TheAcquisition of French in Multilingual Contexts 9781783094530

Citation preview

The Acquisition of French in Multilingual Contexts

SECOND LANGUAGE ACQUISITION Series Editors: Professor David Singleton, University of Pannonia, Hungary and Fellow Emeritus, Trinity College, Dublin, Ireland and Dr Simone E. Pfenninger, University of Zurich, Switzerland This series brings together titles dealing with a variety of aspects of language acquisition and processing in situations where a language or languages other than the native language is involved. Second language is thus interpreted in its broadest possible sense. The volumes included in the series all offer in their different ways, on the one hand, exposition and discussion of empirical findings and, on the other, some degree of theoretical reflection. In this latter connection, no particular theoretical stance is privileged in the series, nor is any relevant perspective – sociolinguistic, psycholinguistic, neurolinguistic, etc. – deemed out of place. The intended readership of the series includes final-year undergraduates working on second language acquisition projects, postgraduate students involved in second language acquisition research, and researchers and teachers in general whose interests include a second language acquisition component. Full details of all the books in this series and of all our other publications can be found at http://www.multilingual-matters.com, or by writing to Multilingual Matters, St Nicholas House, 31–34 High Street, Bristol BS1 2AW, UK.

SECOND LANGUAGE ACQUISITION: 94

The Acquisition of French in Multilingual Contexts Edited by Pedro Guijarro-Fuentes, Katrin Schmitz and Natascha Müller

MULTILINGUAL MATTERS Bristol • Buffalo • Toronto

Library of Congress Cataloging in Publication Data The Acquisition of French in Multilingual Contexts/Edited by Pedro Guijarro-Fuentes, Katrin Schmitz and Natascha Müller. Second Language Acquisition: 94 Includes bibliographical references and index. 1. French language—Study and teaching—Foreign speakers. 2. French language— Grammar—Study and teaching. 3. Second language acquisition. I. Guijarro Fuentes, Pedro, editor. II. Schmitz, Katrin, editor. III. Müller, Natascha, editor. PC2065.A375 2015 448.0071–dc23 2015023402 British Library Cataloguing in Publication Data A catalogue entry for this book is available from the British Library. ISBN-13: 978-1-78309-452-3 (hbk) Multilingual Matters UK: St Nicholas House, 31–34 High Street, Bristol BS1 2AW, UK. USA: UTP, 2250 Military Road, Tonawanda, NY 14150, USA. Canada: UTP, 5201 Dufferin Street, North York, Ontario M3H 5T8, Canada. Website: www.multilingual-matters.com Twitter: Multi_Ling_Mat Facebook: https://www.facebook.com/multilingualmatters Blog: www.channelviewpublications.wordpress.com Copyright © 2016 Pedro Guijarro-Fuentes, Katrin Schmitz, Natascha Müller and the authors of individual chapters. All rights reserved. No part of this work may be reproduced in any form or by any means without permission in writing from the publisher. The policy of Multilingual Matters/Channel View Publications is to use papers that are natural, renewable and recyclable products, made from wood grown in sustainable forests. In the manufacturing process of our books, and to further support our policy, preference is given to printers that have FSC and PEFC Chain of Custody certification. The FSC and/or PEFC logos will appear on those books where full certification has been granted to the printer concerned. Typeset by Techset Composition India (P) Ltd, Bangalore and Chennai, India. Printed and bound in Great Britain by the CPI Books Group.

Contents

Figures and Tables Contributors

viii xiii

Introduction Pedro Guijarro-Fuentes, Katrin Schmitz and Natascha Müller The Present Book Summary of the Contributions Concluding Remarks 1

2

3

Finite Verb Placement in French Language Change and in Bilingual German–French Acquisition Anika Schmeißer and Veronika Jansen Introduction Previous Studies on Finite Verb Placement Predictions The Study: Data and Methods Results and Discussion Conclusion Wh-fronting and Wh-in-situ in the Acquisition of French: Really Variants? Jasmin Geveler and Natascha Müller Introduction Wh-questions in French The Acquisition Data Discussion On the Processing of Subject Clefts in English-French Interlanguage: Parsing to Learn and the Subject Relativizer qui Laurent Dekydtspotter and Kelly Farmer Introduction Parsing to Learn French Relativizer qui Elucidating the Learning Scenario v

1 1 3 8 10 10 14 20 21 27 39 43 43 43 50 61 66 66 68 74

vi

The Acquisit ion of French in Mult ilingual Conte x t s

The Study L2 Sentence Processing in Real Time Parsing and L2 Acquisition 4

5

6

7

Verbal Transitivity Development in First Language Acquisition: A Comparative Study of Russian, French and English Anna Frolova Introduction Null Objects in Adult Grammars Object Omissions in First Language Acquisition The Study Discussion Conclusions Static and Dynamic Location in French and German Child Language Anne-Katharina Harr and Maya Hickmann Introduction Methodology Results Discussion Can L2 Learners Learn New Ways to Conceptualize Events? A New Approach to Restructuring in Motion Event Construal Jeanine Treffers-Daller and Françoise Tidball Introduction Motion Event Construal in English and French Transfer and Simplification in the Expression of Path and Motion in L2 Learners of French Hypotheses of the Current Study Methods Results Discussion and Conclusion A Bidirectional Study: Is There Any Role for Transfer in Adjective Placement? Pedro Guijarro-Fuentes Introduction What Do We Know About Language Transfer? A Sketch The Study Concluding Discussion

78 86 88 94 94 95 100 104 111 116 118 118 126 129 140

145 145 153 158 162 163 166 178 185 185 186 192 209

Content s

8

9

Parameters, Processing and Feature Reassembly in the L2 French Determiner Phrase Julia Herschensohn and Deborah Arteaga Introduction French DP Morphology and Syntax Theoretical Framework The Study Discussion Conclusion

vii

215 215 216 218 224 229 232

Concluding Remarks Katrin Schmitz Introduction What Has Been Done Open Issues in the Contributions What Remains To Be Done

234

References Index

242 265

234 234 235 239

Figures and Tables

Figures Figure 1.1

Word order patterns in German adults’ main clauses

15

Figure 1.2

Language balance based on children’s verb lexicon acquisition in French and German

24

Figure 1.3

Word order in French LC

28

Figure 1.4

Word order in CdD

29

Figure 1.5

Word order in French LC

30

Figure 1.6

Word order in the bilingual children’s French main clauses

30

Figure 1.7

Word order in the monolingual children’s French main clauses

31

Figure 1.8

Word order in the bilingual children’s German main clauses

32

Figure 1.9

Word order in the monolingual children’s German main clauses

33

Figure 1.10 SV in German main clauses matched by MLU

34

Figure 1.11 Position of finite verbs in German subordinate clauses (Alexander)

35

Figure 1.12 Position of finite verbs in German subordinate clauses (Céline)

36

Figure 2.1

Absolute numbers of different types of wh-questions in adult speech addressed to children

50

Figure 2.2

Mean MLU difference in the bilingual children

58

Figure 2.3

French questions in all analyzed subjects

58

viii

Figures and Tables

ix

Figure 2.4

French wh-fronting and wh-in-situ compared for all analyzed subjects

59

Figure 2.5

Development of questions in French over time: Emma

59

Figure 2.6

Development of questions in French over time: Alex

60

Figure 2.7

Development of questions in German all subjects

61

Figure 2.8

Words per minute of the bilingual children in German and French at the age of four

64

Figure 3.1

Partial representation anticipating an operator-variable structure in the processing of subject clefts

73

Figure 3.2

Expected results in msec under structurally governed processing if facilitation is stronger for immediate superordinate categories versus opposites

77

Figure 3.3

Expected results in msec under structurally governed processing if facilitation spreads to all human categories

77

Figure 3.4

Expected results in msec under shallow structure processing if facilitation is stronger for immediate superordinate categories versus opposites

78

Figure 3.5

Expected results in msec under shallow structures if activation spreads to all human categories

79

Figure 3.6

Sample quadruple items

80

Figure 3.7

Results in msec from picture classification pretest task with masked primes

83

Figure 3.8

Profile for human probes in msec: Cross-modal priming experimental task

84

Figure 3.9

Profile for [± human] probes in msec: Cross-modal priming experimental task

86

Figure 5.1

Information locus in the static task: French

129

Figure 5.2

Information locus in the static task: German

130

Figure 5.3

Information locus in the dynamic task: French

131

Figure 5.4

Information locus in the dynamic task: German

133

Figure 5.5

Static task: Semantic components expressed in French verbs

134

Figure 5.6

Static task: Semantic components expressed in other French linguistic devices

134

x

The Acquisit ion of French in Mult ilingual Conte x t s

Figure 5.7

Static task: Semantic components expressed in German verbs

135

Figure 5.8

Static task: Semantic components expressed in other German linguistic devices

136

Figure 5.9

Dynamic task: Semantic components expressed in French verbs

137

Figure 5.10 Dynamic task: Semantic components expressed in other French linguistic devices

137

Figure 5.11 Dynamic task: Semantic components expressed in German verbs

138

Figure 5.12 Dynamic task: Semantic components expressed in other German linguistic devices

139

Figure 6.1

The transfer scenario (A)

151

Figure 6.2

The restructuring scenario (B)

151

Figure 6.3

The creative or hybrid scenario (C)

152

Figure 6.4

The convergence scenario (D)

152

Figure 6.5

Path verbs, deictic verbs, verbs of caused motion and manner verbs (tokens) as a percentage of all motion verb tokens in the speech of learners and native speakers

169

Frequency of four motion verbs among learners and native speakers of French (percentage of total number of motion verbs per group)

171

Different ways to verbalize the boundary crossing event by learners and native speakers (percentages of all expressions used)

172

Figure 6.6

Figure 6.7

Figure A6.1 Erfolglose Anbiederung ‘Unsuccessful ingratiation’

183

Figure A6.2 Unbeabsichtigte Helden ‘Unintentional heroes’

184

Figure 7.1

GJT results

201

Figure 7.2

GJT results

203

Figure 7.3

SCT results

206

Tables Table 1.1

Historical data considered in the present study

22

Table 1.2

Overview of the children analyzed

25

Figures and Tables

xi

Table 1.3

Absolute number of sentence-initial XPs in V2 and V3 main clauses in CdD

29

Table 2.1

Relative frequencies for (WHQ) in previous studies

49

Table 2.2

Analyzed subjects

57

Table 3.1

Reaction times in milliseconds for human probes: Cross-modal priming task

85

Table 3.2

Reaction times in milliseconds for [± human] probes: Cross-modal priming task

86

Table 4.1

Distribution of responses by type of object in the referential context in Russian

108

Table 4.2

Distribution of responses by type of object in the referential context in English

109

Table 4.3

Distribution of responses by type of object in the referential context in French

109

Table 4.4

Distribution of responses by type of object in the non-referential context in Russian

110

Table 4.5

Distribution of responses by type of object in the non-referential context in English

110

Table 4.6

Distribution of responses by type of object in the non-referential context in French

110

Table 5.1

Semantic components expressed in descriptions of static and dynamic location

128

Table 6.1

Overview of informants in the current study

163

Table 6.2

Language proficiency of the learners of French, in comparison with native speakers

164

Table 6.3

Manner verbs (types and tokens) used by learners and native speakers of French found in both stories

167

Table 6.4

Verbs of caused motion (types and tokens) used by learners and native speakers of French

168

Table 6.5

Overview of intergroup comparisons for all motion verbs in the two stories

169

Table 6.6

Overview of intergroup comparisons of the boundary crossing event

174

Table 7.1

Results from the placement task

198

Table 7.2

GJT (intergroup comparisons)

202

xii

The Acquisit ion of French in Mult ilingual Conte x t s

Table 7.3

GJT (intergroup comparison)

204

Table 7.4

GJT (intragroup comparison)

205

Table 7.5

Individual performance

205

Table 7.6

SCT (intergroup comparison)

207

Table 7.7

Group differences between PreAdj vs. PostAdj

208

Table 7.8

Individual performance

208

Table 7.9

Number of participants who performed above chance in the GJT/SCT tasks

209

Table 8.1

Characteristics of subjects

225

Table 8.2

Gender assignment to determiner, three interviews

227

Table 8.3

Gender concord ADJ, three interviews

228

Table A8.1

Suppliance of determiner, determiner accuracy, three interviews

233

Table A8.2

Distribution of determiner error types over three interviews

233

Table A8.3

Number assignment, three interviews

233

Contributors

Deborah Arteaga, full professor in linguistics at the University of Nevada, Las Vegas, NV, holds an MA in French linguistics from the University of Colorado and a doctorate in Romance linguistics from the University of Washington. She has published extensively in several tier-one journals (e.g. The Modern Language Journal, The French Review, Journal of French Studies), many coauthored with Julia Herschensohn. Her publications focus on second language acquisition of Spanish and French, as well as on historical Romance syntax. Laurent Dekydtspotter is professor of French linguistics and second language studies at Indiana University, Bloomington, IN. Models of second language acquisition are his central research interest. His research focuses on problems of learnability, the development of semantic interpretation and the role of the processing system in second language acquisition, with special reference to French. Kelly Farmer holds a PhD in French linguistics from Indiana University, Bloomington, IN. She is the supervisor of the French Language Program at the University of Georgia. Her research interests include sociolinguistics, second language acquisition and applied French linguistics. Anna Frolova completed her PhD in French linguistics at the University of Toronto, Canada in 2014. She studies the syntactic acquisition of verbal transitivity within a generative grammar framework and, in particular, an optional direct object omission stage across child languages, such as French and Russian. Jasmin Geveler is a research assistant and PhD student at the Bergische Universität Wuppertal. Her research deals with the acquisition of syntactic structures in multilingual children with respect to derivational economy and complexity. In her PhD thesis she investigates the bilingual acquisition of wh-questions. xiii

xiv

The Acquisit ion of French in Mult ilingual Conte x t s

Pedro Guijarro-Fuentes is a Profesor Contratado Doctor in Spanish at the Universitat de les Illes Balears, Palma, Spain. His research interests are in the interdisciplinary field of Spanish applied linguistics, sociolinguistics, psycholinguistics and bilingualism. His main current research projects include, among others, the study of copula verbs ser/estar, spatial prepositions, DOM, word order, tense and aspect projects funded by the British Academy, Art and Humanities Research Council, the Deutsche Forschungsgemeinschaft (DFG), and more recently by the Spanish Minister of Economy and Enterprise and the European Commission under the Horizon2020 Research and Innovation Framework Programme. He has authored and co-authored numerous articles and book chapters, and his research has appeared in different international refereed journals such as Language Learning, Bilingualism: Language and Cognition, Cognition, Studies in Second Language Acquisition, Applied Psycho-linguistics, Applied Linguistics and the International Journal of Bilingualism, among others. Anne-Katharina Harr is a research assistant at the Department for German as a Foreign Language (DaF) of the Ludwig-Maximilians-Universität, Munich. Her research focuses on the comparison of language acquisition across languages, including first and second language acquisition and bilingualism, as well as typical and atypical learning situations. In her PhD, which she received from the University of Paris 8 and the Ludwig-Maximilians-Universität, she examined typological constraints in the development of spatial language in German and French adults and children. Currently, she is realizing a project on the language acquisition of blind children. Julia Herschensohn, professor of linguistics at the University of Washington, Seattle, WA, does research on second language acquisition and larger questions of bilingualism, age effects and the processing of first and second languages. She has published seven books, three single-authored and four edited volumes, including the Cambridge Handbook of Second Language Acquisition. She has written dozens of articles, many dealing with French morphosyntactic acquisition, development of second language processing (e.g. agreement, clitic pronouns), and Old French syntax. Maya Hickmann received her PhD from the University of Chicago, IL. She was a member of the Max-Planck Institute for Psycholinguistics (Nijmegen), then of two psychology labs (at the Université René Descartes). She is presently research director at the CNRS in France and co-directs a linguistics lab (Laboratoire Structures Formelles du Langage, Université de Paris 8). Her publications focus on structural and functional determinants of language development, typological constraints on first and second language acquisition, and the relationship between language and cognition. She is chief editor

Contr ibutors

xv

of the journal Language, Interaction and Acquisition (John Benjamins), dedicated to language acquisition. Veronika Jansen was a research assistant for a study of early bilingual language acquisition at the Bergische Universität Wuppertal. Her PhD thesis deals with the acquisition of dislocations in French and German by monolingual and bilingual children. Natascha Müller holds the chair of Romance linguistics at the Bergische Universität Wuppertal. Her research focuses on the syntax of Romance languages and its acquisition. Currently, she works on the simultaneous acquisition of three first languages. Anika Schmeißer is a research assistant and PhD student at the Bergische Universität Wuppertal. Her research focuses on the syntactic and discoursepragmatic developments of Romance languages. In particular, she investigates the historical development of finite verb placement, the acquisition of tense by young children, and child bilingualism. Katrin Schmitz is Junior (i.e. non-tenured) professor of Romance linguistics at the University of Wuppertal. Her research focuses on the development of French, Italian and Spanish, both in a synchronic (in particular language acquisition) and a diachronic perspective. Her current research focus concerns the heritage languages Italian and Spanish spoken in Germany, in particular the phenomena of argument omission, syntax of clitics and differential object marking. Francoise Tidball was senior lecturer in French and linguistics at the University of the West of England, Bristol, until her retirement in 2009. Her main area of research is second language acquisition, in particular measuring vocabulary acquisition in English-speaking learners of French. Jeanine Treffers-Daller is professor of second language education and director of the Centre for Literacy and Multilingualism (CeLM) at the University of Reading. She uses corpora from bilinguals and second language learners to study how they talk about motion through space, and to find out to what extent they differ from monolinguals in their use of conventional (native-like) patterns. She is currently interested in finding out to what extent classroom learners of French and English can engage in statistical learning, and how corpus data can be used to provide evidence for this kind of learning.

