptive element.

Syntax

137

en j a jamais (4.77) a. La mer j , son photographei , elle luii the sea her photographer she to-him of-it has never (HTLD) parlé. talked ‘She never talked about the sea to her photographer.’ b. De la mer j , à son photographei , elle luii en j a of the sea to her photographer she to-him of-it has (CLLD) jamais parlé. never talked ‘She never talked about the sea to her photographer.’ Second, ‘HTLD’ is allowed in embedded clauses (4.78) a. Je pense que son photographei , elle luii en fait voir de I think that her photographer she to-him of-it makes see of (HTLD) toutes les couleurs. all the colours ‘I think she gives her photographer a hard time.’ b. Je pense qu’à son photographei , elle luii en fait voir I think that-to her photographer she to-him of-it makes see de toutes les couleurs. (CLLD) of all the colours ‘I think she gives her photographer a hard time.’ Third, ‘HTLD’ can be resumed by a clitic (as shown in the examples above). Fourth, it is far from clear that HTLD has to precede CLLD. Only half of my seventy-five informants accepted (4.79b), where the left-dislocated element stripped of its dependency marker (allegedly the HTLD) precedes the leftdislocated PP. (4.79a) provides the context for (4.79b). (4.79) a. Vous pouvez dire tout ce que vous voulez à Gertrude sur sa mère. ‘You can tell Gertrude anything you want about her mother.’ en j b. Mais ses frèresi , de leur mère j , on peut pas leuri but her brothers of their mother one can not to-them of-her parler. speak ‘But her brothers, no one can talk to them about their mother.’ (7%; 43%)

Compare this with (4.80b), which contains two unambiguous instances of HTLD. This sentence comes from Delais-Roussarie et al. (2004), who claim it

138

French Dislocation

is grammatical. However, only twenty-four per cent of my seventy-five informants accepted this sentence as grammatical/unmarked. (4.80) a. Il paraît que ton frère a acheté une nouvelle voiture ? ‘Apparently your brother’s bought a new car?’ pas d’elle j . b. Pierrei , sa voiture j , cet idioti ne s’occupe Pierre his car this idiot neg refl-looks-after not of-her ‘Pierre, the idiot doesn’t look after his car.’ (26%; 50%) More quantified data are required before any firm conclusion can be reached. In this context, how should one interpret the data in (4.54), which appear to indicate that CLLD is insensitive to strong islands in spoken French? These data show that native speakers can accept dislocated PPs resumed by an element inside a strong island. To the extent that (i) such dislocated elements are connected to the core of the sentence by the preposition they contain and (ii) HTLD does not display such signs of connectedness, I take these cases to indicate clearly that French CLLD is not constrained syntactically by strong islands: the syntax of spoken French clearly allows such a configuration. Crucially, no property of CLLD vanishes when the resumptive element is contained within an island—and this is true both in cases where ‘connectedness’ is overt and in cases where it is not. The relative marginality of certain PPs dislocated across an island boundary is not sufficient proof that syntactic movement is involved in French CLLD. I do not dispute that it is possible to find cases of PPs dislocated across an island boundary which will be rejected by native speakers. However, this rejection will be due to (i) insufficient likelihood that the dislocated PP is interpreted as a topic (see Appendix for elaboration of this point) and (ii) the fact that left-dislocated PPs resumed by a pronoun or clitic tend to be viewed as marked by most informants (and are extremely rare in corpora of spontaneous production), 39 whether their resumptive element is situated within an island or not. If we were to analyse as HTLD any case of LD which does not match standard assumptions regarding CLLD (in the absence of any independent evidence for an HTLD analysis), we would effectively make the data fit the theory by categorizing as HTLD anything that contradicts what we believe about CLLD. What the facts tell us is that (i) native speakers accept at least some cases of ‘connected’ LD (i.e. CLLD) which span across a strong island, 39 Out of a sample of 4,030 clauses produced by adults (extracted from the York and Cat corpora of spontaneous production), I did not find a single instance of a left-dislocated object PP. A similar observation has been made with respect to other French corpora of spontaneous production by Barnes (1985) and Lambrecht (1981, 1986).

Syntax

139

(ii) the acceptability of left-dislocated structures decreases in the presence of connectivity markers, irrespective of whether the resumptive is inside an island or not, and (iii) discourse factors have a direct impact on acceptability judgements whether or not the dislocated element is a PP. This suggests that narrow syntax alone is not responsible for the well-formedness of dislocated structures. From a strictly syntactic perspective, French dislocation is insensitive to islands. Indeed, as shown at length in the previous sections, it does not display any of the characteristics of a configuration involving movement. I therefore conclude that French LD is not HTLD to the extent that it displays the following properties: (i) it is recursive, (ii) it is allowed in certain embedded contexts (though, as we will see in Section 4.4.3, these have to be root-like), (iii) it can be resumed by a clitic, and (iv) it can (marginally) bear marks of connectivity to the extent that these are visible in spoken French. This conclusion needs qualifying though. If, cross-linguistically, the difference between connected and non-connected LD (i.e. between CLLD and HTLD) translates categorically into different syntactic configurations which are arrived at via different syntactic derivations or mechanisms, an analysis of French LD as involving HTLD in all cases might be more desirable. In other words, it could be that spoken French does not have CLLD (anymore). 40 If this is found to be the case, a revision of the core characteristics of HTLD would be called for (so that it can be recursive, be allowed in certain embedded contexts, and display dependency markers to an extent). Until further crosslinguistic research establishes whether CLLD can hold across strong islands in other languages, it is not possible to choose between these two options. 4.3.7 Which analysis for French RD? In the quest for general principles constraining syntactic derivations, the focus of attention has tended to be on left- rather than right-peripheral elements. 41 As a result, and because of standard assumptions ruling out rightward movement, phenomena involving the right periphery of the clause have tended to be forced into the mould of constraints designed to deal with left-peripheral ‘transformations’. In what follows, I will argue that French LD and RD can 40 Spoken French is already distinguished from other Romance languages by the fact that it is not pro-drop and that postverbal subjects do not appear in thetic contexts. See Lahousse (2003) and De Cat (2002). 41 One of the reasons invoked to explain the relative lack of attention to right-peripheral phenomena has been the assumption that these are less frequent than their left-hand counterparts. This is clearly not true of French RD, which occurs as frequently as LD in corpora of spontaneous production across dialects—see Appendix A for details. In questions, RDs are even more frequent than LDs, as attested in the spontaneous speech of adults from the York and the Cat corpora: in wh-questions, seventy-five per cent of dislocated elements are right-peripheral and in yes-no questions fifty-eight per cent of dislocated elements are.

140

French Dislocation

be accounted for using the same syntactic mechanisms without having to postulate that RD is in fact some kind of LD. The differences between LD and RD will be shown to derive straightforwardly from the analysis proposed, and in particular from the properties of the two sides of the clause periphery. (Cann et al. (2004) do the same in a different framework.) We have seen that in French left-dislocated elements associated with an indirect object are preferably bare, in that they strongly tend to be DPs rather than PPs. By contrast, right-dislocated elements of this type are preferably not bare, as illustrated in (4.81). 42 répondre, ?? (à) Jimi . (4.81) a. Je dois encore luii to Jim I must still to-him reply ‘I must still reply to Jim.’ répondre. b. (?À) Jimi , je dois encore luii to Jim I must still to-him reply ‘I must still reply to Jim.’ Under standard assumptions regarding connectivity, this suggests that RD is derived by movement. Four variants of the movement analysis of RD have been proposed in the literature. The ‘old’ version assumes that rightward movement is part of grammar (as argued for instance by some of the contributors to Beerman et al. 1997). More recently, it has been proposed that the right-dislocated element in fact appears in its argument position with optional clitic doubling and optional dislocation intonation (see e.g. Kayne 1994). Other proposals view RD as a special kind of LD: either RD is situated at the left periphery of an extended projection of V and appears to be rightperipheral because all lexical items that are base-generated under ‘RD’ move (see e.g. Villalba 1998; Cecchetto 1999; Belletti 2001; López 2003), or RD is in fact in the left periphery of the clause but appears to be right-peripheral because of IP inversion (see e.g. Frascarelli 2000; 2004, Samek-Lodovici 2006). I will show that, whatever version one chooses, a movement analysis of French RD makes predictions that are not borne out and should therefore be rejected. 4.3.7.1 French RD is not an LF/PF phenomenon Under Kayne’s assumptions, the only way left-dislocated phrases can be derived is by postulating the presence of a functional projection hosting the left-peripheral element in its specifier. He claims that ‘except for the necessary presence of an abstract 42 This is not to say that PPs can never be stripped of their preposition when right-dislocated: examples are attested in corpora of spontaneous production (see for instance (4.137)).

Syntax

141

X°, [his analysis] is not terribly different from, for example, Cinque’s (1990) treatment of what he calls clitic left-dislocation’. (Kayne 1994: 78–9). Kayne further assumes that right-adjunction is prohibited by UG. He proposes that RD in Romance languages is a unified phenomenon, in which the right-dislocated element occupies the object position with optional clitic doubling and optional dislocation intonation. His line of reasoning proceeds as follows: (i) RD does not depend on the presence of a clitic doubling an object as the clitic can sometimes be omitted; (ii) the doubling clitic does not depend on RD intonation (contrary to what is standardly assumed), which is equivalent to stating that all Romance languages displaying RD allow for clitic doubling. This latter point is argued on the basis of examples like (4.82), in which the indirect object à elle ‘to her’ does not require dislocation intonation. (4.82) Jean lui a parlé à elle. Jean to-her has talked to her ‘Jean talked to her.’

(Kayne 1994: 80)

Consequently, examples like (4.83) are to be treated on a par with cases of obligatory dative clitic doubling in Spanish such as (4.84). Kayne suggests that this justifies analysing right-dislocated phrases as occupying object position. (4.83) Une pierre ∗ (lui) est tombée sur la tête, à Jean. a stone to-him is fallen on the head to Jean ‘A stone fell on Jean’s head.’ (Kayne 1994: 80) (4.84)

∗

(Le) examinaron los dientes al caballo. to-it they-examined the teeth to-the horse ‘They examined the horse’s teeth.’

(Kayne 1994: 80)

The final step in Kayne’s reasoning is to claim that CLLD is derived by movement of the left-dislocated phrase from object position, and that RD is an instance of CLLD at LF. RD intonation is argued to be triggered by an optional feature in the ‘overt syntax’ that would feed both LF (triggering movement) and PF (triggering dislocation intonation). Kayne’s account of RD ignores a vast amount of data: it leaves aside (i) right-dislocated subjects (4.85), (ii) right-dislocated phrases that cannot be argued to occupy the object position (4.86), and (iii) right-dislocated phrases resumed by a non-clitic element (4.87). 43 43 Frascarelli (2004) also shows that in Italian there is no fixed order of topics. Kayne’s analysis predicts the opposite.

142

French Dislocation

(4.85) Ellei ignore les sujets disloqués, son analysei . she ignores the subjects dislocated his analysis ‘His analysis ignores dislocated subjects.’ (4.86) On va lesi manger avec des pommes, [ces petits enfants]i . with some apples these little children one will them eat ‘We’ll eat these little children with apples.’ (4.87) Il aime çai , [la chair fraîche]i . he loves that the flesh fresh ‘He loves fresh flesh.’ In (4.85), the right-dislocated element is a subject. One might choose to assume that the right-dislocated subject is in the VP-internal position where the subject originates (Kayne 1994: 118, 141, following Koopman and Sportiche 1991), and that the object is in [spec,AgroP], thus creating the subject-final order displayed in (4.85). However, such an analysis would not be tenable, at least on the grounds that spoken French does not allow for clitic doubling of the subject, as argued in Section 2.2. In (4.86) the object cannot be in object position as it follows a VP adjunct. As for (4.87), I do not see how such an example can be accounted for in a way that is compatible with Kayne’s analysis of RD, given that the resumptive element is not a clitic, and that under no circumstances can an object be unambiguously doubled by ça ‘that’ (i.e. without dislocation intonation), even in Canadian French, where the use of ça is more pervasive than in European French. (4.88)

∗

On connaît ça les Wombles. one knows that the Wombles

(no dislocation intonation)

Kayne’s claim that French allows for clitic doubling is itself controversial, given that it is only verified in cases where the doubled object is a strong pronoun as, in (4.82), but not in (4.89) or in constructions expressing inalienable possession, as in (4.83). (4.89)

∗

Jean lui a parlé à Gudule. Jean to-her has talked to Gudule

(no dislocation intonation)

If the possession involved is not inalienable, doubling is not possible (4.90a), and in that case it has to be expressed by a possessive determiner (4.90b). (4.90) a.

∗

Une pierre lui est tombée sur la brouette, à Jean. a stone to-him is fallen on the wheelbarrow to Jean

Syntax

143

à Jeani . b. Une pierre est tombée sur sai brouette, a stone is fallen on his wheelbarrow to Jean ‘A stone fell on Jean’s wheelbarrow.’ Note also that the French counterpart of the Spanish clitic doubling construction in (4.84) is ungrammatical. (4.91)

∗

On lui a examiné les dents au one to-it has examined the teeth to-the (no dislocation intonation) cheval. horse

Furthermore, even in a case where the clitic is compulsory, as in (4.83), the alleged doubled object has to bear a dislocation intonation contour, as (4.92) illustrates. In the Spanish example in (4.84), the indirect object has no such intonation. (4.92)

∗

Une pierre lui est tombée sur la tête à a stone to-him is fallen on the head to Jean. (no dislocation intonation) Jean

A clitic doubling analysis of French right-dislocated object can thus only be proposed for cases like (4.93), to the extent that they can be derived from cases like (4.82). Note however that there is a significant difference between (4.82) and (4.93), in that the former but not the latter yields a contrastive reading of the indirect object (Ronat 1979). A sentence like (4.82) can only be felicitous in a context where Jean has talked to a particular woman as opposed to some other person. (4.93) Jean luii a parlé, à ellei . Jean to-her has talked to her ‘Jean talked to her.’ Kayne’s (1994) analysis of right-dislocation as a PF phenomenon is thus untenable on the following grounds: (i) it wrongly predicts that RD of elements doubled by a clitic is optional in French (this is only true in a very limited set of cases) and (ii) it cannot be extended to right-dislocated subjects, XPs that cannot be argued to occupy the object position, or XPs that are resumed by a non-clitic element.

144

French Dislocation

4.3.7.2 French RD is not LD lower in the tree An analysis of RD as LD in a lower Topic position whether a dedicated TopicP (Villalba 1998; Cecchetto 1999; Belletti 2001) or [spec,vP] (López 2003) faces a number of problems, especially with respect to the spoken French data. First, as pointed out by Frascarelli (2004), it requires ad hoc stipulations in order to derive the prosody (something I will not dwell upon here). Second, situating right-dislocated elements at the periphery of vP predicts strict ordering restrictions which are contradicted by the facts. In (4.94a), the conditional clause clearly modifies the whole clause and not just the VP. This indicates that it appears at the periphery of the sentence, in which case the dislocated moi ‘me’ cannot occupy a position in the lower periphery of the clause (be it [spec,AgroP], [spec,vP], or a lower TopicP). If RD was LD in the extended periphery of VP, the dislocated moi ‘me’ would have to precede any sentential modifier, contrary to fact. In (4.94b), the floated quantifier is traditionally considered to occupy the specifier of VP (or vP) and is thus lower than [spec,AgroP]: la bonne ‘the maid’ thus cannot occupy [spec,AgroP]. demain], si tu veux, moi]. (4.94) a. [ CP [ TP J’irai I-will-go tomorrow if you want me ‘I’ll go tomorrow if you want.’ b. Elle les a [ VP tous attrappés], la bonne. all caught the maid she them has ‘The maid has caught them all.’ Third, the above-mentioned analyses all postulate that RD gets to this lower periphery by syntactic movement. This in turn makes predictions that are not borne out by the spoken French data. If the derivation of RD involved leftward movement, it should show at least some of the characteristics in (4.95). (4.95) a. b. c. d. e.

Licensing of parasitic gaps in the IP Weak Crossover effects Relativized Minimality effects with other RD chains Reconstruction effects Sensitivity to islands

Examples (4.96)–(4.98) show that, at the LD ‘stage’, the dislocated element does not license parasitic gaps (4.96), it does not yield Weak Crossover effects (4.97), and it does not show any sign of Relativized Minimality effects (4.98).

Syntax

145

(4.96) Tu lesi as déchirés au lieu de ∗ (les) consulter, les livresi . in place of them consult the books you them have torn ‘You’ve torn the books instead of consulting them.’ Abélardi . (4.97) Sai mère l’i aimait trop, his mother him loved too-much Abélard ‘Abélard’s mother loved him too much.’ Abélardi , son élève j . (4.98) Ili l’ j aimait trop, he him loved too-much Abélard his pupil ‘Abélard loved his pupil too much.’ The impossibility of a narrow-scope interpretation of the right-dislocated element with respect to negation suggests that the dislocated element does not get interpreted in an IP-internal (reconstructed) position. 44 (4.99) Julia, elle ne lesi a pas vendues, toutes ces toilesi -là. all those canvases-there Julia she neg them has not sold ‘Julia didn’t sell any of (all of) those pictures.’ # ‘Julia didn’t sell some of those pictures.’ (4.100) #[Chaque maître]i l’ j a renvoyé, [un de sesi disciples] j . each master him has dismissed one of his disciples Turning now to the last diagnostic (4.95e), French RD does not show any sign of being sensitive to islands in the course of its putative derivation, contrary to what is expected under a (movement-)LD analysis. This is illustrated by (4.101), in which the right-dislocated element under discussion is preceded by another dislocated element associated with the subject of the matrix clause. This indicates that sa fille ‘his daughter’ in (4.101) and aux voleurs ‘to the thieves’ in (4.102) are not attached to the relevant embedded clauses (i.e. we are really looking at instances of right-dislocated elements separated from their resumptives by an island boundary). Aux voleurs ‘to the thieves’ in (4.102) is associated with a resumptive element situated inside an adjunct. This clearly shows that RD is insensitive to islands just as LD is. emmenée,]] moii , sa fillei . (4.101) [Jei connais [l’homme qui l’i a I know the-man who her-has taken-away me his daughter ‘I know the man who took his daughter away.’ 44 Some of the evidence for the absence of reconstruction in LD cannot be adapted to show the absence of reconstruction in RD because the dislocated element is required to introduce a new referent (from a d-linked set), a function which right-dislocated elements cannot perform given that they are introduced/processed after the core of the sentence.

146

French Dislocation

parles]], moii , aux (4.102) [Jei suis partie [avant que tu ne leur j I am left before that you neg to-them speak me to-the voleurs j . thieves ‘I left before you spoke to the thieves.’ It has to be concluded that French RD does not involve an intermediate LD step derived by movement. 4.3.7.3 French RD is not LD+IP-inversion The LD+IP-inversion analysis of RD was proposed by Frascarelli (2004) for Italian. It is however not valid for French, as it would rule out many instances of licit RD (unless one is willing to make a number of ad hoc assumptions). Consider (4.103) as an example. (4.103) Mais le juge j , çai a surpris tout le monde, [qu’elle l j ’ait but the judge it has surprised all the people that-she him-has invité]i . invited ‘But it surprised everybody that she’d invited the judge.’ The right-dislocated clause is associated with the subject position of the matrix clause. On a movement analysis, that clause thus originates in [spec,VP]. Le juge ‘the judge’ could only be extracted prior to the movement of that clause to [spec,IP] (or a strong island violation would occur). The EPP of the matrix I (or T) has to be satisfied as soon as I (or T) is merged, so the clause must raise to subject position before any head can be projected to attract le juge ‘the judge’ out of it. We are thus forced to assume that le juge ‘the judge’ is basegenerated in its left-dislocated position, in which case the derivation proceeds as follows (after raising of the clause to subject position). (4.104) Le juge j , qu’elle l’ j ait invité a surpris tout le monde. the judge that-she him has invited has surprised all the people Next, the sentential subject is left-dislocated over le juge ‘the judge’. invité]i , le juge j , çai a surpris tout le (4.105) [Qu’elle l j ’ait that-she him-has invited the judge it has surprised all the monde. people In the third step, the IP moves leftward by so-called IP-remnant movement, which standardly targets the Focus position in the periphery of the clause.

Syntax

147

It is crucial that this position be higher than the positions hosting the dislocated clause and the dislocated DP in order to ensure they appear rightdislocated. invité tk , le (4.106) [Ça a surpris tout le monde]k , qu’elle l j ’ait it has surprised all the people that-she him-has invited the juge j . judge The last step required to obtain (4.103) is to move le juge ‘the judge’ to sentence-initial position. (It could not have moved together with the IP given that it is a topic and thus cannot be focus-moved.) Given that le juge ‘the judge’ is already in a left-peripheral, topic position in (4.106), it is hard to imagine a principled reason why it should move further up within the same periphery. To my knowledge, topic-to-topic movement has never been attested within the periphery of the same clause. The derivation of sentences like (4.103) by LD+IP-inversion requires a host of ad hoc stipulations which, I believe, render such an analysis implausible. 4.3.7.4 Differences are unexpected if RD = LD Aside from not being able to derive the data in a satisfactory fashion, the LD analysis of RD predicts that LD and RD are much more similar than actually is the case. First, it predicts that the presence of dependency markers such as prepositions should be equally acceptable in either periphery. The contrast in (4.107) is therefore entirely unexpected. (4.107) a. (? À) Clélia, il ne lui écrivait plus.45 to Clelia he neg to-her wrote no-more b. Il ne lui écrivait plus, ??(à) Clélia. he neg to-her wrote no-more to Clelia ‘He didn’t write to Clelia anymore.’ Second, it predicts that any element that can be left-dislocated can be rightdislocated. While this is true in most cases, it is not true (i) when the dislocated element is contrastive and/or emphatic, or (ii) when it identifies a topic among a d-linked set. In both of these cases, only LD is possible. This is illustrated in (4.108) and (4.109). In example (4.108) the first line sets out a context for the (a) and (b) sentences. 45 This sentence tends to be accepted as unmarked if the dislocated element is in clear contrast with another PP in the immediately preceding context.

148

French Dislocation

(4.108) Je ne vois presque jamais Alice. ‘I hardly ever see Alice.’ a. Mais sa sœuri , je lai vois souvent. but her sister I her see often b. #Mais je lai vois souvent, sa sœuri . but I her see often her sister ‘But I often see her sister.’ (4.109) a. Tout ce que tu m’as dit là, j’aimerais mieux all that that you to-me-have said there I-would-like better l’oublier. it-to-forget b. #J’aimerais mieux l’oublier, tout ce que tu m’as dit I-would-like better it-to-forget all that that you to-me have là. said there ‘I’d rather forget what you’ve told me.’ If right-dislocated elements occupy a position in the left-periphery, these differences are unexpected. I will argue in Section 4.4.4 that such contrasts can be derived straightforwardly from the properties of each of the sentence peripheries. In the following section, I show that rightward movement of French RD is impossible. 4.3.7.5 French RD is not subject to the Right-Roof constraint If RD is derived by rightward movement, it should be clause-bound (i.e. subject to Ross’s (1967) so-called Right-Roof constraint). (4.110) a. [Tu es partie [sans ∗ (luii ) parler]], toi, à Davei . you are left without to-him to-speak you to Dave ‘You left without speaking to Dave.’ b. [Tu m’avais dit [que tu ∗ (l’i ) inviterais]], à moi, ta you to-me-had said that you him would-invite to me your mèrei . mother ‘You told me you would invite your mother.’ In (4.110) toi ‘you’ and à moi ‘to me’, which intervene between the embedded VP and the right-dislocated DP, are outside the VP boundary. If this were not the case, such dislocated elements would be parentheticals, and thus should be allowed to precede a VP-internal element such as an object complement.

Syntax

149

However, this is impossible: if the resumptive element is omitted (as indicated by the parentheses above) and the VP-final stress is placed on the PP in (4.110a) or the DP in (4.110b), the sentence becomes ungrammatical. We can safely conclude that à Dave ‘to Dave’ and ta mère ‘your mother’ in (4.110) are right-dislocated outside of the clause containing their resumptive element, as indicated by the bracketing in (4.110). These examples thus show that French RD is not clause-bound, which suggests that it is not derived by rightward movement. 4.3.8 Summary The evidence discussed so far unambiguously points towards a basegeneration analysis of French dislocated elements. It has been argued that syntactic boundaries freely intervene between the dislocated element and its resumptive, and that there are no notable syntactic discrepancies between LD and RD, or between dislocated elements resumed by a clitic and those resumed by a non-clitic. In Section 4.4.4 I will show that the interpretive differences between LD and RD can be derived from the properties of each periphery of the clause.

4.4 A first-merge adjunction analysis of French dislocation 4.4.1 The analysis The analysis I will argue for is summarized in (4.111). It does not involve movement of the dislocated element, 46 (covert) movement of its resumptive, 47 or a dedicated functional projection (such as TopicP). 48 (4.111)

Dislocated elements are adjoined by first-merge to a maximal projection with root properties.

The main points of the proposed analysis are spelled out in (4.112).

46 Analyses arguing for movement of the topic phrase include Iatridou (1995), Zubizarreta (1998), Cecchetto (1999), Grohmann (2003), and López (2003). Some have proposed that the topic phrase moves at PF, see e.g. Kayne (1994), Aoun and Benmamoun (1998). 47 LF-movement of the resumptive element has been argued for by e.g. Demirdache (1991), Anagnostopoulou (1997), Rizzi (1997). 48 Analyses postulating the existence of a Topic projection include Rizzi (1997), Cecchetto (1999), Villalba (2000), Benincà (2001), Frascarelli (2002), Grohmann (2003), Delais-Roussarie et al. (2004) among others. Many alternatives have been proposed that do away with a dedicated Topic Projection, see e.g. Barbosa (2000), Newmeyer (2003), López (2003), Gill and Tsoulas (2004), and Emonds (2004).

150 (4.112)

French Dislocation a. Dislocated elements appear at the edge of Discourse Projections. Discourse Projections are finite root(-like) clauses. b. The numeration is organized in D-subarrays. A D-subarray corresponds to a clause containing a T endowed with a discourse feature. c. When the remaining items of a D-subarray are not visible to Merge, Adjunction applies as a last-resort operation to exhaust the numeration. d. The numeration contains exactly those items to be assembled into sentences/utterances. Consequently, the numeration must consist of items picked from the lexicon according to information structure considerations. e. Topics are licensed and interpreted by a rule of Predication.

Each of these points is argued for in turn below. Section 4.4.2 explores the consequences of the adjunction analysis of French dislocation. 49 Section 4.4.3 explores in more detail the conditions under which dependent clauses can be root-like and hence host a dislocated element. 4.4.1.1 Discourse Projections By definition, topics are associated with a predication (see e.g. Reinhart 1981, Erteschik-Shir 1997). Under the uncontroversial assumption that dislocated structures are the prototypical realization of ‘topic+predication’ structures, the constituent with which the dislocated XP is associated will obligatorily be of a predicative nature. This is true irrespective of the approach adopted, i.e. whether the topic is analyzed as 49 This analysis does not apply to Topicalized structures, in which a (contrastive) left-peripheral element appears without a resumptive. These are best accounted for in terms of syntactic movement. (Note however that it is not necessary to postulate the existence of a TopicP even in such structures, as argued by Lasnik and Saito (1992).) In spoken French, Topicalization is not a topic construction, in spite of the misleading terminology (the same has been argued about Italian by Cinque (1990)). As they are not topics, topicalized XPs are not subject to the specificity and referentiality conditions that apply to leftdislocated XPs. Note that even in English, it is very unclear whether Topicalization can be considered a topic-promoting construction (see in particular Prince (1999)). Reinhart (1981) suggests that in English, topicalized constituents can be topics, but only if they do not receive a focus intonation. Note incidentally that not all LDs without a resumptive are instances of topicalization. Leftperipheral elements not resumed by a clitic should only be analysed as so-called topicalization under strict prosodic and informational conditions, which are defined in Section 2.3.3 and Chapter 3. When the left-dislocated element is clearly a topic (both prosodically and informationally), the ‘resumptive’ can be omitted if and only if this is allowed by the grammar independent of the presence of a dislocated element. In other words, the ‘resumptive’ is subject to the conditions regulating argument omission (for discussion, see e.g. Fónagy 1985; Cummins and Roberge 2004, 2005; Prévost 2006).

Syntax

151

being adjoined to that constituent, as proposed here, or as sitting in the specifier of a dedicated TopicP, the head of which selects the constituent in question. It is however not the case that any predicative XP can be associated with a topic. Dislocated topics are banned from appearing in certain clauses, as illustrated by the contrast below, where the dislocated element can only be hosted by the finite clause in spite of the fact that its resumptive element is inside the non-finite clause. (4.113)

a. #On attendra pour, les photosi , lesi regarder ensemble.50 one will-wait to the pictures them to-look-at together b. Les photosi , on attendra pour lesi regarder ensemble. the pictures one will-wait to them look-at together ‘We’ll wait to look at the pictures together.’

Given that both left- and right-dislocated elements are topics (as argued in Chapter 3), dislocated phrases should equally be banned from the rightperiphery of non-finite clauses. However, this is harder to confirm empirically given the bare nature of the right periphery: a right-dislocated phrase at the end of a non-finite clause could actually be adjoined to the matrix clause (4.114). (4.114) [ CP . . . [CP . . . ] RD ] To the extent that native speakers accept the sentence below without a dislocated element, 51 they also accept (4.115b) but reject (4.115a). The rightdislocated element follows the indirect object of the matrix clause, and hence cannot be in the right periphery of the non-finite clause (which contains its resumptive element). (4.115)

a. #Elle a demandé de lesi équeuter, les haricotsi , à tout le the beans to all the she has asked to them tail monde. people

50

The same judgement obtains if the left-dislocated phrase appears before pour ‘to’. This requires speakers to accept that the non-finite object clause precedes the indirect object instead of appearing to its right. Extraposition is the option preferred by all the informants consulted. 51

(i)

Elle a demandé à tout le monde de les équeuter. she has asked to all the people to them tail ‘She asked everybody to tail them.’

152

French Dislocation b. Elle a demandé de lesi équeuter à tout le monde, les to all the people the she has asked to them tail haricotsi . beans ‘She asked everybody to tail the beans.’

The structural and interpretive reflexes of information structure tend to be absent from embedded clauses which lack root properties (Hooper and Thompson 1973). 52 Dependent clauses with root properties are typically embedded under attitude verbs, which have been argued to introduce quotations (Hooper and Thompson 1973; Tomioka 2001). Emonds (2004) convincingly argues that only Discourse Projections can allow root transformations, and that what counts as a Discourse Projection is parameterized. He proposes that Discourse Projections involve root-like indirect discourse embedding and that this requires them to be finite. Others attribute root-like behaviour of clauses to anchoring properties, i.e. to be root-like a clause has to be interpreted relative to the context, which requires reference to speaker and hearer (e.g. Haegeman 2002, 2003). I would like to propose that dislocated topics can only be hosted by Discourse Projections in the sense of Emonds (2004), and that Discourse Projections contain a T endowed with a Discourse feature. This feature forces the event expressed by the verb in T to be interpreted relative to the discourse context and in particular the topic of the sentence, whose default values correspond to the time and place of utterance (Gundel 1975; Erteschik-Shir 1997) and the speaker. 53 In other words, it forces the truth value of the predication to be evaluated directly with respect to the discourse context, independent of any restriction which might be imposed by the rest of the sentence. This discourse feature has direct interpretive effects in that it endows the root(-like) projection with a Performative feature. Performatives were originally defined by Austin (1962) as speech acts subject to felicity conditions. An utterance has a performative function if it is not simply a potentially true or false description of a situation. 54 The performative function of root-like projections 52

See Heycock (2006) for a comprehensive review of the literature on embedded root phenomena. Encoding this discourse feature on T instead of as a functional projection in its own right is motivated by economy considerations and by the fact that this head/feature does not select the topic nor attract another ‘discourse-marked’ element to the left periphery (in contrast to the F position proposed by Uriagereka 1995, for instance). Positing such a selection process would wrongly predict that topics are merged in [spec,TP], and would not account for the distribution of left- and rightperipheral topics in spoken French. 54 Determining whether it is possibile for dependent clauses to be endowed with a performative function should contribute to understanding why they have limited tolerance to dislocated topics. This is something I am hoping to explore in future research. 53

Syntax

153

will be explored further in Section 5.5, where I investigate complex fragments, i.e. verbless utterances containing what appears to be a dislocated element. Crucially, the discourse feature on T is not a [topic] feature, and it does not require checking in the overt syntax of French: it does not participate in an Agreement relation and hence cannot trigger movement (of the topic XP or its resumptive). In spoken French, an XP in [spec,TP] cannot be interpreted as the topic of the sentence (as shown in Section 2.2). This would be unexpected if T bore a [topic] feature which required checking by an XP topic. It will be shown in Section 4.4.2.1 that the requirement that dislocated topics appear at the edge of Discourse Projections correctly predicts their distribution. 4.4.1.2 D-subarrays Peripheral topics (i.e. dislocated elements) can appear either at the edge of the clause containing their resumptive element (which requires this clause to be a Discourse Projection) or they can appear higher, as in (4.16b), (4.16c), (4.41), (4.52), (4.54), (4.72b), in which case the higher clause (but not the lower one) is a Discourse Projection. The position of the peripheral topic depends on the required interpretation of the sentence: whatever falls within the scope of the topic must be interpreted with respect to that topic. Hence the topic has to be projected high enough to take scope over the (part of the) predication to which it is ‘relevant’, in the sense that it restricts the domain within which that predication is assumed to hold true (cf. Erteschik-Shir 1997). This determines whether the dislocated element appears in the matrix or the embedded clause in a complex sentence. Which clause the dislocated element is relevant to is determined prior to the selection of the numeration, according to the intention of the speaker. The dislocated element is selected as part of the D-subarray to which it is relevant. I would like to propose that D-subarrays are clauses containing a T endowed with a [discourse] feature. 55 The underlying assumption is that the numeration is selected in an information-structure-sensitive fashion. This is not generally acknowledged in the literature, but is necessary under any analysis to account for the role information structure plays in the choice of lexical items. A typical example would be the use of pronominals instead of R-expressions in a context where the referent in question is salient. In a context like (4.116), for instance, the only licit follow-up is (4.116a), not (4.116b). The choice of a pronoun rather than an R-expression in this 55 This allows for the possibility of cross-linguistic variation as to whether an element in vP could fulfil the role of a T with a [discourse] feature, i.e. permitting the vP to host a peripheral topic. This is not possible in spoken French but has been proposed for Italian by Cecchetto (1999) and Belletti (2001), and for Catalan by Villalba (1998) and López (2003).

154

French Dislocation

context is driven by a combination of information structure and economy considerations. (4.116) My friend has two children. a. She feeds them every day. b. My friend feeds her children every day. Under current assumptions, the grammar does not include a rule of pronominalization. Yielding (4.116a) rather than (4.116b) as an output therefore depends entirely on what the numeration contains (given that an informationally inadequate choice would not lead to a derivation crash). The same applies to dislocated elements: their presence in a clause is determined in the numeration. I believe similar conclusions would have to be drawn under a cartographic approach à la Rizzi (1997). 56 The concept of a D-subarray is a straightforward manifestation of the information-structure-sensitive nature of lexical selection in the building of the numeration. 4.4.1.3 Last-resort adjunction Adjunction is the only operation that can be performed ‘blindly’ by the syntactic computational system, as it does not obligatorily involve agreement (see Hoekstra 1991) or selection. 57 This ‘blindness’ is exactly what allows syntax to be freed from the burden of information structure (though not its effects), which is necessary to account for the French dislocation data, as will be shown below. 58 The Extension Condition (Chomsky, 1993; 1995) is met by the fact that adjunction can only exhaust the numeration within a D-subarray, which in effect corresponds to a rootlike projection. The theoretical implications of the base-generated adjunction analysis of French dislocation will be discussed in Section 4.4.5. 4.4.1.4 Topic interpretation The recourse to a rule of Predication to license topics is rooted in the work of Chomsky (1977), Iatridou (1995), É. Kiss (1995), Rizzi (1997), Erteschik-Shir (1997) and Barbosa (2000), among others. Following Erteschik-Shir (1997), I assume that this rule operates in the interpretive component. Recast in terms of the present analysis, the rule of Predication interprets the Discourse Projection as the predicate and the adjoined topic as the subject of predication. The latter is understood as the referent with respect 56 I am not aware of any work that adopts the cartographic approach to the clause periphery which addresses these issues explicitly. 57 A syntactic crash would result if the topic were merged in an argument position because the required resumptive clitic would then be adjoined as a last resort but would not have an appropriate host on which to cliticize. 58 Freeing syntax from information structure features should also apply to Focus for consistency. See Szendr˝oi (2001, 2003) for an analysis eschewing syntactic focus features.

Syntax

155

to which the truth value of the sentence is evaluated (following Reinhart 1981), which in most cases corresponds to what the sentence is about. If the dislocated element cannot be interpreted as the topic, the combination of what the rule of Predication interprets as the ‘subject’ and the ‘predicate’ is anomalous: it gets rejected on interpretive (discourse) grounds. The rule of predication need only apply once, even in sentences containing more than one topic. In such cases, the topics act as multiple restrictors on the world/frame within which the predication is assumed to hold true. 4.4.1.5 On the relation between the dislocated element and its resumptive The present analysis predicts that the dislocated element is not related syntactically to its resumptive element. Instead, this relation is one of discourse coreference, i.e. it is of the same nature as that which holds across sentences. There is no need for the resumptive to be a copy of the dislocated element: a subset relation is sufficient, as illustrated in (4.117) (see e.g. Prince 1981; Ward and Prince 1991). (4.117)

Comme poissons, j’aime bien les sardines. as fish I-like well the sardines ‘As for fish, I like sardines.’