Introduction Pedro Guijarro-Fuentes, Katrin Schmitz and Natascha Müller

The Present Book Spoken by about 110 million people as a mother tongue, French is one of the most important languages in the world, occupying the 11th place on the list of world languages (Crystal, 1997: 288). French is the official language of many countries, such as Belgium, Canada, France, Switzerland and Togo, and its varieties are well studied synchronically as well as diachronically (see, for example, Posner, 1997). Recent years have seen much important work on the acquisition of various languages, among them Romance languages, with respect to the (re)setting of parameters, (in-)complete acquisition, functional categories, features and feature values, and latterly the study of interface phenomena (see, for example, Guijarro-Fuentes & Rothman, 2012; Montrul, 2011, among others). The Romance language which has been the most studied in different constellations is Spanish, mainly as a heritage language in the United States. Many studies have focused on cross-linguistic influence and the effects of language dominance involving various language combinations. The edited volume by Prévost and Paradis (2004) and, more recently, the authored book by Prévost (2009) have bridged the knowledge gap for research on normal and specifically impaired monolingual, simultaneous bilingual (2L1) and second language (L2) acquisition of French with a focus on inflectional morphology and functional categories. Yet there remains much to do. Among these issues is the acquisition of French in combination with languages other than English. The present book will fill this gap, as French acquired together with German or with other Romance languages is emphasized. Furthermore, only a limited number of grammatical phenomena have been studied, the clause being the focus of previous research (such as verb position, object clitics, among others) with respect to different acquisition types to date. The present book will add information about other grammatical domains, like wh-in-situ, subject clefts, verbal transitivity, static and dynamic location, motion event 1

2

The Acquisit ion of French in Mult ilingual Conte x t s

construal, and DP-related domains such as adjective placement and gender and number concord. This edited volume brings together new research from all relevant theoretical paradigms addressing the acquisition of French in various language combinations in bilingual constellations (2L1 and L2) or as a single L1 as compared to other languages, in order to contribute to the general understanding not only of the studied phenomena but also of the particular properties of French by integrating typologically oriented studies. Another task is to integrate further phenomena, in particular some which involve semantic and pragmatic features (either concentrated or in combination with morphosyntactic features). A further set of questions addresses the relationship of features in different acquisition processes, the role of language-external factors for acquisition processes such as frequency (of lexicalization patterns), cognitive (processing) properties and the effects of being bilingual. Recently, much attention has been paid to the role of (LF-interpretable and non-interpretable) features and their combination in the acquisition of the elements bearing them. LF-interpretable features are semantically relevant features (e.g. [telicity] for verbs indicating an event with an inherent goal as in traverser la rue ‘to cross the road’, which is only accomplished when the referent crossed the street). Non-interpretable features, on the contrary, are merely formal features (e.g. person and number on the verb establishing congruency with the subject as in Les enfants jouaient dans la cour ‘The children were playing in the courtyard’). Furthermore, it is useful to study the relationship between acquisition processes and language change (and the related language-internal and external factors) in order to yield a deeper understanding of the evolution of the syntactical structures of French. The aforementioned tasks have been addressed by the papers in the present volume which may thus be understood as a ‘crossroads’ of theoretical and empirical work. The unexpected findings from the studies included in this volume do not support those of previous research and try to further answer crucial questions of formal linguistics such as What constitutes linguistic knowledge? How is such knowledge learned/acquired? (Rothman & Guijarro-Fuentes, 2012: 5) and to explain the observed developmental asymmetries in the types of acquisition processes and grammatical phenomena studied. The contributions, presented in more detail below, basically focus on the acquisition of three domains: first, the highest functional sentence level, the complementizer phrase (CP) with verb placement (V2) and sentence types (subject clefts such as C’est Paul qui est venu ‘It’s Paul who came’ and whquestions as, for example, Qui a vu mon stylo? ‘Who saw my pen?’); secondly, the verbal domain with a focus on the semantics of verbs and the development of transitivity; and finally, the nominal domain – more precisely the determiner phrase (DP) with gender and number concord as well as adjective placement. Most of them involve an intricate interplay of grammatical features which, in the case of bilingual language acquisition, represents an

Introduc t ion

3

even greater challenge than one might be inclined to assume if different feature values from the two languages come into play and provoke crosslinguistic influence (see also the criteria for the probability of cross-linguistic influence by Hulk & Müller, 2000; Müller & Hulk, 2000, 2001). Studies on 2L1 acquisition try, among other aims, to disentangle the effects generated by the properties of the involved languages in interface phenomena from other aspects of bilingualism (e.g. processing and language dominance) in explaining developmental delays as well as sprints in bilingual children. In child and adult L2 acquisition, interface phenomena are studied against the background of their learnability up to ultimate attainment at (almost) native speaker level (see, for example, Montrul, 2011; Slabakova, 2008, 2010), without neglecting the other important research questions mentioned above. The papers in the present volume study the involved phenomena in depth from the perspective of the different descriptive levels. The comparison of the phrase level (in particular the DP as subject or object) with the predicate level (verbal domain or VP) and the sentence level (CP) allows us to compare and to understand different interplays of features and also intraand extra-linguistic factors facilitating or complicating the acquisition task investigated. In bilingual constellations, we can observe the different degrees of cross-linguistic influence. The use of various methodologies such as elicitation tasks, longitudinal studies and parsing experiments further adds to our understanding of how French is acquired in different contexts. Next we will summarize the articles that comprise this collection. They are ordered in terms of domains. The studied phenomena can be roughly associated with CP-related, VP-related and DP-related phenomena as mentioned above. Within each section, first (L1 and 2L1) language acquisition precedes L2 acquisition.

Summary of the Contributions When children begin to acquire their mother tongue(s), they are constrained by limitations of their working memory, perceptual limitations, by the linguistic complexity of the grammatical devices which have to be acquired, and so on. The latter is the focus of all contributions of the present volume, although acquisition of linguistic devices is not treated without recourse to the other capacities. Most research has shown that L1 learners find it particularly difficult to acquire linguistic devices which are related to the functional category C. Indeed, C encodes many language-particular properties and interfaces with other parts of the human knowledge system. The first set of papers deal with CP-related phenomena, starting with the studies using data from 2L1 acquisition with French as one of the languages. Bilingualism is a challenge for cognitive processing, since two languages have to be acquired and linguistic processing has to seek a separate, language-particular linguistic output.

4

The Acquisit ion of French in Mult ilingual Conte x t s

Schmeißer and Jansen’s chapter investigates the language change process of historical French, which once exhibited V2 properties, to Modern French, which is mainly SV-ordered. Adopting Kroch’s (2001) theory, the authors assume that the language change process is driven by intra-individual grammar competition, which itself is contact-induced by nature. For historical French, Schmeißer and Jansen put forward that such a language contact situation with a non-V2-language, namely English, was indeed plausible in former times in France, which must have yielded for the involved generations a bilingual-like coexistence of [±V2]-properties. Following that line of reasoning, they looked into the acquisition of finite verb placement in bilingual children acquiring German (i.e. a V2-language) and Modern French (i.e. a non-V2-language). The study uncovers an increase in SV production from Old to Middle to Modern French, based on the analysis of eight texts produced between the 12th and the 20th centuries. Furthermore, GermanFrench bilingual children generate more SV-structures in German when compared to L1 monolingual children as the authors show in their analysis of longitudinal data of four German-French bilingual and two monolingual children of French and German, respectively. This resemblance implies that language contact between a V2-language and a non-V2language might result in a more frequent use of SV in the V2-language and thus to a preference of SV over V2. The authors argue that this effect may be due to structural priming, which influences the linguistic performance and/ or competence of bilingual speakers. Nevertheless, the strength of such a priming effect in the influenced language hinges on the proficiency level achieved in the prime language. Future research is necessary in order to reveal what ‘highly proficient’ really means and whether there is a threshold for the priming effect to occur or whether this effect is gradual and depends on the gradually higher proficiency in the prime language (French in the present case), by examining many more individuals. Schmeißer and Jansen further presuppose that this type of language interaction in the bilingual individual, which is firstly related to language use, may eventually turn into language change, a qualitative change of the competence system. The coexistence of different variants of wh-questions within one variety of French is addressed in Geveler and Müller’s chapter. More specifically, it deals with French wh-fronting and wh-in-situ in spoken French. The authors depart from the idea that in the speech of adults to children, wh-insitu questions reign. Monolingual French children employ this type of question frequently when their language development is analyzed longitudinally. Nonetheless, the question which then arises is how bilingual children would acquire the two types of questions in French if the other language only displays one option, namely wh-fronting (as far as the questions contain a wh-phrase). To that end, Geveler and Müller study the acquisition of French questions by six German-French bilingual children longitudinally from the age of 1;6 to 5 years. The main result is that the bilingual children abuse

Introduc t ion

5

wh-fronting in French questions, whereas the wh-phrase almost never remains in situ in German questions. Based on the criteria for cross-linguistic influence (cf. Müller & Patuto, 2009), Geveler and Müller suspect the mere occurrence of the influence and hypothesize that German will be the influencing language, and not French. As in the previous study (which relies on a subset of the children analyzed here), the effect of the influence is related to the proficiency or readiness to speak the other language, in their case German: namely, lesser proficiency in the influencing language (i.e. German) ensures a lower probability to overuse a syntactic derivation of this language in the influenced language (i.e. French). Both papers dealing with German-French children thus show an impact of phenomenon-external factors, namely the fact of being bilingual and the proficiency in the involved language. The concluding chapter in this section is by Dekydtspotter and Farmer. Long is the debate about whether L2 acquisition of adults is constrained by universal grammar (UG), and if so how UG is mediated (e.g. via the L1). Researchers have often relied on data which are grammatical or ungrammatical with respect to the target grammar. Dekydtspotter and Farmer analyze a phenomenon-external but possibly crucial factor, namely the processing of subject clefts in English-French interlanguage. In sentence processing, the parser is fed by a grammar. In other words, the architecture of the grammar, more specifically the current grammar of a particular language assumed by L2 learners, can be deduced from observing how subjects process sentences of that language. If there is no difference detectable in sentence processing between natives and L2 learners, then this may constitute an argument in favour of the assumption that, all things being equal, UG constrains grammar in natives as well as in L2 learners. Dekydtspotter and Farmer tread this path. Taking into account the emergence of subject clefts headed by the relativizer qui in the speech of intermediate L2 French learners, the authors examine aspects of the processing of subject clefts by intermediate (n = 30) and advanced L1-English L2-French learners (n = 28) and French native speakers (at the same time being advanced English learners, n = 22) employing a sophisticated cross-modal priming method. Similar priming patterns in all groups indicate that L1-English L2-French learners processed operator-variable structures in real time which reflects that true parsing, namely domain-specific, was available to them (taken as superior to domain-general learning). Dekydtspotter and Farmer’s findings indicate that UG-constrained parses are available as triggers in the acquisition of the subject relativizer qui and allow conclusions for the grammar model assumed where parsing is the union of UG and the parametrized lexicon. All three chapters, as they address variation in the language learner, come to the conclusion that language acquisition theory can only advance if more than the mere acquisition of linguistic devices is taken into account. Still analyzing children’s utterances at the clause level, this path is also taken by

6

The Acquisit ion of French in Mult ilingual Conte x t s

Frolova in the chapter entitled ‘Verbal transitivity development in first language acquisition’. She compares three languages (Russian, French and English) and aims at an expansion on previous research on null objects in child Russian by comparing child- and adult-elicited production data to French and English experimental data in previous recent studies. A modular account including syntactic, pragmatic and semantic factors which in particular distinguishes different types of null objects is employed to the Russian data to account for different types of object omissions found in adult speech, and to forecast L1 child developmental acquisition patterns. The main findings of the actual experiment with three different groups of children (age 3, n = 16; age 4, n = 10; age 5, n = 13) demonstrate that young monolingual Russian children pass through an omission stage, during which they omit direct objects more frequently than adult participants in optional contexts. Based on that, it is presumed that in early stages Russian children, like their French and English peers, may exercise a default cognate null object available in adult grammar. During development they confine the use of this default null object to non-referential contexts. Frolova concludes that the acquisition of Russian grammatical aspect (reinterpreted here as the LF-interpretable feature of perfectivity) may constitute an important factor for object omission in the developing grammars which deserves further study for a better understanding of omissions in child language. The comparison with child French is essential since it underlines the child’s universal tendency to overextend a null object option available in the adult language. The interaction of the acquisition of linguistic devices with typological aspects of language is addressed in the chapter by Harr and Hickmann entitled ‘Static and dynamic location in French and German child language’. The authors take as a point of departure several cross-linguistic studies (e.g. Bowerman & Choi, 2003; Hickmann et al., 2009; Slobin, 2003, among others) according to which lexicalization patterns affect the way children acquire target-like linguistic devices to express certain spatial information. The theoretical background is the assumption of typological differences (Talmy, 1985, 2000a, 2000b) between languages, which determine how speakers, in the case of the present empirical study, French and German, convey location and changes of location. The general assumption examined is that German and French speakers of all ages centre on different types of information because they count on divergent, language-type specific lexicalization patterns to express spatial information. A group of adults and monolingual children (4 and 6 years of age) took part in two tasks in which they had to verbally localize entities (static task) and describe object displacements (dynamic task). The results from both tasks indicate that speakers elect to express different types of information, mostly relying on verbs in French and on satellites in German. French adults typically use verbs to encode manner of attachment whereas German speakers vastly utilize satellites to express path, sometimes with specific verbs providing various types of information

Introduc t ion

7

including posture and manner of causing displacements. Children, on the other hand, unmistakably adopt the pattern of their adult counterparts, even though they also display striking developmental progressions in both languages which become clear from age 4 onwards. Hence, these findings give support to the view that cognitive and typological factors both constrain how children construct the semantics of space during L1 acquisition. The study by Harr and Hickmann is nicely complemented by the chapter entitled ‘Can L2 learners learn new ways to conceptualize events? Evidence from motion event construal among English-speaking learners of French’, by Treffers-Daller and Tidball, which is related to the same theoretical background as the previous chapter, i.e. the typology of lexicalization patterns, associated with the distinction of verb-framed languages (the L2, French) and satellite-framed languages (the L1, here English). Specifically, the authors tested Slobin’s (1996) claim that L2 learners struggle with conceptual restructuring in L2 acquisition. The authors suggest that learners can find themselves in four different reconceptualization situations which measure the way from the L1 construal to the L2 construal: that is, transfer, restructuring, creative/hybrid and convergence, depending on the concept in question. The authors analyze the use and the frequency distribution of path, manner, caused motion and deictic verbs in narratives elicited from 20 intermediate and 21 advanced learners of French, together with 23 native speakers of French and 30 English native speakers, respectively. The results show different results and scenarios for the different learner groups: the intermediate learners’ productions were found to correspond to the creative/hybrid scenario because they contrasted significantly in their motion expressions from both English and French native speakers, except for path, which was verbalized in target-like ways from very early on. Advanced learners, by contrast, were able to reconceptualize motion in the L2, as far as manner and path are concerned, but showed difficulties with deictic verbs and caused motion. The authors found evidence for transfer from the mother tongue in verbalizations among intermediate level learners of events which involved a boundary crossing. The two chapters based on typological properties therefore show that the distinctions investigated are acquired early in the respective languages and can, albeit slowly (depending on the verb type), be acquired by L2 learners as well. Re-interpreting these results in terms of features, one may conceive of [path] as an LF-interpretable feature, whose coding has to be associated with different elements, namely the verbal root (French) and a PP-satellite (German), respectively. While the different monolingual child groups display early target-like feature assembly, the reassembly in L2-acquisition (in terms of Lardiere, 2009a, 2009b) takes more time and effort, but is possible. The next set of chapters deal with DP-related phenomena, more precisely adjective placement and agreement phenomena in the DP in different

8

The Acquisit ion of French in Mult ilingual Conte x t s

groups of L2 learners. Looking at transfer effects in a bidirectional study, Guijarro-Fuentes focuses on the syntactic and semantic correlates in the domain of adjective placement (i.e. pre- and post-nominal noun-adjective collocation), examining four experimental groups – i.e. French-speaking intermediate (n = 12) and advanced (n = 23) learners of Spanish compared to a native-speaker control group (n = 12), together with Spanish-speaking intermediate (n = 33) and advanced (n = 27) learners of French, for each target language compared to a native-speaker control group (n = 15). It is assumed that Spanish and French are by and large very similar, coming from a historically similar system. The results, nevertheless, reveal that L2 acquisition of adjective collocation is influenced by the semantic (dis)similarity of adjective collocation in the L1 and L2; more importantly, salient group and individual level differences depending on the particular L2 language were found, but mainly due to the level of proficiency in the target language. Hence, while it is shown that the acquisition of interpretable features (e.g. focus and contrast) in L2 Spanish is not totally trouble free, the acquisition of French by speakers of Spanish is less problematic and such non-native speakers have little or no difficulty in acquiring both uninterpretable and interpretable features as current generative accounts would advocate. In the same vein, Herschensohn and Arteaga provide a stimulating study in which ideas put forth in feature re-assembly (Lardiere, 2009a) are also put to the test. In particular, they examine the development of the DP in L2 French of three advanced Anglophone learners, considering gender and nominal agreement features, features not found in English, together with [definite], [count], features which do exist in English. 3000 tokens of DP from production data are examined, having been collected from three adult L2 learners over three successive interviews. The results suggest that parametrically different gender features and nominal agreement are available to adult L2 learners, but that L1 transfer nevertheless influences the application of gender assignment and concord, in spite of relatively high levels of accuracy. Moreover, it is also claimed that feature re-assembly in L2 French of features that exist in the L1 is not totally unproblematic, as simple transfer accounts would suggest.

Concluding Remarks To summarize, the present volume offers a series of new chapters studying different linguistic phenomena in L1, 2L1 and L2 acquisition. Investigating the role of cognitive concepts clearly contributes to our understanding of how grammar-internal and grammar-external knowledge interacts and how this interaction influences the developmental path observed by the linguist. The phenomena studied in the presented chapters mostly involve interpretable features, namely linguistic features related to syntax-discourse,

Introduc t ion

9

syntax-semantics and syntax-lexicon. Combined with crucial knowledge on language processing, cross-linguistic influence and language change, new interpretations of the findings have been made possible. In particular, the role of structural priming and language proficiency in bilingual language acquisition (both 2L1 and L2) and even in language change has been highlighted in several contributions. What unites the different studies is the generally easy access of such interpretable features and the surprising ease of their acquisition, although this path is not totally trouble free. This result will enable linguists to take a big step forward in answering the yet unsolved question of which features (uninterpretable versus interpretable) matter to the language acquisition task (Chomsky, 2007). Last, but not least, we would like to acknowledge the valuable contribution and time that all the external reviewers took to read and provide comments on all chapters. In alphabetical order, we would like to thank: Laia Arnaus Gil (University of Wuppertal/Germany), Dalila Ayoun (University of Arizona/USA), Laura Baunaz (University of Geneva/Switzerland), Eve Clark (University of Stanford/USA), Francesca Filiaci (University of Edinburgh/ UK), Alison Gabriele (University of Kansas/USA), Theres Grüter (University of Hawaii/USA), Tanja Ionin (University of Illinois/USA), Scott Jarvis (University of Ohio/USA), Imme Kuchenbrandt (University of Frankfurt, Main/Germany), Eric Mathieu (University of Ottawa/Canada), Asli Ozyurek (MPI Nijmegen/Netherlands), Philippe Prévost (University of Tours/France), Valentina Repetto (Naples/Italy), Leah Roberts (MPI Nijmegen/Netherlands), Petra Sleeman (University of Amsterdam/ Netherlands), Nelleke Strik (University of Toronto/Canada), David Stringer (University of Indiana/USA), Ianthi Tsimpli (University of Thessaloniki/ Greece) and Daniel Valois (University of Montreal/Canada). We are deeply appreciative of their professionalism and their commitment to ensuring the high degree of quality reflected in this edited volume. We would also like to thank David Singleton and Martisse Foster and the team from Multilingual Matters for assisting us in shaping this volume into its final form and David Hanrahan for carefully reading the whole manuscript and correcting our English. A big thank you to you all!

1

Finite Verb Placement in French Language Change and in Bilingual German–French Acquisition Anika Schmeißer and Veronika Jansen

Introduction The idea that language change (LC) is related to language acquisition (LA) has a long tradition and goes back to Neogrammarian works (cf. Paul, 1920) claiming that LC is mainly due to the child’s imperfect learning of the language s/he is exposed to. Research done throughout the last number of decades, however, shows that this assumption is now outdated. More recent approaches which assume a connection between LA and LC are consistent with the generative framework by Chomsky (1982 and subsequent works; cf. Adams, 1987; Clark & Roberts, 1993; Kaiser, 2002; Lightfoot, 1991; Roberts, 2007, among others), according to which children do not inductively learn every single input sentence, but acquire their mother tongue by setting various parameters provided by an innate human language capacity, universal grammar (UG). Following this view, children succeed in acquiring the grammar of their parents, although linguistic input is defective on several levels (cf. Chomsky, 1965), which is commonly referred to as the logical problem of LA. In this regard, the combination of UG and exposure to linguistic input enables the child to acquire a language deductively. Yet, strictly speaking, this idea of a deductive LA process does not allow for LC (cf. Clark & Roberts, 1993). For how can LC happen if children, continually and without effort, eventually acquire the target language, i.e. their parents’ grammars (cf. Niyogi & Berwick, 1995)? Clark and Roberts (1993) refer to this as the logical problem of LC. For parametric LC to occur under these circumstances, the only possible solution to this problem is that the children’s input must have changed in advance. Two possible scenarios have been put forward in the literature to model LC on this basis: the input must either (i) ostensibly 10

Finite Verb Pl acement in French L anguage Change

11

or (ii) actually diverge from the underlying representation of the target grammar. In the case of (i), the change of the input must be considered a surface phenomenon due to changes in phonetics or semantics, for example. These ‘pre’-changes do not affect the underlying syntactic structure, but may render it opaque. Opacity in the input can be understood in terms of ambiguity, in that this structure allows for both options of a parameter, as demonstrated in Examples (1a) and (2a) (the finite verb is marked in bold italic letters). (1) a. b. (2) a. b.

Der The Das The Le The *Le The

Junge boy Buch book garcon boy livre book

liest reads liest reads lit reads lit reads

das the der the le the le the

Buch. book Junge. boy livre. book garçon. boy

German is a V2-language with the finite verb always in second position. This is demonstrated by the German examples in (1), where either the subject (1a) or a non-subject constituent (here: the object) (1b) may precede the finite verb. This diverges from Modern French, a non-V2-language, which only allows the subject constituent to directly precede the finite verb (2a). Hence, sentence (2b) is ungrammatical. Yet, children being solely exposed to (1a) might either deduce a [+V2]- or a [-V2]-grammar, since the sentence is compatible with both underlying representations. The studies by Adams (1987), Clark and Roberts (1993) and Gibson and Wexler (1994), among others, adhere to this line of argumentation. Following these studies, the mentioned structural ambiguity is the crucial point which may allow for LC. The mechanisms which, in this approach, help the children to decide between the two options available are mainly based on principles of economy and the subset principle by Berwick (1985), according to which the language of the subset can be regarded as the default choice in LA (cf. Clark & Roberts, 1993; Gibson & Wexler, 1994). Valian (1990) and Kroch and Taylor (1997), however, prove this approach empirically wrong. This leads us to possibility (ii), which claims that a parameter can only be set in response to an unambiguous input sentence/structure, which serves as a trigger. In this respect, a child acquiring German on the basis of sentences like (1a) alone cannot make a decision concerning parameter setting (cf. Fodor, 1998a; Valian, 1990). Consequently, ambiguous sentences like those in (1a) and (2a) cannot be assumed to trigger LC either. Hence, children must be exposed to unambiguous input of the opposing parameter value in order to acquire a grammar that differs from their parents’. This implies that LC can exclusively occur in language contact situations in which the other language possesses the opposing parameter value. Only

12

The Acquisit ion of French in Mult ilingual Conte x t s

then will triggers in the input convey a different underlying structure, which is a necessary prerequisite to change the target grammar. Yet, Ferraresi and Goldbach (2003) doubt that LC is contact-induced by nature: referring to studies of bilingual (2L1) LA, the authors ask the question of why the second generation of speakers should necessarily end up with a changed grammar instead of becoming bilingual. This is indeed an important question which, however, need not contradict the assumption of contact-induced LC. Instead, bilingual LA might even be part of the whole LC process, as is outlined below.

Kroch’s (2001) syntactic diglossia and the bilingual individual According to Kroch (2001: 731) and his colleagues, it is precisely the additional acquisition of a ‘new’ parameter value which evokes LC. The new parameter value begins to compete with the ‘old’ one on an intra-individual level until one of them finally ‘replaces’ the other. Kroch (2001) describes this competition in the individual as ‘syntactic diglossia’ (SD), which is similar to the bilingual coexistence of two different grammars.1 Following Kroch (2001), Schmeißer (2011) models the entire LC-process, establishing three stages which are relevant for LC to occur on a competence level: (3)

Stage I: parameter value1 L2. This fits in Stage II of the SD model, where both variants co-occur. The concept of attrition, on the other hand, focuses on the changes in the L1 that have occurred during the transition from one stage to the other. Attrition hence implies some kind of LC. Indeed, some studies have already argued for a possible interrelation between attrition and LC (cf. Schmeißer, 2012; Thomason, 2001, 2003). Thomason (2001: 227), for example, describes attrition as contact-induced change which ‘[. . .] would be less likely to happen outside a particular contact situation’. Schmid (2010: 6) argues that such changes most likely occur ‘in lexical areas’. Yet the author also stresses the possibility of morphosyntactic changes in the L1 competence due to influence of the L2. This goes hand in hand with findings on word order changes in individual speakers or language communities whose L1 normally exhibits some non-SVO order, but yet, due to L2 influence, shows a preference for SVO (cf. Sorace, 2000; Thomason, 2001). Sorace (2000), who studies word order patterns in native Italian speakers with L2-English, explains this preference in terms of a destabilization of marked options. She observes that in the speakers’ L1 (Italian) some interpretable features are particularly prone to L1 attrition, for example those associated with the preverbal position, which can be occupied by non-subject and subject constituents alike if indicating topicality. As a result, an underspecification of this marked syntactic position can be found in multicompetent speakers, which on the syntactic level eventually expresses itself in the preference of SV-patterns. The results of studies on language attrition hence equally indicate that more economical syntactic structures are preferred over less economical ones, which is similar to our findings on bilingual LA. In addition to that, Schmid (2010: 6) argues that the grammatical structures of the two languages of a multicompetent speaker have to bear some resemblance in order to be susceptible to language influence on a morphosyntactic level. This is also a prerequisite for linguistic influence in the bilingual children, as argued above. Hence, there is another parallel in the two areas of intra-individual language contact. Last but not least, the prevailing literature on L1 attrition asserts that influence from an L2 onto the L1 is most likely when multicompetent speakers master the respective L2 at a high proficiency level (cf. Sorace, 2000). In conclusion, this leads to the overall assumption that structural priming, as found in the bilingual child, may indeed lead to LC in the domain of finite verb placement.

Conclusion In this chapter, we wanted to reveal the mechanisms prevailing in Stage II of the LC-process in terms of SD, finally leading to a perceptible

40

The Acquisit ion of French in Mult ilingual Conte x t s

change in the underlying system (transition from Stage II to III). We predicted that the interaction of two languages during bilingual first LA equals the assumed grammar competition at Stage II and might explain its mechanisms. To this effect, we compared historical French data with data from bilingual German-French children in the domain of finite verb placement. We argued that, from a theoretical point of view, the existence of V3-patterns found in historical French cannot falsify the hitherto followed assumption of Old and Middle French exhibiting [+V2]-properties. Similarly, we illustrated that language contact with a non-V2-language might indeed have been plausible in Medieval France. This allows for a comparison of grammar competition in historical French with the bilingual acquisition of finite verb placement in German/French. We focused on quantitative observations in LA and LC, namely the increase of SV, which manifests itself in both processes. With regard to LA, those children who master the Romance language at a high proficiency level (i.e. balanced or French-dominant bilinguals) in particular, show a lasting SV-increase and exhibit a high proportion of SV in main clauses even in their highest MLU phases. With respect to subordinate clauses, we demonstrated that Alexander (one of these children) exhibits target-deviant word order. We concluded from this that the Modern French influence must affect the children’s German language competence in the domain of finite verb placement. Suggesting structural priming as the driving mechanism of this influence, we succeeded both in accounting for the existing V2-utterances in the bilingual children’s root clauses and for the observed individual variation among the children investigated. We argued that the French IP functions as a prime during language processing, to the extent that it is hardly inhibited in the bilingual children’s German utterances, leading to SV-IP/V2-IP-structures in their German. The higher the children’s proficiency is in French, the stronger the priming effect is. Therefore, balanced and French dominant children should be particularly prone to this kind of influence. Arguably, it is exactly these children’s underlying representation of German which has been altered. Structural priming might hence result in both: a change in performance and in competence, depending on the degree of proficiency in the respective prime language. Applying these results to LC, we suggest that structural priming is a driving mechanism at Stage II, characterized by SD, and which might eventually lead to Stage III. Assuming that bilingualism in historical France was most likely of an L2 nature, we conclude that the speakers’ multicompetence might have led to structural priming of L2 structures onto the L1 grammar. In this respect, historical French might have lost its [+V2]-property due to structural priming from an SV language. The extensive contact with historical English or Anglo-Norman during the Middle French period might be a possible source for this scenario.

Finite Verb Pl acement in French L anguage Change

41

Historical sources caq

Caquot, A. and de Robert, P. (eds) (1994) Les Livres de Samuel. Geneva: Labor et Fides. CdD (1405) de Pizan, C. Le Livre de la Cité des Dames. In E.J. Richards (ed.) and P. Caraffi (trans.) (1997) La città delle dame/Le Livre de la Cité des Dames. Mailand: Luni Editrice. hon (1570) La Bible qvi est toute la Sainte Escriture: contenãt le Viel & Nouveau Testament. Ov, La Vieille & Nouvelle Alliãce. Auec argumens sur chacun liure. Pour Sebastien Honorati. Lyon. mar (1736) La Sainte Bible qui contient le Vieux & le Nouveau Testament: revûe sur les Originaux & retouchée dans le language. Avec petites notes par David Martin. In C. Marot and T. Bèze (eds) Basle: chés Jean Rudolphe Imhoff avec les caractères de J.H. Decker. 1736. qlr (1170) Li quatre livre des reis. Die Bücher Samuelis und der Könige in einer französischen Bearbeitung des 12. Jahrhunderts. Nach der ältesten Handschrift unter Benutzung der neu aufgefundene Handschriften. In: E.R. Curtius (ed.) (1911) Dresden: Gesellschaft für Romanische Literatur. queste (13th century) Pauphilet, A. (ed.) (1949) La Queste de Saint Graal, roman du XIIIe siècle. Paris: Champion. reg (1494) Le premier volume de la bible en francois. In: F. Regnavlt (ed). (1520) Paris. spt (13th century) Les sept Sages de Rome: roman en prose du XIIIe siècle; d’après le manuscrit no. 2137 de la B.N., C.R.A.L.; Nancy (1981). vil (13th century) de Vileharduyn, J. La Conqueste de Costentinoble: d’après le manuscrit no. 2137 de la B.N., C.R.A.L.; Nancy (1978).

Notes (1) It is possible that the competing grammars presumed by SD may diverge in one parameter value only. (2) The data stem from research done by Leray (see, for example, Leray, 2009, which focuses on other aspects of the same data). (3) Apparent exceptions to this are cases of inversion such as, for example, complex inversion, subject-clitic inversion and stylistic inversion, which primarily occur in interrogatives. The finite verb is then preceded by a [+wh]-element or a restricted set of adverbs. For a detailed description and a syntactic analysis of these Modern French types of inversion, see Roberts (1993). (4) The concrete syntactic analysis of child utterances with postverbal subjects is far from being uncontroversial and will not be discussed in detail here. (5) In this respect, we would like to point to the work by Roeper and Weissenborn (1990), who claim that embedded as opposed to root clauses reflect pure parameter values, as exceptions to the respective parameters are hardly (if ever) found in this domain. (6) In the present chapter, the term historical French generally covers the period of Old and Middle French (i.e. up to approximately 1600). This is legitimate, as we will

42

The Acquisit ion of French in Mult ilingual Conte x t s

show that the change in finite verb placement occurs from Middle to Modern French. Since we want to focus especially on the transition of a V2-to a non-V2-language (i.e. from Middle to Modern French), it is justifiable to refer to Old and Middle French with the term historical French whenever alluding to the former V2-properties of the French language. (7) Most often, language contact situations have been associated with ‘imperfect’ second language learning by adults (cf. Thomason, 2001: 66). (8) The data stem from three different research projects: 1999–2005: ‘Frühkindliche Zweisprachigkeit: Italienisch-Deutsch und Französisch-Deutsch im Vergleich’, Universität Hamburg; 2005–2008: ‘Die Architektur der frühkindlichen bilingualen Sprachfähigkeit: Italienisch-Deutsch und Französisch-Deutsch in Italien, Deutschland und Frankreich im Vergleich’, Bergische Universität Wuppertal; and 2009–2013: ‘Code-switching bei bilingual aufwachsenden Kindern in Deutschland, Italien, Frankreich und Spanien: Italienisch-Deutsch, Französisch-Deutsch, Spanisch-Deutsch, Italienisch-Französisch, Italienisch-Spanisch, FranzösischSpanisch’, Bergische Universität Wuppertal. All projects have been financed by the German Science Foundation (DFG) and have been directed by Natascha Müller. For more information, cf. Cantone et al. (2008), Müller et al. (2015) and Müller et al. (2011). The main research results of the WuBiG are presented in Hauser-Grüdl et al. (2010) and Schmeißer et al. (2015). (9) The historical data which originate from other studies do not exclude these particles and clitics. Notwithstanding, they have been excluded from the data of the present analyses, as they change during the late Middle French period and therefore need to be regarded separately. Schmeißer (2011), however, shows that the distribution of word order patterns is not affected by the inclusion or exclusion of these particles, so that comparability is guaranteed in any case.

2

Wh-fronting and Wh-in-situ in the Acquisition of French: Really Variants? Jasmin Geveler and Natascha Müller

Introduction When asking a question, a speaker of French has several possibilities, among them a wh-question. French wh-words can front to a clause-peripheral position or be positioned in situ. Most syntactic analyses assume that wh-fronting implies movement of the wh-word to the front of the clause, while the wh-word remains in situ in a wh-in-situ question. Wh-in-situ is a familiar phenomenon which has invited a lot of researchers to a discussion about the optionality of movement in French. The second section will summarize this discussion and outline quantitative aspects of the two types of question. Not much is known about how children, especially bilinguals, acquire wh-questions in French. Previous studies on monolingual and bilingual French will be summarized in the third section, which is also devoted to the present study, i.e. how bilingual French-German children acquire whfronting and wh-in-situ in French. The fourth section will discuss the main results of the present study, namely that there is cross-linguistic influence in bilingual acquisition of wh-questions and that wh-fronting is overgeneralized in the French of the bilingual children.

Wh-questions in French Optionality of wh-movement in French? It is well known that French allows two types of wh-questions: one type exhibits fronting or movement of the wh-phrase (WhP) as in (1a); the other allows the wh-phrase to remain in situ as in (1b). Apart from these constructions, French possesses wh-fronting with subject-verb inversion (1c) and wh-fronting with the paraphrase est-ce que, as in (1d). 43

44

The Acquisit ion of French in Mult ilingual Conte x t s

(1) a. [A qui] WhP elle prête sa carte bancaire? To whom she lends her bank card b. Elle prête sa carte bancaire [à qui] WhP? c. [A qui] WhP prête-t-elle sa carte bancaire? d. [A qui] WhP est-ce qu’elle prête sa carte bancaire? The co-existence of wh-fronting and wh-in-situ has led linguists like Aoun et al. (1981) and Lasnik and Saito (1992) to propose a mixed interrogative system for French which combines the English and the Chinese interrogative system. While in English movement of the wh-phrase to the left periphery of the sentence is obligatory, as it is in German, the wh-phrase remains in situ in Chinese.1 Aoun et al. and Lasnik and Saito conclude that wh-movement is optional in French. The question of optionality of wh-fronting has dominated the field for years. Recently, Adli (2004) has defended the view that wh-in-situ interrogatives are a free variant of the wh-fronting type. This view stems from research which shows that both types demand similar cognitive costs in reading experiments (Adli, 2004: 97). Taking the perspective of the architecture of grammar in more general terms, the existence of syntactic options which provide the same interpretation is undesirable. Of course, speakers of French can express a question with both types of wh-questions. Notwithstanding, many linguists have expressed their doubts as to whether the two structures do not differ in subtle syntactic and/or pragmatic aspects. Two mainstreams exist here: one group of linguists argues that wh-in-situ is syntactically restricted. The other group believes that the difference between wh-fronting and wh-in-situ is related to the semantics–pragmatics interface. Linguists who advocate the view that wh-in-situ is syntactically restricted if compared with wh-fronting (Boeckx, 1999; Bošković, 1998, 2000; Chang, 1997; Mathieu, 2002) believe that quantificational elements (other than universal quantifiers) are excluded in wh-in-situ interrogatives in French, like chacun (2a, 2b), negation (2c), modals (2d), quantificational adverbs (2e) and negative quantifiers (2f). The following examples in (2) are taken from Chang (1997: 19): (2) a. *Chaque étudiant a rencontré qui? Every student has met whom b. *Chacun des étudiants a rencontré qui? Each of-the students has met whom c. *Il n’a pas rencontré qui? He NEGCL-has not met whom (NegCL = negative clitic) d. *Il peut rencontrer qui? He can meet whom e. *Il admire toujours qui? He admires always whom f. *Personne n’admire qui? Nobody NEGCL-admires whom

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

45

Bošković (1998) further assumes that French wh-in-situ is restricted to root clauses; it is excluded in embedded sentences, as shown in (3). (3)

*Jean demande à Siria elle a déjà vu quel film John asks (to) Siria she has already seen which movie

Unfortunately, it is far from clear whether these grammaticality judgements reflect the intuitions of native speakers of French. In this light, the observations by Adli (2006), Baunaz (2005, 2011) and Starke (2001) should be taken into consideration. These authors argue that wh-in-situ questions do occur in embedded clauses, with modals, in negative islands and with quantifiers of various types. The issue is far from clear since grammaticality judgements can vary according to language-external factors such as social class and lifestyle. For embedded clauses, Oiry (2011) shows that wh-in-situ appears in adult and child speech productively. She bases her results on an elicited production task. In other words, the non-occurrence of wh-in-situ in French embedded clauses is disproved. Notice that at least some native speakers investigated by Starke (2001), Baunaz (2005, 2011) and Adli (2006) use wh-in-situ productively in the above-mentioned cases which seems to weaken the validity of the syntactic constraints on wh-in-situ. Baunaz (2005, 2011) further argues that the speaker’s grammaticality judgements depend on: (i) the discursive contexts in which the wh-in-situ questions are used; (ii) the type of presupposition they entail; and (iii) the intonation they have. As a result, we can adhere to the following statement: the research literature on syntactic constraints on wh-questions in French seems to agree that it is wh-in-situ, not wh-fronting, which is restricted syntactically, if it is restricted at all. The syntactic constraints are far from clear and results seem to depend on methodological issues. Let us turn to those authors who affirm that the difference between wh-fronting and wh-in-situ must be described in terms of semantics or pragmatics. Again two major groups exist: Some linguists claim that both types of questions are not equivalent from a semantic perspective. They differ in regard to the presuppositions they express (Boeckx, 1999; Chang, 1997; Cheng & Rooryck, 2000, among others). Wh-in-situ questions exclusively appear in ‘strongly presupposed contexts’, while wh-fronted questions can occur in such contexts but are also used out of the blue. Based on the assumption that a presupposition cannot be cancelled (as opposed to an implicature), the test of negation distinguishes between presuppositionality and non-presuppositionality. Therefore, in contexts involving a presupposition the answer to the wh-question cannot be rien ‘nothing’ (Boeckx, 1999; Chang, 1997). Interestingly, although rien is not a possible answer to a wh-insitu question (which is strongly presuppositional (4a)), it is a possible answer to a wh-fronted question as in (4b). Example (4a) presupposes that Mary bought something and asks what. Example (4b) does not presuppose that

46

The Acquisit ion of French in Mult ilingual Conte x t s

Jean has bought anything at all. Rather, it asks whether Jean has bought something and what it is. (4)

Q: Marie a acheté quoi? A: ?? Rien. Marie has bought what? Nothing (Chang, 1997: 42) b. Q: Qu’ a acheté Jean? A: Un livre, une voiture, rien WHCL has bought Jean? A book, a car, nothing What has Jean bought? (WHCL = clitic wh-word) (Boeckx, 1999: 69) a.