4.4.2 Predictions of the adjunction analysis The adjunction analysis predicts that French dislocated elements can appear at the edge of any Discourse projection, and that there is no syntactic constraint on the number of topics allowed or whether they appear in root or embedded clauses. 59 Different predictions are made under an analysis à la Rizzi (1997), which I discuss below. 4.4.2.1 Problematic predictions of the template approach Rizzi’s (1997) cartographic approach constrains the distribution of left-peripheral topics by licensing them only at the edge of designated topic phrases (TopPs). 60 He proposes that TopPs can be projected exclusively on either side of a Focus projection, but that these TopPs are recursive. When the Topic-Focus field is activated, the map of the left periphery is said to be as in (4.118). (4.118)

[ ForceP Force [ TopP Top0 [ FocP Foc0 [ TopP Top0 [ FinP Fin0 [ IP . . . ]]]]]]

59 This is not to say that constraints of another type do not play a role in restricting the distribution of topics in embedded clauses. This issue is addressed in Section 4.4.3. 60 The default is for topics to appear in the specifier of TopP, but Rizzi also allows adverbial topics (such as temporal modifiers) to be adjoined to (a possibly empty) TopP and still check their Topic feature.

156

French Dislocation

However, this rigid structure does not capture the distribution of leftperipheral French topics. 61 Rizzi argues that ForceP is always the top projection of the clause: it encodes clausal type and has to be accessible to selection. He also claims that finite complementizers appear in Force0 and non-finite ones in Fin0 in order to derive fact that TopPs cannot follow a non-finite complementizer. This predicts that attested sentences like (4.119) do not exist, contrary to fact. In (4.119), the PP preceding the embedded clause receives a left-dislocated intonation and clearly modifies the embedded clause rather than the matrix clause. 62 (4.119)

Je pense sur son doigt, [ ForceP qu’ il montrait où est ce que c’ I think on his finger that he pointed where it is that it était sale]. (Catherine, C) was dirty ‘I think he was pointing to where it was dirty on his finger.’

A solution may be to assume that embedded clauses can take two complementizers, one (covert) in Force0 to encode the selectional and force features of the embedded clause, and the other (overt) in Fin0 , below the Topic-Focus field. However, examples such as (4.120) suggest that postulating an empty C in Force0 does not after all rescue Rizzi’s proposed cartography of the left periphery. (4.120) Je sais pas, les clients, [ ForceP ce qu’ils veulent]. I know not the clients that that-they want ‘I don’t know what the clients want.’ In (4.120), the DP les clients ‘the clients’ precedes the embedded object clause. This cannot simply be a parenthetical given the unacceptability of (4.121) (which can be straightforwardly explained if les clients is dislocated given that dislocated elements can only appear at the edge of Discourse projections, which do not include non-propositional complements). (4.121)

∗

Ilsi veulent, les clientsi , des figues. they want the clients some figs

The fact that examples like (4.120) are attested shows that topics must be allowed to be projected higher than ForceP: the dislocated element precedes 61 The cartographic approach also predicts that right-peripheral topics are in fact left-peripheral. This is discussed in Section 4.3.7. 62 This example, along with similar ones, was found in the York corpus.

Syntax

157

the complementizer in a relative clause, which Rizzi argues always sits in Force0 . Under Rizzi’s approach, dislocated topics require the projection of a Topic phrase but ForceP has to be the topmost projection of the clause for selection purposes. A second problem with the cartographic approach is that it does not explain why certain topics cannot occur in the clause containing their resumptive element, as shown by the contrasts between (4.122) and (4.123). In fact, nothing in that approach predicts the unavailability of TopicP in a clause containing a resumptive element, even when the latter is situated inside an island like the complex DP in (4.122a). (4.122) a. [Les malotrus]i , je connais pas grand-monde qui lesi aime. I now not great-people who them likes the louts ‘I don’t know many people who like louts.’ b. [La voiture jaune]i , j’étais triste de lai vendre. yellow I-was sad to it sell the car ‘I was sad to sell the yellow car.’ (4.123) a. ∗ Je connais pas grand-monde qui, [les malotrus]i , lesi aime. them likes I know not great-people who the louts b. ∗ J’étais triste de [la voiture jaune]i , lai vendre. I-was sad to the car yellow it sell A third problem for Rizzi’s analysis is that it requires that dislocated structures be treated as an exceptional type of A -dependency because they do not yield Relativized Minimality effects when they interact with other A -dependencies, including those arising from LD of additional elements. 63 4.4.3 French embedded Discourse Projections The extent to which dislocated elements are allowed at the edge of embedded clauses has, to my knowledge, never been fully investigated in the literature. While an in-depth investigation is beyond the scope of this study, I would like to make a few observations to pave the way for subsequent research. Under the present analysis, French dislocated elements are only allowed to appear at the edge of Discourse Projections. I follow Emonds (2004) in 63 In his original proposal, Rizzi (1997) claimed that adjacency effects resulted from the presence of Topic projections. However, this is contradicted by the data (see e.g. De Cat 2002 for the relevant French data). This proposal was amended in Rizzi (2002).

158

French Dislocation

assuming that only root and root-like clauses contain a Discourse Projection (corresponding to TP). The set of embedded clauses with root properties varies cross-linguistically (a point to which I return below). Emonds notes that such projections are finite, that they are usually complements rather than adjuncts, and that they are governed by V or A (rather than N or P), with some argument of the governing V being animate. I have proposed above that a distinctive trait of Discourse Projections is the [discourse] feature on T. In what follows, I provide a preliminary description of what counts as an embedded root in spoken French on the basis of the dislocation data. In the literature on embedded root phenomena, a number of conditions have been identified for an embedded clause to qualify as root-like. Hooper and Thompson (1973) argue that so-called Root Transformations (i.e. transformations that can only take place in root clauses, following Emonds (1970)) are only possible in embedded clauses to the extent that such clauses can be asserted. Typically, ‘embedded root clauses’ are indicative clauses selected by a verb of saying or a factive verb (see Heycock 2005 for a review of the literature on the subject). The group of embedded clauses allowing a left-dislocated topic in spoken French includes some which are not commonly classified as ‘embedded root clauses’. Left-dislocated topics appear in (i) certain subjunctive clauses (4.124a), 64 (ii) restrictive relative clauses (4.124b), (iii) clauses selected by a negated verb (4.124c), and (iv) clauses that are not assertive (4.124d), all of which fall outside the traditional definition of embedded root clauses. (All the examples in (4.124) come from the York corpus.) (4.124) a. Tu veux que moi, je le dessine? (Catherine, C) you want that me I it draw ‘Do you want me to draw it?’ b. Elle enregistre ce que toi, tu dis. (Denis, F) she records that that you you say ‘It records what you’re saying.’ c. Je savais pas que les cochons, ils avaient des salles de bain, I knew not that the pigs they had some bathrooms moi. (Nelly, B) me ‘I didn’t know pigs had bathrooms.’ 64

See also (4.21a).

Syntax

159

d. Et si moi, je viens et que je casse tous tes jouets, tu and if me I come and that I break all your toys you seras contente? (Nelly, B) will-be happy ‘And if I come and break all your toys, will you be happy?’ However, dislocated topics are banned from embedded non-finite clauses, as predicted by Emonds’ definition of Discourse Projections as inherently finite. 65 (4.125) a. ∗ J’ai dit de, les haricotsi , lesi équeuter. them tail I-have said to the beans b. ∗ J’ai peur de, moi, me couper. I-have fear to me refl cut A reviewer claims only LD is banned from non-finite clauses. It has however never been proved that RD is possible in non-finite clauses: in all instances where this has been claimed to be the case in the literature, nothing indicates that the RD really is in the embedded rather than the matrix clause. (4.126) Je n’ai pas dit [de les effacer], ces gribouillis. I NEG-have not said to them rub-off these doodles ‘I didn’t say to erase these doodles.’ LD and RD share essentially the same properties (as argued at length in this book). The default assumption should therefore be that RD is banned from non-finite clauses to the same extent that LD is. The matrix clause also has an impact on whether the embedded clause can take a dislocated element. In general, an embedded clause tends to have root properties when it conveys indirect discourse, i.e. when the embedding verb introduces reported speech. 66 Emonds’ generalization is that an embedded 65 The dislocated elements in (4.125) can only appear at the edge of the matrix clause (a Discourse Projection).

(i)

a. b.

Les haricotsi , j’ai dit de lesi équeuter. the beans I-have said to them tail ‘I have said to them to tail the beans.’ Moi, j’ai peur de me couper. me I-have fear to refl cut ‘I’m scared of cutting myself.’

66 It is not sufficient for an embedded clause to be selected by a verb like ‘say’ for it to qualify as embedded root, it also has to be finite, as illustrated by (4.125a).

160

French Dislocation

clause will have root properties if the governing verb has an animate argument serving as a subject of consciousness. (4.127) a. #Il faut empêcher que les myrtillesi , ils lesi cueillent toutes all it must impede that the bilberries they them pick aujourd’hui. today b. Ils ont dit que les myrtillesi , ils lesi avaient toutes all they have said that the bilberries they them had cueillies aujourd’hui. picked today ‘They said they had picked all the bilberries today.’ Non-object clauses can also be endowed with root properties. This is true of e.g. conditional clauses (4.124d) and relative clauses. Ease of identification of the topic’s referent facilitates its presence in a relative clause: dislocated elements referring to speaker or hearer are allowed more readily than third-person referents in general. It may well be the case that relevancetheoretic considerations have an impact on the acceptability of topics in such clauses. As a preliminary conclusion, French embedded root clauses do not have exactly the same characteristics as embedded root clauses given their usual definition in the literature. However, this might be due to the fact that embedded root phenomena have mainly been studied with respect to Germanic languages. Further research is clearly necessary to determine the extent of cross-linguistic variation regarding which embedded clauses can be endowed with root properties. The last remaining issue is how the proposed analysis accounts for which side of the clause periphery hosts which dislocated elements in spoken French. 4.4.4 Deriving the differences between LD and RD from the properties of the peripheries Right-peripheral topics are usually considered to have a slightly different status and function to their left-peripheral counterparts. Lambrecht, on the basis of his work on spoken French (Lambrecht 1981, 1986), argues that there are two types of peripheral topics depending on the side of the sentence in which they appear. Right-peripheral topics, which he labels antitopics, are said to differ from left-peripheral topics in several ways. (i) They are higher on the presuppositional scale: either the antitopic is given or it is easily recoverable (more so

Syntax

161

than a left-peripheral topic). (ii) Antitopics can never have a contrastive or an emphatic function, 67 and this is reflected in their lack of stress. (iii) They can have a stylistic function. (iv) They cannot be modified by adverbs like aussi ‘too’, seulement ‘only’, or même ‘even’. (v) They do not have a topic-shifting or topic-creating function. Ashby (1988) argues, against Lambrecht (1981), that all the functions characteristically associated with left-dislocated elements may also be associated with their right-dislocated counterparts. He suggests that right-dislocated elements do not instantiate a different type of topic to left-dislocated elements. 68 I will follow Ashby (1988) in rejecting the notion of antitopic. However, the differences between left- and right-peripheral topics (listed in (i)–(v) above) need to be accounted for. These will be shown to derive directly from the prosodic and the syntactic characteristics of the configurations in which LD and RD occur. 4.4.4.1 Prosodic properties and their consequences Prosodically, left-dislocated elements are particularly salient. They are typically stressed, uttered on a high pitch, and clearly set off from the rest of the utterance (see Section 2.3 for a description and discussion of the relevant diagnostics). By contrast, RD prosody is characterized by the absence of ‘distinctiveness’ of the pitch contour. In most cases, their melody is low and flat. The right-dislocated position is typically non-salient from a prosodic point of view (Ashby 1994; Rossi 1999; Mertens et al. 2001). As a result, any dislocated element requiring prosodic salience will obligatorily appear in the left periphery of the clause. The marking of contrast typically requires prosodic salience. Contrast can be marked in situ by adding stress to the word or syllable in order to emphasize it. In the sentence in (4.128), contrast is achieved simply by emphasizing the word Laken, as indicated by the capitals used for its first syllable. (4.128) Mais non, c’est le brigadier Azewé à LAEken. but no it-is the Brigadier Azewe at Laeken ‘No, it’s Brigadier Azewe at LAEken.’ The sentences in (4.129) illustrate the obligatory LD of contrastive XP topics. These three sentences from the York corpus were uttered one after the 67 Not all contrastive elements are necessarily emphatic: a contrast can arise out of implied comparison with other elements of a given set. 68 An analysis à la Vallduví (1992) is therefore ruled out from the start for the spoken French data. Vallduví (1992) postulates that the post-focal part of the ground, which he labels the tail, does not itself encode topics. Rather, tails provide details of how information related to the topic should be entered on the relevant card (corresponding to the discourse referent currently under discussion). See Chapter 3.

162

French Dislocation

other by a single speaker. Their topics are part of a set defined in the discourse (here, a family). Being mentioned one after the other creates a ‘list’ effect, implying a contrast between the members of the set. (4.129) Le grand frère, c’ est Lucas. L’ autre, c’ est Maxime. Et le the big brother it is Lucas the other it is Maxime and the petit dernier, c’ est Matéo. (A.-Gaël, F) little last it is Matéo ‘The oldest brother is Lucas. The other one is Maxime. And Matéo is the last one.’ Only the topic of the first sentence could appear as a right-dislocated element, provided that some big brother is salient in the context. The second and the third sentences obligatorily require LD. (4.130) C’ est Lucas, le grand frère. # C’ est Maxime, l’ autre. # C’ it is Lucas the big brother it is Maxime the other-one it est Matéo, le petit dernier. is Matéo the little last-one Another instance of clear contrast is that arising from the use of modifiers like aussi ‘too’ and même ‘even’. With one exception (see (4.133)), topics modified by such elements are obligatorily left-dislocated in spoken French. (4.131)

a. Et ceux de Luc aussi, ils sentent le fromage? and those of Luc too they smell the cheese b. #Ils sentent le fromage, ceux de Luc aussi? they smell the cheese those of Luc too ‘And Luc’s smell of cheese too?’

(Nelly, B)

Right-dislocated elements can encode emphasis only to the extent that the contrast can be expressed without the need for prosodic emphasis. This can be done by using a right-dislocated pronoun coindexed with a DP inside the sentence. (4.132) a. Mamani est malade, ellei . Mum is ill her ‘MUM is ill.’ en va pas, ellei . b. Minniei ne s’ Minnie neg refl of-here goes not her ‘MINNIE’s not going away.’

(Nelly, B)

(Nelly, B)

Syntax

163

It is therefore to be expected that right-dislocated pronouns should be the only elements allowing modification by aussi ‘too’. (4.133)

a. Crocroi , ili veut une poire, luii aussi. Crocro he wants a pear him too ‘Crocro too wants a pear.’ b. Ah tu sais ce que c’i est, çai aussi. ah you know that that it is that too ‘You know what that is too.’

(Catherine, C)

(A.-Gaël, F)

(4.134) a. #Ili veut une poire, Crocroi aussi. he wants a pear Crocro too ‘Croco too wants a pear.’ b. #Ah tu sais ce que c’i est, cette imagei aussi. ah you know that that it is that image too The direction of dislocation is thus partly determined by the prosodic requirements of the element. Prosodically salient elements can only be hosted by the left periphery given that this is the only periphery which allows pitch and intensity prominence. 4.4.4.2 General salience and its consequences The prosodic characteristics of left-dislocated elements, coupled with the fact that they are uttered before the core of the sentence, render these elements particularly salient. As a consequence, LD will be the preferred repair strategy in situations where the speaker had originally misjudged the level of recoverability of a topic. This is illustrated in the attested exchange in (4.135). (4.135)

Parent:

Child:

Parent:

Il est où, le bébé? Où il est? he is where the baby where he is ‘Where’s the baby? Where is he?’ Il est là, (l)a tête. he is there the head ‘The head’s there.’ (Picking up a bit of plasticine) Le bébé de Maman, il est où? the baby of Mum he is where ‘Where’s Mum’s baby?’

LD will also be preferred for topics when the speaker judges that the referent of the topic they introduce is not easily recoverable or identifiable.

164

French Dislocation

It is however not true that referents with a low level of salience in the context appear obligatorily in the left periphery. Contrary to what is often assumed in the literature (e.g. Lambrecht 1981; Grobet 2000), new referents can be encoded directly as right-dislocated topics, as in (4.136), for instance, which was uttered in a context where no plane had yet been mentioned. (4.136) Tu sais où il allait, l’avion que tu as vu qui you know where he went the-plane that you have seen that décollait? (Nelly, B) took-off ‘Do you know where the plane you saw taking off was going?’ On the whole, it is less clear what the function of right-dislocated topics is compared with their left-hand counterparts. When presented with minimal pairs involving a left- and a right-dislocated topic, native speakers tend to report an impression of markedness associated with left-dislocated topics. This impression is reported to be particularly strong in wh-questions, which might contribute to explaining why right-dislocation is so prevalent in that utterance type. 4.4.4.3 Linear order and its consequences We have seen that French rightdislocated elements tend to require the presence of syntactic dependency markers but that this is not obligatory. Examples from spontaneous production containing a right-dislocated element without a dependency marker are attested in the literature. Such examples are rare, but one has to bear in mind that right-dislocated elements bearing dependency markers are rare too (as shown in Appendix A). (4.137) a.

b.

Ça nous est égal, nous. that to-us is equal us ‘We don’t mind.’ On peut pas s’en servir, le reste. one can not refl-of-it serve the rest ‘We can’t use the rest.’

(Ashby 1988) (Ashby 1988)

It can be concluded that the presence of dependency markers on rightdislocated elements is preferred, but not obligatory. I will leave for future research investigation into why dependency markers are quasi-obligatory on right-dislocated but not left-dislocated

Syntax

165

elements (a question which, to my knowledge, has not been addressed in the literature). 4.4.5 Theoretical consequences Over the past two decades, a variety of phenomena have been taken to motivate the assumption that syntactic movement could take place via intermediate adjoined positions (for a comprehensive review of the relevant literature, see Sabel (2002)). Postulating the existence of intermediate traces in VP-adjoined positions was shown to explain a variety of phenomena (such as reconstruction effects, locality effects, and the absence of Weak Crossover effects) in certain configurations involving wh-movement and scrambling. Adjunction to VP could not however be left unconstrained. It was established by Chomsky (1986) among others that adjuncts could not be adjoined to adjuncts or to any XP requiring L-marking and that, more generally, movement from [spec,CP] to an adjoined position was ruled out. Sabel (2002) argues that these restrictions do not follow from strictly minimalist assumptions and that they give rise to a number of empirical problems. He defends the idea that movement may not proceed via intermediate adjunction and that successive-cyclic movement only targets specifier positions. I have presented clear evidence to the effect that French dislocated elements are adjoined to Discourse Projections and that, crucially, this type of adjunction does not involve syntactic movement. This has the following consequences. (4.138) a. b.

The ban on adjunction to adjuncts and to object clauses is a derivational constraint, not a representational one. XP-adjuncts need to be distinguished from specifiers: only the latter involve Agreement.

French dislocation has been shown to be possible at the edge of object and adjunct clauses. This, in light of Sabel (2002), suggests that the ban on adjunction to such clauses is not representational in nature (4.138a). The present analysis also provides further empirical support for Hoekstra’s (1991) arguments in favour of maintaining a distinction between XP-adjuncts and specifiers, based on the fact that only the latter involve syntactic agreement (4.138b). The possibility of left- and right-adjunction of dislocated elements does not contravene the Head parameter (which Saito and Fukui 1998 argue applies

166

French Dislocation

to adjuncts too) if the distinction between adjunction by movement vs. base-generated adjunction is maintained: only the former need incorporate the effects of that parameter.

4.5 Reconciling syntax and information structure In Section 4.3.4 I provided evidence that dislocated elements are not obligatorily reconstructed in spoken French, and argued that a movement analysis of French dislocation makes the wrong predictions with respect to Principle C effects. It is however possible to construct sentences which apparently violate Principle C. I will argue below that these are in fact due to clashes in information structure. Three sentences were tested (as part of the second online judgement elicitation task) to try to disentangle apparent Principle C effects from ‘information structural’ effects (henceforth IS effects). A Principle C effect is expected to arise if the dislocated element is reconstructed in a position c-commanded by a pronominal antecedent. The ‘reconstructed’ sentences are given in (b). (4.139) a.

b.

(4.140) a.

b. (4.141) a.

b.

Sans elle, on ne peut pas emmener les enfants de Rosa without her one neg may not take the children of Rosa au parc. to-the park ‘Without her, one cannot take Rosa’s children to the park.’ On ne peut pas emmener les enfants de Rosa au parc one neg may not take the children of Rosa to-the park sans elle. without her Le frère de Madeleine, elle ne parle pas de lui. the brother of Madeleine she neg speaks not of him ‘Madeleine’s brother, she doesn’t speak of him much.’ Elle ne parle pas beaucoup du frère de Madeleine. she neg speaks not much of-the brother of Madeleine Les costumes de Gaston, sans lui, on peut pas les the suits of Gaston without him one may not them repasser. iron ‘Gaston’s suits, you can’t iron them without him.’ On ne peut pas repasser les costumes de Gaston sans lui. one neg may not iron the suits of Gaston without him

Syntax

167

Informants were asked whether they allowed coreference between the proper noun and the pronoun. 69 Assuming a movement analysis of dislocation, coreference should yield a Principle C effect in (4.140a), as the R-expression Madeleine is c-commanded by the pronoun elle ‘she’ in the reconstructed version (4.140b), but there should be no Principle C effect in (4.139a) or (4.141a). The results were as follows. The seventy-five informants overwhelmingly chose disjoint reference for the R-expression and the pronoun for sentences (4.139a): eighty-two per cent and (4.140a): seventy-three per cent. 70 An explanation in terms of Principle C effects could account for this judgement in relation to (4.140a) but not (4.139a). Regarding (4.141a), in which there is no Principle C violation, informants overwhelmingly accepted coreference between the R-expression and the pronoun (seventythree per cent). The contrast between (4.139a) and (4.141a) is important to explain, as neither can be claimed to involve a Principle C effect. In Chapter 3, we saw that topics only need to be expressed with a dislocated element if they are not maximally salient/activated in the context, or if they are part of a set within which the speaker wants to identify a contrast. In sentence (4.140a), the use of a pronoun to designate the referents of the subject and the object indicates that these two referents are considered sufficiently salient by the speaker to be topics. This is verified by the straightforward acceptability of (4.142). (4.142) Madeleinei , son frère j , ellei ne parle pas beaucoup de lui j . of him Madeleine her brother she neg speaks not much ‘Madeleine doesn’t speak about her brother much.’ In (4.140a), the R-expression Madeleine, which is coreferential with the subject elle, is not introduced as a topic in its own right as it is in (4.142); it is embedded inside the DP coreferential with the object. This gives rise to an IS contradiction: on the one hand Madeleine is signalled as not being activated enough because it is encoded with an R-expression, but on the other hand the referent of Madeleine is the most prominent topic of the sentence as it

69

Literally, they were asked whether coreference was allowed, and then they were asked to choose between two options: option A if they thought, for instance, that elle ‘she’ could designate Rosa in (4.139a), and option B if elle ‘she’ had to designate another person not mentioned in the sentence. 70 The difference between these two results is not statistically significant: Chi-square = 1.37028807025059, p < 1. For significance at the .05 level, the chi-square value should be greater than or equal to 3.84.

168

French Dislocation

is coreferential with the subject. 71 This mismatch explains why native speakers are reluctant to allow coreference between Madeleine in the dislocated position and elle ‘she’ in the subject position in (4.140a). In (4.139a), where there is clearly no Principle C violation, IS effects can also explain why native speakers choose disjoint reference between the R-expression and the pronoun. In this case, the referent of Rosa is considered to be activated enough to be encoded directly using a pronoun in the dislocated topic position, but this clashes with the fact that it is encoded using an R-expression later in the main body of the sentence. Contrast this with the grammatical examples given earlier in the text: (4.143) a.

b.

c.

Tes sales petites remarques sur Léoni , ili ne les your dirty little remarks on Léon he neg them apprécierait sûrement pas. would-appreciate surely not ‘Léon would surely not appreciate your dirty little remarks about him.’ Le dernier livre que j’ai prêté à ma sœuri , ellei l’a the last book that I-have lent to my sister she it-has lu en une nuit. read in one night ‘The last book I lent her, my sister read in one night.’ [La meilleure toile d’un artiste j ]i , on ne lui j the best canvas of-an artist, one neg to-him pardonnerait jamais de l’i avoir vendue trop tôt. would-forgive never to it have sold too early ‘One would never forgive an artist for selling his best work too early.’

In these sentences, the R-expression is an adjunct inside a definite dislocated NP. It is also not the only modifier of the head noun. Coreference is allowed with a pronoun inside the sentence, in contrast to (4.140). This is because in (4.143), subordinate update is taking place: the R-expression is introduced into the discourse as part of the dislocated element and this allows it to be resumed by a pronoun later in the sentence. This would not be permitted if the dislocated element appeared in the right periphery of the clause (where subordinate update cannot take place). 71

Recall that the subject is by default the topic of the sentence cross-linguistically.

Syntax

169

Further research is needed to develop this analysis of IS effects. What is clear at this point, however, is that syntax alone cannot explain all the contrasts of acceptability in sentences involving dislocated elements.

4.6 Conclusion French dislocation has been shown to be a syntactically uniform phenomenon encompassing both LD and RD, and dislocations with either a clitic or a nonclitic resumptive. This phenomenon is uniform in the sense that all cases share the following characteristics: (i) the dislocated element expresses the topic of the sentence; (ii) the ‘resumptive’ element can be situated inside an island or inside a clause other than that which hosts the dislocated element; and (iii) dislocated constructions do not display the key properties of movement configurations: they do not license parasitic gaps, do not give rise to Weak Crossover or Minimality effects, and are not interpreted via reconstruction. The term ‘resumptive’ has been argued to be something of a misnomer in this case, given that the clause-internal element which is coreferential with the dislocated XP is not interpreted as a genuine resumptive pronoun in the sense of Sells (1984), nor does it behave as expected following Boeckx (2003). Indeed, there is no syntactic requirement for the dislocated element to be present: any sentence containing a dislocated element would be equally acceptable (from a syntactic point of view) if that dislocated element were deleted. The so-called ‘resumptive’ element is best analysed as a fullfledged (though possibly deficient) pronoun interpreted as a discourse-level anaphor. The analysis proposed is that French dislocated elements are base-generated by adjunction to root-like clauses. The distribution of dislocated elements is determined by their own discourse properties and those of the clausal projection with which they combine to form a Predication in the sense of Chomsky (1977). While there are still some issues concerning root-like embedded clauses that remain to be resolved, an analysis of French dislocation as essentially a root phenomenon seems to be on the right track. Such an approach offers a more principled way of accounting for the distribution of peripheral topics than one which assumes that TopicPs can be projected at the edge of any clause unless they violate a syntactic requirement (such as adjacency). One of the advantages of the proposed analysis is that it allows a more efficient division of labour between syntax and information structure. This is highly desirable given the influence which factors such as

170

French Dislocation

the ease of identification of corresponding discourse referents have on the distribution of dislocated elements. On such an analysis, it is possible to postulate highly specialized components of the language faculty, thus maximizing economy and allowing the potential of the interfaces between these components to be fully exploited. In particular, a direct interaction between information structure and the lexicon is desirable— and perhaps indispensable—if one is to take the Inclusiveness Condition seriously.

5 Acquisition 5.1 Introduction The study of child language and that of the target system build on each other. Child language research relies on an informed study of the target system; in return, because the child’s linguistic production is not entirely target-like but constrained by universal rules, its study provides the researcher with a privileged means of investigating Universal Grammar (UG), hence allowing refinements of the analyses that fed it in the first place. This chapter focuses on the following questions:

r Can the acquisition data contribute evidence for or against an analysis of dislocated structures in the adult grammar?

r How target-like is the early dislocation data? Do children start out with a ‘primitive’ form of predication?

r What does the acquisition of dislocated topics contribute to our understanding of the architecture of the language faculty? (This issue will be approached from the perspectives of both learnability, i.e. the mechanisms underlying language acquisition, and the initial state, i.e. the knowledge available from the outset.) The path to acquisition of dislocated constructions is outlined in Appendix B.6. 1 Although children do not begin by producing all the types of dislocated structures present in the target grammar, they do not produce non-target-like dislocated structures in any systematic way. My assumptions regarding the acquisition process are as follows. One of the principal aspects of language acquisition is the evolution of the child’s grammatical system into the target adult system. I adopt the widely held view that the initial state of language acquisition is more highly constrained than the subsequent states (O’Grady 1997). The child’s successive grammars are all 1 A detailed review of the literature on peripheral elements in child French has not been included here: see De Cat (2002), which contains a more extensive acquisition chapter. Only references directly relevant to the discussion at hand have been included in the present volume.

172

French Dislocation

subsets of the target system. Evolution from one stage to the next is triggered by positive evidence gathered by the child from the input he or she receives. I also adopt a weak version of the Continuity Hypothesis (Pinker 1984), whereby UG is available to the child from the onset, but he or she cannot initially make use of all this knowledge in his/her syntactic representations. Economy is an essential characteristic of the adult grammar (Chomsky 1995): ‘UG is biased toward minimal derivations’ (Pierce 1992: 2). This holds for the developing system of the child as well. As a general principle, I will assume that the child’s grammar is only minimally different from the adult’s: in the absence of evidence to the contrary, I will take the child’s internalized representation of a given construction to be adult-like. The research presented in this chapter is based essentially on the longitudinal study of four children from three different dialectal areas: Max (Canada), Anne (France), and Léa (Belgium) from the York corpus, and Tom (Belgium), from the Cat corpus. Details regarding recording, transcription, coding, and analysis can be found in Appendix A. Information about the children studied (including profiles, ages, and Mean Length of Utterance) can be found in Appendix B, which also contains a list of the transcription conventions used in the examples.

5.2 Identifying early dislocated elements Following the Continuity Hypothesis (Pinker 1984; Lust 1999), I assume that, in the absence of evidence to the contrary, the child’s internalized representation of a given structure that he or she uses productively should be treated as adult-like. The implications for the present study are that the diagnostics used to identify dislocated elements in the adult data should hold for the child data as well, modulo independently explained differences between the two systems. I review below the adult diagnostics in light of the child data. The following have been argued to be reliable diagnostics for dislocated elements in the adult data (see Chapter 2). (5.1)

a. b. c. d. e.

Omissibility of the dislocated element Presence of a resumptive element Word order/intervening material Context Prosody

5.2.1 Omissibility Dislocated elements can be omitted without altering the syntactic or prosodic well-formedness of the sentence (see Rossi 1999 and Chapter 4). This

Acquisition

173

diagnostic should be used with caution when applied to the child data, as it depends on well-formedness judgements. 5.2.2 Resumption The presence of a resumptive element indicates the dislocation of the clauseperipheral phrase coindexed with it. In the early stages, children frequently omit subject and object clitics (for child French, see e.g. Ferdinand (1996); Rasetti (2000); Plunkett (2003) on the omission of subject clitics, and Hamann et al. (1996); Müller et al. (1996); Hamann (2002) on the omission of object clitics). Such omissions occur independent of whether dislocated elements are present (Plunkett and De Cat 2001). Whenever a resumptive element was realized in the child data, a coindexed XP apparently in the clause periphery was considered to be dislocated (on continuity assumptions regarding the child’s grammar). Embryonic forms appearing in a clitic position, such as e in (5.2), are treated here as fully-realized clitics (though they were coded distinctly, which made it possible to confirm that treating them as missing elements did not affect significantly the results of any analysis). (5.2)

Maman, e fait xx, # moi, e fais 0 d(r)apeau mummy she makes xx me e make flag ‘Mummy makes the xx and I make the flag.’

(Anne 2;2.0)

When the child does not utter any resumptive for a possibly dislocated element, as in (5.3), the other diagnostics listed in (5.1) have to be relied upon exclusively. On this basis, I have demonstrated elsewhere (De Cat 2004a) that strong pronouns which apparently occupy subject position, as in (5.3) below, are in fact left-dislocated subjects with a missing resumptive and not true subjects in the default case (contra Schütze 1997 and Wexler et al. 1998). (5.3)

Moi # est capable. me is able ‘I can (do it).’

(Max 2;2.22)

5.2.3 Word order and intervening material Word order allows dislocated elements in child French to be identified in one of two possible ways: either (i) the constituents do not appear in the canonical SVO order, or (ii) some element intervenes between the verb and one of its apparent arguments.

174

French Dislocation

In the first case, the subject may, for instance, appear to be postverbal (like moi ‘me’ in (5.4)). Such word order has been analysed by Friedemann (1992) and Déprez and Pierce (1993) as involving a true postverbal subject, but this analysis has subsequently been shown not to be viable by Ferdinand (1993, 1996) and Labelle and Valois (1996), who have convincingly argued that the apparent postverbal subject is in fact right-dislocated. (5.4)

Prendre, moi, # a balle. e ball take−FIN me ‘I (want to) take the ball.’

(Tom 2;1.11)

There are also instances where the object appears preverbally in the absence of a resumptive clitic, in which case two analyses are possible: either the leftperipheral element has been focus-fronted, or it is dislocated. In the first instance, it should bear focal stress; in the second instance, it should receive dislocation prosody. How one decides between these two analyses is discussed at length in De Cat (2006). (5.5)

Alors, la soupe de poissons, on va mettre là? so the soup of fish one will put there ‘And then the fish soup, shall we put it there?’

(Christine, F)

In the second case, material intervenes between the dislocated element and the apparent subject or object. In (5.6), for instance, there is a moved whword between the clause-initial DP (coreferring with the subject) and the verb. This element is thus obligatorily dislocated, despite the unclear presence of the resumptive clitic. (5.6)

Mes petites sandalesi , où (el)l(e)i est, les my little sandals where it is the +/? (interruption) ‘Where are my little sandals?’

(Anne 3;1.4)

The dislocated element may also be present in a clause other than that in which its resumptive element is expected to appear, as in (5.7a). In some cases, the embedded clause is not pronounced, as in (5.7b). These sentences are targetlike.

Acquisition (5.7)

175

a. Les enfantsi , (il) vaut mieux qu’ ilsi partent à un pas at one not the kids is-worth better that they go (Anne 3;2.29) xx +//. ?? ‘It’s better that the kids go to (??).’ b. Çai , c’ est moi (qui l’i ai fait). (Max 2;0.28) that it is me who it has done ‘It’s me who’s done that one.’

5.2.4 Context The context permits the disambiguation of many otherwise equivocal examples. In (5.8a), Papa ‘daddy’ might at first sight look like a dislocated subject (i.e. the sentence would mean that the dad is not holding something, though we would not know what as the object would be missing), but the context clearly indicates that it is in fact a vocative. In (5.8b), the sentence-initial name could be taken to be the subject of the sentence (i.e. the child would be using her family name to designate herself, and the sentence would mean, ‘Manoliale tells you [something: missing object]’), but the proper name is a focus-fronted element, given in answer to a question from the child’s mother (‘What did you say?’), so it cannot be a dislocated element. (5.8)

a. Papa, i(l) tient pas. (Anne 2;2.0) daddy he holds not ‘Daddy, it won’t stay put.’ (Talking about an alarm clock) b. (Ma)no(l)iale, (je) te dis. (Anne 2;6.1) [her name] (I) to-you say ‘Manoliale, I tell you.’

Similarly, in (5.7b), repeated below as (5.9), the context is that the child is pointing at several pictures and explaining who painted which one. The dislocation is thus not a subject dislocation (i.e. the picture does not represent the child himself) but an object dislocation (i.e. it designates the drawing) linked with an unpronounced dependent clause. (5.9)

Çai , c’ est moi (qui l’i ai fait). that it is me who it has done ‘It’s me who’s done that one.’

(Max 2;0.28)

Discourse factors can also contribute to identifying dislocated elements. For instance, a dislocation analysis will be reinforced by a context which allows a topic reading of the element in question. This diagnostic can usefully

176

French Dislocation

complement others which are not clear-cut, as is the case for (5.10) for instance, where the subject position is filled by an embryonic element. (5.10) Boucle d’or, e voulait aller où? Goldilocks e wanted to-go where ‘Where did Goldilocks want to go?’

(Anne 2;7.1)

5.2.5 Prosody It is traditionally assumed that children master the essential aspects of intonation while still at the prelinguistic stage (e.g. Bever et al. 1971; Dore 1975; Locke 1983; Crystal 1986). A vast body of evidence, presented in the review article by Snow and Balog (2002), suggests that ‘children acquire some core aspects of intonation before the onset of expressive syntax’ (Snow and Balog 2002: 1055), i.e. just before they start combining words. In particular, at this point in their development children have been shown to have mastered the use of prosody to signal utterance boundaries in a clearly distinct fashion to the boundaries of non-utterance-final Intonation Groups. The preliminary prosodic analysis reported in De Cat (2002) suggests that, from the onset of expressive syntax, the children recorded for the York and Cat corpora used prosody in an adultlike fashion to signal dislocated elements. While a full-scale study is necessary to confirm full adult-like mastery of dislocation prosody, I believe this preliminary analysis is in accordance with the above-mentioned findings, and that it suggests that children at the two-word stage use intonation in a target-like fashion to mark Intonation Groups as either dislocated or utterance-final. 2 Typical examples of LD prosody are given in Figures 5.1 and 5.2 for a child (Lisette) and her mother (Audrey) respectively. These are extracted from the Canadian section of the Cat corpus of spontaneous production. The corresponding glossed transcriptions are given in (5.11) and (5.12). In the figures, the thick line represents the pitch (i.e. variations in fundamental frequency) and the thin line the intensity. 3 Each syllable (indicated by vertical lines) is identified separately on the basis of the way the string was pronounced rather than its orthography. In the corresponding text, unpronounced material is given in parentheses. (5.11) Moi, je veux aller chez Solène. [H+H+] [l . . . /l] [h /HH l . . . \l L−L−] ‘I want to go to Solene’s!’