Chang notes that wh-in-situ questions behave like wh-cleft sentences (cf. Boeckx, 1999). Based on the test of negation, both constructions presuppose that Mary has bought something and ask what. Therefore, they cannot appear out of the blue. Example (5) shows that a wh-cleft-sentence, like (4a), is not a possible question for the answer ‘rien’. (5) Q: C’est quoi que Marie a acheté? A: ?? Rien. It is what that Marie has bought? Nothing What is it Marie has bought?

(Chang, 1997: 42)

Boeckx (1999) even goes so far as to propose a syntactic analysis of wh-in-situ questions as disguised clefts. The observation that both structures can only appear in presuppositional contexts would fall out rather naturally. However, as with the syntactic constraints, the picture is far from clear. Hamlaoui (2008), Mathieu (2004) and Starke (2001), among others, claim that there are wh-in-situ questions which can express non-presuppositionality and thus can be used out of the blue. If this observation turns out to be correct more generally, it implicates that there is no relation between clefts and whin-situ (contra Boeckx). The question in (6) illustrates a wh-in-situ question which is not presuppositional. (6)

Fred and Hélène are good friends, and they chat on the phone about the latest gossip. After a while, Fred changes topics and asks: Fred: Ah, au fait, tu vois qui vendredi soir? (Baunaz, 2011: 10) Ah, by the way, you see whom Friday night? By the way who will you hang out with on Friday night?

Let us finally turn to intonation as a feature which was used in order to distinguish wh-fronted and wh-in-situ questions: The existence of a relation between presuppositionality and intonation has already been proposed in some studies. Cheng and Rooryck (2000) find that presuppositions occur coupled with a rising intonation; in other words wh-in-situ questions cannot be found coupled with downfall intonation. Intonation in wh-in-situ

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

47

questions is still a controversial issue today. For example, Adli (2004, 2006) and Mathieu (2002) find that there is much variation in intonational contours of interrogative clauses and conclude that there is no one-to-one relationship between wh-in-situ and a rising intonation. Starke (2001) claims that wh-in-situ questions which are extracted from weak islands occur with a ‘slight accent’ (slight fall rise). Due to the fact that there seem to be different types of wh-in-situ questions, Baunaz (2011, accepted) and Baunaz and Patin (2011) propose a tripartite typology of wh-in-situ questions. Every type is characterized by a different intonation and interpreted differently. Baunaz (2011) discriminates between presuppositional wh-in-situ questions in which the wh-word can have a partitive or specific reading and non-presuppositional wh-questions. A partitive wh-phrase describes an object which belongs to a presupposed set containing more objects and thus all these objects in the set are possible answers. Both dialogue partners have a set of alternative answers to the wh-question in mind. The interlocutor asks for the identification of one object from the set. In contrast, wh-in-situ questions which involve a specific wh-phrase are used when the interlocutor has one familiar entity in mind. Hence, there is only one possible answer, the identification of the referent for the wh-word. While in the first case disjunctive responses can occur, these are excluded in the latter. The third type of wh-in-situ question appears in non-presuppositional contexts. According to Baunaz and Patin ‘non presuppositional (i.e. neither specific, nor partitive) wh-phrases typically appear out of the blue [. . .]’ (Baunaz & Patin, 2011: 97). In this case, there is neither a familiar entity nor a set of possible answers. Baunaz notes that the difference of these three types is reflected in their intonation. As mentioned above, wh-in-situ questions seem to be characterized by different intonation contours (Cheng & Rooryck, 2000; Mathieu, 2002; Starke, 2001). Following Baunaz (2011), the wh-in-situ question in non-presuppositional contexts is accompanied by a rising intonation (contra Cheng & Rooryck, 2000). The in situ question with a partitive wh-phrase has a downfall intonation, while in situ questions with a specific wh-phrase should appear with a slight fall–rise intonation contour. Although Baunaz and Patin (2011: 105) show that ‘speakers discriminate between on the one hand specific and on the other hand partitive and non wh-words’, they do not find a difference between wh-fronted and wh-in-situ questions with respect to the insertion of a sentence accent. This result is contrary to Hamlaoui (2008), who affirms that wh-elements which appear in situ carry the main stress of the sentence and therefore need to be distinguished from wh-fronted elements. In sum, the literature reviewed so far clearly indicates that more research is necessary in order to safely conclude whether wh-in-situ and wh-fronting are real variants. The review of the syntactic, semantico-pragmatic and prosodic constraints discussed in the literature seems to show that wh-in-situ

48

The Acquisit ion of French in Mult ilingual Conte x t s

questions must be described with more syntactic, semantico-pragmatic and prosodic machinery than previously thought. Perhaps, child speech can provide us with further pieces of information which allow a comprehensive analysis of wh-questions in French.

Wh-in-situ and wh-fronting in French: Quantitative aspects The frequency of wh-questions in French has attracted the interest of many linguists (cf. Ashby, 1977; Behnstedt, 1973; Coveney, 1996, 2002, among others). Table 2.1, taken from Coveney (2002: 112), shows the distribution of wh-patterns in several studies. The notation ‘–’ indicates that there was no information about the structure in this study, whereas ‘0’ stands for the absence of the wh-construction. Table 2.1 shows that the distribution of wh-questions is not consistent in the studies. Whereas Ashby finds that leaving the wh-element in situ is one of the favoured patterns, the elderly Belgian couple analyzed by Pohl only use wh-in-situ questions in up to 2% of all utterances. They prefer wh-questions with est-ce que, like the working- and middle-class subjects analyzed by Pohl. Why are the results not consistent? Unfortunately, the methodologies of the studies largely differ from one another. For example, Behnstedts’ amount of wh-in-situ constructions contains echo questions and Coveney does not count them. Coveney (1996) notices that there are some wh-questions which convey a specific pragmatic function like qu’est ce qu’ il y a in the sense of what’s up? Due to this he excludes these wh-questions from the analysis, while other researchers count them. Quillard’s (2000) and Myers’ (2007) quantitative analyses support Coveney’s findings and show that one of the favoured interrogative structures is the wh-in-situ question. In modern-day French, the wh-in-situ question type seems to be on the rise (Coveney, 2002: 284). Since the present study investigates child language and since it is generally conceded that adult speech to children and to adults differs in quantitative ways, we analyzed the children’s input. For the present study, we analyzed the speech of one monolingual French parent2 (cf. Geveler, in progress). We looked at a total of 2092 questions. Wh-fronting with inversion did not occur at all. Wh-in-situ amounted to 51% of all questions. The result that parents use wh-in-situ frequently when speaking to their children is also supported by Strik and Pérez-Leroux (2011). They investigated the input of five children and concluded that the most frequent structure – wh-questions with est-ce que (45%) – is followed by the wh-in-situ construction (35%). Interestingly, in our data wh-in-situ questions are most frequent, followed by interrogatives with est-ce que/est-ce qui (see Figure 2.1). If we consider the frequencies in more detail, another aspect leaps to the eye: dimensions of variation. Studies which take into account variation

87.5 31.6

38.8 10.3 33 2 8.7 12 22 – 12.9

– 6.9

0 – 0 – – 26 9

0

0

QkSV

– 0

0 – 0 – – 6 –

0 – –

QksekSV

– 3.3

0 – 4 – – 3 –

0 – –

QsekSV

–

0 – – – – – –

– – –

.3

seQkSV

– 41.5

38.8 46.5 12 66 53.6 8 18

3 38.9 35.5

QESV 12 48.6 44.3

QV NP

12.5 3.6

2.4 35.5 3 2 28 .5 10.2 9 40

13

50

QV-CL

72 364

85 155 446 184 69 587 ?

4,367 1.515 230

N=

Notes: Taken from Myers (2007: 36,68): Quand venez-vous? [QV-CL] ‘When come you’; Quand est-ce que vous venez? [QESV]; Vous venez quand? [SVQ]; Quand vous venez? [QSV]; Quand que vous venez? [QkSV]; Quand c’est que vous venez? [QsekSV]; Quand que c’est que vous venez? [QksekSV]; C’est quand que vous venez? [seQkSV]; Quand vient Jean? [QV NP]. Source: From Coveney (2002: 112).

7.1 7.7 46 3 27.5 36 11

14.9

25

Formal Middle-class (Behnstedt) Plays (Terry) Belgian TV journalists (Lafontaine & Lardinois) Middle-class (Ashby) Middle-class (Pohl) Colloquial middle class (Behnstedt) Elderly Belgian couple (Pohl) Working-class (Pohl) Working-class (Behnstedt) Belgian 7–12 year-olds (Lafointaine & Lardinois) Montreal 8–18 year-olds (Lefebvre) 9 year-olds (Söll)

QSV 10 12.5 5.3

SVQ

Variety/researcher

Table 2.1 Relative frequencies for (WHQ) in previous studies

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French 49

50

The Acquisit ion of French in Mult ilingual Conte x t s

Sylvain (adult) 1400 1200 1000 800

1317

600 400

599 200

141 0 wh-fronng

wh-in-situ

37 est-ce que

qu'est ce que/i

Figure 2.1 Absolute numbers of different types of wh-questions in adult speech addressed to children

according to social class and speech styles show that the uses of the different types of questions are related to these dimensions of language variation. According to researchers like Coveney, Behnstedt and Adli, wh-in-situ questions characterize the register français familier and even the lower ranked register français populaire. When adults speak to their children in a family situation, they use the aforementioned speech styles. Therefore, it is not surprising that wh-in-situ questions are the preferred question type in the children’s input. In addition to variation according to speech styles, Coveney (2002) and Adli (2004) prove that membership of a particular social class, sex and age are factors which influence the use of wh-questions. In sum, wh-in-situ is used frequently in the children’s input and therefore children should have enough evidence in order to acquire this type of wh-question early in development.

The Acquisition Data Previous studies Previous studies on the acquisition of wh-questions have focused on two aspects. A considerable number of researchers have investigated the order of appearance and acquisition of different wh-structures in monolingual French children. A second group of researchers draw on the results from monolingual children and have studied the effects of cross-linguistic influence in bilingual children. In what follows, we will summarize these studies.

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

51

What type of wh-question is acquired first in monolingual French? Two observations can be distinguished when wh-questions are investigated in monolingual French corpora. In longitudinal studies, most researchers find that wh-in-situ interrogatives are acquired before constructions involving wh-movement (Hamann, 2000, 2006; Hulk, 1996; Plunkett, 1999; Weissenborn et al., 1990). The four children analyzed by Hamann (2000, 2006) use wh-in-situ questions up to 90%, for example. There exist contradictory results if wh-questions are tested in elicited production tasks (Strik, 2007; Zuckerman & Hulk, 2001). Zuckerman and Hulk (2001) test 33 Frenchspeaking children (mean age: 4;7) and compare their data to that of 22 monolingual adults (15 parents and seven other adults). On the basis of the assumption that wh-in-situ is computationally less complex, they predict that children prefer the wh-in-situ option and that they use it more often than their parents/the adults do. Their analysis shows that parents and their children use wh-in-situ infrequently and to the same extent. The children produce wh-in-situ questions in up to 6% of the cases, adults in 5%. This finding contradicts computational complexity as the driving force in early stages of language acquisition. Trying to find an explanation for the unexpected result, Zuckerman and Hulk analyze the data of the adults in detail and observe that only two of the 22 subjects produced a lot of wh-in-situ questions. Eighty-four percent of all wh-in-situ constructions come from these two subjects. The researchers re-evaluate the test by including just the data of the 15 parents. This re-evaluation affects the results immensely: only 1% of all constructions which the parents produced are wh-in-situ questions. Their children use it in up to 8% of all cases. Zuckerman and Hulk argue that these results confirm the predicted increasing use of wh-in-situ in young children due to its lower costs with respect to computational complexity (also cf. Zuckerman, 2001). A second way to re-evaluate the data is to exclude all outliers (adults and children). In this case, the children use wh-in-situ in 3% of cases while the adults leave the wh-element in situ in up to 1%. The researchers believe that in both cases of re-evaluation there is a significant difference between adults and children. They argue that this difference is due to the differing complexity of the question types. In other words, wh-in-situ is the default option in French child grammar due to economy of derivation. Starting from economy of derivation, Strik (2007) makes three predictions for the acquisition of wh-questions in monolingual French children. First, the children should prefer the in situ position in root and long-distance questions. Secondly, root questions should be used before long-distance questions. Thirdly, children should acquire questions without inversion more quickly than questions which exhibit subject-verb inversion. In an elicited production task, Strik investigates the formation of questions in 36 children from the age of three to six years. As a control group, 24 adults were analyzed. Interestingly, Strik shows that the children produced more wh-fronted questions than in situ questions, if object questions of the three-year-old

52

The Acquisit ion of French in Mult ilingual Conte x t s

children are excluded. At first glance, this result contradicts economy of derivation. If root questions with objects and adjuncts are considered, Strik finds a difference between the usage of root questions in which the wh-phrase is an object or an adjunct. Questions with wh-phrases functioning as objects develop as Strik predicts. Thus, at the beginning, the wh-in-situ question is the preferred structure and its use decreases with age. But the wh-element is located in initial position if the wh-phrase functions as an adjunct in the question. An explanation for the preference of wh-fronting in adjunct interrogatives is, in accordance with Strik, that the adjunct (wh-phrase or not) is directly merged in a higher position than the direct object. Thus, the distance between the adjunct and the left periphery (hosting the moved wh-phrase) would be shorter than the distance between the direct object and the left periphery. If this syntactic analysis is on the right track, the more economical structure for adjunct interrogatives would be the wh-fronted question. More evidence for economy of derivation comes from the comparison of root and long-distance questions. Long-distance questions become more frequent with the increasing age of the subjects. The children produce median longdistance questions at early stages; in other words they don’t move the whelement twice just once. Another structure which increases with age in child speech is inversion, which requires movement. In sum, all the above-mentioned studies try to explain the acquisition of wh-questions by economy of derivation. The studies by Zuckerman and Hulk (2001) and Strik (2007), in particular, convincingly illustrate that results may depend on which structures are compared with each other. A considerable number of acquisitionists have formulated doubts as to the precedence of wh-in-situ questions over wh-fronted questions (Crisma, 1992; Oiry & Dermidache, 2006). Crisma (1992) claims that wh-fronting is the preferred structure in monolingual French acquisition.3 In her study she investigates one monolingual child, Philippe, who produces only questions involving wh-fronting. Based on a relatively small database, Crisma concludes that French monolingual children acquire wh-fronted questions first. Further studies (Hamann, 2000; Hulk, 1996) show that Philippe seems to be an exception in terms of the acquisition of wh-questions. In search of an explanation, Philippe’s parental input has been analyzed. His parents never leave the wh-element in situ when speaking to their child. The lack of wh-in-situ questions in Philippe’s corpus may thus be related to their complete absence in the input. Plunkett (1999) observes in another monolingual child, Grégoire, that fronted wh-questions are preferred. She states that the data of Grégoire show a total of less than 20 wh-questions. Moreover, most of these questions seem to be repetitions of adult speech or fixed patterns. Thus, Grégoire should be excluded from analysis and seems to be an exception just like Philippe. Crisma (1992) and Oiry and Demirdache (2006) test monolingual French children on root questions. They find that if there is a preference in root

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

53

questions, it is the fronted type. The 20 analyzed children produce a total of 124 wh-root questions which are classified in three groups: fronted-wh, wh-insitu and in situ/fronted. The last category contains wh-questions which are analyzable as either wh-fronted or wh-in-situ interrogatives.4 Oiry and Demirdache observe that the wh-element is moved to sentence initial position in 65% of all utterances, whereas the children use wh-in-situ questions less frequently, in 22.5% of all cases. Even if one adds up the figures of the third category (in situ/fronted) and those of the wh-in-situ questions (a total of 34%), the fronted wh-interrogatives are still in the majority. Oiry and Demirdache could not find any correlation between the age of the children and a preference for the in situ structure either. These results cast doubt on Zuckerman and Hulk’s assumption that wh-in-situ is the default option in monolingual French. At the same time, they support the claim by Déprez (1999) and Déprez and Pierce (1990, 1993), who observe that wh-movement is implemented at an early age in French child grammar, thus challenging the view of a preference for whin-situ (also cf. Déprez et al., 2013). Since Oiry and Demirdache do not differentiate between adjunct and object wh-phrases, Strik’s proposal that wh-movement is the more economical structure with adjunct wh-phrases cannot be validated. Finally, DeCat (2002) notices that two of four French children acquire fronting and in situ structures simultaneously. Tom and Léa are raised in Belgium and they use both structures right from the beginning. Interestingly, Léa and Tom behave like our analyzed bilingual children. In a nutshell, the results thus far are contradictory and invite the analysis of more corpora of monolingual French children. However, we can safely conclude that the precedence of wh-in-situ questions over wh-fronted questions must be questioned.

Does cross-linguistic influence play a role? While research on monolingual French children has focused on which type of wh-question appears first, the studies of bilingual children are mainly concerned with the question of cross-linguistic influence. The issue of crosslinguistic influence in the acquisition of wh-questions is controversial. Some researchers claim that if children acquire two languages simultaneously, they transfer one question type of language A to language B (Soriente, 2007; Strik & Pérez-Leroux, 2011; Yip & Matthews, 2000, 2007). However, others do not find any influence (Bonnesen, 2005; Mishina-Mori, 2005; Park, 2008). Before we turn to the studies in detail, we will mention the criteria for cross-linguistic influence to appear. Hulk and Müller (2000) and Müller and Hulk (2000, 2001) have defined two conditions under which cross-linguistic influence is likely to occur: (a) The vulnerable grammatical phenomenon is an interface property, e.g. a grammatical property located at the interface between syntax and pragmatics.

54

The Acquisit ion of French in Mult ilingual Conte x t s

(b) The surface strings of the two languages are similar for the expression of the vulnerable grammatical phenomenon. Müller and Patuto (2009) have reformulated Criterion (b) in the following way: (b) The surface strings of the two languages A and B are analyzable in terms of the syntactic derivation of one language (which is less complex). In what follows, we will present and discuss previous studies against this background. We will first present the studies which find a quantitative or qualitative difference between monolingual and bilingual children. Yip and Matthews (2000, 2007) investigate six bilingual English-Cantonese children longitudinally (age span: 1;03–4;06). The children were raised in Hong Kong. Four of the children acquire Cantonese as their dominant language, one child is English dominant and another one seems to be balanced. In English, wh-interrogatives involve wh-fronting, while wh-phrases remain in situ in Cantonese. The difference between the two languages is most clearly found in object questions, because in subject questions the wh-word is located in sentence-initial position in Cantonese as well. The authors find that the four Cantonese dominant children use wh-in-situ questions frequently when speaking English, while the balanced and the English dominant child produce more questions with wh-fronting. Yip and Matthews (2007) argue in favour of cross-linguistic influence due to language dominance. Soriente (2007) analyzes a bilingual Italian-Indonesian child. Her results are similar to Yip and Matthews’ findings. The child frequently produces the wh-in-situ structure when speaking Italian. Soriente argues that this influence is due to language dominance because the child she investigated acquires Indonesian as his dominant language. However, Yip and Matthews (2009) notice that besides language dominance there is another factor which can cause cross-linguistic influence. They admit that ambiguous input could induce the overgeneralization of wh-in-situ in English as well. In this case, English echo questions represent positive evidence for an analysis of English questions in terms of wh-in-situ. But if ambiguous input is the reason for cross-linguistic influence, wouldn’t we expect that wh-fronting is overgeneralized in Cantonese as well? Remember that subject-wh-phrases occur in sentence-initial position in adult Cantonese. Hence, in this case the input must be conceived as ambiguous too. But in accordance with Yip and Matthews no overgeneralization of wh-fronting in Cantonese questions is observed. Rephrasing Müller and Hulk’s Criterion (b) for cross-linguistic influence, the authors add the condition that if there is a surface overlap between the

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

55

two languages, such that language A allows one option and language B allows two, then one of the two options of language B must be isomorphic to language A in order for cross-linguistic influence to occur. If this is on the right track, then there will be no cross-linguistic influence in bilingual children who acquire two languages which do not share the same surface orders (e.g. SVO and SOV languages). We will come back to this explanation later. Meisel (2007) assumes that the findings of Yip and Matthews are no evidence for influence of the two underlying syntactic structures but rather that the influence is due to a processing effect. He proposes that the children may activate the Cantonese grammatical principles while processing English, which is also a type of cross-linguistic influence. In other words, the children activate the Cantonese grammar while speaking English because the wh-insitu question of Cantonese is arguably the more economical structure. In contrast to Yip and Matthews, and similar to Meisel, Strik and PérezLeroux (2011) assume that cross-linguistic influence is rooted in derivational complexity and that mere surface overlap cannot explain cross-linguistic influence adequately. In order to study cross-linguistic influence in detail, they also analyze inversion in wh-interrogatives. The authors investigate 16 bilingual Dutch-French children on the basis of elicited production. All the children were raised according to the principle of one person–one language. They live in France and visit French schools or kindergartens. The children are separated into two groups on the basis of their age. Hence, eight children belong to the younger bilingual group (age 5;0) and eight children who visit elementary school belong to the older bilingual group (age 7;0). Sixteen monolingual Dutch children and 14 monolingual Dutch adults serve as control groups. The data show that for all groups the wh-fronted question with inversion is used most frequently in Dutch. But the bilingual groups produce some nontarget-like questions without inversion and use wh-in-situ questions more frequently than monolingual Dutch children. In particular, three of the eight children belonging to the younger bilingual group produce wh-in-situ questions in Dutch. Note that the mean number of wh-in-situ questions is 2.26, which is significant. Questions without inversion are more prevalent than wh-in-situ interrogatives and they are observed in half of the bilingual population. The authors conclude that since wh-in-situ questions as well as noninverted interrogatives do occur in bilingual Dutch, the influence observed is based on derivational complexity, not on surface overlap. In a nutshell, the studies mentioned so far clearly show that wh-in-situ questions are used in non-in-situ languages like English, Italian and Dutch. There are different explanations for cross-linguistic influence: Soriente, in the vein of Yip and Matthews, argues for language dominance as the source of the influence. Yip and Matthews affirm that isomorphic structures are a necessary condition for cross-linguistic influence. Strik and Pérez-Leroux claim that derivational complexity explains the use of wh-in-situ questions in a wh-fronting language.

56

The Acquisit ion of French in Mult ilingual Conte x t s

Not all researchers found cross-linguistic influence or if they did they did not find the above-mentioned direction of the influence. Mishina-Mori (2005) analyzes the spontaneous speech data of two Japanese-English bilingual children who were raised in the United States. Both bilingual children do not use wh-in-situ questions in English. But interestingly, Mishina-Mori notices that wh-fronting occurs in the bilingual’s Japanese. Park (2008) analyzes Korean-English bilingual children raised in the United States. Korean is an SOV language like Japanese and German. Neither child uses wh-in-situ structures in English, not even in echo questions. Furthermore, Bonnesen (2005) observes that there is no cross-linguistic influence in German-French bilingual children. Trying to explain the different results of these studies, Yip and Matthews (2009) argue that the reason for the absence of cross-linguistic influence is the lack of isomorphic structures. The authors claim that ‘there is no overlap at the level of isomorphic strings’ (Yip & Matthews, 2009: 8), which means that due to the language’s different orderings (SVO and SOV) no influence is observed. Notice, however, that Moshina-Mori’s study proves that cross-linguistic influence can occur although the languages do not have the same surface order. We will come back to this issue when analyzing our data. From what we have seen so far in the literature and taking the criteria for cross-linguistic influence by Müller and Patuto (2009) into account, we can make the following predictions, (1) and (2), for the language combination French-German. Notice that Criterion (a) for cross-linguistic influence is fulfilled, since it seems to be the case, although more research is still necessary, that presuppositionality can be expressed by leaving the wh-phrase in situ. The modified Criterion (b) is fulfilled as well, but only for fronted wh-questions. For a German-French bilingual child, a fronted wh-question in French is analyzable in terms of a fronted wh-question in German. Both languages share the syntactic derivation for fronted wh-questions. The prediction thus would be that: (1) Bilingual French-German children no not use the French wh-in-situ question type when they speak German.5 (2) Bilingual French-German children use wh-fronted questions in French more frequently than monolingual children do. In the following sections, we will present our own study.

The present study For the present study, we investigated the corpora of six bilingual GermanFrench children.6 All children are bilingual from birth and they have been raised following the principle one person-one language according to which each parents speaks her/his mother tongue with the child. The data are part of a larger longitudinal study for which the children were recorded regularly in natural

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

57

Table 2.2 Analyzed subjects Name

Language/s

Age span

Recordings/Transcripts

Amélie

French/German

1;06,12–5;00,16

Alexander

French/German

2;02,06–5;02,21

Emma

French/German

Céline

French/German

1;4,01–4;11,24 1;4,01–4;00,12 2;0,9–5;04,14

Marie

French/German

1;09,19–5;01,23

Caroline

French/German

1;8,23–3;7,28

Madeleine Sylvain (adult) Chantal Annette (adult)

French French German German

1;00,05–3;06,08 – 1;10,18–5;0,11 –

French: 69 German: 69 French: 48 German: 49 French: 73 German: 64 French: 57 German: 56 French: 31 German: 32 French: 33 German: 33 French: 24 French: 38 German: 58 German: 21

settings. The examined time period ranges from the age of 1;08,02 until 5;02,21. The bilingual corpus contains 1360 German and 1282 French wh-questions from the children.7 In addition, we analyzed the data of one monolingual French child named Madeleine from the CHILDES database (Suppes et al., 1973) (244 wh-questions) and one monolingual German child, Chantal, who produced 830 wh-interrogatives. The speech of the parents of one bilingual child was analyzed in order to know how frequent the different types of questions are: 2368 wh-questions of the French father and 1487 interrogatives of the German mother. Amélie, Alex, Céline and Caroline have been raised bilingually in Germany, Emma and Marie in France. For further details see Table 2.2. For the present paper, all wh-questions occurring in each transcript were counted and categorized. In order to categorize the children with respect of language dominance, Figure 2.2 shows the mean difference of the MLU values of all children in both languages, German and French. The figure is based on Arencibia Guerra (2008) and Eichler (2011). Figure 2.2 shows that four of the six analyzed children have a tendency to develop French as their dominant language. The mean MLU difference amounts to 0.6 (words) in Amélie and Emma, while Alexander and Marie show an articulated tendency for French as a dominant language, with a difference of 0.8 (words) and 1.7 (words). In contrast to the four children mentioned so far, Céline develops German as her dominant language. The mean MLU difference is about 1.3 (words). Caroline is a balanced bilingual child. However, she does not speak much during the recordings.

58

The Acquisit ion of French in Mult ilingual Conte x t s

Figure 2.2 Mean MLU difference in the bilingual children

Figure 2.3 gives an overview of the use of all questions which contain a verb in French. Since pourquoi ‘why’ cannot remain in situ in adult French, it is important to show its usage separately. Furthermore, est-ce que questions always involve a fronted wh-phrase in adult French; therefore it is illustrated separately in Figure 2.3. Qu’est-ce (que/i) is the category for a fronted que with est-ce que/est-ce qui. The category est-ce que stands for yes-no questions. Figure 2.2 does not only serve as an overview of the analyzed data, but it also nicely shows that the children know that pourquoi and que with est-ce que/qui must be fronted. Strictly speaking, only wh-fronting and wh-in-situ

Figure 2.3 French questions in all analyzed subjects

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

59

100% 90% 80%

5

58

70%

106 55

60%

139

146

50%

129 1317

Wh-in-situ

40% 30% 20% 10% 0%

8

68

Wh-fronng (- pourquoi and - questceque)

99 22

33

18

15

121

Figure 2.4 French wh-fronting and wh-in-situ compared for all analyzed subjects

should be compared for the purpose of the present paper because the other contexts for questions are restricted to wh-fronting in the adult language. Figure 2.4 compares wh-fronting and wh-in-situ for those questions, for which both structures are in principle possible. The monolingual child Madeleine uses wh-in-situ questions to a large extent. This is also true for the French adult. More than 90% of all questions are characterized by a wh-word which is located in situ. The bilingual child who most resembles the monolingual child is Marie. All other children use

Figure 2.5 Development of questions in French over time: Emma

60

The Acquisit ion of French in Mult ilingual Conte x t s

wh-fronting more often than the monolingual child and Marie. The resemblance to Madeleine is progressive, with Emma being the child who resembles Madeleine most, then come Céline, Amélie and Alex. For Caroline, the data are too scarce to allow any conclusions. An initial glance at Figure 2.2 shows that language dominance cannot explain the more frequent use of wh-fronting in the bilingual children. A first guess would be that the children overuse wh-fronting in French due to their other language, namely German. If language dominance played a role, we would expect that the child with German as the dominant language would use wh-fronting the most. Céline does not fit this picture. Another possible relation would be that the child who is most proficient in French shows the least effect of cross-linguistic influence. Although this is true for Marie, it does not explain why Alex overuses wh-fronting. Due to limits of space, we cannot show the figures for all children which illustrate the development of the different types of questions over time. This analysis shows that neither child seems to prefer wh-in-situ during an early stage of development. Figure 2.5 for Emma and Figure 2.6 for Alex nicely illustrate that wh-fronting is present from very early on. We can conclude that our Prediction (2) is corroborated by the data: bilingual French-German children use wh-fronted questions in French more frequently than monolingual children do. However, some children correspond exactly to monolingual French children. In other words, overuse of wh-fronting is likely to occur but it is not a necessary characteristic of bilingual French. Let us now turn to the bilingual children’s German. Figure 2.7 shows for all bilingual children that wh-in-situ is nearly completely absent in German. As for French, Figure 2.7 is intended as an overview and therefore also shows

Figure 2.6 Development of questions in French over time: Alex

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

61

Figure 2.7 Development of questions in German all subjects

yes-no questions (vSInvers, verb-subject inversion) which are, however, not important for the present study. We can conclude that our Prediction (1) is corroborated by the data: bilingual French-German children do not use French wh-in-situ in German.

Discussion The results of our investigation can be summarized as follows: (1) Bilingual French exhibits more wh-fronting than monolingual French, however, not necessarily. One bilingual German-French child in our study behaved exactly like the monolingual French child in French. Overgeneralization of wh-fronting is independent of language dominance. (2) In addition, our corpora allow the conclusion that wh-in-situ and whfronting are both present from early on. (3) Wh-in-situ does not transfer to German where it is also ungrammatical. Let us start with the second result since it may have consequences for the analysis of wh-questions in adult French. In the first section we have presented the discussion about the question of whether wh-in-situ and wh-fronting are variants, the in situ option representing the less complex construction.

62

The Acquisit ion of French in Mult ilingual Conte x t s

Although the literature does not allow safe conclusions, the results from language acquisition could be interpreted against the view that the two types of questions are variants. If wh-fronting and wh-in-situ are free variants of one another, wouldn’t we expect that the bilingual children would prefer the less complex in situ structure? In fact, some of the analyzed children overgeneralize the more complex structure and move the wh-phrase more frequently to the left periphery of the sentence than monolingual speakers do. In contrast to other studies, we have not found a preference for wh-in-situ questions in early stages. Both question types coexist right from the outset in the German-French bilingual children, which may very tentatively be interpreted in favour of the view that wh-in-situ and wh-fronting are not variants and cannot be compared on the basis of economy of derivation. Further research has to analyze the fronted and in situ wh-phrases in more detail and also consider their grammatical function (object versus adjunct wh-phrase). As to the first and third result, they perfectly fit the criteria for crosslinguistic influence by Müller and Patuto (2009). Criterion (a) is fulfilled. Criterion (b) is fulfilled as well, but only in the case of wh-fronting. With wh-fronted questions, the bilingual child may use the syntactic derivation of either language, German or French, since both will produce the desired surface string. Before we discuss the kind of cross-linguistic influence, we would like to mention that our results disprove the condition of surface overlap as formulated by Yip and Matthews. The authors predict that there is no crosslinguistic influence between an SVO and an SOV language. Overgeneralization of wh-fronting in French, as exhibited by the bilingual German-French children, is in accordance with Mishina-Mori’s observations and represents a clear indication that cross-linguistic influence does occur between an SVO and an SOV language. It is not the language typology, in other words the underlying syntactic structure, which misleads the child, but identical surface strings which can possibly have the same syntactic derivation. Put differently, the bilingual children can use the German syntactic derivation for wh-fronted questions in French. What does this mean exactly? We have shown that wh-in-situ and wh-fronting must be described on the basis of a detailed syntactic, semantico-pragmatic and prosodic analysis. German is not a wh-in-situ language, thus wh-phrases must move to the left periphery, independently of syntactic, semantico-pragmatic and prosodic aspects of the construction in question. Wh-fronted questions are shared by both languages, although it is probable that the functions expressed by wh-fronting in French are only a subset of those expressed by wh-fronting in German (remember that French exhibits wh-in-situ). In essence, we would expect that the bilingual child, due to cross-linguistic influence, also moves the whphrase to the left periphery in French in cases where adults will use the in situ question. Notice that further research will have to analyze the syntactic, semantico-pragmatic and prosodic aspects of wh-fronting and wh-in-situ in

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

63

French questions in order to evaluate the present results against the background of a language acquisition theory. A further implication of our study is that there seems to be a difference between the acquisition of two SVO languages and the simultaneous acquisition of an SVO and an SOV language. Whereas studies investigating the former combination found that wh-in-situ constructions are overgeneralized in the non-in-situ language (e.g. Dutch, English, Italian), the latter came to the opposite conclusion, namely that the children use wh-fronted interrogatives frequently in the in situ language. In other words, the language combination does not serve as a clue for cross-linguistic influence. It is the direction of the influence which seems to be affected by the underlying structure of the simultaneously acquired languages. The effect of the language combination for the direction of the influence falls out naturally from the Criterion (b) of cross-linguistic influence, implying that bilingual children are ready to use syntactic derivations which are shared by the two languages. Both Yip and Matthews and Soriente argue that language dominance plays a major role when it comes to predicting cross-linguistic influence in bilingual first language acquisition. Our results cannot be explained on the basis of dominance. Alexander and Marie behave very differently with regard to cross-linguistic influence, although both of them show an articulated tendency for French as a dominant language. Notice that not all bilingual children showed effects of cross-linguistic influence. Marie did not overgeneralize wh-fronting at all. In contrast to the other children, Marie was raised in France. The language of the environment might play a role. It might be the case that Marie’s knowledge of French is advanced as compared to that of the other children because she had more contact with French. Let us compare the two extreme children, Marie (no overuse of wh-fronting) and Alex (considerably high overuse of wh-fronting), with respect to criteria which reflect that (non-)advanced knowledge of French. Verb types are a good indicator of grammatical development. If we compare the two children at the age of four in French, we can observe no difference between them: Alex reaches 209 types, Marie 207. In other words, both children do have similar knowledge of French verbs, although they are raised in France and Germany, respectively. What about language fluency? We might assume that Marie, because she is raised in France, is more fluent or more ready to speak French than Alex is. Language fluency can be measured in terms of uttered words per minute (Arencibia Guerra, 2008). Figure 2.8 compares the proficiency of the bilingual children in German and French at the age of four. The interesting result is that the ranking according to the overuse of wh-fronting has nothing in common with the children’s fluency in French, but corresponds to the reversed ranking according to their proficiency in German. In other words, the probability of overuse of wh-fronting depends on the degree of proficiency in German: lesser proficiency in German (the influencing language) results in lesser overuse of wh-fronting in French (the influenced language)

64

The Acquisit ion of French in Mult ilingual Conte x t s

Figure 2.8 Words per minute of the bilingual children in German and French at the age of four

(cf. Schmeißer & Jansen, this volume, Chapter 1, for a similar observation with respect to word order). If this analysis is on the right track, we can conclude that cross-linguistic influence can be predicted on the basis of the criteria proposed by Müller and Patuto (2009). The influenced language is French, which is the more constrained language for the grammatical phenomenon in question. The effect of the influence is related to the proficiency or readiness to speak the other language, which is German in our case. Lesser proficiency in the influencing language results in a lower probability to overuse a syntactic derivation of this language in the influenced language. Figure 2.8 picks out one point in development and compares the children. Future research will have to relate the overuse of wh-fronting to mean measures of proficiency calculated for the whole period of investigation.