(Lisette, 2;9)

2 I want to stress that I am not claiming that all aspects of prosody are mastered by children at this point. 3 See Section 2.3 for details.

177

25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 moi 2 1 0 0

80 75 70 65 60 55 50

Intensity (dB)

semitones

Acquisition

45 40 je

veux

a-

ller

chez

So-

35

-lène

30 0.5

1

1.5 time(s)

2

2.5

80

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

75 70 65 60 55 50 45 40 moi

j’aim(e ) rais qu(e) tu

m’ex-

pliques

pa(r)c(e) que... 35

30 0

0.5

1

1.5

time(s)

Figure 5.2. The prosody of left-dislocated moi ‘me’ (spontaneous, Audrey)

Intensity (dB)

semitones

Figure 5.1. The prosody of left-dislocated moi ‘me’ (spontaneous, Lisette)

French Dislocation 85

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

74

63

52

INTENSITY (dB)

PITCH(semitones)

178

41 j(e) (l)a

co(n)-

-nais

pas

hein

moi 30

0

0.5

1 TIME(s)

Figure 5.3. Prosody of a declarative sentence with a right-dislocated element (spontaneous, Chloé)

(5.12) Moi, j’ aimerais que tu m’ expliques [/HH] [l /LL] [\l . . . l HH] parce que (...)4 l...l ‘I’d like you to explain because . . . ’

(Audrey, Mother)

Typical examples of RD prosody in the speech of another child and her mother from a different dialectal area (Belgium) are given in Figures 5.3 and 5.4, with the corresponding glossed transcriptions given in (5.13) and (5.14) respectively. (5.13) Je la connais pas, hein, moi. (Chloé, B, child) I her know not eh me ‘I don’t know her, do I?’ (5.14) Ah ben non ili n’a pas de chapeau, ah well no he neg-has not part hat (Geneviève, B, mother) celui-cii . this-one ‘Ah this one doesn’t have a hat.’ (surprise intonation) 4 The adjoined clause introduced by parce que ‘because’ has not been included in the acoustic analysis because it would make the sentence too long for clear presentation in a single figure. The pitch of the remainder of the sentence is lower than that of expliques ‘explain’, so there is no grouping together of the first part of the sentence (the matrix clause) and the following part to form a larger package.

179 85

20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

74

63

52

INTENSITY (dB)

PITCH(semitones)

Acquisition

41 ah 0

ben

non i(l) n’a pas d(e) cha-

-peau 1

0.5

c(el)ui-

-ci 30 1.5

TIME(s)

Figure 5.4. Prosody of a declarative sentence with a right-dislocated element (spontaneous, Geneviève)

The prosody of the right-dislocated element of the sentence in (5.13), represented in Figure 5.3, exhibits the classic low, flat pitch reported in the literature (e.g. Delattre 1966) accompanied by a decrease in intensity. The dislocated element is preceded by a tag hein ‘eh’, which is typically found between dislocated elements and the sentence nucleus in spontaneous speech. As indicated by the unbroken line representing the pitch contour in Figure 5.3, there is no pause between the dislocated element and the sentence nucleus.

5.3 Dislocations emerge early Dislocated elements appear as early as children start assembling words into utterances (see Appendix B.6). The early emergence of dislocated structures raises important questions regarding their representation, both in the child and the adult grammar. (i) Does the child start with an adult-like representation of these structures? (ii) What kind of positive evidence (in the input) could the child rely on to acquire the target-like representation? (iii) Is there evidence for such a representation on the basis of the early data?

180

French Dislocation

The first question will be addressed in Section 5.5 on the basis of an investigation into the most puzzling type of early data, i.e. dislocated elements appearing in verbless utterances. The answer to the second question depends on the analysis proposed. In languages like French, nothing could possibly indicate to the child that dislocated structures require the involvement of a functional projection in the Cdomain or the v-domain. The head of each of these projections is not filled by any overt element when a dislocated element is present, and there is no overt manifestation of a possible Agree relation between such a head and an element of the topic chain. Indeed, the functional projection analysis is supported by theoretical arguments more than empirical evidence (at least in the case of languages like spoken French). On the other hand, under the adjunction analysis, one could argue that the evidence required consists of configurations in which the resumptive element is situated inside an island. Tokens of this type are however extremely rare in spontaneous production (and unattested in the adult sample considered here) and could therefore be argued to provide insufficient cues for of the target-like representation. However as I suggest in Section 5.6, there are clear indications that adjunction would be postulated by default by the child, thereby resolving the learnability problem. The third question is addressed below.

5.4 Early dislocations and the CP projection In this section, I show that dislocations are used productively by children well before there is any concrete evidence of implementation of the CP layer. This provides further support for the adjunction analysis proposed in this book, and raises further potential problems for analyses that rely on the involvement of functional projections in the C- and/or the v-domains. The non-implementation of a layer does not imply that that layer or the (functional) projections it contains are absent from the child’s grammar (as pointed out by e.g. Hoekstra and Jordens (1994)). Indeed, it is possible that at the initial state the child’s grammar contains all functional nodes courtesy of UG. I will assume however that if there is no evidence that the child is projecting a particular layer (here CP) before time X, it is reasonable to postulate that any functional head within that layer is not yet ‘fully active’, which might be due to the fact that the features of that head are not set yet (along the lines of Ferdinand (1996)). In this spirit, I discuss below a number of diagnostics which have been proposed in the literature for the implementation of CP, and gather evidence of the (gradual) implementation of this functional layer. The results are discussed in Section 5.4.2.

181

Acquisition Table 5.1. First signs of CP implementation Diagnostic

Anne

Max

Tom

Léa

(i) [−T] embedded clauses (ii) Overt [−Fin] complementizers (iii) [+T] embedded adjunct clauses [+T] embedded object clauses (iv) Moved wh-elements (v) Overt [+Fin] complementizers (vi) Inversion (I-to-C) (vii) (wh+)est-ce que questions (viii) Embedded wh

?2;1.19 ?2;4.20 2;1.19 2;6.2 ?2;4.2 2;4.20 ?2;7.16 2;8.20 2;11.2

?2;0.14 2;0.28 2;3.6 2;3.6 ?2;5.1 2;3.6 2;3.20 2;6.12 2;5.1

2;1.13∗ ?2;3.22 2;1.13∗ 2;3.22 2;1.11∗ 2;5.24 — 2;8.0 ?2;3.22

2;8.22∗ 2;8.22∗ 2;8.22∗ 2;8.22∗ 2;8.22∗ 2;8.22∗ 2;10.7 ?2;9.5 ?2;9.21

5.4.1 Eight diagnostics for the implementation of CP There is no possible evidence of the non-projection of CP in child French. The only way to determine if CP is projected is therefore to rely on ‘positive’ diagnostics. Such diagnostics can only indicate that CP is implemented from time X at the latest; they do not provide firm evidence that before time X CP was not projected by the child. What I am trying to determine is the age at which the child is able to produce sentences that clearly involve the CP layer; I am not trying to determine when each construction is used fully productively. Implementation of CP might still be tentative. That being said, when the obligatory contexts for a given diagnostic are present only rarely in the data, that diagnostic will be considered to be a less reliable indicator of earliest genuine implementation of CP. Repetitions of adult utterances will not be taken into account. The diagnostics considered are summarized in Table 5.1. Justification for their use will be provided in the text below. A question mark before the age of the child indicates that the corresponding diagnostic cannot be taken as firm evidence of the projection of (at least some of) CP, as pointed out in the discussion below. An asterisk after the age of the child indicates the first recording session. Note that for Tom there are no data available at present between the ages of 2;10.8 and 3;0.5. I have treated his first two recording sessions as one, given that they take place at only two days interval. The first diagnostic is based on the emergence of non-finite embedded clauses in the data. The motivation behind this diagnostic is that children might truncate root clauses, but not dependent clauses (following Rizzi 1994).

182

French Dislocation

Therefore any non-root clause produced by the child is relevant to this diagnostic. Dependent clauses are all non-finite initially. Before 2;1.19, the only attested example for Anne is that in (5.15). Her first productive examples are given in (5.16). (5.15) Viens voir! come to-see ‘Come and have a look!’ (repeated twice) (5.16) a. (Je) veux voir. I want to-see ‘I want to see.’ b. Je veux fermer. I want to-shut ‘I want to shut (it).’ c. (Je) vais manger # à table. I will eat at table ‘I’m going to eat at the table.’

(Anne 1;11.13) (Anne 2;1.19) (Anne 2;1.19) (Anne 2;1.19)

Max’s first productive non-finite embedded clauses are also his first ones. (5.17) a. (Je) veux (le) flatter. I want him to-stroke ‘(I) want to stroke (him).’ b. xx, moi, (je) (suis) pas capable # de l’ ouvrir. xx me I am not able to it open ‘(I’m) not able to open it.’ c. Moi # c’ est moi # tombE ça. me it is me fall−FIN that ‘I’m the one (who’s made) that fall.’

(Max 2;0.14) (Max 2;0.28) (Max 2;1.16)

Tom’s first non-finite embedded clauses are given in (5.18): 5 (5.18) a. (L)a p(l)asticine a sortir. the plasticine will? take-out ‘The plasticine, we (need to? / will?) take (it) out.’

(Tom 2;1.13)

5 Tom produces quite a few unclear cases like (5.18a) in the first two recordings. Like (5.18a), they contain what looks like a modal verb which is not fully uttered followed by a non-finite verb.

Acquisition b. C’ est mieux # (de) faire # comme ça. it is better to do like that ‘It’s better (to) do like that.’

183

(Tom 2;4.9)

Léa produces non-finite embedded clauses from the first recording. (5.19) Et il faut faire euh [//] aussi il faut en faire un à and it must make er too it must of-it make one to Minnie. (Léa 2;8.22) Minnie ‘And we must make one for Minnie too.’ Using the emergence of non-finite embedded clauses as a diagnostic for the implementation of CP faces (at least) one potential problem: it is not clear how much structure the child projects to produce strings like veux flatter ‘want to stroke’ (5.17a) or vais manger ‘will eat’ (5.18a). These control configurations might be monoclausal structures. 6 If not monoclausal, the absence of any lexical material filling the embedded Comp position and its specifier may suggest initially to the child that such embedded clauses only involve IP. Hence only non-finite clauses with a filled Comp would provide clear evidence that such clauses are CPs. This brings us to the second diagnostic. The second diagnostic requires the child to produce overt non-finite complementizers. Müller (1994) argues that children analyse root clauses as being CPs from the moment they start producing embedded clauses with overt complementizers. She further argues that, in the (German-French) bilingual children she studied, complementizers develop out of prepositions like pour ‘for’, i.e. out of a lexical category. It is not clear however whether pour would even be in C under such a hypothesis, and hence the clause it selects could still be IP. 6 In De Cat (2000) I propose a restructuring analysis of these configurations on the basis of evidence from the distribution of floating quantifiers, which appear to ‘climb’ to the front of the higher nonfinite verb (e.g. Elle a tout voulu planter. ‘Lit: She has all wanted to plant.’). The distribution of clitics, however, cannot be accounted for under such an analysis, as they cannot climb that high in similar structures. A principled explanation for this (and the differences between French and other Romance languages like Italian and Spanish with respect to restructuring) is given in Emonds (2000: Chapter 6). However, until the 16th century, French allowed clitic climbing in these constructions, as in (i) from Les Cent Nouvelles Nouvelles de Vigneulles (Labelle and Hirschbühler in preparation: 430–31).

(i) Il ine le pouvoit oublyer. he neg it could forget ‘He couldn’t forget it.’

(Labelle and Hirschbühler in preparation: 431)

It would be interesting to investigate whether restructuring has entirely disappeared from the grammar of French (as suggested by the clitic data) or whether it is still allowed in some contexts (as suggested by the floating quantifier data).

184

French Dislocation

The use of pour ‘for’ to introduce a non-finite clause is attested before the use of true non-finite complementizers in the speech of Anne and Tom. The first instance is given in (5.20a) for Anne and (5.20b) for Tom. (5.20) a. Pour manger. (Anne 2;4.20) to eat ‘To eat.’ b. Euh pour donner. (Tom 2;3.22) er to give ‘Er to give.’ (answer to the question: C’est quoi, ça? ‘What’s that?’) Max’s first clear non-finite complementizer appears at age 2;0.28 (5.17b), 7 but the next case is not attested until a few months later. The number of obligatory contexts for the use of non-finite complementizers varies greatly from child to child, and makes it difficult to judge when these elements truly emerge in the child’s language. There is no evidence of complementizer de in Tom’s data during the period studied. In Anne’s, there are only two instances before the age of 2;10. These are given in (5.21). Notice that in the first case, the complementizer is omitted in the repetition. (5.21) a. I(ls) n’ ont pas envie # de &man [//] pas envie (de) they neg have not desire to eat not desire to manger. (Anne, 2;5.4) to-eat ‘They don’t want to eat.’ b. c’ est qui, en train de tomber? (Anne 2;8.20) it is who in process to fall ‘Who’s falling?’ Clear non-finite complementizers are productively used by Léa from the start of the recordings (5.22). 8 (5.22) Je l’ essaye de le faire dormir mais elle veut pas I it try to it make sleep but she wants not dormir. to-sleep ‘I try to make her sleep but she doesn’t want to sleep.’

(Léa 2;8.22)

7 I assume that de is a non-finite complementizer in French (e.g. Kayne 1994). This is however not uncontroversial. Kayne himself has held different views over the years. See Kayne (1991) for the view that de is a preposition. 8 In (5.22), the presence of the first object clitic is non-target-like. Notice also the mismatch in gender features between the object clitic and the coreferential subject of the conjoined clause.

Acquisition

185

The third proposed diagnostic for the implementation of the CP layer is the emergence of embedded finite clauses. I consider the dependent clause of a cleft to fall within this category. The first cases attested in Anne’s data are given in (5.23). (5.23) a. C’ est (le) papa, # 0 va manger. it is the dad will eat ‘It’s (the) dad (who) will eat.’ b. C’ est lui qui fait dodo sur lui. it is him who does sleep on him ‘It’s this one who sleeps on that one.’ c. C’ est lui, elle s’ appelle (Pimprenelle). it is him she refl calls Pimprenelle ‘It’s that one who’s called (Pimprenelle).’

(Anne 2;1.19) (Anne 2;4.20) (Anne 2;6.2)

Note that in two of these utterances the complementizer is non-overt. Only (5.23b) features an overt complementizer. All three cases are attempted subject-cleft constructions. Max’s first finite non-root clauses are given in (5.24). (5.24) a. C’ est lui, ça va là? it is him it goes there ‘Is it him, (the one that) goes there?’ b. C’ est quoi, ça # qui est là? it is what that that is there ‘What’s that thing there?’

(Max 2;3.6) (Max 2;3.6)

The sentence in (5.24b) contains a relative clause; the exact structure of (5.24a) is unclear. The prosody and the context suggest that it is an attempted pseudocleft involving the left-dislocation of celui qui va là ‘the one that goes there’. However, only the construction in (5.24b) features a complementizer and constitutes clear evidence for the projection of CP. Tom’s first finite embedded clauses are given in (5.25). 9 (5.25) a. C’ est Bruno [!], # 0 a mise comme ça. it is Bruno has put p. p. f. like that ‘It’s Bruno (who)’s put (it) like that.’

(Tom 2;1.13)

9 I have excluded the following utterance because the right-dislocated tu sors ‘you go out’ is in all likelihood unanalysed by the child, who is asking who tu sors is as if it were a person.

(i) C’ est qui # tu sors? it is who you go-out ‘Who’s that, you go out?’ (clarification question)

(Tom 2;1.14)

186

French Dislocation b. xx [=? veux] voir # quoi # il y a ici. want to-see what it there has here ‘(I) want(?) to see what there is here.’ c. C’ est Mamy [!] 0 a mangé. it is Mamy has eaten ‘It’s Mamy (who) has eaten (it).’

(Tom 2;3.22) (Tom 2;3.22)

The sentences in (5.25a) and (5.25c) are contrastive subject clefts with a missing complementizer. The one in (5.25b) contains an indirect question with targetdeviant fronting of quoi ‘what’. Note that so far, with the exception of (5.25b), the only embedded contexts in the children’s data are adjoined clauses and not selected ones (assuming an adjunction analysis of clefts). Anne’s first finite object clauses are given in (5.26), Max’s in (5.27), 10 and Tom’s in (5.28). (5.26) a. Et on voit (que) le chat est dans la barrière. and one sees that the cat is in the fence ‘And we can see the cat’s in the fence.’ b. E veux (qu’) (elle) y aille # en voiture. e want that she there go-SUBJ in car ‘I want her to go there by car.’ (5.27) a. Moi, xx [?= va] montrer à # Maman # c’ est quoi. me ? will show to Mum it is what ‘I’ll go and show Mum what it is.’ b. On voit pas # qui fait du bruit xx. one sees not who makes some noise xx ‘We can’t see who’s making noise.’ (5.28) a. xx [=? veux] voir # quoi # il y a ici. ? want to-see what it there has here ‘(I) want(?) to see what there is here.’ b. Regarde ce [?] que j’ ai +//. that I have look that ‘Look what I have . . . ’ (self-interruption)

(Anne 2;6.2) (Anne 2;7.1)

(Max 2;3.6) (Max 2;5.1)

(Tom 2;3.22) (Tom 2;5.24)

10 Max also uttered the sentence below, in which it is not clear whether the embedded clause is an object clause that itself contains a missing direct object, or a free relative clause selected by the verb (and interpreted as an indirect question) from which ce ‘it’ is missing.

(i) Comme ça, 0 raconter 0 que moi (j) (ai) vu 0. like that tell−FIN 0 that me (I) (have) seen 0 ‘Like that (I want to?/will?) tell what I’ve seen.’

(Max 2;3.6)

Acquisition c. Je veux (que) # tu # ouvres # le couvercle. I want that you open the lid ‘I want you to open the lid.’

187

(Tom 2;5.24)

The complementizers are generally missing from these first finite object clauses, so it is possible that such clauses do not involve a full CP after all. Léa produces finite embedded clauses from the first recording. These include adjunct clauses (5.29a) and object clauses (5.29b). In (5.29a), Léa uses a target-deviant resumption strategy (which Labelle 1990 argues is commonly used in the development of French relatives) instead of the expected agreeing complementizer qui that would license the gap in subject position (following Rizzi 1991), but even a non-target-like C can be taken to indicate the projection of CP. (5.29) a. C’ est moi que je l’ ai [//] c’ est moi que je les ai it is me that I it have it is me that I them have fermées. (Léa 2;8.22) closed ‘It’s me who’s closed them.’ (She means she has shut the pearl box.) (Léa 2;8.22) b. Il faut bien qu’ on donne une perle, # Mamy. it must well that one give a pearl Mamy ‘We must give a pearl, Mamy.’ The fourth proposed diagnostic relies on the overt movement of wh-elements. I have chosen not to consider in situ wh-elements as a potential diagnostic because of the difficulty of finding tangible evidence that they involve the CP projection in early child language. Before the ages of 2;4.2 and 2;5.1 respectively, Anne and Max use the in situ strategy exclusively to form wh-questions (see Plunkett 1999). Tom and Léa produce wh-questions which involved movement from the first recording. Note however that overt wh-movement is not in itself sufficient evidence that [spec,CP] is projected (contra Crisma 1992). This point is made with respect to child French by Hulk (1996b), who argues that initially children’s wh-movement may target [spec,IP] rather than [spec,CP] (I return to this issue presently). Plunkett (1999) proposes that wh-movement is used as a routine until both strategies (movement and in situ) are implemented by the child. Tom’s first in situ wh-word is attested at 2;1.11; Léa’s is attested at 2;8.22. In both cases this corresponds to the first recording session. However, even target-like movement might not be sufficient evidence that CP is projected in the children’s moved wh-questions: Plunkett (2001) argues that wh-features can sometimes appear on I even in adult French, in cases where the root

188

French Dislocation

complementizer is not projected. Given the above, this diagnostic cannot be taken as firm evidence that CP is projected. The use of finite complementizers as a diagnostic for the implementation of CP (the fifth diagnostic) requires caution. It is well known that the finite complementizer que ‘that’ can be omitted in object clauses in Canadian French. The absence of que ‘that’ from Max’s data in these contexts thus cannot be judged non-target-like. A corollary is that to the extent that the adult clauses without que involve CP, that projection may be implemented by Max earlier than the first attested que in his production. The sentences containing the children’s first two instances of overt finite complementizer are given below (see (5.29) for Léa’s). (5.30) a. C’ est lui qui fait dodo sur lui. it is him who does sleep on him ‘It’s this one who sleeps on that one.’ b. Mettre ça [= clothes] # parce que i(l) s’ put that because he refl habille. dresses ‘(I’ll) put that because he’s getting dressed.’ c. C’ est quoi, ça # qui est là? it is what that that is there ‘What’s that thing there?’ d. Ah oui # tu vois # parce que c’ est vert. ah yes you see because it is green ‘Ah yes you see, because it’s green.’ e. Les petits poissons qui +//. the little fish that ‘The little fish that . . . ’ (self-interruption) f. (Je) veux voir de l’ eau ### qui sort. I want to-see part the water that goes-out ‘I want to see water going out.’

(Anne 2;4.20)

(Anne 2;6.18) (Max 2;3.6) (Max 2;4.4) (Tom 2;5.24) (Tom 2;5.24)

Initially, Anne’s finite object clauses appear without an overt complementizer (the first attested case is given in (5.26a)). The same is true for Tom, although (5.28c) is the only object clause attested in his data before the age of 2;10.7. This might indicate that the embedded CP layer is not projected yet. Max’s first finite object clauses are indirect questions in which he does not realize the complementizer. Doubly-filled Comps are allowed in Canadian French but are not obligatory, so it cannot be concluded from Max’s first complementizer-free embedded questions that he cannot yet realize

Acquisition

189

embedded finite complementizers. The first attested case of a declarative object clause with overt complementizer in Max’s data is given below. (5.31) Il veut pas # que Catherine # le voie. he wants not that Catherine him see ‘He doesn’t want Catherine to see him.’

(Max 2;7.25)

The sixth diagnostic is the presence of subject-verb inversion in the child’s production. It is not clear that the inversion of a verb and a non-clitic subject—the so-called stylistic inversion construction (5.32)—involves the CP projection. 11 (5.32) Où sont les marrons? where are the chestnuts ‘Where are the chestnuts?’ I will therefore only consider simple inversion structures (as in 5.33), which involve a subject clitic but no subject DP. (5.33) Ont-ils des cils? have-they some eyelashes ‘Do they have eyelashes?’ Laenzlinger (1998) argues that French subject clitics are syntactic clitics in proclitic contexts and LF clitics in enclitic contexts (i.e. inversion contexts). Syntactic clitics incorporate overtly into the verb in Agrs◦ . LF clitics incorporate after Spell-Out, i.e. in covert syntax. In the latter case, subject clitics incorporate into the verb in C from the specifier of IP. I will not go into the details of this analysis. What is relevant for the present purpose is that the verb has to move to C in (simple) inversion structures but remains in Infl otherwise. The presence of ‘verb-subject clitic’ strings in child production could thus indicate that C is projected, assuming that the projection of the head C entails the projection of the whole phrase (CP). The simple inversion diagnostic is only relevant to Max and Léa given that inversion is part of the grammar of the input they receive but not of that received by Anne and Tom. 12 Max’s first ‘finite verb-clitic’ string appears at 2;0.14 but this is a one-off. The true emergence of this structure in his production is at 2;3.20. The only

11

It is generally assumed that I-to-C movement does not take place in this construction because unlike other inversions it can freely take place in embedded contexts such as relative clauses. Déprez and Pierce (1993) argue that the inverted subject is situated in [spec,VP]. This type of inversion is in any case not found in the present data. 12 I set aside for the moment the question of whether est-ce que ‘is it that’ is analysed or not in these dialects.

190

French Dislocation

attested cases in Anne’s data are when she is singing a particular song. This might well be an unanalysed, rote-learned expression. (5.34) a. E vois-tu, toi? e see you you ‘Can you see it?’ b. Peux-tu # dessiner? can you draw ‘Can you draw (something)?’ c. Dormez-vous? sleep-you ‘Are you sleeping?’ (from the song ‘Frère Jacques’)

(Max 2;0.14) (Max 2;3.20) (Anne 2;7.16)

There is no instance of simple inversion in Tom’s data. Léa’s first simple inversion structures appear when she is 2;10.7. (5.35) a. Et que fais-tu là? and what do-you there ‘And what are you doing there?’ b. Que fais-tu là, Minnie? what do-you there Minnie ‘What are you doing there, Minnie?’

(Léa 2;10.7) (Léa 2;10.7)

The seventh and eighth diagnostics are based on Hulk’s (1996b) adaptation to child French of a proposal by de Villiers (1991). De Villiers argues that there is a correlation between the appearance of a given wh-word with inversion and its appearance in embedded [spec,CP]. Given that inversion is not part of the grammar of spoken French (at least in the variety spoken by the parents of the CHILDES child Philippe, the subject of Hulk’s study), Hulk argues that a weaker version of de Villiers’ diagnostic should be used for Philippe and other children acquiring the same dialect. She proposes that there is a correlation between the use of a wh-word in embedded clauses and the active use of CP in root wh-questions. Before this point, wh-movement is claimed not to target [spec,CP]. Instead, Hulk proposes that the wh-word is either in [spec,IP] or adjoined to IP. Firm evidence for the projection of CP in child French would therefore be the co-occurrence of a moved whword and a complementizer, as in wh+est-ce que ‘wh+ is it that’, under the assumption that est-ce que is an unanalysed question marker in C. In Philippe’s data, embedded wh+est-ce que appears just before its root equivalent at 3;2.15.

Acquisition

191

Plunkett (2000), on the basis of data from six children, 13 argues that periphrastic questions (i.e. questions with est-ce que) are complex, biclausal structures in European French, and that such structures are not acquired as routines by most children. This explains why such questions are acquired late by these children in spite of their high frequency in the input. The Belgian child Léa, for instance, is shown to fully acquire periphrastic questions at exactly the same time as simple inversion, i.e. at 2;10.7. Philippe would thus be an exception: he starts producing periphrastic questions (albeit exclusively qu’est-ce qu(e)) early on, though as a routine. Plunkett claims that periphrastic questions in Canadian French do not involve inversion and est-ce que is an interrogative complementizer. The fact that Max starts using embedded questions introduced by si ‘if ’ at around the same time as est-ce que supports Plunkett’s proposal. What consequences does this have for the diagnostics under discussion? Under Plunkett’s analysis, the emergence of periphrastic wh-questions in European French is equivalent to the emergence of inverted wh-questions in child English. Given that inversion involves I-to-C movement, the occurrence of such questions indicates that the child is projecting CP. The emergence of periphrastic wh-questions in the speech of children acquiring Canadian French also indicates the projection of the CP layer under the view that estce que is an unanalysed interrogative complementizer. Under both analyses, what we expect to find if de Villiers’s (1991) proposal also applies to French is (approximate) simultaneity between the emergence of embedded wh-questions and that of root questions introduced by wh+est-ce que (or possibly also wh+que in Max’s data). Anne’s first moved wh-element in a non-root context does not appear in a selected clause but in an adjunct clause (5.36). I take the structure of this temporal clauses to be that of a relative. Labelle (1990, 1996) has argued that until the age of approximately six it is not clear whether children use the whmovement strategy to form relative clauses. In the case of (5.36), there is no overt complementizer so it is not clear whether the clause involves a CP. This is even more likely as Anne omits the complementizer que ‘that’ from the top adjunct clause. Notice also the omission of the subject ce ‘it’ from the periphrastic construction est-ce que. At the time she utters (5.38a), Anne is still at the null subject stage. This omission suggests therefore that she does not use est-ce que as an unanalysed question marker [Esk]. 13 The children studied in Plunkett (2000) are Anne, Max, and Léa from the York corpus, and Philippe, Grégoire, and Stéphane from corpora available via CHILDES (the corpora of Leveillé, Champaud, and Rondal respectively).

192

French Dislocation

(5.36) On les met pas # là parce (que) ça fait du bruit quand one them put not there because it makes some noise when (Anne 2;10.18) on les range. one them sorts ‘We don’t put them there because it’s noisy when we put them away.’ In the next recording session Anne utters her first indirect question (5.37). It is consistent with the diagnostic criterion under discussion. (5.37) Tu vois où il va &r [//] il peut rentrer? you see where he will he can enter ‘Do you see where he can get in?’

(Anne 2;11.2)

The first occurrence of a periphrastic wh-question appears in Anne’s data at 2;8.20. However, the first target-like construction of this type is not attested in her production until 3;0.2. (5.38) a. Là, qu’ est-(ce) que c’ est? that it is there what is-it ‘What is that, there?’ b. Eh ben oui mon bébé, qu’ est-ce qu’ i(l) y ah well yes my baby what is-it that it there a? has ‘Yes, my baby, what’s the matter?’

(Anne 2;8.20)

(Anne 3;0.2)

Max starts out by using the in situ strategy exclusively to form embedded questions (5.39). 14 He continues using this strategy alongside wh-movement for some time, as in (5.39b) and (5.39c). (5.39) a. Moi, xx [?= va] montrer à # Maman # c’ est me will? show to Mum it is quoi. what ‘I’ll show Mum what it is.’ b. Je sais pas où il [//] il est où, le I know not where he he is where the tracteur. tractor ‘I don’t know where the tractor is.’ 14

(Max 2;3.6)

(Max 2;6.12)

This strategy can be used in adult Canadian French to form embedded copular questions.

Acquisition c. On va regarder c’ est quoi. one will look it is what ‘Let’s look at what it is.’

193

(Max 2;9.27)

Max’s first object clauses with overt wh-movement are given in (5.40). In (5.40b), Max uses both a moved wh-element and a complementizer. His choice of wh-word is however not target-like (one would expect comme ‘like’). (5.40) a. On voit pas # qui fait du bruit xx one sees not who makes some noise ? ‘We can’t see who’s making noise.’ b. Et regarde euh # comment qu’ il est grand! and look er how that he is big ‘Look how big he is!’

(Max 2;5.1) (Max 2;9.27)

Max’s first periphrastic question appears at 2;6.12, in an embedded clause. 15 This is also his first attested case of an indirect question involving whmovement. (5.41) a. Où est-ce qu’ il est? where is-it that he is ‘Where is it?’ b. Regarde # qu’ est-ce qu’ il y a look what is-it that it there has là-dedans! there-inside ‘Look what there is in there!’

(Max 2;6.12)

(Max 2;10.10)

Tom produces only one indirect question before the age of 2;10.7. This unique indirect question (5.28a) (repeated below as (5.42)) contains a fronted quoi ‘what’, which is not possible in the target language. (5.42) xx [=? veux] voir # quoi # il y a ici. ? want to-see what it there has here ‘(I) want(?) to see what there is here.’

(Tom 2;3.22)

Tom utters two periphrastic questions during this stage but they are both unclear or incomplete. (5.43) a. [?] on va faire un how is-it that one will make a camion? lorry ‘How are we going to make a lorry?’ 15

This is allowed in adult Canadian French.

(Tom 2;8.0)

194

French Dislocation b. Qu’ est-ce que +//. what is-it that ‘What . . . ’ (self-interruption)

(Tom 2;10.7)

Léa’s first attested indirect question is when she is 2;9.21. She might well have uttered such questions before that age, but there are not enough data available to verify this. (5.44) Je ne sais pas où il est, le rideau. I neg know not where he is the curtain ‘I don’t know where the curtain is.’

(Léa 2;9.21)

Her first periphrastic questions are found in the first recording, but this question formation strategy does not seem to be fully productive before the age of 2;10.7. Prior to that, only one of her attested periphrastic questions involves a wh-element (5.45c). (5.45) a. Eh Parrain, est-ce que tu joues aux perles? hey Grandad is-it that you play to-the pearls ‘Hey Grandad, do you play with pearls?’ b. Est-ce que tu as encore soif? it-is that you have still thirst ‘Are you still thirsty?’ c. Qu’ est-ce que t’ as dit? what is-it that you have said ‘What did you say?’ d. Est-ce qu(e) (tu) en as un, de briquet? is-it that you of-it have one of lighter ‘Do you have a lighter?’ e. Minnie, qu’ est-ce que tu fais? Minnie what is-it that you do ‘Minnie, what are you doing?’

(Léa 2;8.22) (Léa 2;8.22) (Léa 2;9.5) (Léa 2;9.21) (Léa 2;10.7)

5.4.2 Discussion The diagnostics used above to identify the first manifestations of CP in the children’s production are compatible with an account in which the implementation of CP emerges gradually. Initially the data appear to suggest that embedded clauses as well as root clauses can be truncated by the child. For instance in (5.26b), repeated below as (5.46), the child embeds a finite clause without overtly realizing the complementizer or the subject. It might thus be postulated that in its early stages the child’s grammar does not require any clause to be a CP (whether root

Acquisition

195

or embedded). The absence of both complementizer and subject from the embedded clause in (5.46) would be the result of truncation. (5.46) E veux (qu’) (elle) y aille # en voiture. e want that she there go.SUBJ in car ‘I want her to go there by car.’

(Anne 2;7.1)

It is however far from clear whether truncation of embedded clauses is a viable hypothesis: the examples in (5.47) both feature a null subject, but in each case the embedded complementizer is overtly realized. I take the presence of a finite complementizer to be incontrovertible evidence of the projection of C. Under the assumption that the projection of a particular layer implies the projection of those beneath it (e.g. Rizzi 1994; Ferdinand 1996), IP is also projected in (5.47a) and (5.47b), and the non-realization of the subject cannot be attributed to the absence of the relevant syntactic host. (5.47) a. Comme ça, 0 raconter 0 que moi (j) (ai) vu 0. like that tell−FIN that me I have seen ‘Like that (I want to?/will?) tell what I’ve seen.’ b. E veux que (tu) arrêtes (de) jouer. e want that you stop to play ‘I want you to stop playing.’

(Max 2;3.6) (Anne 2;7.1)

Even if children do not truncate embedded clauses, the diagnostics above suggest that CP is implemented in a gradual fashion by the child. Non-finite complementizers are found in Max’s production before finite complementizers. In Anne and Tom’s production, however, there is no clear evidence that non-finite complementizers appear first. Instances of non-finite complementizers are initially very scarce in the data though, so it is possible that their total absence from Anne and Tom’s data is due to a sampling effect. If that proved to be the case, the use of non-finite complementizers before their finite counterparts would support Rizzi’s (1997) analysis of what he labels the Csystem, under a Structure Building approach to language development: evidence for the projection of FinP (the lowest layer of the C-system, which hosts non-finite complementizers) would be attested before evidence of ForceP (the highest layer, which hosts finite complementizers). Additional data from a larger group of children is required to prove or disprove this hypothesis. An important observation is that children continue omitting complementizers even after having shown signs that they can realize them. It is therefore possible that the apparent implementation of layers hosting non-finite complementizers before finite ones is also ascribable to the very small size of the sample. In the early stages, children produce very few finite embedded clauses

196

French Dislocation

and the overt realization of the CP layer might be possible but tentative for a few months. As a result, only some complementizers would be overt, and their finiteness might have no impact on their realization. A more in-depth study is necessary to determine when the use of complementizers is fully mastered by the child and whether non-finite complementizers emerge first. These data also suggest that in the three youngest children finite adjunct clauses appear before finite object clauses. Initially, all instances of finite complementizers are in adjunct clauses. Finite object clauses remain rare in the data during the period investigated here and only become frequent from approximately 3;2 for Anne, and 2;10 for Max and Tom. A possible explanation is that children are initially unable to master the implementation of CP under selection. De Villiers (1991) aimed to identify the true emergence of CP, not just the first signs of it. 16 Taking the age at which children produce both periphrastic wh-questions (or instances of doubly-filled Comp) and their first indirect questions with overt wh-movement should therefore give us the turning point in the children’s mastery of CP implementation. This turning point appears to be around 2;10 for Max, Tom, and Léa, and around three for Anne (who is a slower developer in other respects too). From that point onwards, the four children produce periphrastic wh-questions and indirect questions productively and in a target-like fashion. Going back to the initial question that led to this investigation, it is now clear that children use dislocated structures before there is any evidence that any part of the CP layer is implemented. The earliest possible (though suspect) evidence of CP implementation is found at the ages of 2;1.19 and 2;0.14 in the youngest children, and dislocated elements are clearly attested before that. Under a Rizzi-style analysis, dislocated structures would be the only constructions manifesting CP overtly at that point in the child’s development. While this does not constitute evidence against such an analysis, it undermines it on developmental grounds.

5.5 Sentence fragments: mini root projections From the earliest recordings, children produce verbless utterances consisting of two constituents, one of which appears to be the topic and the other the comment. One might hypothesize (following e.g. Gruber 1967) that this is a ‘primitive’ form of predication, a precursor to the canonical ‘subject-verb’ 16

De Villiers (1991) argues that inversion is acquired individually for each wh-word.

Acquisition

197

combination, as claimed by e.g. van Kampen (2004). Here are a few typical examples. (5.48) a. Et moi, une fille. and me a girl ‘And me, (I’m) a girl.’ b. Moi, pas l’hippopotame. me not the-hippopotamus ‘I (do) not (want to draw) the hippopotamus.’ c. Lui, ici. him there ‘That one (goes) there.’

(Tom 2;1.11) (Anne 2;2.20) (Max 2;0.28)

Verbless utterances of this type are not attested exclusively in the very early data. They are found throughout the period studied here (5.49) and also in the adult data (5.50). (5.49) a. De la tomate, maintenant. (Anne 2;10.18) part the tomato now ‘Now (let’s add) some tomato.’ (Max 2;8.9) b. Toi, cette côté-là, moi, cette côté-là. you that side-there me that side-there ‘I (do) this side, you (do) that side.’ c. Quoi, ma gauche? (Léa 2;9.21) what my left ‘What (is the matter with) my left (side)?’ d. Là, encore une étiquette. (Tom 3;0.6) there again a sticker ‘(There’s) another sticker there.’ (5.50) a. Toujours, moi. (Nelly, B) always me ‘Me, (I am) always (hungry).’ b. Deux pattes, le canard? (A.-Gaël, F) two legs the duck ‘The duck (has) two legs?’ c. Beaucoup de boutons, hein, le pull de Mamy, hein many of buttons eh the jumper of Mamy eh oui? (Nelly, B) yes ‘Mamy’s jumper (has) lots of buttons, doesn’t it?’