Notes (1) For reasons of space, we cannot present the details of the English and Chinese interrogative systems. For example, English allows wh-in-situ in multiple questions in which more than one wh-phrase exists and in so-called echo questions. Chinese does not completely disallow wh-fronting. (2) Sylvain lives in Paris with his family. He doesn’t speak any German. (3) A similar conclusion is drawn by Soares (2003) for the acquisition of European Portuguese. Soares claims that the in situ structure is acquired later than the whfronted question. (4) Note that we will not consider these utterances in our analysis since this type of question is ambiguous. (5) Note that in both languages echo questions are less frequent in the parental input. Sylvain produces two echo questions in French and Annette only one in German.

Wh-f ront ing and Wh-in-situ in the Acquisit ion of French

65

(6) The data stem from three different research projects: 1999–2005: ‘Frühkindliche Zweisprachigkeit: Italienisch-Deutsch und Französisch-Deutsch im Vergleich’, Universität Hamburg; 2005–2008: ‘Die Architektur der frühkindlichen bilingualen Sprachfähigkeit: Italienisch-Deutsch und Französisch-Deutsch in Italien, Deutschland und Frankreich im Vergleich’, Bergische Universität Wuppertal; 2009–2013: ‘Code-Switching bei bilingual aufwachsenden Kindern in Deutschland, Italien, Frankreich und Spanien: Italienisch-Deutsch, Französisch-Deutsch, SpanischDeutsch, Italienisch-Französisch, Italienisch-Spanisch, Französisch-Spanisch’, Bergische Universität Wuppertal. All projects have been financed by the German Science Foundation (DFG) and have been directed by Natascha Müller. For more information cf. Cantone et al. (2008), Müller et al. (2011) and Müller et al. (2015). The main research results of the WuBiG are presented in Hauser-Grüdl et al. (2010) and Schmeißer et al. (2015). (7) The numbers include questions which do not contain a verb and which we will not analyze here. They serve as an overview for the reader.

3

On the Processing of Subject Clefts in English-French Interlanguage: Parsing to Learn and the Subject Relativizer qui Laurent Dekydtspotter and Kelly Farmer

Introduction Second language (L2) sentence processing has become the subject of intense empirical investigation and of increasing conceptual interest to the field of L2 acquisition in general (Gregg, 1996, 2003). This is because the targetlanguage input does not directly trigger language acquisition – rather, acquisition is mediated by a parser, a mechanism for grammatically structuring the input into intake. As parser–grammar relations have become better articulated in the generative paradigm, elements of a considerably more precise theory of grammatical change in an L2 have begun to emerge. A body of research on first language (L1) sentence processing shows that representations in various modules are computed concomitantly in real time (Crocker, 1996). Such sentence processing crucially exhibits an instinct to structure the input immediately according to specific structural strategies implicating the computations of phrase structure representations, of movement chains, of thematic role assignments, and of anaphoric dependencies inter alia (e.g. Fodor & Inoue, 1998, 2000; Frazier, 1987; Frazier & Fodor, 1978; Frazier & Rayner, 1988). Syntactic parsing seems to involve the incremental satisfaction of grammatical requirements (Pritchett, 1992), in a computational system for human language (CHL), obeying principles of economy (Chomsky, 1995). Clearly, sentence processing involves more than syntactic computations; the central claim of the modularity hypothesis includes an informationally encapsulated parser 66

On the Processing of Subjec t Clef t s in English-French Interl anguage

67

(Fodor, 1983, 2000) that interfaces with other modules in specific ways. This view seems well supported by the evidence of an economy-driven parser that fails to learn from mistakes: Taking the same routes over and over again, the sentence processor fails to change the path of analysis in view of experience and context. This does not exclude the possibility of the early use of pragmatic information, animacy, frequency and plausibility effects, as phrasestructure computations interface with processing in other modules (within the limits allowed by universal grammar (UG)) at every stage of processing. As van Gompel and Pickering (2007) point out in a recent survey of the field, modularity has not gone unchallenged by researchers examining various interventions on the parse from other sources of knowledge in sentence processing, although there are clear limits as to what such information can do. L2 sentence processing research has shown that the resolution of grammatical dependencies is automatic, suggesting a reflex to structure the input as quickly as L2 lexical access allows (Hoover & Dwivedi, 1998; Juffs & Harrington, 1995; Williams et al., 2001). Thus, the processing of filler-gap dependencies in an L2 is compatible with the active filler strategy, according to which chain dependencies are postulated at the first opportunity. However, the nature of the computations performed by L2 learners is very much in dispute. Following Juffs and Harrington (1995) and allied work, some authors argue that L1 and L2 sentence processing shares the same performance theory (Dekydtspotter et al., 2008; Frenck-Mestre, 2005; Frenck-Mestre & Pynte, 1997; Juffs, 1998, 2006), crucially including a reflex to apply grammatical constraints incrementally (Pritchett, 1992). However, a body of research disputes the claim that L2 learners can compute detailed syntactic representations during real-time sentence processing (Felser et al., 2003, 2009; Marinis et al., 2005; Papadopoulou & Clahsen, 2003, inter alia). Thus, L1 sentence processing includes syntactic computations as well as independent meaningbased computations (Townsend & Bever, 2001), which may occasionally take over (Ferreira et al., 2002; Patson & Ferreira, 2009). Clahsen and Felser’s (2006a, 2006b) shallow structure hypothesis claims, however, that meaningbased computations become the primary route in the incremental processing of the L2. In comparison to native speakers, L2 learners typically over-rely on lexico-semantic, thematic, pragmatic and contextual cues, as a result of a compromised ability to compute detailed syntactic representations. The hypothesis that UG constitutes the knowledge base of the parser (Fodor, 2000; Schwartz, 1999) provides an acquisition scenario for the development of UG-constrained interlanguage grammars in which sentence processing plays a crucial role. This hypothesis requires a specific mental architecture, in which the parser is constituted by the grammatical system for human language, language-specific information, and a general performance theory that deploys the knowledge. Parsing involves the generation of representations that must be quickly licensed by the current grammar (Gorrel, 1995). Parsing routines arise at the union of universal performance

68

The Acquisit ion of French in Mult ilingual Conte x t s

principles and grammatical knowledge. As Fodor (1998b) points out, the parser seems to be designed in such a way that no parsing routines need to be learned in addition to the acquisition of grammars. In the specific parser-grammar organization sketched above, domainspecific acquisition – in contradistinction to domain-general learning – resides in parsing within the constraints of UG (Fodor, 1998b). Because the knowledge base of the syntactic parser includes CHL together with a parameterized lexicon, the parser can compute UG-constrained representations of the input beyond those that the current grammar can license. Licensing failure triggers the need to change the grammar, so that the acquisition of grammar is failure driven (Clark & Roberts, 1993). This view of parser-grammar relations naturally makes full transfer/full access (FT/FA) (Schwartz & Sprouse, 1994, 1996) the theory of interlanguage grammars. The learning-theoretic implications of a domain-specific processing module for L2 acquisition were anticipated in Schwartz (1987) and further expounded in Schwartz (1999). Crucially, parses will not trigger changes in the grammatical states, unless they incorporate syntactic details that the current grammatical state does not license. If the relevant triggers are not computed, the acquisition mechanisms are bypassed. Under these conditions, L2 knowledge is expected to have a singularly different character. This has indeed been argued (BleyVroman, 1989, 2009; Clahsen & Muysken, 1989; Meisel, 1997, among others). Given the fact that only what can be processed in domain-specific ways stands a chance of being acquired (rather than learned), we examine the processing of French subject clefts: ‘C’est XP qui . . .’ ‘it’s XP that’ and potential implications for the acquisition of the subject-linked relativizer qui in L1-English L2-French interlanguage development. Furthermore, we investigate whether the processing preconditions are met for the UG-constrained acquisition of the subject relativizer qui. Two learning scenarios are contrasted. In the first scenario, learners develop hypotheses on the basis of linear characteristics of the distribution of qui (Hawkins, 1989). In the second scenario, the input is parsed with detailed syntactic representations (Rizzi, 1990; Sportiche, 2011), leading to the acquisition of the subject-linked relativizer qui in the FT/FA model. L2 acquisition might involve a competition between these scenarios. However, evidence suggests that intermediate and advanced L1-English L2-French learners alike computed detailed structural processing of subject clefts. Implications for the acquisition of the subject-linked relativizer qui in English-French interlanguage grammars are examined.

Parsing to Learn French Relativizer qui In French cleft sentences and relative clauses, the form qui appears in two capacities. The first is as a complement to a preposition, qui is an XP-level

On the Processing of Subjec t Clef t s in English-French Interl anguage

69

wh-operator restricted to animate antecedents (1a) in parallel to English who. Qui thus participates in a complex distribution with other XP relativizers, including a series of DP-level pronouns bearing gender and number features that do not distinguish between animate and inanimate antecedents. The use of relativizer quoi, restricted to inanimate antecedents, in (1b) is criticized, but is occasionally found in native speech. (1) a. b.

C’est la personne à laquelle/à qui Paul tient le plus. It’s the person to whom Paul is most attached to. C’est la chose à laquelle/?à quoi Paul tient le plus. It’s the thing to which Paul is most attached to.

The second capacity is as a morphologically simple, short relativizer, in which qui links the cleft noun to the subject role with no regard to animacy (2a, 2b); qui is a syntactic head compatible with inanimate or animate antecedents, as are the X0 relativizers que and dont (Godard, 1992; Kayne, 1976). As an X0, relativizer qui is unlike the English relative pronoun who. Relativizer qui simply links the cleft noun phrase to the subject function of an upcoming predicate in the complement clause, which must have the appropriate verbal morphology. (2) a. b.

C’est la fille qui est arrivée. It’s the girl REL arrived. C’est la lettre qui est arrivée. It’s the letter REL arrived.

Subject clefts are particularly interesting in studying the possible triggers for the acquisition of relativizers. A number of studies have brought to light uses of c’est clefts in intermediate learners of French (Ferdinand, 2002; Kerr, 2002; Sleeman, 2004; Watorek, 2004). Presumably, as the result of restrictions on phonological focus, cleft sentences constitute a highly idiomatic device at the interface of syntax with information structure in French (Lambrecht, 1981, 1987). Subject clefts as in (2) seem most frequent. Donaldson’s (2008) examination of 10 native speakers and 10 near-native speakers on a range of conversation topics found similar percentages of the use of subject clefts: 86% of clefts (n = 46) in native speakers and 85.1% (n = 32) in near-native speakers were subject clefts. Clefts produced by both groups exhibited a similar distribution of the information status of the referents of cleft NPs: textually or situationally evoked referents were preferred over inferable or unused referents (Chafe, 1976). Clefts may incur a lesser processing load than relative clauses, which involve the composition of two clauses. On most analyses, the cleft noun phrase identifies a (null) wh-operator that derives a predicative clause (CP). How this relation is mediated by the matrix clause is in dispute. Following

70

The Acquisit ion of French in Mult ilingual Conte x t s

Chomsky (1977), Heggie (1988) argues that the copula selects a CP-level small clause. Others generate the cleft noun as an argument of the copula (Delahunty, 1882; Williams, 1980). In this vein, Kiss (1998) argues that the cleft noun is generated in a focus projection headed by the copula taking a predicative CP as its complement. This captures the observation that the copula provides identificational focus. Hedberg (2000) notes that cleft pronouns and predicative clauses, like definite descriptions, require a presupposition of existence and exhaustiveness. She proposes that the syntax of cleft sentences involves a predicative clause adjoined to the cleft noun phrase at the DP level, whereas semantically the predicative clause identifies the restriction of the cleft pronoun, which semantically functions as a definite determiner. We assume that, in the cleft frame, the copula takes two arguments – the cleft noun and a referential chain implicating the pronoun and the predicative clause as its restriction. The cleft noun is identified by the copula as being the unique entity that satisfies the CP predicate. Functionally, relativization relates a semantic role in the matrix clause with another in the relative clause. Given that there are a priori fewer referents to keep track of in the derivation of clefts, it seems that, from the point of view of the ability to devote as many resources as possible to the form of relativizers in acquisition, clefts could be triggers of choice. This raises the question of the manner in which subject clefts are processed by intermediate learners (and beyond), given that the acquisition of the relativizer qui seems to take place in learners of intermediate proficiency. Evidence that subject clefts are processed in domain-specific ways in real time by intermediate learners would highlight the potential of cleft sentences as providing triggering input in the acquisition of the relativizers. Crucial aspects of the acquisition of the relativizer qui in relative clause constructions are well known: English-speaking learners of French first restrict qui to animate NPs on a par with the English pronoun who. On a cloze test task, further supported by other tasks, Hawkins (1989) found that lower L1-English L2-French intermediate learners supplied the form qui with inanimate antecedents at very low rates (23%). Higher intermediate learners, however, appeared to have revised their treatment of qui so that it could co-occur with a verb immediately to its right irrespective of the nature of the referent of the noun to the left. Rates of relativizer qui with inanimate antecedents rose sharply in this group (>90%). Hawkins (1989) noted that the acquisition can be described in terms of minimal adjacency constraints. In Stage I, the relativizer qui must be adjacent to an animate noun phrase (3). In Stage II, it must be adjacent to a verb, given that the NP subject is missing (4). (3) Stage I: Rel → qui/NP [+animate] (4) Stage II: Rel → qui/ [NP e ] V

On the Processing of Subjec t Clef t s in English-French Interl anguage

71

Hawkins (1989) shows that configuration, rather than a ranked hierarchy of grammatical functions of subject, direct object, indirect object, oblique object, genitive, etc. (Gass, 1979; Hyltenstam, 1984; Schachter, 1974; Tarallo & Myhill, 1983), plays a determinant role in capturing the acquisition of relative clauses. The cornerstone of Hawkins’ argument is provided by the acquisition of genitive relatives introduced by the relativizer dont, showing clear structural asymmetries in subject and object positions across the noun complement function. Thus in (5a) and (5b), the grammatical function is constant, but configurations differ. The ranked hierarchy of grammatical functions predicts similar outcomes in (5a) and (5b). If differences arise in L2 acquisition, these differences must be positional. Hawkins (1989) found that learners controlled the relative configuration in (5b) in advance of the relative configuration in (5a). (5) a. b.

Le visiteur dont j’avais oublié le nom est là. The visitor whose name I had forgotten is here. Le visiteur dont le nom m’était inconnu est là. The visitor whose name was unknown to me is here. (Hawkins, 1989: 172, ex. 12a, b)

If configuration plays a necessary role in the case of relativizer dont, then the whole sequence can be profitably captured in terms of position, given that grammatical functions can be defined configurationally. Hawkins (1989) provides additional evidence from the interaction of relativization with the verb-subject word order in stylistic inversion. This also receives a configurational explanation. We set it aside here, given the rarity and formal status of stylistic inversion. Configuration captures the entire sequence by reducing it to processing complexity, whereas an account based on a hierarchy of grammatical functions constitutes a descriptive generalization and still additionally requires position to account for relativizer dont. Hawkins’ (1989) data established that configuration and processing complexity plays a crucial part in the development of knowledge of relativization in French. Hawkins (1989) provides a plausible, even an elegant, explanation of the manner in which learners might gain a workable entry point into the complex phenomenon of relativization in French with relatively little expenditure of resources, on the (not uncontroversial) assumption that less structural computing is better when resources are stretched. Thus, for Hawkins, ‘learners first appear to construct rules for RRCs [restrictive relative clauses] on the basis of adjacent categories in surface configurations; rules required by the target language which relate non-adjacent categories only appear subsequently’ (Hawkins, 1989: 176). Hawkins (1989) assumes that L1-English L2-French learners have the knowledge to eventually compute operator variable structures in the relevant way across positions – knowledge afforded them by UG and also put to use in English relative

72

The Acquisit ion of French in Mult ilingual Conte x t s

clauses. The claim that learners’ knowledge first reflects strictly linear cooccurrence constraints on adjacent expressions is compatible with recent claims about the inability of (most) L2ers to compute detailed representations in real time (Bley-Vroman, 2009; Clahsen & Felser, 2006a, 2006b; Meisel, 2009), perhaps with the plausible exception of very advanced learners who achieve superior command in the L2, but seem exceptional. Indeed, Hawkins and Casillas (2008) argue that learners first acquire morphological paradigms as co-occurrence constraints on lexical items. If the relevant features have been selected in L1 acquisition, input patterns can be eventually re-analyzed in the constraints of these features and UG-given operations on them. The learning of linear configurations considered in Hawkins (1989) is achievable by domain-general means, and therefore, does not require domainspecific knowledge. Learners seem guided by a general requirement of conservatism: the Stage I hypothesis is derived by conservatively assuming that French is much like English in the experience of English-speaking learners. However, finding qui after inanimate nouns in French input disabuses the learner of this belief. The change from the Stage I hypothesis to the Stage II is still guided by conservatism: having failed to establish a proper cooccurrence constraint (immediately) to the left of qui, L1-English L2-French learners then establish co-occurrence constraints (immediately) to the right of the relativizer matching the target language. Stage II is a hypothesis derived by small input-driven adjustments and constrained by assumptions of local co-occurrence relations. From this perspective, the change seems relatively straightforward. For this generalization to hold, parenthetical information must be disregarded, given that it is irrelevant to the predication relation. Hence, various types of information must be taken into consideration as in hypothesis formation, which is the hallmark of general cognition according to J. Fodor (1983, 2000). In addition to a general ability of adult L2 learners to formulate strategies based on linear configurations as in (3) and (4) as argued in Hawkins (1989), there is also presumably an automatic CHL response to the input. Under a reflex to compute a partial syntactic representation incrementally (Fodor, 1998b; Fodor & Inoue, 2000, and very many others), the string ‘C’est Aurélie . . .’ is immediately structured as a cleft involving a construal with a CP-level predicate, the contents of which may be contextually supplied. If not, then an operator variable structure implicating the subject position, by virtue of the active filler strategy, is postulated at once as in the partial structure. In the structure in Figure 3.1 there are few choices for the form qui. It may be an XP wh-Operator or a head: In this latter case, qui is either the exponent of an agreeing head-governing complementizer (Rizzi, 1990) or a wh-clitic of category D under C licensing a null DP wh-Operator in the subject position (Sportiche, 2011). We remain agnostic as to which analysis is to be preferred on UG-theoretic grounds. On the FT/FA model (Schwartz & Sprouse, 1994, 1996) which is extended to the lexicon (Sprouse, 2006), we

On the Processing of Subjec t Clef t s in English-French Interl anguage

73

Figure 3.1 Partial representation anticipating an operator-variable structure in the processing of subject clefts

expect that L2 learners will first analyze French qui as a wh-Operator analogous to English who, since this allows the state of grammar to remain fundamentally unchanged. French input, such as ‘C’est la letter qui . . .’ ‘It’s the letter which/that’ is incompatible with the analysis. In the morphological paradigm of French in which quoi and qui as DP wh-Operator contrast in animacy, such input induces a parse in which qui is a head in a predicative C – a representation that the English-like interlanguage state cannot license. This failure to license what the parser computes constitutes the disconfirmation of the English-like interlanguage grammar. Changing the state of grammar means the adoption of lexical specifications that license qui as a head in a (predicative) C (Lardiere, 2009a, 2009b). Felix (1988) surmised that UG knowledge is engaged only when general cognitive systems fail to offer competing accounts: cognitive strategies in the sense of Felix (1985) or Clahsen and Muysken (1989) block UG access in those structural domains in which surface structure clues tend to obscure deeper regularities. For example, one might imagine that word order regularities, which in most languages are superficially ‘visible’, might ‘evoke the operation of the relevant general cognitive mechanisms’ (p. 99). The linear rules in (3) and (4) present a potential test case. However, current understanding of the role of CHL in cognition undermines this line of thought: CHL provides processing efficiencies in the conceptual-intentional systems that make it a privileged route. For instance, the hypothesis that syntactic computations are privileged in language design is crucial to the current understanding of Principle B in language design (O’Grady, 2000; Reinhart & Reuland, 1993; Reuland, 2001) and delay in the acquisition of Principle B (Chen & Wexler, 1990). Efficiencies of specialized computations in L2 acquisition would lead to the allocation of resources to the computation of analyses in CHL . CHL-derived parses that the current state of grammar does not license would feed UG-constrained

74

The Acquisit ion of French in Mult ilingual Conte x t s

acquisition, tapping into a second language instinct (Schwartz, 1998). Cognitive competition has largely been set aside in the generative paradigm. The purely formal role of de in quantification structures in French induces highly domainspecific interpretive effects at the syntax–semantic interface in L1-English L2-French interlanguage (Dekydtspotter, 2009). L2 learners are also able to acquire new semantic values for morphemes guided by the hierarchy of categories and mapping principles (Slabakova, 2006). Last but not least, the acquisition of word order offers poverty of the stimulus arguments indicative of UG-constraints (Schwartz & Sprouse, 2000). Nevertheless, if domain-general and domain-specific systems can separately engage aspects of the input, learners’ real-time analyses of input can be empirically investigated as a window onto constraints on acquisition, given that what is detected in the input limits what can be acquired.