198

French Dislocation d. (C’ est un) avec un bec bleu, lui.17 it is one with a beak blue him ‘That’s one with a blue beak.’

(Catherine, C)

Two questions come to mind: (i) How much structure is projected for these utterances in the adult representation? and (ii) Should we assume that the children’s representation of these structures is adult-like? The prosody and the relation to the context suggest that these utterances involve a peripheral element, in spite of the fact that the ‘nucleus’ of the utterance does not consist of a full clause. This nucleus contributes the essential information of the utterance and bears the main stress: it corresponds to the focus. The peripheral-looking element (or, more descriptively, the ‘satellite’) can appear to the left or to the right of the nucleus. On the surface, we have the following structures, in which XP is the utterance’s nucleus. (5.51) a. [ Focus XP ] satellite b. satellite [ Focus XP ]

(as in e.g. (5.49a), (5.50)) (as in e.g. (5.48), (5.49b))

The satellite is not dislocated in the traditional sense of the term, as it is not associated with a position or a resumptive element inside a clause. However, it displays a number of the core characteristics of dislocated elements in spoken French. First, it receives LD prosody when it appears on the left of the nucleus and RD prosody when it appears on the right. Second, it can be omitted without altering the discourse contribution of the utterance (provided its referent is salient enough in the context). Third, it appears to fulfil the same informational function as a dislocated element: either it expresses what the utterance is about (as in (5.52)), or it restricts the (temporal or spatial) domain within which the predication holds (as in (5.53)). In each case, the verbless utterance behaves like a fragment in the sense that it is interpreted as if it were a full proposition with assertoric force, the unpronounced (yet understood) surrounding linguistic environment somehow being retrieved from the context. (This has been observed by e.g. Hankamer 1979; Barton 1990; Ginzburg 17 The utterances immediately preceding are as follows, in a context where Catherine and Max are identifying parrots in a picture.

(i)

∗ CAT: ∗ CAT: ∗ CAT: ∗ CAT: ∗ CAT: ∗ CAT:

Il y a des belles couleurs. ‘There’s some nice colours.’ Pi regarde! ‘And look!’ Il y en a un autre là. ‘There’s another one there.’ Avec un bec bleu, lui. ‘With a blue beak, that one’. Ah pi un autre ici. ‘Ah and another one here.’ Mais c’ est pas tout des mêmes. ‘But they’re not all identical.’

Acquisition

199

and Sag 2000; Stainton 2004; and Merchant 2005 with respect to fragments without satellites). (5.52) a. Les voilà, les petits copains. (Dominique, B) them presentative the little friends ‘Here (are) the little friends.’ b. Ils sont là, les petits copains. they are there the little friends ‘The little friends are here.’ (5.53) a. De la tomate, maintenant. (Anne 2;10.18) part the tomato now ‘Now (let’s add) some tomato.’ b. On met de la tomate, maintenant. one puts part the tomato now ‘Now let’s add some tomato.’ Therefore, in spite of lacking a verb, fragments behave like full clauses from an interpretive point of view and in terms of their ability to take clause-level modifiers. On the assumption that syntactically non-propositional XPs cannot be mapped onto semantic propositions, Merchant (2005) proposes that fragments are in fact fully propositional in syntax, but that most of the structure remains unpronounced at PF. He argues that the fragment occupies the specifier of a left-peripheral phrase (which he suggests may be FocusP) whose head is endowed with an E feature. This feature has two functions: (i) it instructs PF not to parse its complement (and hence not to pronounce it), and (ii) it consists of a ‘partial identity function over propositions’, which is supposed to ensure that the complement of an E-endowed head has an appropriate antecedent in the discourse (essentially, that the content of the unpronounced structure is identifiable/recoverable). Merchant’s analysis is illustrated by the tree in (5.54). (5.54)

FP

F

DP

a fancy treei

F[E ]

TP → not pronounced I want ti

200

French Dislocation Table 5.2. The age at which overtly moved wh-elements and fragments are first attested

Anne Max

First moved WH

First fragments

2;4.2 2;5.1

2;0.13 1;11.0

In principle, such an analysis could account for fragments with satellites if one assumed that satellites are dislocated elements situated at the periphery of the clause (e.g. they can be adjoined to FP). Merchant’s ellipsis analysis of fragments faces a number problems with respect to the adult data, which I will not go into here (see De Cat and Tsoulas (2006) for details and for a discussion of the theoretical implications of such an approach). What I will show here is that this analysis is implausible from an acquisition perspective for two reasons. The first relates to learnability considerations. When presented with fragments in the input, all the child hears is a verbless XP (on its own or accompanied by a satellite). Nothing indicates that it is part of a larger syntactic structure: the fragment is not endowed with a distinctive prosody that would indicate that some structure has not been pronounced, 18 nor with any kind of morphology that would force the child to assume that the nucleus XP has been fronted. The second is that fragments are heavily used by children from the onset of expressive syntax, well before the first instances of wh-fronting are attested (see Section 5.4). Children in the York corpus do not front wh-phrases before the age of 2;4 (Anne) or 2;5 (Max). 19 Evidence of focus fronting appears at around the same time, but attested occurrences remain very rare throughout the period of study, as well as in the adult data. Under Merchant’s analysis, the only evidence of complex syntax involving the left periphery would in fact be invisible/inaudible as most of the structure is unpronounced. The alternative I will argue for is what could be called the WYSIWYG approach, i.e. What You See Is (almost) What You Get. On this view, fragments only contain as much structure as evidenced by their overt structure 18 The prosody of the nucleus is in most cases identical to that of VP-final focus as derived by the Nuclear Stress Rule (Cinque 1993), as if the top of the clause had been truncated. 19 Data from Tom and Léa are not relevant here because they started recording later than the other children and they produced fronted wh-elements in the first recording session.

Acquisition

201

(in the spirit of e.g. Barton 1990; Ginzburg and Sag 2000; Stainton 2004). In some cases, this is uncontroversially the only possible analysis as the fragment cannot be reconstructed into a full structure containing a verb (see e.g. (5.52a) and possibly (5.49c)). I argue below that WYSIWYG is the most appropriate analysis for all types of fragments. Dislocated elements have been shown to be restricted to clauses with root properties, which correspond to discourse properties (see Chapter 4). If a satellite that appears with a fragment is equivalent to a dislocated element, this analysis predicts that such a fragment will have root properties. I proposed in Section 4.4 that the discourse properties of root-like projections are performative in nature. Given that only a fragment endowed with root properties can take a satellite, it is predicted that only fragments that can stand alone (e.g. in answer to a wh-question) may appear with a satellite; those that provide a continuation of the previous utterance may not. The data verify this prediction. In (5.55), for instance, the DP answering the question can be accompanied by a satellite (here pour le moment ‘at the moment’, which would be a stage topic). (5.55) A: Qui est-ce qui te manque le plus ? who is-it that to-you misses the most ‘Who do you miss the most?’ B: Pour le moment, Marie-Hélène.20 for the moment Marie-Hélène ‘At the moment, Marie-Hélène.’ By contrast, in (5.56), the DP contributed by speaker B as a continuation of speaker A’s utterance cannot appear with a satellite, even though the full utterance would allow for one (5.57). (5.56) A: Qui est-ce qui manque le plus à . . . who is-it that misses the most to ‘Who does . . . miss the most?’ B: #Ma sœur(, pour le moment) ? my sister for the moment (5.57) Qui est-ce qui manque le plus à ta sœur, pour le moment? who is-it that misses the most to your sister at the moment ‘Who does your sister miss most at the moment?’ 20

A right-peripheral PP would also be acceptable.

202

French Dislocation Table 5.3. Fragment utterances with satellite in the child data Child

Spread of data

File

Number of fragments with satellite

Anne Max Léa Tom

[1;10.12–3;1.15] [1;9.19–2;10.24] [2;8.22–2;10.7] [2;1.11–2;10.7]

28 28 4 13

111 40 5 36

9

192 59

TOTAL Adults

In both the adult and the child data, fragments with satellites clearly fulfil the performative requirement: they may be questions as in (5.49c) and (5.50b), exclamations as in (5.50c), interdictions as in (5.58a), suggestions as in (5.49a), and statements as in (5.50d) and (5.58b). (5.58) a. Manger, non. Goûter, oui. (Nelly, B) to-eat no to-taste yes ‘Not to eat (it) but to taste (it).’ b. La boutonnière de ce bouton-là, la voilà. (Nelly, B) the buttonhole of that button-there it presentative ‘There is that button’s buttonhole.’ In fragments, only the focus is pronounced (leaving aside for the moment the possible presence of a satellite). Given that the focus is by default the most deeply embedded constituent, fragments could therefore be considered an extreme form of truncation. Truncation has been independently argued to play a role in child language (e.g. to account for root infinitives, see Rizzi 1994). Fragments with satellites are produced by adults as well as children, as shown in Table 5.3. The frequency of their use strongly depends on the type of discourse situation. Situations strongly favouring the use of fragments include doing jigsaws, interaction based on books or images, and language games, i.e. situations in which the discourse context is externally restricted (see below). Frequency counts have not been included below as the counts do not include fragments without satellites (which is likely to skew the picture). 21 21

Each file corresponds to one recording session.

Acquisition

203

Fragments are uttered by adults in contexts in which they feel the omitted information is easily recoverable by their addressee. This is also what makes fragments potentially highly ambiguous, and therefore means that they are not always sufficient to comply with Grice’s Quantity Maxim (which explains why we do not talk in fragments all the time). Children also appear to strive to meet the Quantity requirement, but they are independently constrained as to how much structure they can produce (Pinker 1984). The combination of fragment+satellite allows them to be maximally efficient informationally in spite of restrictions on how much they can pronounce: the nucleus expresses the essential information (the focus) while the satellite identifies what this information applies to (i.e. the topic), thereby to some extent disambiguating the possible interpretation of the nucleus. Most of the time, the child seems to be successful in conveying the information intended, as far as one can judge by the reaction of their addressee. Another example of successful exploitation of the ‘topic+incomplete structure’ strategy comes in the form of children’s early attempts at clefts, as illustrated in (5.59). In these examples, the embedded clause of the cleft is not pronounced, but its content is made more easily recoverable by the presence of a dislocated element which is coreferential with the would-be object of the embedded clause. (5.59) a. (C’) est Mamani , (quii le j fait) ça j . who it does that it is Mum ‘It’s Mum (who does) that.’ b. Ça j , c’ est moii , (quii l’ j ai fait). that it is me who it have done ‘It’s me (who’s done) that.’ c. [De l’ eau à chauffer] j , c’ est moii some the water to heat it is me (quii le j fais). who it do ‘It’s me (who’s) heating the water.’ d. Non c’ est moii (qui l’ j ai), le pingouin j . no it is me who it have the penguin ‘No, I’m the one (who has) the penguin.’

(Anne 1;11.13) (Max 2;0.28)

(Tom 2;1.11) (Tom 2;1.14)

With respect to fragments with satellites, there is no indication that a child’s underlying representation is different from that of adults. In spite of disagreement regarding their exact nature, the existence of constraints on the structures which children are initially able to express is uncontroversial. It is therefore not surprising that children should exploit early

204

French Dislocation

on the availability in the target grammar of a highly efficient device that maximizes the interpretive potential of a syntactically simple structure (by relying heavily on the ability of the hearer to identify unexpressed information). Indeed, children have been shown to make informationally judicious choices regarding what they (do not) express even at the one-word stage (Greenfield et al. 1985).

5.6 Primitives, learnability, and early discourse competence I have argued that French-speaking children can encode XP topics in a targetlike fashion as soon as they start combining words. Their utterances may be target-deviant in some respects (e.g. the resumptive may be missing (5.60a); the verb may be non-finite in a context requiring a finite verb and the clause may be truncated below CP, as under the standard analysis of root infinitives (5.60b); there may be the odd agreement error (5.60c)), but this does not hinder the dislocation process. ‘Errors’ such as those illustrated below are also attested in the absence of dislocated elements, and are uncontroversially independent phenomena. 22 (5.60) a. (C’) est quoi, ça là? it is what that there ‘What’s that?’ b. Le camion, 0 mettre là. the lorry put−INF there ‘The lorry, (I’ll) put (it) there.’ c. Et moi, (j’) a gagné. and me (I) have-3 p.sg. won ‘I’ve won.’

(Max 2;0.28) (Anne 2;6.2) (Tom 2;4.8)

However, the presence of dislocated elements in the early data is not in itself sufficient evidence that their intended interpretation as topics is target-like. Indeed, it is commonly assumed in the literature that discourse-related abilities emerge more slowly than syntactic abilities (e.g. Hamann 2002). This delay has been invoked to explain target-deviance in relation to pronominal interpretation (Avrutin 1999), the licensing of null subjects (Schaeffer et al. 2002), and other phenomena. Some claim the delay is due to the very nature of pragmatic competence, which is assumed to be cognitive rather than linguistic (see e.g. Schaeffer and Matthewson 2005). Whatever the reason, many acquisition studies have reported that children are initially poor at discourse 22

These ‘errors’ are widely reported in the acquisition literature.

Acquisition

205

integration, as evidenced by their underuse of anaphoric reference in narrative production (Karmiloff-Smith 1981; Hickmann 1982, 2003), their overuse of definite DPs to introduce new referents (Maratsos 1974; Emslie and Stevenson 1981), their interpretation of indefinites as referring to previously introduced referents (Krämer 2000), and their target-deviant interpretations of scrambled indefinites (Unsworth 2005). However, it is important to realize that not all discourse competence is initially defective. If this were the case, children would be much poorer communicators than they actually are. Instead, even in the early stages when they omit large parts of most sentences, children are judicious in the information they choose to utter as it tends to be sufficient for successful interaction with their addressee. At the one-word stage, children have been shown to have already mastered the contrast between old and new information (Baker and Greenfield 1988). A certain degree of competence with respect to the structuring of information must therefore be available from a very early age. In particular, I will argue below that, from the onset of expressive syntax, children master the notion of topic. 5.6.1 Early discourse competence Children’s spontaneous language production can provide us with various types of indication that they have mastered the notion of topic at an early age. I review them in turn. 5.6.2 Absence of violations of the relevant discourse rules In the two corpora under investigation, all the dislocated elements occurred in a context which allowed them to be interpretated as the topic of the utterance. This was confirmed by the analysis in context of 1,249 clauses, each containing a dislocated element. This observation is not sufficient to prove the point, however: evaluating the salience and relevance of a referent in spontaneous discourse is a subjective task and may not reflect children’s own evaluation of such criteria. Compatibility with a topic interpretation therefore does not provide conclusive proof of children’s target-like use of dislocated elements. 5.6.2.1 ILP subjects Individual Level Predicates (ILPs) provide a better way of testing children’s knowledge that XP topics have to be dislocated. The rationale is as follows. ILPs force a topic interpretation of their subject, which must then be dislocated in spoken French (provided it is non-weak and not in contrastive focus; see Chapter 3). Therefore, what needs to be shown is that: (i) children show clear signs of compliance with the ‘ILP requirement’, i.e. they do dislocate heavy elements expressing the subject of ILPs; and

206

French Dislocation

(ii) children do not violate the ‘ILP requirement’, i.e. they do not utter sentences containing an ILP with a heavy subject unless there is clear narrow focus on that subject. The sentences in (5.61) prove the first point. In all cases, a property reading of the predicate was clearly intended. (5.61) a. Les sucettes, ça finit pas. the lollipops it finishes not ‘Lollipops don’t end.’ b. Le poisson là, ça, c’ est un poisson. the fish there that it is a fish ‘That fish is a fish.’ c. Ça, c’ est un dinosaure. that it is a dinosaur ‘That one is a dinosaur.’ d. Le coca, ça soigne le hoquet aussi. the coke it cures the hiccough too ‘Coke cures hiccough too.’ e. Et les vaches, # elles mangE # de l’ herbe. and the cows they eat some the grass ‘And cows eat grass.’ (Clearly generic context)

(Anne 3;1.15) (Max 2;5.1) (Max 2;8.9) (Léa 2;9.5) (Tom 2;1.13)

The corpora were then searched for heavy subjects. A total of 186 cases was found in the child data. Only one case (5.62) involves an ILP, and it is targetlike: the subject is in narrow (contrastive) focus. The other cases clearly do not involve ILPs; they typically contain eventive predicates, as in (5.63). (5.62) Mon papa aussi est gentil. (Anne 2;8.3) my dad too is nice ‘My dad’s nice too.’ (5.63) a. Ah et tout le monde est tombé. (Anne 2;7.1) ah and all the people is fallen ‘Ah everybody’s fallen over.’ b. (La) vache # mange # (des) ca(r)ottes. (Max 2;5.29) the cow eats some carrots ‘The cow’s eating carrots.’ (Clear ongoing interpretation) c. Parrain, Luc s’ est fait mal. (Léa 2;8.22) grandad Luc refl is done pain ‘Grandad, Luc’s hurt himself.’

Acquisition d. Et pourquoi # les lunettes s’ étaient xx and why the glasses refl were ?

207

(Tom 2;6.12)

envolées? flown ‘And why did the glasses fly away?’ Children’s spontaneous production thus clearly suggests that they know that ILPs cannot appear in thetic sentences, which indicates that they are able to identify and encode topics in a target-like fashion. 5.6.2.2 Dislocated indefinites It has been claimed in the literature (e.g. ZribiHertz 1994) that children acquiring French allow quantified DPs to be resumed by a subject clitic. 23 In fact, dislocated quantified expressions are only illicit as topics under an existential reading (see Section 3.2.2). A topic interpretation of indefinites requiring a d-linked interpretation is allowed. In the corpora under investigation, whenever the child uses a universal quantifier, the interpretation is clearly that of a set of individuals either mentioned in the immediately preceding discourse or clearly salient in the context. In (5.64a), Max is referring to the cars that have just been fixed by the interviewer; in (5.64b), Anne is referring to the group of little figures she has been playing with; in (5.64c) she is referring to the four members of a puppet family she has just put to bed. In each case, the quantified expression is a pragmatically adequate topic because the referent is clearly identified in the context. (5.64) a. Toutes les autos, elles sont réparées. (Max 2;4.18) all the cars they are fixed ‘All the cars are fixed.’ b. Les copains, tous les copains, i(ls) veulent entrer. (Anne 2;7.1) the friends all the friends they want to-enter ‘All the friends want to come in.’ 23 Zribi-Hertz (1994: 461–2) does not identify the source of her alleged child French examples, which I cite below. (i) a. Personne i(l) m’aime. nobody he me-loves ‘Nobody loves me.’ b. Personne i(l) veut m’aider. nobody he wants me-to-help ‘Nobody wants to help me.’ c. Personne il a rien dit. nobody he has nothing said ‘Nobody said anything.’

To my knowledge, no example featuring dislocated personne ‘nobody’ has been attested in any corpus of spontaneous production by French-speaking children.

208

French Dislocation c. Tout le monde, i(l) dort. all the people he sleeps ‘Everybody’s sleeping.’

(Anne 2;8.20)

I have found four instances of apparently non-target-like use of dislocated indefinites in the data, given in (5.65)–(5.70) below. However, in their context, none of them could be claimed to violate the rules of topic encoding. (5.65) Un petit bébé, je sais pas c’ est quoi. a little baby I know not it is what ‘I don’t know what the little baby’s called.’

(Max 2;7.25)

The sentence in (5.65) is uttered after the child has ‘cured’ a little baby and the interviewer has said (5.66). (5.66)

∗

CAT:

∗

CAT:

Tu as guéri le petit bébé? ‘You cured the little baby?’ Comment il s’ appelle, le petit bébé? ‘What is the little baby called?’

It is clear from the context that Max’s utterance does not mean that he does not know what a little baby is. The most likely interpretation is that he does not know what the baby is called. His sentence would thus be a truncated version of (5.67). (5.67) Un petit bébé, je sais pas c’ est quoi qu’ il s’ appelle. a little baby I know not it is what that he refl calls ‘I don’t know what the little baby is called.’ The intended interpretation of the dislocated un petit bébé ‘a little baby’ is definite and this DP is clearly the topic of Max’s sentence. The use of an indefinite instead of a definite article could either be a performance error or a sign that the child has not yet fully mastered the definite/indefinite distinction. 24 Further research is needed to decide between these two analyses. What is relevant for the present purpose is that the child does not dislocate indefinites with an existential interpretation and hence does not contravene pragmatic requirements on topics. The dislocation of an indefinite in (5.68) is also target-deviant but not for the same reasons as (5.65). (5.68) Des cochons, ils font 0 [=! she swirls]. some pigs they go ‘Pigs go . . . ’ (She demonstrates what pigs do) 24

(Léa 3;0.5)

It is also possible that un ‘a’ is in fact a mispronunciation of le ‘the’ with the consonant omitted.

Acquisition

209

The interpretation here is clearly generic, as Léa is demonstrating how pigs fly to her grandmother; she is not referring to any pigs in particular. What is target-deviant in this sentence is the use of the indefinite determiner des instead of the dummy les, which is required by plural NPs interpreted generically. A performance account is probably appropriate in this case, given that in the following utterance this child correctly uses the dummy determiner les: 25 (5.69) C’ est comique, hein, les cochons, de # faire +//. it is funny eh the pigs to do ‘It’s funny, isn’t it, for pigs to do . . . ’ (self-interruption) Consider now the sentences in (5.70). (5.70) a.

b.

C’ est où, le l’ &her [//] le l’ herbe [//] it is where the the grathe the grass de l’ herbe? part the grass ‘Where’s “the grass”?’ Oui un œil, (ç)a va où? yes an eye it goes where ‘Where does (that) eye go?’

(Anne 2;6.2)

(Max 2;4;4)

In (5.70a), the child is trying to find in a picture the grass mentioned earlier by her mother. The interpretation of the dislocated element is clearly that of a topic. In (5.70b), the child is doing a jigsaw. The interviewer has identified a piece with a yellow eye (saying just un œil ‘an eye’) and the child is asking her where that eye could be placed. The indefinite receives a specific interpretation in this context and is clearly the topic of the sentence. All the other instances of dislocated indefinites I have come across in the child data from the York and the Cat corpora are target-like. I conclude that children’s dislocation of quantified and indefinite expressions is pragmatically target-like in the sense that the dislocated expression always expresses the topic of the sentence. 5.6.3 Positive evidence for the relevant pragmatic competence With respect to positive evidence, there is clear proof that children use leftand right-dislocated elements in the same way as adults. While this does not automatically imply that the required pragmatic knowledge underlies their use of LD vs RD, the null hypothesis should be that children’s linguistic 25

Sentence (5.69) is also odd, but for reasons irrelevant to the present discussion.

210

French Dislocation

Proportions of all dislocated elements

100% 90% 80% 70% Subject

60%

Other selected XP la`

50%

Unselected XP

40% 30% 20% 10% 0%

Max (1) Max (2) Anne (1) Anne (2) Tom (1)

Tom (2)

Le´a

Figure 5.5. Functions associated with dislocated elements in the child data

competence is no different to that of adults unless there is evidence to the contrary. In the absence of proof that the necessary pragmatic knowledge does not underlie the dislocation of XPs in child language, one should thus assume that children have that knowledge. I take the following facts to suggest that this is the case. First, most of the children’s dislocated XPs are coreferential with the (intended) subject of the sentence, as is the case in the adult language. It is therefore reasonable to assume that the well-documented cross-linguistic association between the pragmatic function of topic and the grammatical function of subject is encoded in the children’s system from the earliest point at which evidence is available. The graph in Figure 5.5 gives for each child the total proportion of dislocated elements according to the function they are associated with inside the clause. 26 For comparison, the corresponding frequencies in the adults’ speech distinguished by dialect are given in Figure 5.6. Apart from Léa (whose data was collected from a later age and consequently appears more adult-like), the children produced even more dislocated subjects than the adults. In the early recordings, almost all of the dislocated elements in their speech express the subject. Second, from the start children seem to be aware of the differences between left- and right-dislocated elements. Whenever a contrast is implied between the topic and another discourse referent in the context or between the members of a set which the child is commenting upon, the topic in question is 26

For a definition of the two periods of development of Anne, Max, and Tom, see Appendix B.4.

Acquisition

211

Proportions of all dislocated elements

100% 90% 80% 70% 60%

Subject Other selected XP

50%

la`

40%

Unselected XP

30% 20% 10% 0%

Canada

France

Belgium

Figure 5.6. Functions associated with dislocated elements in the adult data

left-dislocated, in line with the principles that rule the adult language. This is illustrated by (5.71a), where the child clearly establishes a contrast between her mother and herself. The contrast is expressed by means of two LDs. (5.71) a.

Maman, e fait yy [%pho: atu] # moi, e fais mum e makes me e make (le) drapeau. the flag ‘Mum makes the xx and I make the flag.’

(Anne 2;2.00)

Another illustration is provided in (5.72) below, where the left-dislocated elements designate the various members of a set in sentences revealing the origin of each individual (here drawings), thus implying a contrast. (5.72)

∗

MAX:

∗

CAT:

∗

MAX:

∗

MAX:

Ça, c’ est moi [?] (qui l’ai fait). that it is me who it-has done Ça, c’ est toi qui l’ as fait? that it is you who it has done ‘You’re the one who’s done that?’ Ça, c’ est Maman (qui l’a fait). that it is mum who it-has done Ça, c’ est moi (qui l’ai fait). that it is me who it-has done

(Max 2;0.28)

Another type of example which shows that the child must have knowledge of the different pragmatic functions of left- and right-dislocated elements is the repair strategy illustrated in (5.73). First, the child asks a question in which she encodes the topic by RD. Then she realizes, given the caregiver’s

212

French Dislocation

follow-up question, that the referent of her topic was not salient enough in the context to be properly identified. She rectifies this by using the more salient LD construction to repeat her question while still encoding the topic as a dislocated element. (5.73)

∗

ANN:

∗

TAT:

∗

ANN:

I(l) est où, mon [=? ma] cass(er)ole? it is where my saucepan ‘Where is my saucepan?’ Qu’ est-ce que tu cherches? what is-it that you search? ‘What are you looking for?’ Mon cass(er)ole, (e)lle [=? (i)l] est où? my saucepan it is where ‘Where is my saucepan?’

(2;4.2)

This use by children of both LD and RD is also typical of the adults’ language, as illustrated in the exchange in (5.74) from the York corpus. (5.74)

∗

FAT:

∗

MOT:

∗

ANN:

∗

MOT:

Il est où, le bébé? he is where the baby ‘Where’s the baby?’ Où il est? where he is ‘Where is he?’ Il est là, (l)a tête. he is there the head ‘The head’s there.’ (Picking up a bit of plasticine) Le bébé de Maman, il est où? the baby of Mum he is where ‘Where’s Mum’s baby?’

Dislocated subjects appear in the right periphery of adult dislocated whquestions seventy-eight per cent of the time in the sample under investigation. The same proportion is observed in the data of Max, Anne, and Tom, as shown in Table 5.4. 27 5.6.4 Learnability considerations It is a widely held assumption that the child’s grammar is initially less complex than the adult’s (e.g. Lebeaux 1988; Powers 2001). On this view, the presence of dislocated constructions in the very early data indicates that the mechanisms 27 With respect to Léa, the preference for RD in wh-questions is even more dramatic as she does not produce a single left-dislocated wh-question in any of the recording sessions up to age 3;0.5.

Acquisition

213

Table 5.4. Frequency of wh-questions involving a subject dislocation in child French (Anne, Max, and Tom)

Period 1 Period 2 All

Left-dislocation

Right-dislocation

28% (54/190) 17% (24/140) 24% (78/330)

72% (136/190) 83% (116/140) 76% (252/330)

it involves cannot be too demanding in terms of derivational complexity or in terms of processing. In particular, I have argued that dislocation of topics is unlikely to require that the child has the ability to implement functional layers within (or coextensive with) CP. I have proposed that French dislocated elements are adjoined to a maximal projection with root properties. There is independent evidence that adjunction is one of the earliest operations to be mastered by the child. It has even been argued that adjunction is available as a default in the child’s grammar (Lebeaux 1988; de Villiers 1991; Roeper 1992). This would explain why dislocated structures can appear so early. On the adjunction analysis, the child does not need (to be sensitive to) any evidence in the input that peripheral topics have to check a feature against a Topic head because there is no such head. No evidence is necessary to trigger an adjunction analysis of dislocated elements because adjunction is what the child is likely to postulate for dislocated structures by default. The very early availability of the pragmatic knowledge required to encode XP topics could be due to the fact that the notion of topic is a primitive, as argued independently by É. Kiss (1995).

5.7 Conclusion The present investigation into the acquisition of dislocated elements in child French has revealed that children do not produce non-target-like dislocated structures in any systematic way. The positive findings are the following: (i) dislocated constructions are produced by the child from the start of the multi-word utterance stage; (ii) dislocations are used productively by the child well before there is evidence that the CP functional layer is (fully) implemented; and (iii) from a very early point, the child has mastered whatever discourse/pragmatic knowledge is required to encode XP topics as peripheral

214

French Dislocation

elements, even in verbless utterances or fragments; this knowledge could well be present from the start. This study has also confirmed previous findings regarding the early mastery by the child of the prosodic rules governing the dislocation of elements, at least in terms of variations in fundamental frequency (Crystal 1973; Allen 1983). Regarding the architecture of the language faculty, the early acquisition data call for two observations to be made. First, discourse competence and syntactic competence develop relatively independently of each other, at least with respect to the phenomena investigated here. Indeed, children start using dislocated constructions when their syntax is still very immature, and well before the C-domain (which has been argued to have a major role to play in encoding discourse phenomena) is fully operational. Second, children are able to exploit the interface between various components (syntax, discourse, phonology) efficiently from the onset of word combination. This indicates that interface phenomena are not automatically too costly for immature production systems.

6 Concluding remarks One of the clear implications of this investigation is that spoken French should be regarded as a discourse-configurational language, i.e. one in which ‘primary sentence articulation is motivated by discourse-semantic, rather than theta role or case, considerations’ (É. Kiss 1995). In spoken French, any non-weak constituent expressing a topic has to be dislocated. The high frequency of dislocated XPs does not indicate an evolution towards allowing argument doubling (accompanied by loss of the argumenthood of resumptive clitics) nor that dislocation is becoming ‘grammaticalized’ (implying a loss of interpretive effects), but rather it is a manifestation of discourse configurationality. This property manifests itself not only in the language’s overt topic-comment articulation, but also in its expression of focus: focus strongly tends to be encoded by clefts and pseudo-clefts in spoken French. (This is something I have only had the chance to allude to in this book, but hope to develop in future work.) In spoken French, dislocated elements are entirely dispensable from a syntactic and a prosodic point of view. They do not have argument status and need only be overt when the topic referent is not fully salient/active in the context (i.e. when a mere pronoun would not be sufficient to identify the topic referent) or to convey contrast. Against the grain of many Minimalist analyses, which often rely on an extensive array of functional projections, I have proposed a syntactically parsimonious account of dislocated structures. Postulating the existence of a dedicated functional projection for topics would be uneconomical: no movement whatsoever is involved in the derivation of dislocated structures, peripheral topics can appear in the right periphery of the clause (which would prevent the checking of a functional feature), and their distribution is freer than a cartographic approach would predict. The analysis proposed is based on the interaction of specialized components: a syntactic component relying on a default adjunction mechanism to exhaust the numeration, but which is blind to information-structural configurations; a discourse component licensing dislocated structures (which

216

French Dislocation

explains why illicit structures are anomalous rather than straightforwardly ungrammatical); a PF component responsible for the prosodic encoding of dislocated structures; and a lexicon in which the relevant cross-linguistic variation is located. In other words, French dislocation is by essence an interface phenomenon. Supporting evidence for this account comes from the early presence of dislocated structures in children’s speech—at a time when the clausal functional domain is still under-exploited—and from learnability considerations (including the independently attested use of adjunction as a default mechanism in early grammars). The very early use of dislocation by children also shows that interface phenomena are not necessarily delayed in first language acquisition, and that they do not automatically increase the computational load excessively for immature production systems. Dislocated elements have been argued to be licensed only at the periphery of root-like projections, which are known to have discourse properties. I have proposed that root-like projections can be defined in terms of Performativity. Data from child language (which allows dislocated elements in truncated structures such as Root Infinitives) and complex fragments found in adult and child language suggest that root-like projections need not always be CPs (or even contain a verb) as long as they are Performative. As a final note, it is important to highlight the fact that the selection of the numeration plays a significant role in ensuring that licit structures are produced, whether or not they involve a dislocation. In particular, decisions such as whether to use pronouns or full DPs, overt (i.e. dislocated) topics, clefts, etc. are made during this selection process, and as a consequence we have to acknowledge that it is sensitive to and informed by information structure considerations. This is by no means unique to the present account and requires further investigation.

Appendix A Adult data This appendix gives a brief overview of the spontaneous dislocation data which is the main empirical basis of this book. 1 The description below is based on the detailed analysis of a random sample of the adult data from the York and Cat corpora.

A.1 Brief description of the corpora used The York and Cat corpora contain longitudinal data which chart the language development of children, gathered from their spontaneous interaction with adults who come from roughly the same broad dialectal area as themselves. The children’s names are all pseudonyms. The York corpus was collected under the direction of Bernadette Plunkett between 1997 and 1999. 2 It contains data from three children: Léa from Liège (Belgium), Max from Montreal (Canada), and Anne from Paris 3 (France). The children were recorded for half an hour every fortnight, at home or in another familiar environment, over a total period of approximately eighteen months. The Cat corpus was collected by myself between 1998 and 2001. Data from only one child have so far been transcribed in this corpus: Tom from Brussels (Belgium). Tom was recorded by his grandmother at more irregular intervals, once and three times a month, over slightly more than a year. Each recording lasts forty-five minutes on average. Additional longitudinal data for the Cat corpus were also collected in Brussels (Belgium). These charted the language development of Chloé and were 1 Dialectal variation has not been treated separately in this overview but is mentioned whenever relevant. 2 The research project within which the York corpus was collected and coded was funded by the Economic and Social Research Council (grant R000 22 1972), and was entitled The acquisition of whquestions in French: a cross-dialectal comparison. I was the research assistant on that project. I became very familiar with the data as I checked, tagged, and partially coded all the transcriptions. I would like to take this opportunity to thank Bernadette Plunkett for generously allowing me to use the York corpus in my research. 3 Anne’s parents are from the Pyrenees region.

218

Appendix A

recorded by her mother, Geneviève. These data have been used exclusively for prosodic analyses (the sound quality of the other two corpora was not sufficient for that purpose). The Cat corpus also contains cross-sectional data from two children and four adults, collected over a period of four weeks in May and June 2000, in Montreal (Canada). 4 The speakers included are: Catherine (the interviewer who followed the child Max for the York corpus), Renée, Jade, and Audrey (adults), and Lisette and Pauline (children). Catherine and Renée were recorded for a total duration of two hours, Jade for one hour, Audrey and Lisette for two and a half hours, and Pauline for three hours. All the data collected in Montreal for the Cat corpus was recorded on minidisc. 5 In the case of the York corpus, with the exception of the second half of the Montreal corpus, the transcriptions were made by the interviewer on the basis of the audio tape. They were subsequently checked against the video and coded by myself, a native speaker of (Belgian) French. Transcriptions for the Cat corpus as well as all the coding for both corpora were done by myself. The same transcription and coding methods were applied to both corpora, largely following the conventions of CHILDES, 6 with some additions to facilitate the study of (child) French. This extra set of conventions is discussed in De Cat and Plunkett (2002) and in Plunkett (2002). Two sets of data have been considered in the present research: (i) a subset of the adult data and (ii) a subset of the child data, both from the York and Cat corpora. Within these two large subsets, all the data was manually coded for use with the Varbrul programs. 7 The coding system was designed especially for this study. It is based on the definition of twenty groups of variables relevant to dislocated constructions and their acquisition. The groups of variables enabled the identification of each clause with respect to the following factors: identity of the speaker, information about the clause (type, polarity, function), information about dislocated elements (structural, functional and pragmatic description) and about their resumptive element (nature, function, realization), information about the verb forms (finiteness, overt agreement marking), information about the arguments of the verb (subject, object, attribute), 4 Additional funding for this research was provided by the Economic and Social Research Council and by the Communauté française de Belgique, for which I am most grateful. 5 The equipment used was a Sony MZ-R91 minidisc recorder and a Sony ECM-F8 boundary microphone. I had no choice as to the location for recording spontaneous data collection, and hence no control over echo levels. 6 CHILDES stands for CHIld Language Data Exchange System (see MacWhinney 2000). Details can be found at http://childes.psy.cmu.edu. 7 Varbrul (©Susan Pintzuk 1988) is a group of DOS programs designed to perform variable rule analyses of linguistic data.

Adult data

219

and information specific to questions (wh-movement, inversion, question type, etc.). Also coded was the status of the sentence when it could not be taken to be genuinely representative of spontaneous speech or truly characteristic of spoken French (repetition of another speaker’s utterance, quote from a song or a story, reading from a book, etc.). Within each group, a number of factors were defined to describe the data in enough detail so as to allow the subsequent analysis of precise phenomena. Given that one of the main aims was to study language development, care was taken to ensure that deviance from the target grammar could always be identified. The adult sample contains 5,613 tokens, of which 4,030 are clauses and 1,583 are verbless utterances. This sample corresponds to three recording sessions per country, selected randomly but from across the whole time span of the recording period so as to counter as much as possible any effect that the children’s limited linguistic abilities might have on the adults’ speech. The sample of child data corresponds to what I consider to be the period leading up to full acquisition of the CP layer, as argued in Section 5.4.1. For Max, Tom, and Léa, this stage lasts until approximately 2;10. For Anne, this period extends until just after the age of three. The data set considered for Léa is small (comprising only four recording sessions) because she was already 2;8.22 when recording began. For Max, Tom, and Anne, the data set corresponds fairly closely to the null subject stage. Léa, who consistently seems to be more advanced, has passed that stage by the start of the recordings. The ages and Mean Length of Utterance in Words (MLUw) for each child are given in Appendix B.3. Slightly different data sets are used for particular analyses. 8 In general, two substages are distinguished to give a more accurate picture of development. The relevant turning point is characterized by the following changes in the child’s production: (i) there is a sharp rise in the overall number of sentences produced per recording session, (ii) the proportion of null subjects falls below fifteen per cent in obligatory finite contexts. For Anne and Max, a levelling of the proportion of utterances which involve a dislocated element is also observed at this point. 9 As prosody is a key diagnostic for the identification of dislocated elements, data from sources other than the York and Cat corpora could not be used in the present study. The French corpora available via CHILDES come without sound files, and the transcription conventions followed in these corpora do not allow reliable identification of dislocated elements. 8

Justifications for this are provided where appropriate. In the main, this is likely to be an effect of the rise in the number of full sentences per session. The greater number of utterances most probably diminishes the token-type effect which accounts for variability in the proportion of dislocations. 9

220

Appendix A

A.2 What gets dislocated and where Approximately a fifth to a quarter of clauses contain a dislocated element in the sample. 10 As shown in Section 3.2, most dislocated elements express the topic of the sentence or clause in spoken French, i.e. they express what the sentence is about. Elements that are not traditionally regarded as topics but that are typically dislocated are exemplified in (A.1)–(A.4). 11 Such elements are usually clause-level adjuncts, including temporal (A.1a) or locative modifiers (A.1b), if-clauses (A.2) and other types of clause modifiers (A.3), or dislocated infinitival clauses (A.4). (A.1) a. Maintenant, tu flottes. (A.-Gaël, F) now you float ‘Now you (can) float.’ b. Oui juste à côté, c’ est vert. (Catherine, C) yes just to side it is green ‘Yes, next to it, it’s green.’ (A.2) Donc si on dit un petit nain, ça, on appelle ça un so if one says a little dwarf that one calls that a pléonasme. (Denis, F) pleonasm ‘So if you say a little dwarf, that’s what we call a pleonasm.’ (A.3) Malgré mon rhume, je sens (que) ça sent bon. (Dominique, B) despite my cold I smell that it smells nice ‘Even with my cold, I can smell it smells nice.’ une (A.4) Ben décidément, Léa, tu ne penses qu’ à ça, toi, faire well decidedly Léa you neg think only to that you to-make a queue. (Nelly, B) tail ‘Really, Léa, you can’t think about anything except making a tail.’ Not all dislocated phrases are resumed by an element inside the clause. In most cases, this is due to the fact that the dislocated XP is not associated with an argument of the verb, as in most cases in (A.1)–(A.3). 12 In cases where the 10 A slight variation is observed between the dialects, but it is too small to be significant: the proportion of clauses involving a dislocation in the sample is twenty-two per cent, eighteen per cent, and twenty-four per cent in the varieties from Belgium, Canada, and France respectively. 11 Some of these dislocated elements have been argued to be stage topics, i.e. elements defining the spatio-temporal window within which the predication holds true. See Section 3.1.5. 12 This is true except for the two dislocated objects, ça ‘that’ in (A.2) and the infinitival clause in (A.4), and the dislocated subject toi ‘you’ in (A.4).

Adult data

221

dislocated XP is associated with an argument but is nonetheless not resumed by anything inside the clause, the absence of a resumptive tends to yield a particular reading, different from that of the corresponding sentence with a resumptive. For instance, in (A.5a) and (A.5c) but not in (A.5b) and (A.5d), the dislocated element receives a generic reading (see Section 3.2.2), i.e. it is not interpreted as referring to an individual (or set of individuals) in particular but rather to a type of individual. (A.5) a. Tu aimes bien, # les colliers? you like well the necklaces ‘Do you like necklaces?’ (generic reading) b. Tu lesi aimes bien, les colliers?i you them like well the necklaces ‘Do you like the necklaces?’ (specific reading) c. Le jus de fruits aussi, j’ aime bien. the juice of fruit also I like well ‘I like fruit juice too.’ d. Le jusi de fruits aussi, je l’i aime bien. the juice of fruit also I it like well ‘I like the fruit juice too.’

(Nelly, B)

(Léa 2;9.5)

The absence of a resumptive does not always yield a generic reading though. It can simply mark the dislocated element as the topic of the sentence, as in (A.6), from which the object clitics are omitted. (A.6) a. Tu ne sauras pas écrire, ça. (Dominique, B) you neg will-know not to-write that ‘You won’t be able to write that.’ b. Toi, tu regardes, toi, le film. (Nelly, B) you you watch you the film ‘You’re watching the film.’ c. Tu fais, toi, des petits boudins pour les bras? (A.-Gaël, F) you make you some little sausages for the arms ‘Would you roll some little sausages for the arms?’ When a dislocated element that is not resumed by a clitic appears in the left periphery, the interpretation is usually contrastive. I have treated cases like (A.7) on a par with other dislocations, i.e. not as instances of focus fronting. (A.7) a. Non non ça, tu laisses comme ça. no no that you leave like that ‘No, that one, you leave like that.’

(Bruno, B)

222

Appendix A b. Alors, la soupe de poissons, on va mettre so the soup of fish one will put

(Christine, F)

là? there ‘And then the fish soup, shall we put it there?’ In the examples in (A.7), the left-dislocated element is contrasted with other elements in the context. Sometimes contrast creates a ‘list’ effect (whether or not the dislocated element is associated with a resumptive element), as in (A.8). (A.8) Pi ça, c’ est moi, # ça, c’ est toi, # pi ça, c’ est [/] c’ est and that it is me that it is you and that it is it is 13 (Max 2;4.18) moi. me ‘And that’s me, that’s you, and that’s me.’ In other cases, the absence of a resumptive is due to the omission of an embedded clause, as in (A.9). (A.9) Ah oui Luc, c’ est normal. ah yes Luc it is normal ‘Oh yes, it’s normal (that) Luc (should be like that).’

(Nelly, B)

The absence of a resumptive can also be due to the fact that the dislocated element is only loosely connected to the sentence, as in (A.10). This has led researchers to call such dislocated elements Hanging Topics (see Section 4.3.6 for discussion). (A.10) a. Ah ben ça, où est elle, (il) paraît? ah well that where is she it seems ‘Where is it?’ b. Maman, ça a l’ air d’ aller, oui. Mum it has the air to go yes ‘As for Mum, things seem to be going OK.’ c. C’ est fini, robinet. it is finished tap ‘That’s enough with the tap.’

(Dominique, B)

(Nelly, B)

(Jaco, B)

13 The child is pointing in turn at two monkeys in a picture, pointing at himself when he says the second c’est moi ‘that’s me’.

223

Adult data

Table A.1. Distribution of dislocated elements expressing selected constituents across dialects of spoken French

Belgian French Root clauses Dependent clauses Total Canadian French Root clauses Dependent clauses Total French French Root clauses Dependent clauses Total

Left-Dislocated

Right-Dislocated

Total

46% (143/314) 62% (8/13) 46% (151/327)

54% (171/314) 38% (5/13) 54% (176/327)

96% (314/327) 4% (13/327)

53% (138/258) 74% (17/23) 55% (155/281)

47% (120/258) 26% (6/23) 45% (126/281)

92% (258/281) 8% (23/281)

50% (153/308) 29% (5/17) 49% (158/325)

50% (155/308) 71% (12/17) 51% (167/325)

95% (308/325) 5% (17/325)

In most cases, the dislocated element is associated with a position inside a root clause. Table A.1 represents the distribution of dislocated elements according to the type of clause with which they are associated. 14 Right-dislocated elements are counted separately. In most cases, it is impossible to tell whether a right-dislocated element associated with a position inside an embedded clause appears in the embedded or the matrix clause. In Table A.1, no distinction has been made between dislocated elements appearing in the root clause, as in (A.11a), or in the dependent clause as in (A.11b), in cases where a left-dislocated element is associated with a position inside a dependent clause. Both cases are treated as involving dislocated elements associated with a dependent clause. (A.11) a. Ah Mamani , regarde qu’ est-ce qu’ ellei a look what is-it that she has ah Mum fait! done ‘Ah look at what Mum’s done!’

(Catherine, C)

14 Note that in Table A.1, the figures do not represent the number of clauses but the number of dislocated elements. This means that clauses involving more than one dislocated element are counted multiple times.

224

Appendix A b. Je savais pas que les cochonsi , ilsi avaient des salles de they had some rooms of I knew not that the pigs (Nelly, B) bain, moi. bath me ‘I didn’t know that pigs had bathrooms.’

On the whole, the proportion of left- and right-dislocated elements is the same across dialects. 15 Most dislocated elements (around ninety-five per cent) are associated with a position inside a root clause. There is no significant difference between dialects with respect to the direction of the dislocation when the dislocated element is associated with a position inside a root clause. The proportion of left- vs. right-dislocated elements is roughly the same in such cases. However, when the dislocated element is associated with a position inside a dependent clause, there is a strong preference for LD in Belgian French and Canadian French. The opposite is true in French French. The relevant data set is too small to conclude anything from this contrast. All dislocated elements appear in finite clauses in the present sample. When the resumptive is situated in a (dependent) non-finite clause, the dislocated element always appears in the matrix clause. I found two examples in the present sample (A.12), both from the Belgian French data. faut il lei mettre? (Nelly, B) (A.12) a. Le biberon,i où where must expl it to-put the bottle ‘Where does the bottle go?’ b. Ah mais [les petites bêtes qui marchent sur Maman]i , il ne it neg ah but the little beasties that walk on Mum (Nelly, B) faut pas lesi mettre sur Mamy, hein! must not them to-put on Mamy eh ‘Eh but don’t put on Mamy the little beasties that are walking on Mum!’ In most cases, there is one dislocated element per utterance (or sentence). In the sample under consideration, a maximum of four dislocated elements can be found in a single clause. The distribution of (multiple) dislocations across dialects is given in Table A.2. When there are two or more dislocated elements, they can appear either in the same periphery, as in (A.13a) and (A.13b), or at each end of the sentence, as in (A.13c).

15

The direction of the dislocation varies across utterance types though, as shown in Section A.5.

225

Adult data Table A.2. Number of dislocated elements associated with a single clause in the adult data

Belgium Canada France

One dislocated element

Two dislocated elements

Three or more dislocated elements

79% (257/327) 80% (224/281) 80% (261/325)

19% (61/327) 20% (57/281) 19% (61/325)

3% (9/327) 0% (0/281) 1% (3/325)

(A.13) a. Comme toi, ton prénom, c’ est Léa. (Dominique, B) like you your first name it is Léa ‘It’s the same with you: your first name is Léa.’ (Catherine, C) b. Çai nage dans l’ eau, çai , les dauphinsi ? it swims in the water that the dolphins ‘Do dolphins swim in water?’ (Denis, F) c. Ben Gaëtani , ili l’ j a pris avec luii , le livre j . well Gaëtan he it has taken with him the book ‘Gaëtan has taken the book away.’ In most cases, one of the dislocated elements is coreferential with the subject, which is to be expected given that overall most dislocated elements are coreferential with the subject (see Section 3.1.5). Multiple dislocated elements are sometimes coreferential with each other, but not necessarily so. In (A.14a), each of the three dislocated elements expresses something different (subject, object, clause modifier). In (A.14b), the subject dislocation is duplicated (on the left and the right). In (A.14c), two dislocated elements express the object, the second one (on the right) disambiguating the referent in question. (A.14) a. Toi aussi, tu en fais, des cadeaux, Anne, you too you of-them make indef presents Anne partout? (A.-Gaël, F) everywhere ‘Do you also draw presents everywhere, Anne?’ (drawing activity) b. Toi, tu regardes, toi, le film. you you watch you the film ‘As for you, you’re watching the film.’

(Nelly, B)

226

Appendix A c. Oui moi, l’ autre, c’ est Dingo, ## l’ autre que j’ yes me the other it is Dingo the other that I (Pol, 4;10.26) ai. 16 have ‘The other one I have is Dingo.’

When a second dislocated element situated in the right periphery of the sentence is coreferential with (but not identical to) the preceding dislocated element, it is often preceded by a pause. It could be that right-dislocated elements of this type, which disambiguate the reference of another element in the clause, are instances of so-called afterthought. 17 Example (A.15) illustrates the possibility of having four dislocated elements within a single clause. In this example, we find two dislocated adverbs (après ‘then, after’ and alors ‘so, then’), a dislocated subject (Gaëtan), and a dislocated object (les bougies ‘the candles’). The dislocated object is not resumed by an element inside the clause, but the fact that it follows an IP modifier forces a RD analysis of it. 18 (A.15) Et après, Gaëtani , ili a soufflé, alors, les bougies? and then Gaëtan he has blown then the candles ‘And did Gaëtan blow out the candles, then?’

(Denis, F)

I have found no strict restrictions on the ordering of dislocated elements relating to the function of their resumptive element, contra Larsson (1979). 19 In particular, dislocated subjects can appear before or after other dislocated elements, regardless of whether both elements are in the left 16 The adult-like utterance in (A.14c) has not been included in the counts reported in this section as it is uttered by a child. The reason it is included here is that I wanted to show that coreferential dislocated elements can also be found in the Canadian data, in spite of their absence from the adult sample of that dialect. 17 Besides the presence of a pause, the prosody of such elements is no different to that of other rightdislocated elements. I have treated cases that might be analysed as afterthoughts as right-dislocated. 18 The prosody of this example clearly indicates that alors ‘then, after’ is not parenthetical: the main stress falls on soufflé ‘blown’, indicating the end of the sentence nucleus. See Section 2.3.1. 19 More precisely, Larsson (1979: 120) claims that left-dislocated elements allegedly derived by movement cannot precede base-generated left-dislocated elements. Larsson claims that only elements bearing a dependency marker (like a preposition, see Section 4.3.6) are dislocated by a movement operation. Hence (i) is claimed to be ungrammatical, contrary to fact.

(i)

De ce truci -là, David j , j’ose pas trop lui j eni parler. of that thing-there David I-dare not too-much to-him of-it talk ‘I’m not sure I dare talk to David about that thing.’

In a context where David is salient, most native speakers accept (i) and other such sentences.

Adult data

227

periphery (A.16), the right periphery (A.17), or both peripheries (A.18) and (A.13c). (A.16) a. Le rosei , ici, ili va peut-être là. the pink here he goes may be there ‘Maybe the pink one over there goes here.’ b. Ici, çai , c’i est la [//] le mur. the wall here that it is the ‘Here is the wall.’

(Catherine, C)

(Catherine, C)

(A.17) a. Est-ce que tui la j trouves, la tête j du lion, toii ? (Sarah, F) is-it that you it find the head of-the lion you ‘Can you find the lion’s head?’ b. Tui vas la j chercher, toii , Anne, la feuille j de you go it to-fetch you Anne the leaf of capucine? (Denis, F) nasturtium ‘Can you go and get the nasturtium leaf, Anne?’ (A.18) Les moutons j , jei les j ai déjà mis, moii . I them have already put me the sheep ‘I’ve already put the sheep (in the jigsaw).’

(Cécile, B)

However, this absence of restrictions does not mean that the order is completely free. Indeed, Vion (1992) reports that in her data, the S(V)O order of NPs prevails even when dislocations are taken into account: NPs associated with the subject tend to appear before NPs associated with other arguments. In the sample data investigated here, the order SV and SO have also been found to be prevalent. More complex cases are not uncommon, for instance where a dislocated clause itself contains a dislocated element, as in (A.19). (A.19) Et si moii , jei viens et que je casse tous tes jouets, tu seras and if me I come and that I break all your toys you will-be (Dominique, B) contente? happy ‘And if I come and break all your toys, will you be happy?’ In this brief overview, we have seen that not all dislocated elements are topics, but most of them are. The commonest case is a sentence containing a single dislocated element, usually coreferential with the subject. The maximum number of dislocated elements attested in a single clause is four in the present

228

Appendix A

sample. There does not seem to be a preference for the direction of the dislocation when all the data are considered as a whole, nor do there seem to be any restrictions on the ordering of dislocated elements. All the observations so far hold for all three dialects under investigation. Approximately a quarter of clauses contain a dislocation in spoken French, as it is represented in the present sample. I now turn to describing the variation observed in the dislocation data. The following criteria were used to characterize the variation observed: nature of the dislocated element, function of the resumptive element or empty position within the clause associated with the dislocated element, position of the dislocated element, and direction of the dislocation across clause types.

A.3 Nature of the dislocated element and function it is associated with In the sample under investigation, most dislocated elements are DPs. Most dislocated DPs are definite, as in (A.20). (A.20) Et les jouetsi , qu’ est-ce qu’ ilsi font là-dedans? what is-it that they do in-there and the toys ‘And what do the toys do in there?’

(Nelly, B)

In some cases, the DP only consists of a proper name (A.21). (A.21) Tu sais qui c’ est, Dominique. you know who it is Dominique ‘You know who Dominique is.’

(Nelly, B)

Some dislocated DPs are demonstrative. (A.22) Eh il est gros, ce lapin-là hein? eh he is fat that rabbit-there eh ‘That’s a fat rabbit, isn’t it?’

(Catherine, C)

Some are complex DPs, i.e. DPs modified by a relative clause. (A.23) Eh c’ est bizarre, ça, une voiture qui a une casquette. (A.-Gaël, F) eh it is weird that a car that has a cap ‘Eh that’s weird, a car with a cap!’ Dislocated indefinite and quantified DPs are also found, but never with an existential reading, as demonstrated in Section 3.2.

Adult data

229

Table A.3. The functions associated with dislocated DPs and strong pronouns in spoken French Function of the resumptive Subject Other selected constituent DP modifier Clause modifier No function

Dislocated DPs

Dislocated 3 person pronouns

Dislocated 1,2 person pronouns

86% (315/366) 11% (42/366) 0% (0/366) 2% (6/366) 1% (3/366)

89% (156/175) 7% (13/175) 0% (0/175) 1% (1/175) 3% (5/175)

98% (125/128) 0% (0/128) 2% (3/128) 0% (0/128) 0% (0/128)

(A.24) Tout le mondei , le soir de la fête, oni était tousi all the people the evening of the party one was all (Denis, F) fatigués. tired ‘We were all tired on the night of the party.’ Most dislocated DPs (eighty-six per cent, see Table A.3) express the subject of the clause containing their resumptive element, as in the examples above. In other words, the function of their resumptive element is that of subject. A small proportion (eleven per cent) of dislocated DPs express the direct object (e.g. the right-dislocated element in (A.13c)). Other cases in which the dislocated element expresses an indirect object (A.25a), an attribute (A.25b), 20 or a temporal adjunct (A.25c), are very few and far between. mettre la main (A.25) a. Et le petit bonhommei , tu vas luii the hand and the little man you will to-him put gauche? (A.-Gaël, F) left ‘Are you going to put the little man’s left hand on?’ b. Ah oui il y en a, des phoques. (Catherine, C) ah yes it there of-them has indef seals ‘Oh yeah there are seals.’ c. Et ce matin, on a vu ## le soleil. (Jaco, B) and this morning one has seen the sun ‘And this morning we saw the sun.’ 20 In (A.25b), there is no thematic subject and the dislocated element is associated with the complement position rather than the subject position.

230

Appendix A

There are also cases where the dislocated DP does not correspond to a function in the clause (A.26). This is sometimes due to the omission of a clause, as in (A.26b). (A.26) a. Maman, ça a l’ air d’ aller, oui. mum it is the air to go yes ‘Yes, Mum seems ok.’ b. Eh ben le papa, je sais pas. ah well the dad I know not ‘As for the dad, I don’t know (where he is).’

(Nelly, B)

(A.-Gaël, F)

The second most common type of dislocated elements are strong pronouns. Most instances are singular, as in (A.27), but some rare cases involving a first person plural pronoun are also attested in the sample (A.28). (A.27) Mais ili est habitué, luii , de marcher dans l’ eau. him to walk in the water but he is used ‘But he’s used to walking in the water.’ (A.28) Ben nous, on va faire le chat. well us one will make the cat ‘We’ll make the cat.’

(Catherine, C)

(A.-Gaël, F)

Dislocated first and second person pronouns are almost exclusively associated with the subject of the clause. Third person pronouns are more similar to dislocated DPs as far as the function they are associated with is concerned. PPs are less often dislocated than DPs and strong pronouns, but they do nonetheless occur in both peripheries. Most dislocated PPs (sixty-seven per cent of cases, see Table A.4) correspond to non-selected constituents, as in (A.29). In such cases, the PP can be for instance a locative adjunct (A.29a), an Table A.4. The functions associated with dislocated PPs in spoken French Function of the resumptive

Dislocated PPs

Selected constituent Clause modifier Non-selected constituent DP modifier

17% (16/93) 44% (41/93) 23% (21/93) 16% (15/93)

Adult data

231

instrumental adjunct (A.29b), a temporal adjunct (A.29c), or another type of non-selected constituent (A.29d). (A.29) a. Pi dans son chapeau # il y avait un verre d’ and in his hat it there was a glass of eau. (Catherine, C) water ‘And in his hat, there was a glass of water.’ b. Avec ça ici là, # avec ça # et avec ça, # je pense qu’ with that here there with that and with that I think that on pourra faire une petite maison. (Catherine, C) we will-be-able to-make a little house ‘I think we can make a little house with this, this, and this.’ c. Pi après ça, on va revenir jouer dans la and after that one will return to-play in the chambre. (Martine, C) room ‘And after that we’ll come back to play in the bedroom.’ (Jaco, B) d. Comme ça, tu es caché? that you are hidden like ‘You’re hidden, like that?’ Dislocated PPs can also express the indirect object, as in (A.30). This can happen even in the absence of a resumptive, as seen in (A.30b). (A.30) a. On va luii faire des moustaches, au one will to-him make indef moustaches to-the monsieuri . mister ‘We’ll make the man a moustache.’ b. Aux dominos, tu ne veux plus jouer? to-the dominoes you neg want no-more to-play ‘Don’t you want to play dominoes anymore?’

(A.-Gaël, F)

(Nelly, B)

Another function associated with dislocated PPs is that of DP modifier, as in (A.31). Configurations where a right-dislocated element is resumed by the clitic en have been extensively studied (see e.g. Kayne 1975; Milner 1978; Emonds 2000). I have not investigated them in detail in this book. as, ## de doigtsi ? (A.31) a. Toi, combien t’ eni you how-many you of-them have of fingers ‘How many fingers do you have?’

(Catherine, C)

232

Appendix A b. Il y en a uni qui veut pas se laver les mains, it there of-them has one who wants not refl wash the hands des nainsi ? (Denis, F) of-the dwarfs ‘There’s a dwarf who doesn’t want to wash his hands?’ c. Avec le chandail rouge, c’ est qui? (Catherine, C) with the jumper red it is who ‘Who is it with the red jumper?’

The remainder of the dislocated elements are adverbs and (finite or non-finite) clauses. 21 In most (i.e. 115 out of 116) cases, dislocated adverbs are, as expected, not associated with a function within the clause. I have found one exception to this in the present sample, where the adverb is associated with the subject of the clause (A.32). (A.32) Cei n’ est pas un bon jour, aujourd’huii , hein? eh it neg is not a good day today ‘Today’s not a good day, is it?’

(Nelly, B)

As for dislocated clauses, they can express the subject (A.33), the direct or indirect object (A.34), or be clause modifiers of various sorts (A.35). Whenever a dislocated clause expresses the subject, it appears with a copular main clause in which the impersonal subject clitic ce ‘it’ is its resumptive. (A.33) a. Parce que [se donner en spectacle comme ça]i , hein, cei n’ est that eh it neg is because refl give in show like vraiment pas beau, sais-tu. (Dominique, B) really not handsome know-you ‘Because it really doesn’t look good to show off like that, you know.’ (A.-Gaël, F) b. C’i est ça, [ce que tu me dis]i ? it is that that that you to-me say ‘Is that what you’re telling me?’ 21

(i)

There is also one (target-like) instance of a dislocated adjective in the children’s data. Coiffé, il est plus beau, hein. hairdressed he is more handsome eh ‘He looks better with his hair done, doesn’t he?’

(Léa, 3;8.26)

The prosody of coiffé is clearly that of a left-dislocated element. It is interpreted as a stage topic. This adjective in effect stands for an if-clause (‘when his hair is done, he looks nicer’) and is interpreted generically, as expected under a dislocation analysis (see Chapter 3).

Adult data

233

Dislocated clauses expressing the direct object are always coindexed with the strong pronoun ça in the sample. As shown in Section 3.2.2, this forces a particular interpretation of the dislocated clause. (Catherine, C) (A.34) a. Il j aime çai , Crocro j # [qu’ on le caresse]i . he likes that Crocro that one him stroke ‘Crocro likes to be stroked.’ b. Ben décidément, Léa, tu ne penses qu’ à çai , toi, [faire well decidedly Léa you neg think only to that you to-make (Nelly, B) une queue]i . a tail ‘Really, Léa, the only thing you can think about is making a tail.’ In the sample, dislocated clauses that are not associated with a function inside the sentence are either temporal (A.35a) or locative adjuncts (A.35b), or they express the premise of a reasoning (A.35c), a comparison (A.35d), an aim (A.35e), or a condition (A.35f). (A.35) a. Tu seras déjà grande, sais-tu, quand tu auras dix you will-be already big know-you when you will-have ten ans. (Dominique, B) years ‘You’ll already be grown up, you know, when you’re ten.’ b. Tu sais, Max, où est-ce que Catherine habite you know Max there is-it that Catherine lives là +//. (Catherine, C) there ‘You know, Max, where Catherine lives, . . . ’(self-interruption) c. Puisque Gaëtan avait sept ans, le gâteau, il a eu sept as Gaëtan had seven years the cake he has had seven ans. (A.-Gaël, F) years ‘Since Gaëtan was seven, the cake was seven too.’ d. Puis elle va avoir mal aux dents, comme Maman l’ a Mum it has then she will have pain to-the teeth as expliqué tantôt. (Nelly, B) explained earlier ‘And then she’ll get a toothache, as Mum just explained.’

234

Appendix A e. Parce que pour faire le cochon, j’ ai besoin de beaucoup de because to make the pig I have need of much of (Nelly, B) plasticine, hein? plasticine eh ‘Because to make the pig, I need lots of plasticine.’ f. On pourrait mettre des choses dans le sac pi les piger, one could put indef things in the bag and them pick (Catherine, C) si tu veux. if you want ‘We could put some things in the bag and fish for them, if you want.’

Table A.5 summarizes the various functions with which dislocated elements are associated in the sample according to the nature of the dislocated element. The percentages in the central part of the table are calculated on the basis of the rows: fifty-two per cent of dislocated elements associated with the subject are DPs, etc. The figures in the rightmost column give the total of the elements in each row divided by the total sum of all tokens. This column thus gives the proportion of dislocated elements associated with each function in the data

Table A.5. The nature and function of dislocated elements in spoken French Grammatical function

Subject Other argument

Nature of the dislocated element DP

Pronoun

52% (315/602) 56% (42/75)

47% (281/602) 17% (13/75) 16% (3/19)

5% (6/110) 2% (3/126)

5% (6/126)

21% (16/75) 79% (15/19) 37% (41/110) 17% (21/126)

39% (366/932)

33% (303/932)

10 % (93/932)

DP modifier Clause modifier No function Total

PP

Clause

Adverb

Total

1% (5/602) 5% (4/75)

0% (1/602)

65% (602/932) 8% (75/932) 2% (19/932) 12% (110/932) 14% (126/932)

3% (3/110) 33% (41/126)

5% (1/19) 55% (60/110) 44% (55/126)

6% (53/932)

13% (117/932)

Adult data

235

as a whole: sixty-five per cent of dislocated elements are associated with the subject in the present sample, etc. The figures in the row at the bottom of the table give the total of the elements in each column divided by the total sum of all tokens. They thus give the proportion of each type of dislocated element, according to their nature, in the data as a whole: thirty-nine per cent of dislocated elements are DPs in the present sample, etc.

A.4 ‘Locality’ of the dislocated element We have seen above that most dislocated elements are ‘local’ in the sense that, in the majority of cases, they appear in the clause containing their resumptive element (or to which their function is related). Locality is also observed inside the clause itself: in most cases, no element intervenes between the dislocated element and the subject in the left periphery, or between the dislocated element and the right boundary of the VP. In Table A.6, such cases are identified as ‘local’. ‘Non-local’ cases are those in which an overt element intervenes between the dislocated element and the IP (or VP) boundary. In most cases, the intervening element is itself a dislocated element, as in (A.36). (A.36) a. Et [ce qui reste là sur la table]i , là, c’i est and that that remains there on the table there it is quoi? (Nelly, B) what ‘And what’s that stuff left there on the table?’ (Catherine, C) b. Il aime çai , Crocro # [qu’ on le caresse]i . he likes that Crocro that one him stroke ‘He likes being stroked, Crocro.’ c. Tu fais, toi, [des petits boudins pour les bras]? (A.-Gaël, F) you make you some little sausages for the arms ‘Would you roll some little sausages for the arms?’ The intervening element can also be a moved wh-element (A.37a), sometimes together with a complementizer (A.37b), 22 an interjection or tag (A.38a), or a vocative (A.38b), (A.38c). (A.37) a. Les autres garçonsi , comment ilsi s’ appelaient? (A.-Gaël, F) they refl called how the other boys ‘What were the other boys called?’ 22 Plunkett (2000) argues that in Canadian French est-ce que is an unanalysed interrogative complementizer. See Section 5.4.1 for details.

236

Appendix A Table A.6. Locality of dislocated elements associated with an argument of the verb Relative location of dislocated element and resumptive Same clause Local

Different clauses

Non-local

Left-dislocated elements 88% (408/462) 9% (41/462)

3% (13/462)

Right-dislocated elements 87% ? (406/466) 13% ? (59/466)

0% ? (1/466)

b. Monsieur Patatei , où est-ce qu’ ili est? Mister Potato where is-it that he is ‘Where is Mister Potato?’

(Catherine, C)

(A.38) a. C’ est un problème, hein, les boutonnières, hein? (Nelly, B) it is a problem eh the buttonholes eh ‘They’re tricky, buttonholes, aren’t they?’ (Nelly, B) b. Çai ne pleure pas, hein, chouchou, une plantei . it neg cries not eh darling a plant ‘Plants don’t cry, darling.’ c. Je vais te montrer quelque chose, Max, avec I will to-you show some thing Max with les lunettes. (Catherine, C) the glasses ‘I’ll show you something with the glasses, Max.’ Table A.6 contains information about the locality of the dislocated element, i.e. whether it and its resumptive appear inside the same clause or different clauses, and in the case of the former if they are ‘local’ or ‘non-local’. 23 The data considered here are exclusively sentences in which dislocated elements are associated with an argument of the verb. When the dislocated element appears on the right, it is generally impossible to determine whether it is situated in the matrix clause or in the embedded 23 It is not always the case that the clause in question effectively contains a resumptive element, as we have seen earlier in the text. I have treated such cases as if they do (e.g. by assuming a null pronominal) for the present purpose.

Adult data

237

clause (hence the question marks in Table A.6) as there is typically no marking of clause-final boundaries. When a right-dislocated element associated with the matrix clause appears after an embedded clause, I have coded the dislocated element as adjoined to the matrix clause. In (A.39), for instance, the right-dislocated element is associated with the subject of the matrix clause, and clearly not with any element of the embedded clause. (A.39) Jei sais pas [laquelle va où], moii , hein? I know not which goes where me eh ‘I don’t know which one goes where, you know.’

(Dominique, B)

In examples like (A.40), I have analysed the dislocated element as being associated with the subject of the matrix clause, in spite of the fact that it is also coindexed with the PRO subject of the embedded clause. 24 (A.40) Jei sais pas vraiment comment PROi faire, moii . to-do me I know not really how ‘I don’t know how to do this.’

(Catherine, C)

This analysis is supported by the fact that a dislocated element associated with a constituent in an embedded clause is typically realized in the matrix clause, as in (A.41). (A.41) Le biberoni , où faut il lei mettre? the bottle where must it it to-put ‘Where does the bottle go?’

(Dominique, B)

Hence even in (A.42), where the right-dislocated element is clearly associated with an element in the embedded clause (and not with one in the matrix clause), this element could be situated in the matrix clause in spite of there being no clear indication that it is. (A.42) Ah tu sais ce que c’ est, ça aussi. ah you know that that it is that too ‘Ah you know what this is, too.’

(A.-Gaël, F)

Other potentially ambiguous cases involving right-dislocated elements are of the type shown in (A.43). 24 Further research is required to determine whether dislocated elements can appear inside an embedded non-finite clause.

238

Appendix A

(A.43) Ellei va pas être contente, [Sarah]i , [si tu lui racontes rien], she will not be happy Sarah if you to-her tell nothing hein. (Denis, F) eh ‘Sarah won’t be happy if you don’t tell her anything.’ I have treated such cases as involving two right-dislocated elements even though, in principle, Sarah could be a left-dislocated element associated with the dependent clause. Two facts motivate a RD analysis: (i) the prosody of Sarah is typical of a right-dislocated element (it is uttered on a low, flat pitch; see Section 2.3.1), and (ii) most dislocated elements are associated with a subject rather than an indirect object, which favours an analysis of Sarah as associated with elle ‘she’ in the matrix clause rather than lui ‘to her’ in the dependent clause. We have seen that, in general, when a dislocated XP is associated with an element inside a dependent clause, that XP appears in the matrix clause (see Table A.6). Examples of this configuration are given in (A.44). assoie en face. (A.-Gaël, F) (A.44) a. Papai , il faut qu’ ili s’ daddy it must that he refl sits in front ‘Daddy needs to sit opposite (you).’ b. Ah Mamani , (re)garde qu’ est-ce qu’ ellei fait! (Catherine, C) ah Mum look what is-it that she does ‘Ah look at what Mum’s doing!’ c. çai , tu ne crois pas que c’i est la même that you neg think not that it is the same (Dominique, B) couleur? colour ‘Don’t you think that this is the same colour?’ This is not because it is impossibile for a dislocated element to appear in a dependent clause, as attested by (A.45). (A.45) a. On regarde dans le livre [pour voir Monsieur Patatei , [où we look in the book to see Mister potato where (Catherine, C) est-ce qu’ ili est]]? it-is that he is ‘Shall we look in the book to see where Mister Potato is?’ (Denis, F) b. Elle enregistre [ce que toii , tui dis]. she records that that you you say ‘It records what you say.’

Adult data

239

c. Ah je pense [que le crapaud j , je lei sais, moi, # [où il j where he ah I think that the toad I it know me va]i ]. (Catherine, C) goes ‘I think I know where the toad goes.’ Note however that in (A.45c), while the dislocated element le crapaud ‘the toad’ appears in an embedded clause (i.e. the object clause je le sais ‘I know it’), this is not the clause which contains its resumptive (i.e. the dislocated clause où il va ‘where he goes’). From these examples, we can observe that (i) left-dislocated elements can appear in object clauses (A.45c), but the preferred option might be for them to appear in the matrix clause; (ii) left-dislocated elements associated with an element in an embedded wh-clause do not appear inside that clause in the present sample, but rather to the leftmost edge of it (before the embedded wh-element and complementizer (A.45a)) or in a higher clause (A.44b); (iii) left-dislocated elements are attested inside relative clauses (A.45b). All this suggests that there may be some restrictions on the presence of (left-)dislocated elements in dependent clauses. This is discussed in Section 4.4. Another construction which rarely allows left-dislocated elements is a matrix wh-question with a fronted question word (A.46). In the present sample, left-dislocated elements are only attested after the wh-word pourquoi ‘why’. (A.46) Mais pourquoi ces deux-lài , ilsi vont ensemble? but why these two-there they go together ‘But why do these two go together?’

(Sarah, F)

With respect to all other fronted wh-words, the dislocated element appears either before the question word (A.47a) or as a right dislocation (A.47b). (A.47) a. Les yeuxi , comment je vais faire çai ? the eyes how I will do that ‘How shall I make the eyes?’ b. Comment ili fait, le canardi ? how he does the duck ‘What does the duck do?’

(Catherine, C)

(A.-Gaël, F)

This apparent restriction is most probably not due to the fact that only a portion of wh-questions have a fronted question word in spoken French (which would reduce the chances of finding left-dislocated elements following moved wh-words in any sample), given that a similar restriction is attested in many

240

Appendix A Table A.7. Direction of dislocation in root wh-questions across dialects of spoken French

Belgium Canada France Overall

Left-dislocated elements

Right-dislocated elements

21% (18/84) 28% (10/36) 26% (26/100) 25% (54/220)

79% (66/84) 72% (26/36) 74% (74/100) 75% (166/220)

other languages that do not display ‘optionality’ of wh-movement to the extent that French does. The ban on the order wh-element+topic has been observed in Modern Greek (Iatridou 1995), Italian (Rizzi 1997), Spanish (Zubizarreta 1998), European Portuguese (Barbosa 2000), Ma’di (Nigel Fabb, p.c.), and many other languages.