Elucidating the Learning Scenario Echoing Gregg’s (2003) discussion of the state of the development of a transition theory for L2 acquisition, it could be pointed out that a blind processing module quickly computing (immediately knowing) the type of representation involved in ‘C’est la lettre qui . . .’ ‘It’s the letter which/that’ might be expected to lead L2 learners to soon acquire the lexical specifications for French relativizer qui. Robust knowledge of relativizer qui, however, is far from immediate in L2 learners’ behaviour, despite explicit instruction and training. This might be taken as evidence for the greater plausibility of a domain-general approach reliant on offline pattern extraction over many episodes. To account for the distribution of relativizer qui, intermediate learners would compute linear arrangements of lexical items and develop linear co-occurrence constraints (perhaps only provisionally) through the domaingeneral power to abstract generalizations constrained by conservatism. Alternatively, the cognitive plausibility of storing huge amounts of linguistic details could be questioned on the grounds that less is better. Depending on one’s perspective, one might rebut that assuming a domain-specific computational system is not assuming less. These two scenarios can only be advanced by considering learners’ ability to parse. More specifically, this allows an exploration of the existence of potential triggers as learners engage in the task of sentence processing. The nature of the parse will determine what can be learned. In either scenario, input strings such as ‘C’est la fille qui . . .’ ‘It’s the girl who’ trigger no change: these data cannot disabuse L1-English L2-French learners of the hypothesis that qui is marked for animacy rather than subjecthood like the English ‘who’. However, the available parses in L1-English L2-French acquisition will determine what can be learned when aspects of the parses required by the target-language input fail to be licensed by the L1 grammar.

On the Processing of Subjec t Clef t s in English-French Interl anguage

75

Our investigation of learner computations takes advantage of concomitant processing of operator-variable structures at three levels. In syntax, movement chains implicating traces involve the activation and reactivations of referents as the semantic dependencies that movement chains instantiate are computed. Thus, syntactic movement dependencies have been documented to affect referential dependencies in discourse-semantic systems. Indeed, referential features induce (re)activations of referents (Bever & McElree, 1988; Frazier & Clifton, 1989; Gibson & Warren, 2004; Love, 2007; Love & Swinney, 1996; Nicol & Swinney, 1989). As new referents are introduced, the activation faces decay (Gibson, 1998, 2000): structurally induced (re)activation help refresh the referent (Gibson & Warren, 2004). Reactivations of referents facilitate their maintenance in the face of decay. Thus, while an operator-variable structure is computed in the processing of (6), chain-theoretic reactivations of the individual concept Aurélie will induce further conceptual structure computations mediated by the (re)activation of the discourse referent. (6) C’est Aurélie, dit Paul, qui lundi matin aura déjà vu Jean-Marie. It’s Aurélie, said Paul, who Monday morning will-have already seen Jean-Marie. In the conceptual structure component, the activation of an individual referent in discourse representation leads to the activation of related concepts in the semantic web (Meyer & Schvaneveldt, 1971, and many others). Thus, the activation of the referent Aurélie activates the properties Girl, Human and Female, as well as Animate. Degrees of activation depend on what is deemed contextually relevant. Crucially, the conceptual structure component constitutes an interface between the linguistic domain and visual information (Jackendoff, 1987; Marr, 1982), so that concomitant linguistic and visual processing can interact. We, therefore, consider the predictions resulting from cross-modal priming (Love & Swinney, 1996), exploring the interaction of concomitant linguistic and visual processing at the interface between linguistic and visual processing in the conceptual structure component. Thus, while respondents listen to sentences, picture probes (related to the syntactic processing or not) appear on a computer screen, timed either to coincide with a crucial processing moment to be investigated or a previous control position shortly before this point. Respondents make basic classificatory decisions on these probes and response times are measured. Faster classification times are expected for those picture probes processed concomitantly with congruent conceptual structure activations than for those picture probes unrelated to concomitant processing in the conceptual-intentional system. Thus, in computing an operator-variable structure as detailed in Figure 3.1 for the input in (5) in which an operator binds a trace, left behind after

76

The Acquisit ion of French in Mult ilingual Conte x t s

syntactic movement, certain syntactic operations are mandatory. The integration of the auxiliary verb aura into the partial structure requires that an agreement relation be computed in the Spec-head configuration involving the (referential) trace in the subject position. On standard minimalist assumptions about the nature of natural language grammars, the verb form spells out uninterpretable number and person features associated with the category Inflection (INFL or I). Values of uninterpretable number and person features of I receive their value from the (referential) ϕ-features of the subject (trace) in the Spec-head feature-checking configuration (Chomsky, 1995). This computation of agreement dependent on ϕ-features implicated in reference naturally reactivates the referent, with conceptual structure implications that are induced by the purely formal grammatical processing and unexpected otherwise. The classification of a picture probe matching or mismatching the referent of the cleft noun at different moments during the processing of the sequence qui lundi matin aura is expected to be affected by linguistic processing only under specific hypotheses about the nature of this processing. If the integration of the auxiliary verb into the structure involves an agreement relation implicating the (referential) ϕ-features of the subject trace, then faster classification times can be expected for those pictures matching the cleft noun aligned with aura than for such pictures aligned with the earlier matin ‘morning’, although they appear sooner. Two scenarios are possible depending on the relative facilitations to superordinate categories. If activation spreads most strongly for the immediately dominating superordinate category Human & Female for Aurélie relative to the opposite Human & Male, the profile on the classification is expected to be as in Figure 3.2, in which the picture classifications aligned with the auxiliary will be facilitated for matching human probes only. If activation spreads maximally to all (relevant) superordinate categories, then all classifications of human pictures aligned with the auxiliary will be facilitated, yielding the pattern in Figure 3.3. Crucially, the patterns in Figures 3.2 and 3.3 do not follow from the computational load associated with linear distance or the salience of the referent. Greater salience of the referent would account for faster classifications across target and control positions. The processing load associated with linear distance does not explain the pattern either: the smallest reaction times (RTs) are associated with the greatest linear distance between an NP and its matching probe. Thus, facilitations in the classification of antecedent-related picture probes concomitant with the auxiliary are suggestive of reactivations due to specific computations involving referential features. In fact, if L2 learners process (5) without computing syntactic chain dependencies on line, but instead learn procedures constraining word co-occurrences, the same type of interactions between the processing of the sequence qui lundi matin aura and conceptual structure processing are not expected.

On the Processing of Subjec t Clef t s in English-French Interl anguage

77

Figure 3.2 Expected results in milliseconds (msec) under structurally governed processing if facilitation is stronger for immediate superordinate categories versus opposites

In this kind of processing, the cleft noun contributes a discourse referent that must remain active until all semantic relations in the sentence have been resolved, including those beyond the auxiliary. Agreement does not involve a computation with a (referential) trace but consists of a constraint on the form of the upcoming verb. The structurally guided reactivations were missing in the processing of indirect object relatives in L2 English, during which no effect of position could be found. In the absence of chain dependencies,

Figure 3.3 Expected results in msec under structurally governed processing if facilitation spreads to all human categories

78

The Acquisit ion of French in Mult ilingual Conte x t s

the referents must be maintained until semantic relations are satisfied (Felser & Roberts, 2007). In this case, picture classifications could be sensitive to the salience of the focused element yielding the pattern in Figure 3.4, as in Felser and Roberts (2007). RTs could be fast across the board as in Figure 3.5, if activation spreads equally to all superordinate categories of which the referent is a member. We therefore propose investigating the processing of subject relatives in English-French interlanguage by considering evidence of a syntactically derived referential dependency between the parser and conceptual structure computations presumed to involve the mediation of referential features in the processing of subject clefts. Beyond the question of the degree to which and the conditions under which L2 learners compute UG-sanctioned syntactic representations at a given level of proficiency, an investigation of L2 processing provides a crucial window into the way in which the acquisition can proceed.

The Study Experimental procedures Our study involved several steps. Respondents first filled out a questionnaire providing their language background. A pretest was then given to the respondents in which they were trained to categorize pictures as either [±human]. The task was designed to familiarize respondents with all the

Figure 3.4 Expected results in msec under shallow structure processing if facilitation is stronger for immediate superordinate categories versus opposites

On the Processing of Subjec t Clef t s in English-French Interl anguage

79

Figure 3.5 Expected results in msec under shallow structures if activation spreads to all human categories

pictures used in the study as well as to train them on the categorization process that would be used in each of the subsequent tasks. After six examples, the pictures were presented in random order. Pictures were all preceded by a * that appeared for 833 milliseconds (msec). During this time, respondents were asked to fixate upon the * and then as soon as the picture appeared, they were to classify it as quickly as possible by pressing appropriate buttons on the keyboard. The left and right arrow buttons were designed as [−human] and [+human] responses, respectively. Respondents first trained with six items to ensure that they understood the instructions. Following the pretest, respondents next performed a masked prime picture classification task. The same pictures used in the pretest were now introduced by the sequence ######, which functioned as a forward mask. Although it was imperceptible to the respondents, a name – which was also used in the critical conditions in the main experiment (e.g. Paul/Aurélie) – appeared for 80 msec behind the forward mask. Immediately thereafter, a picture appeared to be classified by respondents as [±human]. Primes were considered either a match, where the prime Paul was followed by the picture of a boy, or a mismatch, where the prime Paul was followed by either a picture of a horse, or a picture of a girl. For all intents and purposes, this task appeared identical to the previous task; respondents were told that we were again making sure that they were absolutely clear how to respond to pictures on the keyboard. There were six training items, followed by 40 pictures. Each picture was preceded by a matching/mismatching masked prime. No one reported being aware of the primes in debriefings.

80

The Acquisit ion of French in Mult ilingual Conte x t s

Picture classification with masked primes allows for an independent consideration of the degree to which [+human] decisions on pictures overshadow the gender distinction. It is expected that seeing the name Aurélie for 80 msec would speed the facilitation of a matching girl picture with respect to a mismatching male picture or animal picture, as a result of congruent conceptual structure activations. Whereas primes generally trigger facilitations on related categories, inhibitions are observed in some individuals, presumably as a function of lexical access and maintained focus of attention on the referent. Individuals who, for whatever reason, cannot show priming with the names and pictures used in the main experiment, cannot reasonably provide the opportunity for a viable test of the hypotheses in the cross-modal priming task. We therefore retained only those individuals that showed priming, putting aside individuals that failed to show any priming on this screening task. This was operationalized by requiring that respondents showed speeded classifications of at least 10 msec for pictures with matching primes relative to pictures with mismatching primes. Thus, the purpose of the masked primed task was to screen individuals with inhibitory patterns where facilitation patterns were expected. The main experimental task was a cross-modal picture classification in which respondents classified pictures timed with control and targeted critical moments of the parse. It included 20 critical sentences organized in a Latin square as exemplified in Figure 3.6: four conditions crossed control positions (notated **) aligned with the onset of the second temporal noun (e.g. matin ‘morning’, après-midi ‘afternoon’, or soir ‘evening’) versus target positions (notated *) aligned with the onset of the auxiliary verb aura ‘willhave’ with probes matching either the cleft NP or an intervening NP that appeared in a parenthetical. The control positions occurred within a temporal modifier that offered a protective zone from any trace temporarily associated with the complement position of the verb. The control probe appeared just over a second

Figure 3.6 Sample quadruple items

On the Processing of Subjec t Clef t s in English-French Interl anguage

81

before the target position. Pictures were presented for 800 msec. RTs for ([±human] classifications were measured. This was controlled by DMDX (Forster & Forster, 2003). The stimulus was placed in a context in which the repeated use of subject clefts would be possible. Respondents were introduced to character Jacques Delaschmolle writing a spoof of Le Petit Nicolas using school playground drama. They were told that they would be listening to Jacques as he reminded himself of events that took place right around the one-week spring vacation. Respondents were told that, as they listened to Jacques, photographs would appear and to quickly classify as [±human] with left-arrow or right-arrow button presses as before. They were also told that, right after Jacques had spoken, a sentence would appear on the computer screen, which they needed to acknowledge as a true with the Y key for ‘yes’ or reject with the N key for ‘no’. Respondents were all right-handed. They used their right hand to classify pictures and their left-hand to respond to the comprehension checks. For 20 items, probes depicting school-aged girls or boys were balanced between males and females and were always related to one of two noun phrases in the sentence. Relatedness of picture probes to expressions in the sentence controls a crucial variable, since mismatched probes are also unrelated to the sentence. When mismatching probes are always unrelated to the sentence, the possibility of respondents adopting a task-taking strategy independent of structure is introduced. Yet, since classificatory decisions on probes involve [+human] value and both male and female referents are associated with the sentence, humanness might become the central property of referents, so that the male–female distinction is overshadowed. In view of this possibility, 10 foils in which animals unrelated to noun phrases in the sentences cycled in control and target positions provided alternative mismatching probes. This allowed us to examine the classification of gendermatched human probes in control and target position with the classification of (mismatching) animal probes in the same positions. Given that the animal pictures should be unaffected by the human names in the sentence, this provided a second test of the hypothesis. Twenty-two distracter items (also subject clefts) completed the experiment: they included both human and animal probes, so that the task was balanced between animal pictures and human pictures. Crucially, distracter items included non-human primate probes, the goal of which was to focus the attention of respondents away from the male–female distinction in the human probes. The number of (non-human) primate and nonprimate animal probes was balanced. The task was, consequently, seamless from the perspective of respondents and was short enough to sustain their attention by avoiding fatigue and boredom. In constructing items, care was taken to avoid common animal names. Popular female and male names were selected and appeared in both positions throughout. Familiarity with the names used in the experiment was pilot-tested to ensure that they

82

The Acquisit ion of French in Mult ilingual Conte x t s

would be known by most learners of intermediate proficiency. Results could not be expected if learners, for the most part, did not know the gender of the names appearing in the stimulus. The familiarity of our participants with the names was also tested as the last step in the data collection process with a questionnaire on 20 French names used in critical items. Our respondents indicated whether the name was a female name or a male name, with few errors. In the event that working memory (WM) in the L2 plays a role (i.e. the ability to process the L2 input and keep information in memory), respondents did a version of Harrington and Sawyer’s (1992) WM reading span test modelled for L2 learners of English after Daneman and Carpenter (1980). This follows Felser and Roberts’ (2007) study of the processing of indirect object relative clauses in L2 English. Respondents read aloud 42 sentences organized in 12 sets increasing in size from two to five sentences per set. They were asked to judge the grammaticality of the sentence, while maintaining the last word of each sentence in memory. After each set, respondents recorded these words in order on an answer sheet. WM scores represent the total number of recalled words out of 42, and the number of completely correct sets out of 12. In the native speakers’ WM task, the sentences were longer by five to seven words, making them comparable in length to Daneman and Carpenter’s (1980) task (we gave them the same test as the non-natives). WM measures in the L2 may be clouded by proficiency and often fail to be a useful indicator of L2 syntactic processing (Juffs & Harrington, 2011), as was the case in Felser and Roberts’ (2007) study. For this reason, we substituted the masked prime classification task as a screening test, but included WM scores for comparisons.

Respondent characteristics The subject pool included lower intermediate L1-English learners (n = 30) in their second and fourth semesters of college French, as well higher intermediate L1-English learners (n = 28), in their sixth and eighth semesters of college French. In addition, we tested advanced L1-English learners (n = 22) who were graduate students in French studies currently teaching French at the college level. Finally, we tested native speakers (n = 14), who were both graduate students and faculty members. The advanced learners, who had studied French for an average of 9.2 years (S.D. = 3.56), had spent an average of 10 months (S.D. = 2.61) in France or other French-speaking countries. These learners had also started learning French in school, at or after the age of 12. Predictions of faster processing for pictures matching the referent only hold to the extent that respondents can successfully maintain information in memory in a manner that produces facilitations (rather than inhibitions or no priming) by gendermatched primes. As expected, some subjects in each group either produced

On the Processing of Subjec t Clef t s in English-French Interl anguage

83

inhibitions, showing increased times for matching pictures, or were flat. Twenty-four out of the 30 lower intermediate learners showed priming, with WM scores (M = 2, 0–7/M = 23, 15–34). Out of the 28 lower intermediates, 18 individuals showed priming, with WM scores (M = 3, 1–5/M = 24, 16–34). For the 22 advanced learners, 16 showed priming, with WM scores (M = 6, 1–10/M = 30, 9–40). Lastly, 10 out of the 14 native speakers showed priming on this task, with WM (M = 8, 7–10/M = 37, 35–40). WM scores obtained in French were essentially tied to L2 proficiency, a factor that clearly interferes with the effectiveness of the task at predicting processing behaviour. The respondents’ characteristics on the picture classification task with masked primes can be seen by group in Figure 3.7. An ANOVA revealed no effect of group or qualification by group. There was a main effect of condition, F(1, 64) = 17.956, p < .0005. Specifically, gender-matched human probes were classified more quickly than animal probes, t(67) = 4.815, p < .0005, and gender-matched human probes were classified more quickly than gendermismatched human probes, t(67) = 2.649, p < .05. On this task, the gender of the names played a significant role.