A.5 Direction of the dislocation We have seen that overall the proportion of LD and RD is almost identical. However, grouping all the data together hides variation across clause types: in particular, there is a striking preference for RD in wh-clauses, especially in root wh-questions (whether wh-words are moved or appear in situ). Table A.7, shows the distribution of LDs and RDs in root wh-questions for each dialect. In yes/no questions, there is also a slight preference for right-dislocated elements, as shown in Table A.8. What this table does not indicate is that the preference for RD in this utterance type relates essentially to dislocated subjects. Table A.8. Direction of dislocation in root yes/no questions across dialects of spoken French

Belgium Canada France Overall

Left-dislocated elements

Right-dislocated elements

51% (20/39) 33% (15/45) 42% (21/50) 42% (56/134)

49% (19/39) 67% (30/45) 58% (29/50) 58% (78/134)

Adult data

241

Table A.9. Direction of dislocation in root declarative clauses across dialects of spoken French

Belgium Canada France Overall

Left-dislocated elements

Right-dislocated elements

57% (107/187) 64% (115/179) 68% (113/166) 63% (335/532)

43% (80/187) 36% (64/179) 32% (53/166) 37% (197/532)

In declaratives sentences, left-dislocated elements are more common than right-dislocated elements overall. Again, most dislocated elements express the subject, and dislocated subjects show the clearest directional preference. Table A.10. Proportion of root clauses involving dislocation across utterance types in spoken French With a dislocated element

Without a dislocated element

Wh-questions Belgium Canada France Overall

56% (84/150) 38% (36/96) 43% (100/233) 46% (220/479)

44% (66/150) 63% (60/96) 57% (133/233) 54% (259/479)

Yes/no-questions Belgium Canada France Overall

18% (39/222) 17% (45/267) 21% (50/239) 18% (134/728)

82% (183/222) 83% (222/267) 79% (189/239) 82% (594/728)

Declaratives Belgium Canada France Overall

31% (187/608) 29% (179/610) 33% (166/503) 31% (532/1721)

69% (421/608) 71% (431/610) 67% (337/503) 69% (1189/1721)

242

Appendix A

A.6 Proportion of dislocations across utterance types The proportion of root clauses containing a dislocated element is given in Table A.10 for each utterance type. On average, almost half of all wh-questions involve a dislocated element. The proportion of declaratives containing a dislocated element is a third, while a fifth of all yes/no questions involve dislocation.

Appendix B Child data B.1 Transcription conventions Below are the transcription conventions used in the examples. # -E

, () 0 e

xx yy +/? +//. & [//] [!] [=? . . . ] [?] [: . . . ]

indicates a short pause (repetition of this code is proportional to the duration of the pause) at the end of a verb stands for the [e] sound, which is ambiguous between infinitival and participial morphology in verbs of the first class indicates a syntactic juncture indicates an unpronounced string (inside the parentheses) indicates a missing element stands for an ‘embryonic’ element, i.e. a sound roughly corresponding to a schwa in the child’s pronunciation, which appears in a slot normally occupied by a function word (see Bottari et al. 1992; Bohnacker 1998; Peters 2001) stands for an unintelligible word stands for an unintelligible word and is followed by a rough phonetic transcription indicates an interruption in a question indicates a self-interruption in a declarative indicates a self-interrupted word indicates the beginning of a reformulation indicates emphasis on the previous word (after an unintelligible word) indicates a possible interpretation indicates that the preceding word is unclear; this applies to a string of words if they appear inside triangular brackets after a child form (e.g. mispronounced), indicates the target form

244

Appendix B

B.2 Information about the children studied Anne (York corpus) was born in Paris, France, on 18 June 1995. She was also brought up in that city. She is the eldest of two children. Her parents are originally from the Pyrenees region, where the family spends most of the summer holiday every year. Until the age of 3;2, Anne was looked after by a child-minder on weekdays, together with another little girl who was a few months older than her. The birth of Anne’s brother (when Anne was 2;6) seems to have triggered a slight regression in Anne’s language development and she was frequently reprimanded by her parents for speaking ‘like a baby’ at the time. Some recording sessions were made by Anne’s parents in the absence of Sarah, the interviewer. Lisette (Cat corpus) was born in Montreal in October 1997. She was an only child at the time of the recordings. Léa (York corpus) was born on 17 May 1994. She was brought up in Liège, Belgium. She is the eldest of two children. Her brother is one and a half years younger than her. Léa was already attending nursery school when the recordings began. The sessions took place either at Léa’s home or at her maternal grandmother’s. All the recording sessions were conducted by Léa’s grandmother with the collaboration of either her husband or her daughter (Léa’s mother). Max (York corpus) was born on 27 March 1995. He was brought up in Montreal, Canada. He is the youngest of two children (he is two and a half years younger than his brother Pol). All recording sessions took place at Max’s home. His mother was often present. The interviewer was Catherine. Pol (York corpus) is Max’s older brother. He was born in Montreal (Canada) on 21 August 1992. Tom (Cat corpus) was born on 25 June 1996. He was brought up in Brussels, Belgium. He is the youngest of two children. His sister is two and a half years older than him. Tom was a relatively ‘late’ talker: the first recording took place when he was aged 2;1.11, two weeks after his parents reported he had started talking. No delay in his development was observed compared to that of the other children at similar ages. Until the age of 2.6, Tom attended a nursery on weekdays. He then started nursery school. All the recording sessions were conducted by his paternal grandmother, usually at his grandmother’s home.

B.3 Ages and Mean Length of Utterance in words The children’s ages at the time of each recording are given in Tables B.1 and B.2. Only those recordings included in the data set which was studied for this book are included in the tables. 1 1

There is no Session 5 for Max.

Child data

245

Table B.1. Max and Anne’s ages at each recording session Max 1 2 3 4 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

15-JAN-1997 30-JAN-1997 13-FEB-1997 27-FEB-1997 27-MAR-1997 10-APR-1997 24-APR-1997 13-MAY-1997 22-MAY-1997 5-JUN-1997 18-JUN-1997 3-JUL-1997 17-JUL-1997 31-JUL-1997 14-AUG-1997 28-AUG-1997 11-SEP-1997 25-SEP-1997 9-OCT-1997 24-OCT-1997 7-NOV-1997 21-NOV-1997 6-DEC-1997 19-DEC-1997 8-JAN-1998 23-JAN-1998 6-FEB-1998 20-FEB-1998

Anne 1;9.19 1;10.3 1;10.17 1;11.0 2;0.0 2;0.14 2;0.28 2;1.16 2;1.25 2;2.9 2;2.22 2;3.6 2;3.20 2;4.4 2;4.18 2;5.1 2;5.15 2;5.29 2;6.12 2;6.27 2;7.11 2;7.25 2;8.9 2;8.22 2;9.12 2;9.27 2;10.10 2;10.24

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

30-APR-1997 14-MAY-1997 31-MAY-1997 16-JUN-1997 1-JUL-1997 15-JUL-1997 6-AUG-1997 18-AUG-1997 7-SEP-1997 17-SEP-1997 3-OCT-1997 20-OCT-1997 7-NOV-1997 22-NOV-1997 6-DEC-1997 20-DEC-1997 5-JAN-1998 19-JAN-1998 3-FEB-1998 21-FEB-1998 10-MAR-1998 2-APR-1998 6-MAY-1998 20-MAY-1998 20-JUN-1998 30-JUN-1998 22-JUL-1998 2-AUG-1998

1;10.12 1;10.26 1;11.13 1;11.29 2;0.13 2;0.27 2;1.19 2;2.0 2;2.20 2;2.30 2;3.15 2;4.2 2;4.20 2;5.4 2;5.18 2;6.2 2;6.18 2;7.1 2;7.16 2;8.3 2;8,20 2;9,15 2;10.18 2;11.2 3;0.2 3;0.12 3;1.4 3;1.15

The data set from Léa is very small because the recordings started when she was already 2;8.22. The blank cells in Tom’s table correspond to recordings that have not been transcribed. The Mean Length of Utterance in words (MLUw) is given for each child in Tables B.3–B.6. 2 2

Utterances consisting of only oui ‘yes’ or non ‘no’ were excluded from the counts.

246

Appendix B Table B.2. Tom and Léa’s ages at each recording session Léa 1 2 3 4

8-FEB-1997 22-FEB-1997 10-MAR-1997 24-MAR-1997

Tom 2;8.22 2;9.5 2;9.21 2;10.7

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

4-AUG-1998 7-AUG-1998 8-AUG-1998

2;1.11 2;1.13 2;1.14

17-OCT-1998 2-NOV-1998

2;3.22 2;4.8

3-NOV-1998 19-DEC-1998 19-DEC-1998

2;4.9 2;5.24 2;5.24

6-JAN-1999 16-JAN-1999 25-FEB-1999 13-APR-1999 2-MAY-1999

2;6.12 2;6.22 2;8.0 2;9.19 2;10.7

B.4 Language development stages The data from the three youngest children has been subdivided into stages for the various analyses presented in Chapter 5. Two subdivisions have been made: one according to the Root Infinitive phenomenon and one according to the null subject phenomenon. Nothing in the analyses presented in this book hinges on these subdivisions of the data into periods. What is gained, I Table B.3. MLUw and age at each recording of Léa Session

1 2 3 4

Age

Number of utterances

Number of words

Words per utterance

Standard deviation

2;8.22 2;9.5 2;9.21 2;10.7

373 259 362 318

1,820 884 2,217 1,554

4.879 3.413 6.124 4.887

3.467 2.643 4.982 4.581

247

Child data Table B.4. MLUw and age at each recording of Max Session

1 2 3 4 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

Age

Number of utterances

Number of words

Words per utterance

Standard deviation

1;9.19 1;10.3 1;10.17 1;11.0 2;0.0 2;0.14 2;0.28 2;1.16 2;1.25 2;2.9 2;2.22 2;3.6 2;3.20 2;4.4 2;4.18 2;5.1 2;5.15 2;5.29 2;6.12 2;6.27 2;7.11 2;7.25 2;8.9 2;8.22 2;9.12 2;9.27 2;10.10 2;10.24

123 129 185 261 171 199 218 228 244 197 201 209 209 157 192 198 185 231 249 269 119 204 212 231 181 210 238 240

141 137 250 430 291 364 413 491 572 554 505 582 532 496 632 762 627 809 934 875 362 544 629 694 627 654 832 842

1.146 1.062 1.351 1.648 1.702 1.829 1.894 2.154 2.344 2.812 2.512 2.785 2.545 3.159 3.292 3.848 3.389 3.502 3.751 3.253 3.042 2.667 2.967 3.004 3.464 3.114 3.496 3.508

0.376 0.241 0.87 1.031 1.037 1.103 1.389 1.544 1.85 2.361 2.032 2.215 2.237 2.677 3.07 3.203 2.62 2.824 2.9 2.664 2.548 2.398 2.6 2.445 2.614 2.341 2.557 2.672

believe, is a more accurate picture of development (though one which is not overwhelmed with detail). B.4.1 The Root Infinitive stage

The term Root Infinitive stage designates the period of development during which children produce non-finite verb forms in contexts where a finite form

248

Appendix B

Table B.5. MLUw and age at each recording of Anne Session

Age

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

1;10.12 1;10.26 1;11.13 1;11.29 2;0.13 2;0.27 2;1.19 2;2.0 2;2.20 2;2.30 2;3.15 2;4.2 2;4.20 2;5.4 2;5.18 2;6.2 2;6.18 2;7.1 2;7.16 2;8.3 2;8.20 2;9.15 2;10.18 2;11.2 3;0.2 3;0.12 3;1.4 3;1.15

Number of utterances

Number of words

Words per utterance

Standard deviation

255 302 211 129 77 170 214 172 201 187 40 241 172 298 371 295 280 209 196 281 391 253 403 381 349 246 344 294

468 579 472 239 173 442 578 426 457 481 96 687 524 807 1,240 1,002 931 612 679 810 1,440 910 1,478 1,508 1,431 837 1,394 1,023

1.835 1.917 2.237 1.853 2.247 2.6 2.701 2.477 2.274 2.572 2.4 2.851 3.047 2.708 3.342 3.397 3.325 2.928 3.464 2.883 3.683 3.597 3.667 3.958 4.1 3.402 4.052 3.48

1.18 1.031 1.277 0.997 1.095 1.657 1.632 1.515 1.285 1.582 1.356 1.658 1.804 1.543 2.065 2.141 1.876 1.971 2.363 1.763 2.078 2.44 2.875 2.64 2.691 2.546 2.666 2.223

would be required in the adult language (Rizzi 1994). Root Infinitives (RIs) disappear from children’s production in a gradual fashion. In the early stages, RIs represent a high proportion of verbs. After the core of the RI stage, examples of RIs continue to appear in the data for a period of time that varies from child to child but which can be relatively long (i.e. up to a few months).

249

Child data Table B.6. MLUw and age at each recording of Tom Session

Age

Number of utterances

Number of words

Words per utterance

Standard deviation

1 2 3 5 6 8 9 10 12 13 14 15 16

2;1.11 2;1.13 2;1.14 2;3.22 2;4.8 2;4.9 2;5.24 2;5.24 2;6.12 2;6.22 2;8.0 2;9.19 2;10.7

295 353 315 196 237 228 205 353 624 325 349 166 205

671 862 746 479 535 587 496 960 1,530 890 1,102 397 719

2.275 2.442 2.368 2.444 2.257 2.575 2.42 2.72 2.452 2.738 3.158 2.392 3.507

1.369 1.397 1.466 1.651 1.41 1.608 1.702 1.913 2.033 1.85 1.987 2.488 3.346

Collapsing into one group all the data from the period when RIs are attested would not give an accurate picture of the child’s evolving grammar. In particular, it would mask the fact that RIs initially represent a high proportion of verbs. For this reason, I have divided the RI data into two periods, one representing the core RI stage and another the part following it. In this second period, the proportion of RIs is very low. The cut-off point I have chosen is five per cent of RIs in contexts requiring a finite verb. RIs continue to appear at about the same level of frequency until the end of the period considered here, and even a little longer in the case of some children. The core RI stage ends at the following ages for each child: 2;8.3 for Anne (= file t), 2;4.4 for Max (= file n), and 2;4.9 for Tom (= file h). 3 Léa had already passed the RI stage by the time recordings with her began: of 259 root verbs, she only produces eleven potential cases of RIs. The graphs in Figures B.1–B.3 represent the distribution of finite vs. nonfinite root verbs in the children’s data.

3

The letters designating each file are identifiable on the x-axis of the graphs in Figures B1–B6 .

250

Appendix B

120

Raw numbers

100 80 Finite root verbs

60

Non-finite root verbs

40 20 Time [recording sessions]

0 a b c

d e

f

g h

i

j

k

l m

n

o p q

r

s

t

u v w x

z A

y

Raw numbers

Figure B.1. Distribution of root verbs in Max’s data

280 260 240 220 200 180 160 140 120 100

Finite root verbs Non-finite root verbs

80 60 40 20 0 a b c

d e

f

g h

i

j

k

l m

n

o p q

r

s

t

u v w x

y

z A

Time [recording sessions]

Figure B.2. Distribution of root verbs in Anne’s data

B.4.2 Two periods based on the null subject phenomenon

The null subject phenomenon is particularly relevant to the present study, given that most dislocated elements are coreferential with the subject, and that the presence of a resumptive is one of the essential diagnostics for dislocated structures. Like RIs, null subjects do not disappear overnight. The rate of omitted subjects on the whole reduces gradually, but there seems to be at least one turning point in this process. Towards the end of the overall period of development

Child data

251

180 160

Raw numbers

140 120 100

Finite root verbs

80

Non-finite root verbs

60 40 20 0 a=

=b=

=c

e

f=

=h

i=

=j

l

m

n

o

p

Time [recording sessions]

Figure B.3. Distribution of root verbs in Tom’s data

studied in this book, the proportion of root verbs requiring a subject becomes very small: less than fifteen per cent of subjects are missing from obligatory contexts. As was the case for RIs, this small proportion persists for a while, as if the child has reached a sort of plateau in acquisition. It is at this point that the child also suddenly starts to produce many more utterances containing a verb than previously.

120

Raw numbers

100 Finite verbs needing overt subject

80 60

Overt subjects

40 20 0 a b c d e f g h i

j k l m n o p q r s t u v w x y z A

Time [recording sessions]

Figure B.4. Distribution of root verbs and their subjects in Max’s data

252

Appendix B

240 220 200

Raw numbers

180 160

Finite verbs needing overt subject

140 120

Overt subject

100 80 60 40 20 0 a b c

d e

f

g h

i

j

k

l m

n

o p q

r

s

t

u v w x

y

z A

Time [recording sessions]

Figure B.5. Distribution of root verbs and their subjects in Anne’s data

These changes are visible in the graphs in Figures B.4–B.6. Period 2 starts at the following ages for each child: 2;6.27 for Anne (= file u), 2;5.29 for Max (= file q), and 2;5.24 for Tom (= file i). Again, Léa had already passed the null subject stage by the time recording with her began: she only omits the subject in six cases in total during the short period considered here. 160 140

Raw numbers

120 100 80

Finite verbs needing overt subject

60

Overt subjects

40 20 0

a=

=b=

=c

e

f=

=h

i=

=j

l

m

n

o

p

Time [recording sessions]

Figure B.6. Distribution of root verbs and their subjects in Tom’s data

253

Child data

B.5 Additional tables relating to the child data

Table B.7. Direction of dislocation in the children’s declarative sentences Subject dislocation

Max (1) Max (2) Anne (1) Anne (2) Tom (1) Tom (2) Léa

Other dislocation

LD

RD

LD

RD

86 (77%) 153 (62%) 160 (52%) 158 (53%) 13 (17%) 33 (50%) 61 (44%)

10 (9%) 34 (14%) 82 (26%) 56 (19%) 39 (52%) 15 (23%) 23 (16%)

14 (13%) 51 (21%) 45 (15%) 73 (24%) 14 (19%) 9 (14%) 42 (30%)

2 (2%) 9 (4%) 23 (7%) 12 (4%) 9 (12%) 9 (14%) 14 (10%)

Total

112 247 310 299 75 66 140

Table B.8. Direction of dislocation in the children’s yes/no questions Subject dislocation

Max (1) Max (2) Anne (1) Anne (2) Tom (1) Tom (2) Léa

Other dislocation

LD

RD

LD

RD

9 (56%) 4 (40%) 10 (45%) 3 (21%) 0 (0%) 1 (17%) 2 (22%)

7 (44%) 1 (10%) 9 (41%) 4 (29%) 5 (71%) 2 (33%) 2 (22%)

0 (0%) 3 (30%) 2 (9%) 2 (14%) 1 (14%) 0 (0%) 2 (22%)

0 (0%) 2 (20%) 1 (5%) 5 (36%) 1 (14%) 3 (50%) 3 (33%)

Total

16 10 22 14 7 6 9

254

Appendix B Table B.9. Direction of dislocation in the children’s wh-questions Subject dislocation

Max (1) Max (2) Anne (1) Anne (2) Tom (1) Tom (2) Léa

Other dislocation

LD

RD

LD

RD

24 (39%) 8 (21%) 30 (32%) 14 (20%) 0 (0%) 2 (3%) 0 (0%)

32 (52%) 27 (71%) 61 (65%) 36 (51%) 41 (93%) 53 (87%) 7 (100%)

4 (7%) 2 (5%) 2 (2%) 11 (16%) 0 (0%) 0 (0%) 0 (0%)

1 (2%) 1 (3%) 1 (1%) 9 (13%) 3 (7%) 6 (10%) 0 (0%)

Total

61 38 94 70 44 61 7

Table B.10. Nature of the dislocated elements in the child data

Max (1) Max (2) Anne (1) Anne (2) Tom (1) Tom (2) Léa

1/2 person

3 person

DPs

Adverbs, PPs

Predicates

46% (101/218) 36% (119/327) 22% (100/454) 22% (88/400) 24% (31/129) 18% (24/133) 24% (41/170)

30% (66/218) 33% (108/327) 12% (56/454) 21% (84/400) 10% (13/129) 23% (31/133) 8% (14/170)

18% (39/218) 18% (59/327) 52% (238/454) 34% (134/400) 53% (69/129) 42% (56/133) 34% (58/170)

4% (8/218) 11% (37/327) 12% (56/454) 20% (81/400) 11% (14/129) 12% (16/133) 24% (41/170)

2% (4/218) 1% (4/327) 1% (4/454) 3% (13/400) 2% (2/129) 5% (6/133) 9% (16/170)

Table B.11. Nature of the dislocated elements in the adult data 1/2 person

3 person

Canada 32% (89/281) 16% (46/281) France 12% (39/324) 10% (34/324) Belgium 12% (40/327) 17% (56/327)

DPs

Adverbs, PPs

Predicates

30% (84/281) 18% (50/281) 4% (12/281) 52% (170/324) 20% (64/324) 5% (17/324) 35% (113/327) 29% (94/327) 7% (24/327)

Child data

255

B.6 Emergence of dislocated structures This section outlines the emergence of dislocated elements in the child data, including details regarding the nature of the dislocated elements and the functions with which they are associated inside the clause. B.6.1 Very first dislocations

The earliest data from the corpora under investigation come from the York corpus (Anne and Max). It is very difficult to tell whether the children’s first attested dislocations are truly productive or not as in most cases they might be unanalysed (near-)repetitions of adult utterances. During the first three recording sessions, Max does not say much, and most of his utterances comprise a single word. No dislocated element is attested before the fourth recording session, during which only a small proportion of his utterances involve more than one word. He has clearly only reached the multi-word stage (Nice 1925; Brown 1973) 4 at the age of approximately 2;1 (as shown in Table B.4). Max’s first dislocated utterances are given in (B.1). The dislocated elements in these utterances all express the subject, even in a verbless sentence like (B.1c). Of these three examples, only (B.1c) is unambiguously not an attempted repetition of what an adult said in the preceding discourse. (B.1)

a. Moi, 0 vu. me seen ‘I’ve seen (one).’ (Talking about dalmatians) b. 0 s’ appelle, lui? refl calls him ‘What’s that one called?’ c. 0 minou (à) Yolande, ça. cat to Yolande that ‘(That’s) Yolande(’s) cat.’

(Max 1;11.0) (Max 1;11.0) (Max 1;11.0)

Anne seems to be more advanced than Max at the time of her first recording session: most of her utterances are already multi-word (see Table B.5). Dislocated elements are attested in Anne’s production from the start of data collection, but most of them are unclear or repetitions of adult utterances. Typical examples are given in (B.2). Most of Anne’s dislocated elements 4 Nice (1925) is the first researcher to have relied on the Average Length of Sentences to evaluate a child’s language development.

256

Appendix B

express the subject but we also find some instances of stage topics, 5 all in the form of a dislocated là ‘there’ referring to the situation on which Anne is commenting, as in (B.2b). In common with Max, some of Anne’s early dislocations occur in verbless utterances (B.2c). (B.2)

a. Mimi, e(lle) va toutoutou toutoutoutou. (Anne 1;10.12) mimi she goes tootootoo tootootoo ‘Mimi goes tootoot.’ (Imitating a train) b. Là, (elle) (est) p(r)ête. (Anne 1;10.12) there she is ready ‘There, she’s ready.’ (Commenting on a picture of a mouse all dressed up) c. Dans les b(r)iques, le monhomme in the bricks the man [: bonhomme]. (Anne 1;10.26) ‘The man(’s) in the bricks.’ (Answering a question as to where the man is)

The first recording of Tom (from the Cat corpus) took place two weeks after his parents reported that he had started talking. At that point, Tom was 2;1 and had already clearly passed the one-word stage, as shown in Table B.6. Tom’s first dislocations are clearly productive. They are for the most part not repetitions of adult utterances. Typical examples are given in (B.3). They involve subject and object dislocations. (B.3)

a. 0 est pas une fille, Isabelle. is not a girl Isabelle ‘Isabelle’s not a girl.’ b. L’ herbe, # (v)a sécher, moi. the grass will dry me ‘I’ll dry the grass.’ c. (Je) peux le casser, (le) lapin? I can it break the rabbit ‘Can I break the rabbit?’

(Tom 2;1.11) (Tom 2;1.11) (Tom 2;1.13)

These observations suggest that the dislocation data from Anne and Max’s very early recordings (Max 1–4, Anne 1–3) must be treated with caution, as they are likely not to reflect the children’s true competence at that stage. The

5

Stage topics are defined in Section 3.1.5.

Child data

257

dislocated utterances they produce then may consist entirely of (incomplete) repetitions in which dislocations are not analysed as such by the child. B.6.2 Nature of the dislocated elements

Examining the nature of dislocated elements provides another example of what could be labelled ‘casual variation’. Max’s early dislocations almost exclusively involve strong pronouns. He also uses dislocated moi ‘me’ more than other children. This is likely to be due to Max’s more prevalent recourse to deixis to identify the referents he is talking about. The type of activity in which Max and the interviewer take part may also have an effect: they mostly do jigsaws or play with a set of characters, whereas the other children often play with plasticine or draw during recording sessions and consequently do not need to identify particular individuals or particular locations as much as Max does. The relatively high proportion of dislocated elements referring to the speaker and the hearer in Max’s data are also possibly an effect of the activity type. Max and Catherine (the interviewer) comment on their respective roles, suggesting that each other do this or that and announcing what they will do. Subject clitics are not salient enough to express contrasts of this sort and as a result the strong pronouns moi ‘me’ and toi ‘you’ are used. This interpretation of the data predicts that a similarly high proportion of dislocated strong pronouns will be found in the speech of adults who interact with Max, and indeed this is confirmed by the data. The distribution of dislocated elements according to their nature is represented in Figure B.7 for the children and in Figure B.8 for the adults. These graphs correspond to Tables B.10 and B.11 respectively.

Proportions of all dislocated elements

100% 90% 80% 70% Clauses and predicates Adverbs and PPs Nominals 3 person pronouns 1 or 2 person pronouns

60% 50% 40% 30% 20% 10% 0% Max (1) Max (2) Anne (1) Anne (2) Tom (1) Tom (2)

Le´ a

Figure B.7. Nature of the dislocated elements in the child data

258

Appendix B

Proportions of all dislocated elements

100% 90% 80% 70% Clauses and predicates Adverbs and PPs Nominals 3 person pronoun 1 or 2 person pronoun

60% 50% 40% 30% 20% 10% 0% Canada

France

Belgium

Figure B.8. Nature of the dislocated elements in the adult data

Adverbs, PPs, and clauses represent a small proportion of peripheral elements, especially in the early data. This proportion rises noticeably from period 1 to period 2. 6 Léa, the oldest and most advanced child, produces these elements the most, proportionally speaking. However, notice that on the whole the proportion of these elements is greater for Belgian adults than it is for Canadian or French adults. As an illustration of this type of dislocated element, the first dislocated adverbs attested in the children’s data are given in (B.4) below. (B.4)

6

a. (Il) (y) a p(l)us (de) bébé, maintenant. it there has no-more of baby now ‘There’s no baby left, now.’ b. Après, e l’ aut(r)e, alors. after e the other-one then ‘After that (I’ll have?) the other one then.’ c. Ap(r)ès, # c’ est fini, après. after it is finished after ‘After that it’s finished.’ d. Tantôt, # on a vu # un accident. earlier one has seen an accident ‘Earlier we saw an accident.’ e. Et maintenant, c’ est moi. it is me and now ‘And now it’s my turn.’

The two relevant periods are defined in Appendix B.4.1.

(Anne 1;11.13) (Anne 2;0.27) (Max 2;1.15) (Max 2;5.1) (Tom 2;3.22)

Child data f.

C’ est quoi, ici? it is what here ‘What’s this here?’ g. Mais # tu l’ as prise, déjà. but you it have taken already ‘But you’ve already taken it.’ h. Après, je reviens. after I come-back ‘I’ll come back afterwards.’

259

(Tom 2;3.22) (Léa 2;8.22) (Léa 2;8.22)

The first dislocated PPs are given in (B.5). 7 (B.5)

a. C’ est ça, le [///] # sur le chemin. (Anne 2;6.2) it is that the on the path ‘That’s “on the path”.’ (Pointing, to show she’s understood) b. Comme Papa, e donne les range. (Anne 2;6.2) like Dad e gives them put-away ‘Like Dad, (I) give (and you) put away.’ c. Ici # en haut # c’ est quoi? (Max 2;2.9) here in top it is what ‘What’s that at the top?’ d. C’ est quoi, dedans ta bouche à toi? (Max 2;3.6) it is what within your mouth to you ‘What’s that in your mouth?’ (Tom 2;1.13) e. Avec du piano, ça sonne. with some piano it rings ‘With the piano, it rings.’ f. Et dans la rue, j’ ai revenu. (Tom 2;10.7) and in the street I have come-back ‘And in the street I came back.’ (Meaning he rode back on the street) g. Pour l’ instant, je vais donner à boire aux for the moment I will give to drink to-the

7 Two of Léa’s utterances are target-deviant. From (B.5g), it is clear that she does not fully understand the meaning of pour l’instant ‘for the moment’ yet because she uses it with a verb expressing future time as if it means ‘presently’. In (B.5h) Léa’s use of the infinitival is surprising. This might be due to a performance error. The target might be either pas beaucoup à boire ‘not much to drink’ or pas beaucoup bu ‘not drunk much’. Tom’s use of the auxiliary avoir ‘have’ in (B.5f) is also target-deviant. The expected form is suis ‘am’.

260

Appendix B

Proportions of all dislocated elements

100% 90% 80% 70% Subject

60%

Other selected XP la`

50%

Unselected XP

40% 30% 20% 10% 0%

Max (1) Max (2) Anne (1) Anne (2) Tom (1)

Tom (2)

Le´a

Figure B.9. Functions associated with dislocated elements in the child data

plantes. plants ‘At the moment I’m going to water the plants.’ h. Chez Mamy, elle a pas beaucoup boire. at Mamy she has not much to-drink ‘At Mamy’s, she doesn’t have much to drink.’

(Léa 2;8.22) (Léa 2;8.22)

Regarding the emergence of dislocated clauses, see Section 5.4. B.6.3 Functions associated with the dislocated elements

Given the distribution of dislocated elements in the target language, it is to be expected that the majority of dislocated topics will be coreferential with the subject in the child language too, on the assumption that child language differs only minimally from the adult language. This is confirmed by the early data. The graph in Figure 5.5, repeated here as Figure B.9, represents the dislocated elements in the child data according to the function with which they are associated inside the clause. As mentioned earlier, dislocation of the subject prevails in the early recordings. In one child’s data (Max) dislocated elements start out expressing only the subject (before he is 2;0.28). However, Max does not appear to use dislocations productively at this point (see Appendix B.6). Furthermore, dislocated elements are so rare initially in the speech of very young children 8 that the exclusive presence of subjects in some of the early recordings might simply be the product of a sampling effect: dislocation of the subject is by far the most 8

This is shown in Appendix B.6.4.

Child data

261

common dislocation, even in the speech of adults. If only a few dislocations are attested, they are extremely likely to be coreferential with the subject as a result. The dislocation of XPs coreferential with a non-subject constituent selected by the verb is much rarer, both in the child and adult data. In the period studied here, all such cases involve a dislocated element coreferential with the object of the verb (direct or indirect). Initially, the youngest children’s dislocated objects are not resumed by a clitic. In some cases this is due to the omission of the clause that would contain the resumptive, as in (B.6). (B.6)

a. (C’) est Maman, ça. it is Mum that ‘It’s Mum (who does) that.’ b. Ça, c’ est moi. that it is me ‘It’s me (who’s done) that.’

(Anne 1;11.13) (Max 2;0.28)

In other cases, the absence of a resumptive is target-like when the dislocated element appears on the left and there is a contrast between the dislocated element and other members of a discourse-defined set, as in (B.7a) and (B.7b), or when the interpretation is generic, as in (B.7c). Non-target-like omissions of the resumptive element are difficult to identify because in most cases the absence of the resumptive is target-like, as in (B.7a), (B.7b), and (B.7c). The sentence in (B.7d) is one of the rare instances where the omission of the resumptive element is clearly not target-like. (B.7)

a. Ça, on va mettre [//] euh faire ça comme that one will put euh do that like Maman. Mum ‘That one, we’ll do that, like Mum.’ b. Et ça aussi, on peut mettre. and that too one can put ‘That too, we can put (there).’ c. Il aime bien, les haricots? he likes well the beans ‘Does he like beans?’ d. 0 met là, (les) petites f(l)eurs. put+T there the little flowers ‘I/one/you put (them) there, the little flowers.’

(Max 2;3.20) (Tom 2;8.0) (Anne 2;10.18) (Anne 2;6.2)

262

Appendix B

The first object resumptive appears in Max’s data at 2;2.9, in Anne’s at 2;10.18, and in Tom’s, at 2;1.13, (at the start of the recording sessions). 9 As expected, object resumptives are attested from the earliest recordings of Léa’s production (at 2;8.22). (B.8)

a. Moi, # (je) veux l’ essayer, lui. me I want him to-try him ‘I want to try that one out.’ b. Celui-là, on le voit. that-one-there one it sees ‘That one, we (can) see it.’ c. (Je) peux le casser, (le) lapin? I can it break the rabbit ‘Can I break the rabbit?’

(Max 2;2.9) (Anne 2;10.18) (Tom 2;1.13)

Dislocated elements expressing the indirect-object argument of a three-place predicate are very rare in the child (and adult) data. An example is given in (B.9). (B.9)

Chez Mamie, je suis allée trois jours. at Grandma I am gone three days ‘I went to Grandma’s for three days.’

(Anne 3;1.4)

In Figure B.9, I have distinguished the dislocation of là ‘there’ from that of other stage topics because some children (especially Anne) rely quite heavily on là ‘there’ to disambiguate what they are talking about. This use of deixis to narrow the search-space for the referent of the topic is reminiscent of Max’s predominant use of dislocated strong pronouns. Neither Tom nor Léa relies heavily on either of these strategies. I have collapsed into one category all other types of dislocated elements. They include stage topics other than là ‘there’ and adverbs which do not express the topic. Stage topics are typically sentence modifiers like adverbs (B.10a), temporal clauses (B.10b), purpose clauses (B.10c), and quantified expressions generalizing over situations (B.10d). They are absent from the very early data. (B.10) a. Après, c’ est le clown. after it is the clown ‘After that, it’s the clown.’

(Anne 2;2.30)

9 It would take me too far afield to investigate the reason why Anne’s first object resumptive clitic appears so late in her data. She does however use object clitics from the age of 2;2.00. Max’s first object clitic appears at 2;0.28.

Child data

263

b. Quand [/] quand ma maman était malade, elle prend aussi des when when my mum was ill she takes also some médicaments, ma maman. (Tom 2;10.7) medicine my mum ‘When my mum was ill she also took some medicine.’ c. Pi a(rr)ête là [//] pour a(rr)êter, peser su(r) and stop there to stop press on the bouton là. (Max 2;2.9) l(e) button there ‘To stop (we need to) press that button.’ d. Eh ben de tous les soirs, eh ben tu vas dans ton lit eh well of all the evenings eh well you go in your bed grand. (Léa 2;10.7) big ‘Every evening you go into your big bed.’ Also included in this last category are dislocated elements modifying a DP (B.11) and Hanging Topics (B.12), i.e. dislocated elements that are not in direct relation with any element in the sentence; they do not have the function of a stage topic, but can nonetheless be included in the broad category of topic for they refer to the situation as a whole or to the point of view of the speaker or hearer. (B.11) Le monsieur, ben ses mains, elle sont dans les airs. the man well his hands they are in the air ‘The man’s hands are in the air.’

(Max 2;5.1)

(B.12) a. Non toi, si on met le narbre comme ça, c’ était pas no you if one puts the tree like that it is not joli. (Anne 2;11.2) pretty ‘No, if we put the tree like that it’s not pretty.’ b. Moi, tu sais quoi? (Max 2;5.29) me you know what ‘You know what?’ This exploration of the various types of syntactic elements dislocated by children does not support the claim that dislocation of the subject is the only type of dislocation allowed in the early stages of acquisition, or that the dislocation of other types of elements is more ‘costly’ to the child. The dislocation almost exclusively of subjects in the earliest recordings could be due to a sampling

264

Appendix B

100% 90% 80% 70% 60%

sentences with a dislocated element

50% 40% 30% 20%

37%

10%

29% 19%

24%

24%

20%

20%

Tom (2)

Le´a

0% Max (1)

Max (2) Anne (1) Anne (2) Tom (1)

Figure B.10. Proportion of sentences containing a dislocated element in the child data

effect. Stage topics and dislocated objects are attested from very early on, but target-like resumption in the case of the latter construction is delayed until later in the child’s development. Full mastery of the resumption strategy is thus not necessary for the production of dislocated constructions. This is particularly clear from data which involve the dislocation of subjects and objects in the absence of the required resumptive clitics. I take this to support the view that the encoding of XP topics as peripheral elements does not involve demanding computation or highly complex syntactic mechanisms. B.6.4 Evolution of the proportion of dislocated utterances

Overall, the proportion of sentences containing a dislocated element is very similar in children and adults. There is a slight decrease in the proportion of dislocated clauses found in the data of the three children from period 1 to period 2, as illustrated by the graph in Figure B.10. 10 This decrease represents a significant difference in Max ( p<0.001, ˜ 2 = 84.438) and in Anne ( p<0.01, ˜ 2 = 9.516), but not in Tom ( p<0.1, ˜ 2 = 3.513). Does this statistically significant result with respect to the younger children have any linguistic significance? In other words: does the observed decrease in the proportion of dislocated clauses indicate a change in the children’s grammar? The encoding of XP topics is not a syntactic requirement. Nonetheless, differences in the rate at which such topics appear could be due to the child using pragmatic means to compensate for deficiencies in their syntactic system. A decrease in the proportion of dislocated utterances might reflect a 10

The two relevant periods are defined in Appendix B.4.1.

Table B.12. Presence of a dislocated subject and realization of the subject according to the finiteness of the verb Period 1

154 (68%) 214 (62%) 368

+FIN −SUBJ

RI

49 (22%) 24 (11%) 87 (25%) 44 (13%) 136 68

Total Period 2

227 S. dis. 345 No S. dis. Total

+FIN +SUBJ

+FIN −SUBJ

RI

145 (92%) 8 (5%) 5 (3%) 619 (89%) 31 (4%) 45 (6%) 764 39 50

Total

158 695

219 (67%) 91 (28%) 495 (59%) 249 (30%) 714 340

16 (5%) 326 S. dis. 99 (12%) 843 No S. dis. 115 Total

291 (97%) 9 (3%) 0 300 976 (96%) 34 (3%) 10 (1%) 1,020 1267 43 10

62 (63%) 24 (24%) 275 (61%) 100 (22%) 337 124

13 (13%) 78 (17%) 91

112 (98%) 1 (1%) 432 (93%) 24 (5%) 544 25

99 S. dis. 453 No s. dis. Total

1 (1%) 7 (2%) 8

114 463

Child data

Max Subject dislocation No subject dislocation Total Anne Subject dislocation No subject dislocation Total Tom Subject dislocation No subject dislocation Total

+FIN +SUBJ

265

266

Appendix B

turning point in development, indicating that the child no longer relies on the dislocation strategy to compensate for e.g. a syntactic deficiency. For instance, it is possible that children use dislocated subjects more during the null subject stage because they want to ensure that the referent of these subjects is identified. This hypothesis can be tested by investigating whether the proportion of dislocated subjects decreases as subject clitics come to be realized more often, i.e. if there is a correlation between non-realization of the subject and subject dislocation. Table B.12 gives the proportion of tokens which contain a (left- or right-) dislocated subject according to (i) whether the subject is realized or not and (ii) the finiteness of the verb. Only verbs requiring an overt subject are considered here: imperatives and other constructions licensing subject traces (e.g. subject wh-questions, subject relatives, and subject clefts) are excluded. Also excluded are any repetitions of adult utterances and all cases in which it is unclear whether a verb or a subject is present. ‘+Fin’ stands for finite verb (i.e. [+T]) and ‘RI’ stands for Root Infinitive (including any RI with an embryonic modal). The difference in the rate of subject dislocations when the subject is realized compared to when it is omitted is not statistically significant for any of the children in either period of development. I therefore conclude that children do not show a stronger tendency to encode the subject referent by means of a dislocated topic when they omit the subject. Looking at it from another perspective, children do not omit the subject more in the presence of a subject topic. This result is particularly interesting given the current trend in analyses of the null subject phenomenon for attributing a significant proportion of subject omissions to pragmatic factors (e.g. Schaeffer et al. 2002). If no correlation is observed between the encoding of a subject topic and omission of the subject in a language like French, where information structure is particularly transparent, postulating such a correlation for other languages should be done with caution and be backed up by strong arguments, something which Schaeffer et al. (2002) do not do. More generally, the above discussion suggests that variation in the proportion of sentences containing a dislocated element does not have any particular syntactic significance. An explanation as to what motivates such variation could be the greater need that children might initially feel to encode the topic of the sentence overtly. Alternatively, variation could be due to a token-type effect: it may be that the children are uttering only a few types of sentences a great number of times in exactly the same way, and it just so happens that a high proportion of these sentences contain a dislocated element. I will not explore these possibilities further here.

Appendix C Judgement elicitation This section provides further details of the methodology used to elicit native speaker judgements experimentally.

C.1 Methodology Three points must be taken into account when eliciting judgements on dislocation data. First, it is important to bear in mind that dislocated elements encode the topic of the sentence, and that this requires the satisfaction of various felicity conditions. A context has to be provided to ensure that the referent in question is sufficiently salient. In this context, it has to make sense to say something about that referent. To ensure informants were not confused about what they had to judge, the context was provided in written from. Second, prosody is one of the key distinctive features of dislocated constructions. For this reason, recorded test sentences had to be used to ensure that the prosody was exactly the same each time the test was performed. This also delimited (as much as possible) the extent to which informants could assign a contrastive interpretation to the dislocated element, given that it is directly proportional to the prosodic saliency of that element. Informants were invited to click on a button to hear the test sentence after reading the context provided. Third, one has to ensure that speakers provide judgements on spoken French, not written French. Dislocations are used consistently in spoken French to encode topics, as the present study demonstrates. Information structure is encoded differently in written French (see Lahousse 2003). 1 In test situations, native speakers tend to be influenced by the ‘norm’ (i.e. prescriptive grammar) which is strongly promoted in schools and in the media. In order to avoid the influence of written French as much as possible, the data was presented in audio form only. Informants were told that there was 1 Thetic sentences are encoded using postverbal subjects in written French but not in spoken French. In spoken French, the canonical subject position can only host subjects if they are in focus (whether in narrow focus or in a thetic structure). In written French, subjects in this position can receive a topic interpretation.

268

Appendix C

no ‘correct’ answer and that the study aimed to survey the way people really speak. They were invited to choose a judgement from a pull-down menu (the default setting was ‘choose’, so any item skipped could be excluded from the analysis). The options to choose from were as follows. (C.1)

a. I could say that sentence. b. I could say that sentence but in another context. c. I could never say a sentence like that, but I know that other French speakers could. d. That sentence is too weird. No French speaker talks like that.

Options (a) and (b) were considered to indicate acceptability: the informant identified the sentence as something they could say themselves. Option (c) was taken to indicate markedness rather than unacceptability because the informant accepted that this sentence is possible in spoken French (more on this below). Option (d) was taken to indicate unacceptability. Justification for this interpretation of the results is provided by the judgements given for sentences predicted to be (i) acceptable and unmarked, and (ii) strongly marked or ungrammatical. 2 The unmarked sentence used as a baseline in test 1 is given in (C.2). (C.2)

Non, il y a personne qui est venu. no it there has nobody who has come ‘No, nobody came.’

(0%; 0%)

Ninety-seven per cent of the thirty two informants gave an (a) rating to this sentence and three per cent gave it a (b) rating. This sentence does not violate any syntactic or information-structural requirement. A (b) rating should thus not be interpreted as a significant restriction on the acceptability of a sentence. Note that (b) ratings were always significantly fewer than (a) ratings for each of the sentences tested. In test 2, the grammatical and unmarked sentence used as a benchmark was the following. (C.3)

Si tu veux, on ira à Rio voir le carnaval. if you want we will-go to Rio to-see the carnival ‘If you want, we’ll go to Rio to see the carnival.’

(0%; 0%)

2 The use of uncontroversially ungrammatical and marked sentences as benchmarks was refined in test 2. See below.

Judgement elicitation

269

Eighty-nine per cent of the seventy-five informants rated it as (a), and eleven per cent as (b). On this basis, options (a) and (b) were collapsed into the category ‘grammatical’. Indeed, the aim of this judgement elicitation task was to identify what is allowed by the grammar of spoken French, and not what is or is not expected in a given context. The strongly marked sentence used as a baseline in test 1 is given in (C.4). (C.4)

À quii est-ce que tu ne sais pas [ce [qu’elle lit to whom is-it that you neg know not that that-she ti ]]? reads

(41%; 19%)

The ratings for (4) were as follows: (a) twenty-five per cent, (b) sixteen per cent, (c) nineteen per cent, (d) forty-one per cent. As pointed out in Section 4.3.5, this sentence was unfortunately given in a context favouring a d-linked interpretation, which is known to alleviate island effects. This is reflected in the relatively high acceptability ratings. However what is crucial here is that forty-one per cent of informants categorically rejected this sentence: this constitutes a significantly higher proportion than was observed for any of the test sentences involving dislocation (where the highest rejection rate was nineteen per cent). In test 2, the clearly ungrammatical sentence was (C.5), and the marked sentence was (C.6). (C.5) (C.6)

∗

Qu’i est-ce qu’elle avait [un sac [qui contenait ti ]]? (81%; 17%) what-is-it that-she had a bag that contained Mais je ne sais pas [qu’est-ce qu’elle leur a dit]. (7%; 61%) but I neg know not what-is-it that-she to-them has said ‘But I don’t know what she said to them.’

Sentence (C.5) was accepted, i.e. rated (b), by one speaker only; no speaker chose option (a). The vast majority (eighty-one per cent) considered it ungrammatical. Sentence (C.6) instantiates a structure only allowed in certain dialects (i.e. a doubly-filled Comp in an embedded clause). It was considered grammatical by thirty-two per cent of informants (thirty-one per cent gave it an (a) rating (including all the Canadian informants) and one per cent gave it a (b) rating). Seven per cent of informants considered it ungrammatical, while the majority of informants (sixty-one per cent) considered it acceptable for speakers other than themselves, option (c). On this basis, option (c)—but not option (b)—could be interpreted as indicating (an impression of) inter-speaker variation. I have used the term ‘marked’ to refer to this

270

Appendix C

in the text. While I agree with Haspelmath (2006) that the term markedness tends to be overused and should be replaced with alternatives wherever possible, the data from this online task is insufficient to elucidate what it means exactly for a speaker to rate a sentence as ‘I could never say a sentence like that, but I know that other French speakers could’. A full-scale sociolinguistic study would be necessary to identify the relevant factors underlying this variation. Possible factors include age, sex, education, and social class.

C.2 Results C.2.1 Test 1

The seven sentences testing whether dislocated elements can be separated from their resumptive element by a strong island boundary are given below in the (b) examples; their context is given in (a). Acceptability ratings are given in Table C.1. a. J’ai recommencé les Tintin d’avant 1960. ‘I’ve started reading the pre-1960 Tintins again.’ je vais attendre [avant de lesi relire]. b. Les autresi , before to them to-reread the other-ones I will wait ‘I will wait before reading the other ones again.’ (C.8) a. On s’attendait à voir toutes sortes de gens saugrenus à la fête pour sa sortie de prison. ‘We were expecting to see all kinds of weird people at her release party.’ b. Mais le juge j , çai a surpris tout le monde, [qu’elle l’ j but the judge it has surprised all the people that-she him ait invité]i . has invited ‘It surprised everyone that she should have invited the judge.’ (C.9) a. Elle est en train de lire les Contes de Mille et Une Nuits aux plus grands. ‘At the moment, she’s reading 1,001 Nights to the big ones.’ b. Aux petitsi , je sais pas [ce [qu’elle leuri lit]]. to-the little-ones I know not that that-she to-them reads ‘I don’t know what she’s reading to the little ones.’ (C.10) a. Je veux plus les voir. Plus jamais. ‘I never want to see them again. Ever.’

(C.7)

Judgement elicitation

271

b. Ta mèrei , je ferai tout pour être parti [quand ellei your mother I will-do all to be gone when she viendra]. will-come ‘I’ll do all I can to be away by the time your mother comes.’ (C.11) a. On va d’abord interroger les témoins directs. ‘We’ll interrogate the immediate witnesses first.’ parler]. on va attendre [avant de leuri b. Aux autresi , before to to-them to-speak to-the other-ones one will wait ‘We’ll wait before talking to the other ones.’ (C.12) a. Ça me dit rien d’aller au cinéma, ce soir. ‘I don’t feel like going to the cinema tonight.’ déjà vu [tous les extraits [qu’on b. Ce film-lài , j’ai that film-there I-have already seen all the extracts that-one eni montre]]. of-it shows ‘I’ve already seen all the trailers shown for that film.’ (C.13) a. Je sais pas ce que François a reçu. ‘I don’t know what François got.’ b. Mais à Alicei , je connais le livre qu’ils luii ont donné. but to Alice I know the book that-they to-her have given ‘But I know the book they’ve given to Alice.’ In Table C.1, the numbers in the second and third columns indicate the proportion of informants who gave an (a) or a (b) rating for each sentence, indicating acceptability; the numbers in the fourth column correspond to the proportion of (c) ratings, indicating markedness; the numbers in the fifth column correspond to (d) ratings, indicating unacceptability. One informant omitted to provide a judgement for (C.11) so the proportions for that sentence were calculated with respect to thirty-one informants rather than thirty-two. The poor rating of (C.13) compared with (C.9) which is syntactically similar (i.e. both cases involve a dislocated PP that is coreferential with a cliticized indirect object inside a complex DP), suggests that non-syntactic factors can have a significant effect on acceptability. It is probable that informants considered the test sentence to be an unlikely follow-up in the context provided. A modified version (given in (C.14)) was presented to six native speakers (three of whom had taken part in the original task). It received the following ratings: (a) 4, (b) 1, and (c) 1.

272

Appendix C Table C.1. Results of the first Internet judgement elicitation task Example number (C.7) (C.8) (C.9) (C.10) (C.11) (C.12) (C.13)

(a)

(b)

(c)

(d)

91% 75% 66% 56% 45% 34% 16%

3% 6% 9% 13% 10% 13% 9%

6% 19% 25% 28% 32% 47% 56%

0% 0% 0% 3% 13% 6% 19%

(C.14) a. Je sais pas ce qu’ ils ont donné à François. ‘I don’t know what they gave François.’ ont vu le livre qu’ils luii b. Mais à Alicei , j’ai but to Alice I-have seen the book that-they for-her have acheté. bought ‘But I saw the book they’ve bought Alice.’ There are two key differences between (C.13) and (C.14): first, the dislocated PP is contrasted with a PP in the preceding sentence; second, the predicates in the main clauses are semantically closer in (C.14) than in (C.13), which seems to render (C.14b) a more plausible follow-up to (C.14a) than (C.13b) is to (C.13a). The results for sentence (C.15) in test 2 confirms these findings. C.2.2 Test 2

(C.15) a. C’est une femme étrangement secrète. ‘She’s a strangely secretive woman.’ b. Par exemple, elle raconte tout à son mari. Mais à son by example she tells everything to her husband but to her a jamais [plein de trucs [qu’elle ne luii psyi , il y a analyst it there is lots of things that-she neg to-him has never dits]]. said ‘For instance, she tells her husband everything. But there are lots of things she’s never told her analyst.’

Judgement elicitation

273

(C.16) a. Alors, tu as lu ce que je t’ai envoyé ? Qu’est-ce que tu en penses ? ‘So have you read what I sent you? What do you think of it?’ b. Je n’ai rien lu encore de ton article. Mais de ton I neg-have nothing read yet of your paper but of your [j’avais le temps] et [j’eni ai lu la romani , hier, novel yesterday I-had the time and I-of-it have read the moitié]. half ‘I haven’t read any of your paper yet. But yesterday, I had the time and I read half of your novel.’ (C.17) a. Il paraît que ton frère a acheté une nouvelle voiture ? ‘Apparently your brother’s bought a new car?’ pas d’elle j . b. Pierrei , sa voiture j , cet idioti ne s’occupe Pierre his car this idiot neg refl-looks-after not of-her ‘Pierre, the idiot doesn’t look after his car.’ (C.18) a. C’est fou, cette sécheresse. En plus de ça, la météo raconte n’importe quoi. ‘This draught is unbelievable. And on top of that, they just talk rubbish on the weather forecast.’ b. La pluiei , [j’ai espéré toute la semaine] mais [ellei n’est but she neg-is the rain I-have hoped all the week jamais venue]. never come ‘I hoped all week but the rain never came.’ (C.19) a. Je sais que tu veux tout vendre, mais ne sois pas trop pressé. ‘I know you want to sell everything, but don’t be in too much of a rush.’ l’i achèterai]. b. Ta voiturei , [j’économiserai] puis [je te I-will-save-up then I to-you it-will-buy your car ‘I’ll save up and then I’ll buy your car.’ (C.20) a. Vous pouvez dire tout ce que vous voulez à Gertrude sur sa mère. ‘You can tell Gertrude anything you want about her mother.’ en j b. Mais ses frèresi , de leur mère j , on peut pas leuri but her brothers of their mother one can not to-them of-her parler. speak ‘But her brothers, no one can talk to them about their mother.’ (C.21) a. Madame, je comprends votre agacement face au comportement de ma fille à l’école. Mais je peux vous assurer qu’elle fait de son

274

Appendix C

b.

(C.22) a.

b.

(C.23) a.

b.

(C.24) a.

b.

mieux. Son problème, c’est qu’elle est extrêmement distraite. ‘Madam, I understand you are losing patience with my daughter’s behaviour at school. But I can assure you that she is doing her best. The problem is that she is extremely absent-minded.’ [elle fait pas exprès] mais [elle lesi Ses devoirsi , her homework-pl. she makes not purpose but she them oublie toujours]. forgets always ‘She doesn’t do it on purpose but she always forgets her homework.’ Dans la phrase suivante, est-ce que ‘elle’ peut désigner ‘Rosa’? Pour Rosa, choisissez (a). Pour une autre personne, choisissez (b). ‘In the following sentence, can ‘she’ designate ‘Rosa’? For Rosa, choose (a). For another person, choose (b). Sans elle, on ne peut pas emmener les enfants de Rosa without her one neg may not take the children of Rosa au parc. to-the park ‘Without her, one cannot take Rosa’s children to the park.’ Dans la phrase suivante, est-ce que ‘elle’ peut désigner ‘Madeleine’? Pour Madeleine, choisissez (a). Pour une autre personne, choisissez (b). ‘In the following sentence, can ‘she’ designate ‘Madeleine’? For Madeleine, choose (a). For another person, choose (b). Le frère de Madeleine, elle ne parle pas de lui. the brother of Madeleine she neg speaks not of him ‘Madeleine’s brother, she doesn’t speak about him much.’ Dans la phrase suivante, est-ce que ‘il’ peut désigner ‘Gaston’? Pour Gaston, choisissez (a). Pour une autre personne, choisissez (b). ‘In the following sentence, can ‘he’ designate ‘Gaston’? For Gaston, choose (a). For another person, choose (b). Les costumes de Gaston, sans lui, on peut pas les the suits of Gaston without him one may not them repasser. iron ‘Gaston’s suits, you can’t iron them without him.’

In Table C.2, the numbers in the second and third columns indicate the proportion of informants who gave an (a) or a (b) rating, indicating acceptability;

275

Judgement elicitation Table C.2. Results of the second Internet judgement elicitation task Example number

(a)

(b)

(c)

(d)

(C.15) (C.16) (C.17) (C.18) (C.19) (C.20) (C.21) (C.22) (C.23) (C.24)

63% 72% 12% 65% 60% 43% 61% 17% 25% 74%

15% 7% 12% 9% 21% 8% 3% 83% 75% 26%

23% 22% 50% 23% 19% 43% 35%

0% 0% 26% 3% 0% 7% 1%

the numbers in the fourth column correspond to the proportion of (c) ratings, indicating markedness; the numbers in the fifth column correspond to (d) ratings, indicating unacceptability. In the bottom three rows, an (a) rating indicates coreference and (b) indicates disjoint reference. The results of test 2 confirm the conclusions of test 1: (i) spoken French allows the resumptive element of dislocated PPs or DPs to be situated inside a strong island, and (ii) information structure has a clear impact on the acceptability of sentences involving dislocated elements.

C.3 A reply to Delais-Roussarie et al. (2004) Delais-Roussarie et al. (2004) claim that CLLD (to the extent that it can be distinguished from HTLD) is sensitive to islands in spoken French. They attribute the alleged difference between (C.25a) and (C.25b) to a difference in structure: the former is claimed to instantiate CLLD (which is traditionally assumed to be sensitive to strong islands cross-linguistically), and the latter HTLD (which is not). ∗

A Mariei , je connais le flic qui luii a retiré son permis. to Marie I know the cop who her has taken her licence b. Mariei , je connais le flic qui luii a retiré son permis. Marie I know the cop who her has taken her licence ‘I know the cop who took Marie’s licence away.’

(C.25) a.

276

Appendix C

Delais-Roussarie et al. (2004) claim that the contrast in (c.26) proves that there is a correlation between island sensitivity and the presence of a preposition in the dislocated element, which is taken to be the one infallible diagnostic for CLLD. ∗

Je crois qu’à Mariei , je connais le flic qui luii a retiré I believe that-to Marie I know the cop who her has taken son permis. her licence b. Je crois qu’à Mariei , le flic luii a retiré son permis. I believe that-to Marie the cop her has taken her licence ‘I think the cop took Marie’s licence away’.

(C.26) a.

What Delais-Roussarie et al. (2004) omit to say is that stripping the preposition from the dislocated element in (C.26a) (which they predict would turn it into HTLD) does not improve acceptability: 3 (C.27)

∗

Je crois que Mariei , je connais le flic qui luii a retiré son I believe that Marie I know the cop who her has taken her permis. licence

I have shown experimentally that left-dislocated PPs resumed by a clitic inside a complex DP (exactly as in (C.26a)) are accepted by native speakers if presented in a context where they are plausible topics. The relative unacceptability (or markedness) of (C.26a) is most probably due to discourse factors, as has been shown with respect to (C.13). Further indication of the impact of non-syntactic factors on the acceptability of dislocated structures is provided by Delais-Roussarie et al. themselves when they claim that left-dislocated PPs resumed by a clitic argument are more acceptable in embedded than in root contexts. They attribute the original observation to Larsson (1979: 76–8), whose informants found such dislocated PPs ‘not very natural’ in root contexts (C.28a) but ‘normal’ in embedded contexts (C.28b). (C.28) a. ?À ce confort, on s’y habitue très to that comfort one refl-to-it accustoms very

3 The use of an asterisk in (C.25a), (C.26a), and (C.27) is too strong in my view: some native speakers report an impression of markedness rather than unacceptability. I have nonetheless used an asterisk in (C.27) to show that this sentence is as marked as Delais-Roussarie et al.’s example (C.26a).

Judgement elicitation

277

vite. (Larsson 1979: 77) fast ‘One quickly gets used to comfort.’ b. Je crois qu’à ce confort, on s’y habitue très I think that-to that comfort one refl-to-it accustoms very (Larsson 1979: 77) vite. fast ‘I think one quickly gets used to comfort.’ Aside from the fact that I have not been able to replicate this contrast (none of my informants ever reported a difference between matrix and embedded leftdislocated PPs), there is no syntactic reason why embedding should improve a structure involving CLLD: on the contrary, non-root contexts tend to be hostile to the interpretive reflexes of information structure (Hooper and Thompson 1973; Heycock 2006). The alleged contrast between matrix and embedded clauses is therefore most likely to be another manifestation of the impact that non-syntactic factors have on the acceptability of dislocated structures.

References Alexopoulou, T., Doron, E., and Heycock, C. (2004). ‘Broad Subjects and Clitic Left Dislocation’. In Adger, D., De Cat, C., and Tsoulas, G., editors, Peripheries: Syntactic edges and their effects. Kluwer: Dordrecht. 329–58. Allen, G. D. (1983). ‘Some suprasegmental contours in French two-year-old children’s speech’. Phonetica 40: 269–92. Anagnostopoulou, E. (1997). ‘Clitic Left Dislocation and Contrastive Left Dislocation’. In Anagnostopoulou, E., van Riemsdijk, H., and Zwarts, F., editors, Materials on Left Dislocation. John Benjamins: Amsterdam. 151–192. Aoun, J. and Benmamoun, E. (1998). ‘Minimality, reconstruction, and PF movement’. Linguistic Inquiry 29: 569–97. Ashby, W. (1988). ‘The syntax, pragmatics, and sociolinguistics of Left- and RightDislocations in French’. Lingua 75: 203–29. Ashby, W. J. (1994). ‘An acoustic profile of Right-Dislocations in French’. Journal of French Language Studies 4: 127–45. Auger, J. (1993). ‘More evidence for verbal agreement-marking in colloquial French’. In Ashby, W., Mithun, M., Perissinotto, G., and Raposo, E., editors, Linguistic perspectives on Romance languages. John Benjamins: Amsterdam. 177–98. Auger, J. (1994). Pronominal Clitics in Québec Colloquial French: A Morphological Analysis. Doctoral dissertation, University of Pennsylvania. Austin, J. (1962). How to Do Things with Words. Oxford University Press: Oxford. Avrutin, S. (1999). Development of the Syntax-Discourse Interface. Kluwer: Dordrecht. Baker, M. (1996). The Polysynthesis Parameter. Oxford University Press: New York. Baker, N. and Greenfield, P. (1988). ‘The development of new and old information in young children’s early language’. Language Sciences 10: 3–34. Barbosa, P. (2000). ‘Clitics: a window into the null subject property’. In Costa, J., editor, Portuguese Syntax: New Comparative Studies. Oxford University Press: Oxford. 31–93. Barnes, B. (1985). The pragmatics of Left detachment in spoken standard French, Pragmatics & Beyond, vol. VI.3. John Benjamins: Amsterdam. Barton, E. (1990). Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. John Benjamins: Amsterdam. Bauche, H. (1926). Le langage populaire. Payot: Paris. Beerman, D., Leblanc, D., and van Riemsdijk, H., editors (1997). Rightward movement. John Benjamins: Amsterdam. Belletti, A. (1999). ‘Italian/Romance clitics: Structure and derivation’. In van Riemsdijk, H., editor, Clitics in the Languages of Europe. Mouton de Gruyter: Berlin. 543–80.

References

279

Belletti, A. (2001). ‘Aspects of the low IP area’. In Rizzi, L., editor, The Structure of CP and IP: The Cartography of Syntactic Structure, vol. 2. Oxford University Press: Oxford. 16–51. Benincà, P. (2001). ‘The position of topic and focus in the left periphery’. In Cinque, G. and Salvi, G., editors, Current Studies in Italian Syntax: Essays offered to Lorenzo Renzi. Elsevier: Amsterdam. 39–64. Bever, T., Fodor, J., and Weksel, W. (1971). ‘Theoretical notes on the acquisition of syntax: a critique of “contextual generalization” ’. In Bar-Adon, A. and Leopold, W., editors, Child Language: A Book of Readings. Prentice-Hall: Englewood Cliffs, NJ. Blanche-Benveniste, C. (1997). Approches de la langue parlée en français: L’essentiel français. Ophrys: Paris. Blasco, M. (1993). Les énoncés à ‘redoublement’ et ‘dislocation’ en français contemporain: Analyse en ‘double marquage’. Thèse de doctorat, Université de Provence. Bloom, L. (1970). Language Development: Form and Function in Emerging Grammars. MIT Press: Cambridge, MA. Boeckx, C. A. (2003). Islands and Chains: Resumption as Stranding. John Benjamins: Amsterdam. Bohnacker, U. (1998). Icelandic plus English: Language differentiation and functional categories in a successively bilingual child. Doctoral dissertation, University of Durham. Bottari, P., Cipriani, P., and Chilosi, A. (1992). ‘Proto-syntactic devices in the acquisition of Italian free morphology’. Geneva Generative Papers. 1–2: 83–101. Bresnan, J. and Mchombo, S. (1987). ‘Topic, pronoun, and agreement in Chichewa’ ˆ . Language 63: 741–82. Brown, R. (1973). A first language: The early stages. Harvard University Press: Cambridge, MA. Büring, D. (1999). ‘Topic’. In Bosch, P. and van der Sandt, R., editors, Focus. Cambridge University Press: Cambridge. Cadiot, P. (1992). ‘Matching syntax and pragmatics: A typology of topic and topicrelated constructions in spoken French’. Linguistics 30: 57–88. Campion, E. (1984). Left Dislocation in Montreal French. Doctoral dissertation, University of Pennsylvania. Cann, R., Kempson, R., Marten, L., Otsuka, M., and Swinburne, D. (2004). ‘On the Left and On the Right’. In Adger, D., De Cat, C., and Tsoulas, G., editors, Peripheries: Syntactic edges and their effects. Kluwer: Dordrecht. 19–47. Carlson, G. (1977). Reference to Kinds in English. Doctoral dissertation, University of Massachussets. Carroll, S. (1982). ‘Les dislocations ne sont pas si populaires que ça’. In Lefebvre, C., editor, La syntaxe comparée du français standard et populaire: approches formelle et fonctionnelle. Gouvernement du Québec, Office de la langue française: Québec. 211–46.

280

References

Cecchetto, C. (1999). ‘A comparative analysis of Left and Right Dislocation in Romance’. Studia Linguistica 53: 40–67. Cecchetto, C. (2000). ‘Doubling structures and reconstruction’. Probus 12: 93–126. Chafe, W. (1976). ‘Givenness, contrastiveness, definiteness, subjects, topics, and point of view’. In Li, C., editor, Subject and Topic. Academic Press: New York. 27–55. Cheung, L. Y. (1997). A Study of Right Dislocation in Cantonese. MPhil dissertation, Chinese University of Hong Kong. Chierchia, G. (1995). ‘Individual Level Predicates as inherent generics’. In Carlson, G. N. and Pelletier, F. J., editors, The Generic Book. University of Chicago Press: Chicago. 175–223. Chomsky, N. (1965). Aspects of the Theory of Syntax. MIT Press: Cambridge, MA. Chomsky, N. (1971). ‘Deep Structure, Surface Structure, and semantic interpretation’. In Steinberg, D. and Jacobovits, L., editors, Semantics. Cambridge Univesity Press: Cambridge. 183–216. Chomsky, N. (1977). ‘On wh-movement’. In Culicover, P., Wasow, T., and Akmajian, A., editors, Formal Syntax. Academic Press: New York. 71–132. Chomsky, N. (1986). Barriers. MIT Press: Cambridge, MA. Chomsky, N. (1993). ‘A minimalist program for linguistic theory’. In Hale, K. and Keyser, S. J., editors, The View from Building 20: Essays in Linguistics in honor of Sylvain Bromberger. MIT Press: Cambridge, MA. 1–52. Chomsky, N. (1995). The Minimalist Program. MIT Press: Cambridge, MA. Cinque, G. (1977). ‘The movement nature of Left Dislocation’. Linguistic Inquiry 8: 397–412. Cinque, G. (1983). ‘ “Topic” constructions in some European languages and “connectedness” ’. In Ehlich, K. and van Riemsdijk, H., editors, Connectedness in Sentence Text and Discourse. KBU: Tilburg. 7–42. Cinque, G. (1990). Types of A-bar dependencies. MIT Press: Cambridge, MA. Cinque, G. (1993). ‘A null theory of phrase and compound stress’. Linguistic Inquiry 24: 239–97. Contreras, H. (2004). Topic as an interpretive notion. Paper presented at the Lisbon workshop on alternative views on the functional domain. Côté, M.-H. (1999). Issues in the analysis and acquisition of clitics in (Spoken) French. Manuscript, MIT. Côté, M.-H. (2001). ‘On the status of the subject clitics in child French’. In Almgren, M., Barreña, A., Ezeizabarrena, M.-J., Idiazabal, I., and MacWhinney, B., editors, Proceedings of the 8th Conference of the IASCL. Cascadilla Press: Somerville, MA. 1314–30. Crisma, P. (1992). ‘On the acquisition of wh-questions in French’. Geneva Generative Papers 1–2: 115–22. Crystal, D. (1973). ‘Non-segmental phonology in language acquisition: a review of the issues’. Lingua 32: 1–45. Crystal, D. (1986). ‘Prosodic development’. In Fletcher, P. and Garman, M., editors, Studies in First Language Development. Cambridge University Press: New York. 174–97.

References

281

Culicover, P. and Jackendoff, R. (2005). Simpler Syntax. Oxford University Press: Oxford. Cummins, S. and Roberge, Y. (1994). ‘A morphosyntactic analysis of Romance clitic constructions’. In Mazzola, M. L., editor, Issues and theory in Romance linguistics. Georgetown University Press: Washington, DC. 239–57. Cummins, S. and Roberge, Y. (2004). ‘Null objects in French and English’. In Auger, J., Clements, J. C., and Vance, B., editors, Contemporary Approaches to Romance Linguistics. John Benjamins: Amsterdam. 121–38. Cummins, S. and Roberge, Y. (2005). ‘A modular account of null objects in French’. Syntax 8: 44–64. De Cat, C. (2000). ‘Towards a unified analysis of French floating quantifiers’. Journal of French Language Studies 10: 1–25. De Cat, C. (2002). French Dislocation. Doctoral dissertation, University of York. De Cat, C. (2004a). ‘Apparent non-nominative subjects in L1 French’. In Prévost, P. and Paradis, J., editors, The Acquisition of French in Different Contexts: Focus on Functional Categories. John Benjamins: Amsterdam. 60–115. De Cat, C. (2004b). ‘A fresh look at how young children encode new referents’. International Review of Applied Linguistics 42: 111–27. De Cat, C. (2005). ‘French subject clitics are not agreement morphemes’. Lingua 115: 1195–219. De Cat, C. (2006). ‘Do Root Infinitives ever have an overt subject in child French?’ In Unsworth, S., Parodi, T., Sorace, A., and Young-Scholten, M., editors, Paths of development in L1 and L2 acquisition. John Benjamins: Amsterdam. 55–76. De Cat, C. (in press). ‘French dislocation without movement’. Natural Language and Linguistic Theory. De Cat, C. and Plunkett, B. (2002). ‘Qu’ est ce qu’ i(l) dit, celui+là ? Notes méthodologiques sur la transcription d’un corpus francophone’. In Pusch, C. D. and Raible, W., editors, Romanistische Korpuslinguistik: Korpora und gesprochene Sprache/Romance Corpus Linguistics: Corpora and Spoken Language. Narr: Tübingen, CD-ROM edition. De Cat, C. and Tsoulas, G. (2006). Complex fragments. Manuscript, University of Leeds and University of York. de Villiers, J. (1991). ‘Why questions?’ In Maxfield, T. and Plunkett, B., editors, Papers in the Acquisition of WH. UMOP: University of Massachussets, Amherst. 155–73. Delais-Roussarie, E., Doetjes, J., and Sleeman, P. (2004). ‘Dislocation’. In Corblin, F. and de Swart, H., editors, Handbook of French Semantics. CSLI Publications: Stanford, CA. 501–28. Delattre, P. (1966). ‘Les dix intonations de base du français’. French Review 40: 1–14. Demirdache, H. (1991). Resumptive chains in restrictive relatives, appositives and dislocation structures. Doctoral dissertation, MIT. Déprez, V. and Pierce, A. (1993). ‘Negation and functional projections in early grammar’. Linguistic Inquiry 24: 25–67.

282

References

Deshaies, D., Guilbault, C., and Paradis, C. (1993). ‘Prosodie et dislocation à gauche par anaphore en français québécois spontané’. In Crochetière, A., Boulanger, J.-C., and Ouellon, C., editors, Actes du XVe Congrès international des linguistes vol. II. P.U.L.: Québec. 31–4. Diesing, M. (1989). ‘Bare plural subjects, inflection, and the mapping to LF’. In Bach, E., Kratzer, A., and Partee, B. H., editors, Papers on Quantification. University of Massachussets: Amherst. Diesing, M. (1992). Indefinites. MIT Press: Cambridge, MA. Dore, J. (1975). ‘Holophrases, speech acts and language universals’. Journal of Child Language 2: 21–40. É. Kiss, K. (1995). ‘Introduction’. In É. Kiss, K., editor, Discourse configurational languages. Oxford University Press: Oxford. 3–28. É. Kiss, K. (1998). ‘Identificational focus versus information focus’. Language 74: 245–73. Emonds, J. (1970). Root and structure-preserving transformations. Doctoral dissertation, MIT. Emonds, J. (1979). ‘Appositive relatives have no properties’. Linguistic Inquiry 10: 211–43. Emonds, J. (2000). Lexicon and Grammar: The English Syntacticon. Mouton de Gruyter: Berlin. Emonds, J. (2001). ‘La relation entre la dislocation à droite et le clitique franco-italien en/ne’. Journal of the Linguistic Society of Japan 119: 1–32. Emonds, J. (2004). ‘Unspecified categories as the key to root constructions’. In Adger, D., De Cat, C., and Tsoulas, G., editors, Peripheries: Syntactic edges and their effects. Kluwer: Dordrecht. 75–120. Emslie, H. and Stevenson, R. (1981). ‘Pre-school children’s use of the articles in definite and indefinite referring expressions’. Journal of Child Language 8: 313–28. Erteschik-Shir, N. (1986). ‘Wh-questions and focus’. Linguistics and Philosophy, 9: 117–49. Erteschik-Shir, N. (1997). The Dynamics of Focus Structure. Cambridge University Press: Cambridge. Escobar, L. (1997). ‘Clitic Left Dislocation and other relatives’. In Anagnostopoulou, E., van Riemsdijk, H., and Zwarts, F., editors, Materials on Left Dislocation, vol. 14. John Benjamins: Amsterdam. 233–74. Ferdinand, A. (1993). ‘Subject dislocations in French child language’. HIL Manuscripts: 54–64. Ferdinand, A. (1996). The Development of Functional Categories: The Acquisition of the Subject in French, Holland Academic Graphics: The Hague. Firbas, J. (1964). ‘On defining the theme in functional sentence analysis’. Travaux Linguistiques de Prague 1: 267–80. Firbas, J. (1975). ‘On the thematic and non-thematic section of the sentence’. Style and Text. 314–34.

References

283

Fónagy, I. (1985). ‘J’aime_, je connais_: Verbes transitifs à objet latent’. Revue Romane 20: 3–35. Frascarelli, M. (2000). The syntax-phonology interface in focus and topic constructions in Italian. Kluwer: Dordrecht. Frascarelli, M. (2002). Topicalization as Merge. A comparative analysis of Left and Right Dislocation in Italian. Manuscript, Università degli Studi Roma Tre. Frascarelli, M. (2004). ‘Dislocation, clitic resumption and minimality’. In Bok-Bennema, R., Hollebrandse, B., Kampers-Manhe, B., and Sleeman, P., editors, Romance Languages and Linguistic Theory 2002 John Benjamins: Amsterdam. 99–118. Frei, H. (1929). La grammaire des fautes. Geuthner: Paris. Friedemann, M.-A. (1992). ‘The underlying position of external arguments in French’. Geneva Generative Papers 1–2: 123–44. Gadet, F. (1989). Le français ordinaire. Armand Colin: Paris. Gadet, F. (1998). ‘Le “français avancé” à l’épreuve de ses données’. In Bilger, M., van den Eynde, K., and Gadet, F., editors, Analyse linguistique et approches de l’oral: recueil d’études offert en hommage à Claire Blanche-Belveniste. Peeters: Paris. 59–68. Gadet, F. (2000). ‘Des corpus pour ne . . . pas’. In Bilger, M., editor, Corpus: méthodologie et applications linguistiques, Champion: Paris. 156–67. Gill, K.-H. and Tsoulas, G. (2004). ‘Peripheral effects without peripheral syntax’. In Adger, D., De Cat, C., and Tsoulas, G., editors, Peripheries: Syntactic edges and their effects. Kluwer: Dordrecht. 121–41. Ginzburg, J. and Sag, I. (2000). Interrogative investigations: The form, meaning and use of English interrogatives. CSLI: Stanford, CA. Givón, T. (1976). ‘Topic, pronoun and grammatical agreement’. In Li, C., editor, Subject and Topic. Academic Press: New York. 149–211. Godard, D. (2004). ‘French negative dependency’. In Corblin, F. and de Swart, H., editors, Handbook of French semantics. CSLI Publications: Stanford, CA. Greenfield, P., Reilly, J., Leaper, C., and Baker, N. (1985). ‘The structural and functional status of single-word utterances and their relationship to early multiword speech’. In Barrett, M., editor, Children’s single-word speech. Wiley: New York. 233–67. Grobet, A. (2000). L’identification des topiques dans les dialogues. Thèse de doctorat, Université de Genève. Grohmann, K. (2000). ‘Copy Left Dislocation’. In Billerey and Lillehaugen, editors, Proceedings of WCCFL 19. Cascadilla Press: Somerville, MA. 139–52. Grohmann, K. (2003). Prolific Domains: On the Anti-Locality of Movement Dependencies. John Benjamins: Amsterdam. Gruber, J. (1967). ‘Topicalisation in child language’. Foundations of Language 3: 37–65. Guilbault, C. (1993). ‘Dislocation à gauche et prosodie en québécois spontané’. Dialangue 4: 34–8. Gundel, J. (1974). The role of topic and comment in linguistic theory. Doctoral dissertation, University of Texas at Austin.

284

References

Gundel, J. (1975). ‘Left Dislocation and the role of topic-comment structures in linguistic theory’. OSU WPL 18: 72–131. Haegeman, L. (2002). ‘Anchoring to speaker, adverbial clauses and the structure of CP’. Georgetown University Working Papers in Theoretical Linguistics 2: 117–80. Haegeman, L. (2003). The syntax of adverbial clauses and its consequences for topicalisation. Manuscript, Université Charles de Gaulle, Lille III. Halle, M. and Marantz, A. (1993). ‘Distributed morphology and the pieces of inflection’. In Halle, K. and Keyser, S., editors, The View from Building 20. MIT Press: Cambridge, MA. 111–76. Halliday, M. A. K. (1967). ‘Notes on transitivity and theme in English II’. Journal of Linguistics 3: 199–244. Hamann, C. (2002). From Syntax to Discourse. Pronominal Clitics, Null Subjects and Infinitives in Child Language, Kluwer: Dordrecht. Hamann, C., Rizzi, L., and Frauenfelder, U. H. (1996). ‘On the acquisition of subject and object clitics in French’. In Clahsen, H., editor, Generative Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations and Crosslinguistic Comparisons. John Benjamins: Amsterdam. 309–34. Hankamer, J. (1979). Deletion in coordinate structures. Garland: New York. Haspelmath, M. (2006). ‘Against markedness (and what to replace it with)’. Journal of Linguistics 42: 25–70. Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases. Doctoral dissertation, University of Massachussets. Heim, I. (1983). ‘File change semantics and the familiarity theory of definiteness’. In Baüerle, R., Schwarze, C., and von Stechow, A., editors, Meaning, Use, and Interpretation of Language. de Gruyter: Berlin. 164–89. Heycock, C. (2006). ‘Embedded root phenomena’. In Everaert, M. and van Riemsdijk, H., editors, The Blackwell Companion to Syntax, vol. II. Blackwell: Oxford. 174–209. Heycock, C. and Kroch, A. (1999). ‘Pseudocleft connectedness: Implications for the LF interface level’. Linguistic Inquiry 30: 365–97. Heyd, S. and Mathieu, E. (2005). The semantics of [de N] structures, incorporation and referential intentions. Talk presented at the Brussels International Conference on Indefinites and Weak Quantifiers. Hickmann, M. (1982). The development of narrative skills: Pragmatic and metapragmatic aspects of discourse cohesion. Doctoral dissertation, University of Chicago. Hickmann, M. (2003). Children’s discourse: Person, space and time across languages. Cambridge University Press: Cambridge. Higgins, R. (1973). ‘On J. Emonds’s analysis of extraposition’. Syntax and Semantics, vol. 2. Academic Press: New York. 149–95. Hirschbühler, P. (1974). ‘La dislocation à gauche comme construction basique en français’. In Rohrer, C. and Ruwet, N., editors, Colloque franco-allemand de grammaire transformationnelle, vol. 1. Tübingen. Hirschbühler, P. (1975). ‘On the source of lefthand NPs in French’. Linguistic Inquiry 6: 155–65.

References

285

Hoekstra, E. (1991). Licensing Conditions on Phrase Structure. Groningen Dissertations in Linguistics (GroDiL), Groningen. Hoekstra, T. and Jordens, P. (1994). ‘From adjunct to head’. In Hoekstra, T. and Schwartz, B., editors, Language Acquisition Studies in Generative Grammar. John Benjamins: Amsterdam. 119–49. Hooper, J. B. and Thompson, S. A. (1973). ‘On the applicability of root transformations’. Linguistic Inquiry 4: 465–97. Hulk, A. (1986). ‘Subject clitics and the pro-drop parameter’. In Coopmans, P., Bordelois, I., and Dotson-Smith, B., editors, Going Romance: Formal Parameters of Generative Grammar II. Foris: Dordrecht. 107–20. Hulk, A. (1995). ‘L’acquisition du sujet en français’. Recherches Linguistiques de Vincennes 24: 33–55. Hulk, A. (1996a). ‘The delay of inversion questions in the acquisition of French’. In Parodi, C., Quicoli, C., Saltarelli, M., and Zubizarreta Maria, L., editors, Aspects of Romance Linguistics. Georgetown Washington, DC. 241–52. Hulk, A. (1996b). ‘The syntax of wh-questions in child French’. In Philip, W. and Wijnen, F., editors, Connecting Children’s Language and Linguistic Theory, Amsterdam Series in Child Language Development, vol. 5. Institute for General Linguistics, University of Amsterdam: Amsterdam. 123–72. Hyams, N. (1986). Language Acquisition and the Theory of Parameters. Reidel: Dordrecht. Iatridou, S. (1995). ‘Clitics and islands effects’. In Izvorski, R. and Tredinnick, V., editors, U.Penn Working Papers in Linguistics. 2: 11–30. Jaeggli, O. (1982). Topics in Romance Syntax. Foris: Dordrecht. Jakubowicz, C., Nash, L., Rigaut, C., and Gérard, C. L. (1998). ‘Determiners and clitic pronouns in French-speaking children with SLI’. Language Acquisition 7: 113–60. Jäger, G. (2001). ‘Topic-comment structure and the contrast between Stage Level and Individual Level Predicates’. Journal of Semantics 18: 83–126. Kadmon, N. (2001). Formal Pragmatics. Blackwell: Oxford. Kaiser, G. (1994). ‘More about Inflection: The acquisition of clitic pronouns in French’. In Meisel, J., editor, Bilingual first language acquisition: French and German grammatical development. John Benjamins: Amsterdam. Karmiloff-Smith, A. (1981). ‘The grammatical marking of thematic structure in the development of language production’. In Deutsch, W., editor, The Child’s Construction of Language. Academic Press: London. 121–47. Kayne, R. (1975). The Transformational Cycle. MIT Press: Cambridge, MA. Kayne, R. (1984). ‘Binding, quantifiers, clitics and control, In Kayne, R. editor, Connectedness and binary branching. Foris: Dordrecht. Kayne, R. (1991). ‘Romance clitics, verb movement and PRO’. Linguistic Inquiry 22: 647–86. Kayne, R. (1994). The antisymmetry of syntax. MIT Press: Cambridge, MA. Koopman, H. and Sportiche, D. (1991). ‘The position of subjects’. Lingua 85: 211–58.

286

References

Krämer, I. (2000). Interpreting Indefinites. An experimental study of children’s language comprehension. Doctoral dissertation, University of Utrecht. Kratzer, A. (1989). ‘Stage-Level and Individual-Level Predicates’. In Bach, E., Kratzer, A., and Partee, B. H., editors, Papers on Quantification. University of Massachussets: Amherst. 147–222. Krifka, M., Pelletier, F. J., Carlson, G. N., ter Meulen, A., Link, G., and Chierchia, G. (1995). ‘Genericity: An introduction’. In Carlson, G. N. and Pelletier, F. J., editors, The Generic Book. University of Chicago Press: Chicago. 1–124. Kuno, S. (1972). ‘Functional sentence perspective: A case study from Japanese and English’. Linguistic Inquiry 3: 269–320. Kuroda, S. Y. (1972). ‘The categorical and the thetic judgment: Evidence from Japanese syntax’. Foundations of Language 9: 153–85. Labelle, M. (1985). ‘Caractère post-lexical de la cliticisation française’. Linguisticæ Investigationes 9: 83–96. Labelle, M. (1990). ‘Predication, wh-movement and the development of relative clauses’. Language Acquisition 1: 95–119. Labelle, M. (1996). ‘The acquisition of relative clauses: Movement or no movement?’ Language Acquisition 5: 65–82. Labelle, M. and Hirschbühler, P. (in preparation). Syntaxe du français: L’universel et le particulier dans la langue. Manuscript, Université du Québec à Montréal and Université d’Ottawa. Labelle, M. and Valois, D. (1996). ‘The status of post-verbal subjects in French child language’. Probus 8: 53–80. Laenzlinger, C. (1998). Comparative Studies in Word Order Variation. John Benjamins: Amsterdam. Lahousse, K. (2003). ‘NP-subject inversion in French and (preposed) adverbs’. In Pérez-Leroux, A. and Roberge, Y., editors, Romance Linguistics: Theory and acquisition. John Benjamins: Amsterdam. 181–96. Lambrecht, K. (1981). Topic, antitopic and verb agreement in non-standard French, Pragmatics & Beyond vol. II: 6. John Benjamins: Amsterdam. Lambrecht, K. (1986). Topic, focus, and the grammar of spoken French. Doctoral dissertation, University of California. Lambrecht, K. (1994). Information Structure and Sentence Form. Topic, focus, and the mental representation of discourse referents. Cambridge University Press: Cambridge. Lambrecht, K. (2001). ‘Dislocation’. In Haspelmath, M., König, E., Oesterreicher, W., and Raible, W., editors, Language Typology and Language Universals. An International Handbook, vol. 20.2. Mouton de Gruyter: Berlin. 1050–78. Lamiroy, B. (1991). ‘Coréférence et référence disjointe: les deux pronoms en et y’. Travaux de Linguistique 22: 41–67. Larsson, E. (1979). La dislocation en français: étude de syntaxe générative. Gleerup: Lund. Lasnik, H. and Saito, M. (1992). Move alpha. MIT Press: Cambridge, MA. Lasnik, H. and Stowell, T. (1991). ‘Weakest crossover’. Linguistic Inquiry 22: 687–720.

References

287

Lebeaux, D. (1988). Language acquisition and the form of the grammar. Doctoral dissertation, University of Massachussets. Li, C. N. and Thompson, S. A. (1976). ‘Subject and topic: A new typology of language’. In Li, C., editor, Subject and Topic. Academic Press: New York. 457–89. Locke, J. (1983). Phonological Acquisition and Change. Academic Press: New York. López, L. (2003). ‘Steps for a well-adjusted dislocation’. Studia Linguistica 57: 193–231. Lust, B. (1999). ‘Universal Grammar: The strong continuity hypothesis in first langugae acquisition’. In Ritchie, W. C. and Bhatia, T. K., editors, Handbook Of Child Language Acquisition. Academic Press: San Diego. 111–55. McNally, L. (1998). ‘On the linguistic encoding of information packaging instructions’. In Culicover, P. and McNally, L., editors, The Limits of Syntax. Academic Press: San Diego. 161–83. MacWhinney, B. (2000). The CHILDES Project: Tools for Analyzing Talk. Lawrence Erlbaum Associates: Mahwah, NJ. Maratsos, M. (1974). ‘Preschool children’s use of definite and indefinite articles’. Child Development 45: 446–55. Merchant, J. (2005). ‘Fragments and ellipsis’. Linguistics and Philosophy 6: 661–738. Mertens, P. (1987). L’intonation du français. De la description linguistique à la reconnaissance automatique. Doctoral dissertation, Katholieke Universiteit Leuven. Mertens, P. (1993). Intonational grouping, boundaries, and syntactic structures in French. In House, D. and Touati, P., editors, Proc. ESCA Workshop on prosody, Lund Working Papers 41: 156–9. Mertens, P. (1997). ‘De la chaîne linéaire à la séquence de tons’. Traitement Automatique des Langues 38: 27–51. Mertens, P., Goldmann, J.-P., Wehrli, E., and Gaudinat, A. (2001). ‘La synthèse de l’intonation à partir de structures syntaxiques riches’. Traitement Automatique des Langues 42: 142–95. Miller, P. and Monachesi, P. (2003). ‘Les pronoms clitiques dans les langues romanes’. In Godard, D., editor, Les langues romanes: problèmes de la phrase simple. Editions du CNRS: Paris. 67–123. Milner, J.-C. (1978). De la syntaxe à l’interprétation: quantités, insultes, exclamations. Seuil: Paris. Milner, J.-C. (1979). ‘Le système de la négation en français et l’opacité du sujet’. Langue Française 4: 80–105. Milsark, G. (1974). Existential Sentences in English. Doctoral dissertation, MIT. Moreau, M.-L. (1976). C’est. Etude de syntaxe transformationnelle. Doctoral dissertation, Université de Mons. Müller, N. (1994). ‘Parameters cannot be reset: Evidence from the development of COMP’. In Meisel, J. M., editor, Bilingual First Language Acquisition: French and German Grammatical Development. John Benjamins: Amsterdam. 235–69. Müller, N., Crysmann, B., and Kaiser, G. A. (1996). ‘Interaction between the acquisition of French object-drop and the development of the C-system’. Language Acquisition 5: 35–63.

288

References

Newmeyer, F. J. (2003). ‘Against the split-CP hypothesis’. In Banski, P. and Przepiorkowsk, A., editors, Proceedings of GLiP-5. Polish Academy of Sciences: Warsaw. 81–93. Nice, M. (1925). ‘The length of sentences as a criterion of child’s progress in speech’. Jounal of Educational Psychology 16: 370–79. O’Grady, W. (1997). Syntactic Development. University of Chicago Press: Chicago. Pesetsky, D. (1986). ‘Wh-in situ: Movement and unselective binding’. In Reuland, E. and ter Meulen, A., editors, The Representation of (In)definiteness. MIT Press: Cambridge, MA. 98–129. Peters, A. (2001). ‘Filler syllables: What is their status in emerging grammar?’ Journal of Child Language. 28: 229–42. Pierce, A. (1992). Language Acquisition and Syntactic Theory: A Comparative Analysis of French and English Child Grammars. Kluwer: Dordrecht. Pierce, A. (1994). ‘On the differing status of subject pronouns in French and English child language’. In Lust, B., Hermon, G., and Kornfilt, J., editors, Syntactic theory and first language acquisition—Crosslinguistic perspectives: Volume 2—Binding, dependencies, and learnability. Lawrence Erlbaum Associates: Hillsdale, NJ. 319–33. Pinker, S. (1984). Language Learnability and Language Development. Harvard University Press: Cambridge, MA. Plunkett, B. (1999). Periphrastic questions as inversion and their acquisition in European French. Poster presented at the IASCL conference, San Sebastian, Basque Country, Spain. Plunkett, B. (2000). ‘The emergence of periphrastic questions in child French’. In Perkins, M. and Howards, S., editors, New Directions in Language Development and Language Disorders. Plenum Press: New York. 105–17. Plunkett, B. (2001). ‘Attract and covert merge: Predicting interrogative variation’. In Alexandrova, G. M. and Arnaudova, O., editors, The Minimalist Parameter. John Benjamins: Amsterdam. 159–72. Plunkett, B. (2002). ‘Null Subjects in Child French Interrogatives: A view from the York Corpus’. In Pusch, C. and Raible, W., editors, Romanistische Korpuslinguistik: Korpora und gesprochene Sprache. Narr: Tübingen. 441–52. Plunkett, B. (2003). ‘Null subjects and the setting of subject agreement parameters in child French’. In Pérez-Leroux, A. and Roberge, Y., editors, Romance Linguistics: Theory and acquisition. John Benjamins: Amsterdam. 351–66. Plunkett, B. and De Cat, C. (2001). ‘Root specifiers and null subjects revisited’. Proceedings of the 25th Boston University Conference on Child Language Development. Cascadilla: Somerville, MA. 611–22. Portner, P. and Yabushita, K. (1998). ‘The semantics and pragmatics of topic phrases’. Linguistics and Philosophy 21: 117–57. Postal, P. (1970). Cross-Over Phenomena. Holt, Rinehart and Winston: New York. Powers, S. (2001). ‘A minimalist account of phrase structure acquisition’. In Alexandrova, G. M. and Arnaudova, O., editors, The Minimalist Parameter. John Benjamins: Amsterdam. 33–50.

References

289

Prévost, P. (2006). ‘The phenomenon of object omission in child L2 French’. Bilingualism: Language and Cognition 9: 263–80. Prince, E. (1999). How not to mark topics: ‘Topicalization’ in English and Yiddish. Manuscript, University of Pennsylvania. Prince, E. F. (1981). ‘Topicalization, focus-movement, and Yiddish-movement: A pragmatic differentiation’. Proceedings of the Seventh Annual Meeting of the Berkeley Linguistics Society: 249–64. Rasetti, L. (2000). ‘Null subjects and Root Infinitives in the child grammar of French’. In Friedemann, M.-A. and Rizzi, L., editors, The Acquisition of Syntax: Studies in Comparative Developmental Linguistics. Longman: New York. 236–68. Reinhart, T. (1981). ‘Pragmatics and linguistics: An analysis of sentence topics’. Philosophica 27: 53–94. Reinhart, T. (1996). ‘Interface economy: Focus and markedness’. In Wilder, C., Gaertner, H.-M., and Bierwisch, M., editors, The Role of Economy Principles in Linguistic Theory. Akademic Verlag: Berlin. 146–69. Rizzi, L. (1986). ‘On the status of subject clitics in Romance’. In Jaeggli, O. and SilvaCorvalan, C., editors, Studies in Romance Linguistics. Foris: Dordrecht. 391–419. Rizzi, L. (1990). Relativized Minimality. MIT Press: Cambridge, MA. Rizzi, L. (1991). ‘Residual verb second and the wh-criterion’. Technical Reports in Formal and Computational Linguistics, 2: 1–28. Rizzi, L. (1994). ‘Early null subjects and root null subjects’. In Lust, B., Hermon, G., and Kornfilt, J., editors, Syntactic theory and first language acquisition— Crosslinguistic perspectives: Volume 2—Binding, dependencies, and learnability. Lawrence Erlbaum Associates: Hillsdale, NJ. 249–72. Rizzi, L. (1997). ‘The fine structure of the left periphery’. In Haegeman, L., editor, Elements of grammar: A handbook of generative syntax. Kluwer: Dordrecht. 281–337. Rizzi, L. (2002). ‘Locality and left periphery’. In Belletti, A., editor, Structures and Beyond: The Cartography of Syntactic Structures, vol. 3. Oxford University Press: Oxford. 223–51. Roberge, Y. (1986). ‘Subject doubling, free inversion, and null argument languages’. Canadian Journal of Linguistics 31: 55–79. Roberge, Y. (1990). The syntactic recoverability of null arguments. McGill-Queen’s University Press: Montréal. Roberts, C. (1998). Information Structure in Discourse: Towards an Integrated Formal Theory of Pragmatics. Manuscript, Ohio State University. Rochemont, M. (1986). Focus in generative grammar. John Benjamins: Amsterdam. Rodman, R. (1974). ‘On Left-Dislocation’. Papers in Linguistics 7: 437–66. Roeper, T. (1992). ‘From the initial state to V2: Acquisition principles in action’. In Meisel, J., editor, The Acquisition of Verb Placement: Functional Categories and V2 Phenomena in Language Acquisition. Kluwer: Dordrecht. 333–70. Ronat, M. (1979). ‘Pronoms topiques et pronoms distinctifs’. Langue française 44: 106–28.

290

References

Rooth, M. (1992). ‘A theory of focus interpretation’. Natural Language Semantics 1: 75–116. Ross, J. R. (1967). Constraints on Variables in Syntax. Doctoral dissertation, MIT. Rossi, M. (1971). ‘Le seuil des glissando ou seuil de perception des variations tonales pour les sons de la parole’. Phonetica 23: 1–33. Rossi, M. (1972). ‘Le seuil différentiel de la durée’. In Valdman, A., editor, Papers in linguistics and phonetics to the memory of Pierre Delattre. Mouton: The Hague. 435–50. Rossi, M. (1973). ‘L’intonation prédicative dans les phrases transformées par permutation’. Linguistics 103: 64–94. Rossi, M. (1981). ‘Intonation, énonciation, syntaxe’. In Rossi, M., Di Cristo, A., Hirst, D., Martin, P., and Nishinuma, Y., editors, L’intonation: de l’acoustique à la sémantique. Klincksieck: Paris. 184–234. Rossi, M. (1999). L’intonation. Le système du français: description et modélisation. Ophrys: Paris. Rowlett, P. (1998). Sentential Negation in French. Oxford University Press: Oxford. Rowlett, P. (1999). ‘French ne in non-verbal contexts’. In D’Hulst, Y., Rooryck, J., and Schroten, J., editors, Romance Languages and Linguistic Theory 1999: Selected papers from Going Romance. John Benjamins: Amsterdam. 335–53. Ruwet, N. (1990). ‘En et y: deux clitiques pronominaux anti-logophoriques’. Langages 97: 51–81. Sabel, J. (2002). ‘Intermediate traces, reconstruction and locality effects’. In Alexiadou, A., editor, Theoretical Approaches to Universals. John Benjamins: Amsterdam. 259–313. Saito, M. and Fukui, N. (1998). ‘Order in phrase structure and movement’. Linguistic Inquiry 29: 439–74. Samek-Lodovici, V. (2006). ‘When Right-Dislocation meets the Left-periphery’. Lingua 116: 836–73. Schachter, P. (1973). ‘Focus and relativization’. Language 49: 19–46. Schaeffer, J., Gordishevsky, G., Hadar, G., and Hacohen, A. (2002). ‘Subjects in English-speaking children with Specific Language Impairment’. In Costa, J. and Freitas, M. J., editors, Proceedings of the GALA 2001 Conference. Associação Portuguesa de Linguística, Lisboa. 223–31. Schaeffer, J. and Matthewson, L. (2005). ‘Grammar and pragmatics in the acquisition of article systems’. Natural Language and Linguistic Theory 23: 53–101. Schütze, C. T. (1997). INFL in Child and Adult Language: Agreement, Case, and Licensing. Doctoral dissertation, MIT. Sells, P. (1984). Syntax and Semantics of Resumptive Pronouns. GSLA: Amherst. Simon, A.-C. (2001). ‘Le rôle de la prosodie dans le repérage des unités textuelles minimales’. Cahiers de Linguistique française 23: 99–125. Snow, D. and Balog, H. (2002). ‘Do children produce the melody before the words? A review of developmental intonation research’. Lingua 112: 1025–58.

References

291

Sportiche, D. (1988). ‘A theory of floating quantifiers and its corollaries for constituent structure’. Linguistic Inquiry 19: 425–49. Sportiche, D. (1995). ‘Clitic constructions’. In Rooryck, J. and Zaring, L., editors, Phrase Structure and the Lexicon. Kluwer: Dordrecht. 213–76. Stainton, R. (2004). ‘In defense of non-sentential assertion’. In Szabo, Z., editor, Semantics vs. Pragmatics. Oxford University Press: Oxford. 383–457. Stalnaker, R. (1974). ‘Pragmatic presuppositions’. In Munitz, M. and Unger, P., editors, Semantics and Philosophy. New York University Press: New York. 197–213. Strawson, P. F. (1964). ‘Identifying reference and truth values’. Theoria 30: 96–118. Szendr˝ oi, K. (2001). Focus and the Syntax-Phonology Interface. Doctoral dissertation, UCL. Szendr˝ oi, K. (2003). ‘A stress-based approach to the syntax of Hungarian focus’. The Linguistic Review 20: 37–78. Tanaka, H. (2001). ‘Right-Dislocation as scrambling’. Journal of Linguistics 37: 551–79. Tellier, C. (2001). ‘On some distinctive properties of parasitic gaps in French’. In Culicover, P. and Postal, P., editors, Parasitic Gaps. MIT Press: Cambridge, MA. 341– 67. Tomioka, S. (2001). ‘Information structure and disambiguation in Japanese’. In Megerdoomian, K., and Bar-el, L., editors, Proceeding of WCCFL, 20. Cascadilla Press: Somerville, MA. 552–64. Unsworth, S. (2005). Child L2, Adult L2, Child L1: Differences and Similarities. LOT: Utrecht. Uriagereka, J. (1995). ‘An F position in Western Romance’. In É. Kiss, K., editor, Discourse configurational languages. Oxford University Press: Oxford. 153–75. Vallduví, E. (1992). The Informational Component. Garland: New York. van Haaften, T., Smits, R., and Vat, J. (1983). ‘Left Dislocations, connectedness and reconstruction’. In Ehlich, K. and van Riemsdijk, H., editors, Connectedness in Sentence, Discourse and Text. KUB: Tilburg. 133–54. van Kampen, J. (2004). ‘The learnability of syntactic categories’. In van Kampen, J. and Baauw, S., editors, Proceedings of GALA 2003. 245–56. van Riemsdijk, H. and Zwarts, F. (1974). Left Dislocation in Dutch and the Status of Copying Rules. Unpublished manuscript, MIT and University of Amsterdam. Vat, J. (1981). ‘Left Dislocation, connectedness and reconstruction’. Groninger Arbeiten zur Germanistischen Linguistik 20: 80–103. Vecchiato, S. (2000). ‘The ti/tu interrogative morpheme in Quebec French’. Generative Grammar in Geneva 1: 141–63. Vendryes, J. (1920). Le langage. Albin Michel: Paris. Villalba, X. (1998). ‘Right Dislocation is not Right Dislocation’. In Fullana, O. and Roca, F., editors, Studies on the Syntax of Central Romance Languages. Universitat de Girona: Girona. 227–41. Villalba, X. (2000). The Syntax of Sentence Periphery. Doctoral dissertation, Universitat Autònoma de Barcelona.

292

References

Vion, M. (1992). ‘The role of intonation in processing Left and Right Dislocations in French’. Journal of Experimental Child Psychology 53: 45–71. von Fintel, K. (1994). Restrictions on Quantifier Domains. Doctoral dissertation, University of Massachussets. Ward, G. and Prince, E. F. (1991). ‘On the topicalization of indefinite NPs’. Journal of Pragmatics 15: 167–78. Wexler, K., Schütze, C. T., and Rice, M. (1998). ‘Subject case in children with SLI and unaffected controls: Evidence for the Agr/Tense omission model’. Language Acquisition 7: 317–44. Zribi-Hertz, A. (1984). ‘Prépositions orphelines et pronoms nuls’. Recherches Linguistiques 12: 46–91. Zribi-Hertz, A. (1994). ‘The syntax of nominative clitics in standard and advanced French’. In Cinque, G., Koster, J., Pollock, J.-Y., Rizzi, L., and Zanuttini, R., editors, Paths Towards Universal Grammar: Studies in Honor of Richard S. Kayne. Georgetown University Press: Washington, DC. 453–72. Zubizarreta, M. L. (1998). Prosody, Focus and Word Order, MIT Press: Cambridge, MA. Zwicky, A. and Pullum, G. (1983). ‘Cliticization vs. inflection: English n’t’. Language 59: 502–13.

Index aboutness 64, 76 see also topic adjacency 157 adjunction 101, 103, 105, 141, 150, 154, 165, 166, 180, 213 adverbs 72, 112, 155, 232, 234, 258, 262 affixes 10–11, 13, 15, 16, 18–19, 32, 34 agree 100, 130–1, 180 agreement 7, 10–13, 32, 34, 98, 101, 130–1, 153, 154, 165 agreement morpheme 10–11, 13, 15, 21, 26–31, 33 anti-reconstruction 123 see also reconstruction antitopic 66, 160, 161 Arabic 64, 98, 102, 104, 125, 133 c-command 19, 123, 166, 167 Case 10, 31, 136 case 43, 101, 106, 108, 135–6, 173 Catalan 99, 102, 108, 153 chain 103–4, 118, 121, 129–30, 144, 180 Chichewa ˆ 27–9 clefts 95, 96, 130, 185–6, 203, 215, 266 pseudo-clefts 91–3, 96, 185, 215 clitics see pronouns complementiser 8, 99–100, 130–1, 156–7, 183–91, 193–6, 237, 241 acquisition of 180–1, 183, 185, 187–91, 194–6, 221 Complementiser Phrase (CP) 8, 20, 100, 101, 213 doubly-filled Comp 13, 188, 196, 269 connectedness see connectivity connectivity 101, 103, 106, 108, 138–40 contrast contrastive focus 25, 95–6, 205

contrastive topic 147, 161–2, 167, 210–11, 215, 224 d-linking see indefinites (d-linked) dialects 6–9, 13–14, 21, 23, 85, 126, 128, 139, 172, 178, 210, 226–7, 230, 241, 269 see also French Discourse Projections 98, 150, 152–9, 165 dislocation, Clitic Left Dislocation (CLLD) 98, 102, 104–5, 107–11, 118–19, 121, 125, 131, 135–9, 141, 275–277 Clitic Right Dislocation (CLRD) 104, 110 Hanging Topic Left Dislocation (HTLD) 99–100, 106–8, 135–9, 224, 267, 283–4 left dislocation (LD) 9, 34, 44, 98, 106, 107, 110–11, 115, 125, 129, 132, 134, 136, 139, 147, 157, 159, 161, 163–4, 176, 226, 242 right dislocation (RD) 34–7, 39–41, 98, 99, 105, 110, 112, 115–17, 139, 141–9, 159, 161, 164, 178, 212, 242 very local RD 112–17 doubling 10–11, 19, 102, 105, 140–3 doubly-filled Comp see complementiser dummy determiner 83, 209 ellipsis 200 embedded clauses 17, 20, 78, 100, 107, 118, 135, 137, 152, 153, 155–60, 181–3, 185, 187, 190, 193–5, 203, 224, 225, 239–41, 276, 277 embryonic elements 173, 176, 266 English 23, 48, 73, 86, 89, 104, 132, 150, 191

294

Index

f-structure 30, 63, 73, 89, 90 features associated with clitics 10–11, 14, 26, 31–3 associated with dislocated topics 98–102, 105, 141, 153, 155, 156 of Discourse Projections 152, 153, 158 file cards 64, 66, 70 focus 22–6, 29, 30, 65–7, 71, 78, 81, 94–6, 99–101, 103–4, 154, 155, 198–200, 202–3, 206 contrastive focus 25, 95, 96, 205 fragments 198–203, 214 French advanced French 7–8 colloquial French 6, 8, 13 informal French 6 non-standard French 6, 9 standard French 4–6, 8, 32 Fundamental frequency (F0) 34, 36, 45, 48–50, 55, 176, 214 generic reading 41, 69, 78, 82–6, 108, 209, 221 see also indefinites Germanic languages 135, 160 Greek 64, 98, 101–2, 125, 240 hanging topics see dislocation (HTLD) heavy subjects 44, 47, 48, 50–1, 53–62, 77–8, 206 indefinites d-linked 42, 86, 89, 92, 94, 127, 145, 147, 207, 269 generic reading of 21, 41, 79, 82–3, 86, 261 Individual Level Predicates (ILPs) 25–6, 72–3, 77–81, 84, 87, 205–7 Intonation Group (IG) 37, 45–6, 48–51, 53–4, 56, 61, 176 Irish 130

Italian 63, 98–9, 101–2, 104, 106, 108, 123, 125, 141, 146, 150, 153, 183, 240 left dislocation (LD) see dislocation Ma’di 240 Mean Lenth of Utterance (MLU) 172, 219, 245, 249 negation 4, 6, 11, 15, 17, 93, 121, 123, 128, 145 Negative Polarity Items (NPIs) 124 ordering restrictions 100, 115, 144, 229 out-of-the-blue contexts 68, 78, 95 package 45, 46, 50, 51, 54, 56 parasitic gaps 104, 118–20, 144, 169 parenthesis intonation 35 pause 35, 44, 47, 50, 107, 179, 228 Perception Unit 45, 49 performative requirement 152, 201–2, 217 permanent registry 68–9 pitch 34–40, 44, 45, 48–53, 55–9, 61–2, 161, 163, 176, 179, 238 poset (partially ordered set) 69, 88, 97 pragmatics 63, 72, 102, 108, 204, 208–11, 213, 266 presentational constructions 88–90, 94–6 presupposition 65–6, 160 Principle C effects 121–3, 166–8 pronouns clitics object clitics 10–11, 15, 18, 173 subject clitics 7–10, 32–3, 43, 80, 189, 207, 257, 266 strong pronouns 57, 107, 142, 173, 229–30, 233, 257, 262 pseudo-clefts see clefts quantified expressions 81, 89, 207, 262 Question-Under-Discussion 71

Index reconstruction 104–5, 118, 121–3, 144, 145, 165, 169 relative clauses 92, 128, 132, 157–8, 160, 191, 239 Relativised Minimality 118, 120, 144, 157 relevance 70, 71, 76, 97, 160, 205 resumptive 24, 25, 41, 58, 62, 80, 84–6, 98, 101–5, 107–8, 111–2, 118–21, 123, 128–36, 138, 139, 142, 145, 148–50, 153, 155, 157, 169, 172–3, 204, 215, 223–4, 229–32, 250, 261–4 epithet 109 non-clitic resumptive 98, 107, 109, 110, 136, 141, 143, 189 see also pronoun (clitic) right dislocation see dislocation Right Roof constraint 139 rightward movement 100, 139, 140, 148–9 Romance languages 105, 107, 141 Root Infinitives 202, 204, 217, 247–52, 269 salience 68–70, 97, 164, 205 Spanish 23, 102, 118, 125, 141, 143, 183, 240 Stage Level Predicates (SLP) 72–3, 78–9 stage topics 72–4, 76, 78, 81, 90, 96, 112, 201, 256, 262–4 stress 22, 37, 46–8, 50, 66, 149, 161, 174, 198 subordinate update 88–90, 92–4, 97, 132, 168

295

Swedish 113, 131 theme 66 thetic interpretation 24, 25, 30, 74, 81, 84, 94–6, 139, 207, 267 topic 20–6, 28–33, 45, 50, 58, 61, 63–74, 76–8, 81–3, 86–94, 96, 99, 101–3, 109, 132, 150–1, 153–5, 160, 161, 163–4, 167, 196, 204, 205, 207–10, 213, 215 aboutness topic 24–5, 74, 76, 81, 96 contrastive topic 147, 161–2, 167, 210–11, 215, 221, 267 Topic Phrase(s) 99, 100, 103, 144, 149, 151, 157, 169 lower TopicP 100, 144 topicalization 104, 150 truncation 195, 202 variable 23–4, 92–4, 103, 119, 121, 124, 131–2 verbless utterances 180, 196–8, 214, 221, 255–6 see also fragments weak crossover 118, 144, 165, 169 wh-configurations 13–14, 22, 100, 127, 132, 164–5, 187, 190–4, 196, 200, 212, 240–2, 254 wide scope 32, 92, 93, 121, 123

OXFORD STUDIES IN THEORETICAL LINGUISTICS published 1 The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis by Jason Merchant 2 Questions and Answers in Embedded Contexts by Utpal Lahiri 3 Phonetics, Phonology, and Cognition edited by Jacques Durand and Bernard Laks 4 At the Syntax–Pragmatics Interface: Concept Formation and Verbal Underspecification in Dynamic Syntax by Lutz Marten 5 The Unaccusativity Puzzle Explorations of the Syntax–Lexicon Interface edited by Artemis Alexiadou, Elena Anagnostopoulou, and Martin Everaert 6 Beyond Morphology Interface Conditions on Word Formation by Peter Ackema and Ad Neeleman 7 The Logic of Conventional Implicatures by Christopher Potts 8 Paradigms in Phonological Theory edited by Laura Downing, T. Alan Hall, and Renate Raffelsiefen 9 The Verbal Complex in Romance by Paola Monachesi 10 The Syntax of Aspect: Deriving Thematic and Aspectual Interpretation edited by Nomi Erteschik-Shir and Tova Rapoport 11 Aspects of the Theory of Clitics by Stephen Anderson 12 Canonical Forms in Prosodic Morphology by Laura J. Downing 13 Aspect and Reference Time by Olga Borik 14 Direct Compositionality edited by Chris Barker and Pauline Jacobson

15 A Natural History of Infixation by Alan C. L. Yu 16 Phi Theory: Phi-Features Across Interfaces and Modules edited by Daniel Harbour, David Adger, and Susana Béjar 17 French Dislocation: Interpretation, Syntax, Acquisition by Cécile De Cat published in association with the series The Oxford Handbook of Linguistic Interfaces edited by Gillian Ramchand and Charles Reiss [Published in association with the series] in preparation The Logic of Pronominal Resumption by Ash Asudeh Inflectional Identity edited by Asaf Bachrach and Andrew Nevins Phi Syntax: A Theory of Agreement by Susana Béjar Tense, Aspect, and Indexicality by James Higginbotham The Syntax of Sentential Stress by Arsalan Kahnemuyipour Conditionals by Angelika Kratzer Stratal Optimality Theory by Ricardo Bermúdez Otero The Ecology of English Noun-Noun Compounding by Ray Jackendoff Adjectives and Adverbs: Syntax, Semantics, and discourse edited by Louise McNally and Chris Kennedy Null Subject Languages by Evi Sifaki and Loanna Sitaridou