Analytical procedures Only correct classifications were considered. Measures that were two standard deviations beyond the means were discarded and replaced by means. For each respondent, mean RTs per condition were obtained. The specific patterns targeted as a function of the manner in which activation spreads across categories were presented in Figures 3.2 and 3.3. Faster classifications of gender-matched probes in target position (C1) than of 485 480 475 470 465

Lower Intermediates, n = 24 Higher Intermediates, n = 18 Advanced Speakers, n = 16 Native Speakers, n =10

460 455 450 445 440 Human/ Match

Human/ Mismatch

Animal/ Mismatch

Figure 3.7 Results in msec from picture classification pretest task with masked primes

84

The Acquisit ion of French in Mult ilingual Conte x t s

gender-mismatched probes in the same position (C2) would indicate greater activation of immediate superordinate categories in the semantic network. Faster classifications of gender-matched probes in target (C1) versus control positions (C3) would indicate the mediation of a referential chain in agreement. Possible interactions were examined with a mixed design ANOVA examining the behaviour across groups. We then examined possible asymmetries between C1 and C2 as well as between C1 and C3, with α = 0.05 for theoretically driven planned t-tests. Asymmetric predictions warrant onetailed significance. Two-tailed values are provided unless otherwise specified. Given that RTs are positively skewed, statistics were run after log transformations. Reports of statistical significance are followed by reports of effect sizes with partial eta squared values for ANOVA results and Cohen’s d values for t-test results.

Results Figure 3.8 graphs respondents’ RT profiles for gender-matched versus gender-mismatched human probes on the cross-modal picture classification task. The descriptive statistics are provided in Table 3.1. For gender-matched versus gender-mismatched human probes, an ANOVA revealed no main effect of group, F(3, 64) = 1.557, p = 0.209. Nor was there any interaction with group. The ANOVA, however, revealed a main effect of position, F(1, 64) = 25.892, p < .0005, η2partial = 0.059. The prediction of faster classifications of gender-matched probes in target (C1) versus control positions (C3) was verified: t(67) = 3.512, p = 0.001, d = 0.430. Classification times for human probes aligned with the auxiliary were faster than decision times for

710 690 670 650 Lower Intermediates, n = 24

630

Higher Intermediates, n = 18

610

Advanced Speakers, n = 16 590

Native Speakers, n = 10

570 550 530 Control/ Match

Control/ Mismatch

Target/ Match

Target/ Mismatch

Figure 3.8 Profile for human probes in msec: Cross-modal priming experimental task

On the Processing of Subjec t Clef t s in English-French Interl anguage

85

Table 3.1 Reaction times in milliseconds for human probes: Cross-modal priming task Position

Lower intermediate Higher intermediate Advanced speakers Native speakers M

Control Gender match Gender mismatch Target Gender match Gender mismatch

S.D.

M

S.D.

M

S.D.

M

S.D.

675.16 125.314

647.15

53.266

648.43 99.852 637.71 90.619

695.50

98.937

634.89

84.345

636.08 97.225 618.30 83.888

646.42

66.838

620.25

87.902

598.00 70.677 607.25 84.648

634.58

77.659

633.27

52.532

600.60 75.935 599.25 71.446

human probes in the control position. This was independent of gender match. This general effect is consistent with the pattern in Figure 3.3 in which attention resources are focused on the entire category human, rather than restricted to the complex category human female or male which requires dampening of the opposite category. Crucially, on general grounds of linear distance between name and matching probe, target position probes matching cleft nouns might have induced the longest latencies. This is not what happened. The classification of all human probes was speeded by the processing of the auxiliary verb. Figure 3.9 graphs respondents’ RT profiles for gender-matched (human) probes versus mismatched animal probes. Classification times for gendermatched human probes in control and target positions are contrasted with classification times for animal probes in the same positions. Table 3.2 provides the descriptive statistics. There was again no effect of group in the ANOVA, F(3, 64) = 1.862, p = 0.145. A main effect of position was found, F(1, 64) = 7.007, p < .05, η2partial = 0.099, as well as a crucial interaction of probe with position, F(1, 64) = 6.226, p < .05, η2 partial = 0.089. Thus, classification times for gender-matched human probes aligned with the auxiliary were faster than for animal probes as predicted, t(67) = 1.829, p < .05 (one-tailed) , d = 0.224. This happened when classification times in control positions were no faster for gender-matched human probes than for animal probes, t(67) = 0.646, p = 0.521. Recall that classification for gender-matched human probes aligned with the auxiliary were faster than for gender-matched human probes in control position, with p < .005, as indicated earlier. Hence, it is clear that there was no evidence of activations maintained continuously across the sequence: specific moments of processing, here agreement, induce specific processing in conceptual structure.

86

The Acquisit ion of French in Mult ilingual Conte x t s 700 680 660 640 620 Lower Intermediates, n = 24 600

Higher Intermediates, n = 18 Advanced Speakers, n = 16

580

Native Speakers, n = 10

560 540 Control/ Human Match

Control/ Animal Mismatch

Target/ Human Match

Target/ Animal Mismatch

Figure 3.9 Profile for [± human] probes in msec: Cross-modal priming experimental task Table 3.2 Reaction times in milliseconds for [± human] probes: Cross-modal priming task Position

Lower intermediate Higher intermediate Advanced speakers Native speakers M

Control Human match Animal mismatch Target Human match Animal mismatch

S.D.

M

S.D.

M

S.D.

M

S.D.

675.16 125.314 647.15

53.266

648.43

99.852

637.71 90.619

689.65 113.386 644.78

60.808

616.89

49.363

604.97 99.887

646.42

66.838

620.25

87.902

598.00

70.677

607.25 84.648

654.14

71.789

630.24

61.380

614.73

71.522

644.20 71.611

L2 Sentence Processing in Real Time Human probes were classified more quickly in target position at the onset of the auxiliary, as opposed to the control position at the onset of the temporal noun. Animal probes, in contrast, were classified in target and control positions at similar speeds. There was no evidence of maintained activation or effects of linear distance. The processing of the auxiliary induced faster classifications of (matching) human probes. We submit that this constitutes prima facie evidence of structurally constrained conceptual

On the Processing of Subjec t Clef t s in English-French Interl anguage

87

structure processing mediated by a syntactically derived referential chain, also implicated in agreement. Indeed, if respondents were continuously maintaining referents under a domain-general linear strategy, then they might have been faster at classifying a matching probe at the first opportunity, that is to say, at the onset of the temporal noun. The evidence that probe classifications were speeded by the processing of the auxiliary, therefore, highlights an association between agreement and discourse/conceptual structure. This association is a hallmark of the syntactic system: a presumed reflex of the fact that ϕ-features computations are involved in referential chain-dependencies as well as in agreement. Syntactic chains thus implicate (interpretable) ϕ-features providing the values of uninterpretable ϕ-features associated with the category I, and spelled out in verbal morphology. In this grammatical organization, the referential chain derived in syntax mediates between the integration of the auxiliary verb and conceptual structure processing. Activations in conceptual structure concomitant with the grammatical processing of the auxiliary follow naturally from the role of referential ϕ-features in subject-verb agreement. We conclude that there is little reason for L2 learners to converge on such a processing profile unless they are guided to do so by virtue of mental design in the domain of language. This suggests that L2 sentence processing shares UG-governed representations and principles of performance with L1 sentence processing (e.g. Dekydtspotter et al., 2006; Frenck-Mestre, 2005; Frenck-Mestre & Pynte, 1997; Juffs, 1998, 2006; Juffs & Harrington, 1995; Williams et al., 2001). Evidence of UG-computations in real-time sentence processing across groups is important. The processing profiles that we have uncovered point to a specific grammatical architecture in which syntactic operations mediate between morphological paradigms and conceptual representations. The literature on relative clause acquisition suggests that indirect object relative clauses are harder to process and are acquired later. From the acquisition perspective, the difference between subject and indirect object dependencies is not surprising when bearing in mind Hawkins’ (1989) conjecture that L2 learners’ acquisition of relativization (and by extension clefts) is guided by configuration and processing costs. It is, therefore, likely that processing load might have yielded the results of Felser and Roberts (2007) for indirect object relative clauses in L2 English, in which no effect of structure was in evidence. In examining the processing mechanism in the cross-modal priming modality, the effects of structural complexity and processing load due to structure and task must be minimized. Subject clefts were, therefore, selected with these issues in mind. Since intermediate learners produce them, their complexity is clearly less, as is expected from acquisition research. Furthermore, it appears that Felser and Roberts’ (2007) conclusions must be further qualified: for indirect relative clauses, Miller (2011) presents evidence showing a priming effect in advanced L1-English L2-French acquisition, using picture classifications during reading. There appears to be a range of possible reasons

88

The Acquisit ion of French in Mult ilingual Conte x t s

why learners would produce distinct patterns including processing load due to structure and to task.

Parsing and L2 Acquisition In addition to the nature of representations computed in real time in L2 sentence processing, these priming patterns are relevant to the nature of the transition between grammatical states involved in the acquisition of an L2. Turning first to the issue of the grammatical states involved, evidence that French subject clefts are parsed in domain-specific ways by L2 learners points to an explanation of the acquisition of the relativizer qui in terms of FT/FA through feature reassembly (Lardiere, 2009a, 2009b; Schwartz & Sprouse, 1994, 1996). With UG limiting the morpho-syntactic treatment available to the relativizer qui, the sequence of grammatical states is straightforward: the relativizer qui is first mapped to the formal representation of who in English. Guided by representational constraints, the relativizer qui is then analyzed as a head (Rizzi, 1990; Sportiche, 2011). On this view, a grammar is constituted by the union of UG and the parametrized lexicon. The parser is constituted by this grammatical state augmented with a theory of performance. This is the simplest scenario involving UG-constrained states compatible with the evidence. It might, however, be pointed out that the computation of operatorvariable structures in subject clefts in real time does not rule out the development of linear strategies by domain-general mechanisms. Therefore, one may still want to entertain the possibility that, despite structural processing, a set of linear constraints is nevertheless what intermediate learners have extracted. In the spirit of Hawkins and Chan (1997) as well as Hawkins and Hattori (2006), it might thus be argued that certain (re)configurations of features detected by the parser cannot trigger a grammatical update after puberty and require various ersatz strategies. In this account, the accessibility of certain ϕ-feature (bundles) in the wh-marked C-head or wh-marked D-head in the case of a clitic treatment is limited to parsing and fails to extend to exponents of the functional lexicon. Knowledge of the relativizer qui then would take on the distinctly different character of linear co-occurrence statements. Such ersatz knowledge could not be effectively deployed in real-time processing. However, the feasibility of UG-constrained sentence processing failing to trigger the UG-constrained acquisition of a functional lexicon is conceptually problematic: if a grammatical representation is available in real time, it is difficult to see how it could be unavailable for the reassembly of feature matrices for categories and their morphological realizations (Dekydtspotter & Renaud, 2009; Lardiere, 2009a, 2009b). Given a grammatical architecture in which a largely universal parser mediates L2 acquisition, the task of explaining the nature of transitions

On the Processing of Subjec t Clef t s in English-French Interl anguage

89

between grammatical states in L2 acquisition resides in a much greater understanding of the changes to the (parametrized) lexicon in response to licensing failure. The motivation for updating the grammatical state seems clear: licensing failures are costly. Updating the grammatical state means (re)packaging features and/or feature values into the matrices of functional categories and relevant exponents with broad repercussions. Lardiere (2009a, 2009b) argues that feature reassembly is a complex operation that requires the accumulation of evidence. In the case of the French relativizer qui, the initial state of L1-English L2-French learners is well supported by the many sentences of French input that do not trigger the need for an update in the interlanguage grammar and de facto support the grammatical status quo potentially delaying grammatical change. Thus, in the Donaldson corpus of unguided conversations by native and near-native speaker dyads, who were either friends, co-workers or spouses, native speakers produced 46 subject clefts with 40 animate referents, which were also human, and only six of which involved inanimate referents. The productions of near-natives reproduce a similar asymmetry: there were a total of 32 subject clefts, 23 of which involved animate and human cleft nouns and nine involving inanimate cleft nouns (Donaldson, pc.). It seems that the data provided by natural communicative exchanges are not as robust as one might expect.

Acknowledgements The authors gratefully acknowledge the detailed comments and evaluation of two anonymous reviewers and as well the feedback received at the 11th Generative Approaches to Second Language Acquisition (GASLA) conference held at the University of Washington in Seattle. All remaining errors are the authors’ own.

Appendix: List of Sentences (1) C’est Lucas, dit Aurélie, qui mardi matin, aura déjà aidé Célestine. It is Lucas, says Aurelie, who Tuesday morning, will have helped Célestine. (2) C’est François, dit Lucie, qui mercredi matin, aura déjà vu Céline. It is François, says Lucie, who Wednesday morning, will have seen Céline. (3) C’est Romain, dit Lydie, qui vendredi matin, aura déjà attendu Guillaume. It is Romain, says Lydie, who Friday morning, will have waited for Guillaume. (4) C’est Rémi, dit Sophie, qui dimanche après-midi, aura déjà accompagné Marie-Claire.

90

The Acquisit ion of French in Mult ilingual Conte x t s

It’s Rémi, says Sophie, who Sunday afternoon, will have accompanied Marie-Claire. (5) C’est Benjamin, dit Magalie, qui jeudi après-midi, aura déjà écouté Jean-Luc. It’s Benjamin, says Magalie, who Thursday afternoon will have listened to Jean-Luc. (6) C’est Aurélie, dit Paul, qui lundi matin, aura déjà vu Jean-Marie. It’s Aurelie, says Paul, who Monday morning, will have seen Jean-Marie. (7) C’est Léa, dit Fred, qui jeudi matin, aura déjà cherché Henri-Luc. It’s Lea, says Fred, who Thursday morning, will have looked for Henri-Luc. (8) C’est Lucie, dit Guillaume, qui samedi matin, aura déjà embrassé Pierre-Jean. It’s Lucie, says Guillaume, who Saturday morning, will have hugged Pierre-Jean. (9) C’est Constance, dit Fabien, qui lundi après-midi, aura déjà appelé Jean-François. It’s Constance, says Fabien, who Monday afternoon, will have called Jean-François. (10) C’est Madeleine, dit Thomas, qui mardi après-midi, aura déjà rencontré Maxime-Henri. It’s Madeleine, says Thomas, who Tuesday afternoon, will have met Maxime-Henri. (11) C’est Vincent, dit Magalie, qui vendredi après-midi, aura déjà retrouvé Anne-Laure. It’s Vincent, says Magalie, who Friday afternoon, will have found Anne-Laure. (12) C’est Olivier, dit Marie, qui lundi soir, aura déjà contacté Marie-Chantal. It’s Olivier, says Marie, who Monday evening, will have contacted Marie-Chantal. (13) C’est Jacques, dit Chloé qui mardi soir, aura déjà dupé Marie-France. It’s Jacques, says Chloé, who Tuesday evening, will have fooled Marie-France. (14) C’est Hugo, dit Emma, qui mercredi soir, aura déjà aura oublié Marie-Christine. It’s Hugo, says Emma, who Wednesday evening, will have forgotten Marie-Christine. (15) C’est Fabien, dit Charlotte, qui jeudi soir, aura déjà critiqué Jean-Christophe. It’s Fabien, says Charlotte, who Thursday evening, will have criticized Jean-Christophe. (16) C’est Fanny, dit Robert, qui mercredi après-midi, aura déjà entendu Jean -Bertrand. It’s Fanny, says Robert, who Wednesday afternoon, will have heard Jean-Bertrand. (17) C’est Cécile, dit Quentin, qui samedi après-midi, aura déjà présenté Anne-Laure.

On the Processing of Subjec t Clef t s in English-French Interl anguage

91

It’s Cécile, says Quentin, who Saturday afternoon, will have introduced Anne-Laure. (18) C’est Marie, dit Quentin, qui dimanche après-midi, aura déjà embêté Jean-Pierre. It’s Marie, says Quentin, who Sunday afternoon, will have bothered Jean-Pierre. (19) C’est Alexandra, dit Nico, qui vendredi soir, aura déjà ennuyé Anne-Sophie. It’s Alexandra, says Nico, who Friday evening, will have annoyed Anne-Sophie. (20) C’est Charlotte, dit Mattéo, qui samedi soir, aura déjà soutenu Anne-Marie. It’s Charlotte, says Mattéo, who Saturday evening, will have supported Anne-Marie. (21) C’est Clément, dit Juliette, qui dimanche soir, aura déjà dénoncé Marie-Anne. It’s Clément, says Juliette, who Sunday evening, will have denounced Marie-Anne. (22) C’est Paul, dit Emma, qui mardi soir, aura déjà peint Marie-Jeanne. It’s Paul, says Emma, who Tuesday evening, will have painted Marie-Jeanne. (23) C’est Enzo, dit Clara, qui mercredi soir, aura déjà abandonné Marie-Ange. It’s Enzo, says Clara, who Wednesday evening, will have abandoned Marie-Ange. (24) C’est Nathan, dit Manon, qui jeudi soir, aura déjà protégé Marie-Claire. It’s Nathan, says Manon, who Thursday evening, will have protected Marie-Claire. (25) C’est Christophe, dit Maddy, qui mardi matin, aura déjà surpris Chantal-Marie. It’s Christopher, says Maddy, who Tuesday morning, will have surprised Chantal-Marie. (26) C’est Simone, dit Enzo, qui lundi soir, aura déjà étudié Jean-Paul. It’s Simone, says Enzo, who Monday evening will have studied Jean-Paul. (27) C’est Inès, dit Eliott, qui vendredi matin, aura déjà suivie Marie-Line. It’s Inès, says Eliott, who Friday morning, will have followed Marie-Line. (28) C’est Audrey, dit Mathis, qui samedi matin, aura déjà félicité Jean-Georges. It’s Audrey, says Mathis, who Saturday morning, will have congratulated Jean-Georges. (29) C’est Sarah, dit Serge, qui dimanche matin, aura déjà accueilli Paul-Henri. It’s Sarah, says Serge, who Sunday morning, will have welcomed Paul-Henri. (30) C’est Sandrine, dit Paul, qui lundi matin, aura déjà accusé Marie-Madeleine. It’s Sandrine, says Paul, who Monday morning, will have accused Marie-Madeleine. (31) C’est Paul, dit Aurélie, qui mercredi matin, aura déjà sollicité Jean-Baptiste.

92

The Acquisit ion of French in Mult ilingual Conte x t s

It’s Paul, says Aurélie, who Wednesday morning, will have questioned Jean-Baptiste. (32) C’est Romain, dit Lydie, qui dimanche matin, aura déjà persuadé Anne-Laure. It’s Romain, says Lydie, who Sunday morning, will have persuaded Anne-Laure. (33) C’est Rémi, dit Sophie, qui mardi après-midi, aura déjà accompagné Marie-France. It’s Rémi, says Sophie, who Tuesday afternoon, will have accompanied Marie-France. (34) C’est Sophie, dit Lucas, qui jeudi matin, aura déjà assisté Anne-Josée. It’s Sophie, says Lucas, who Thursday morning, will have assisted Anne-Josée. (35) C’est Céline, dit François, qui vendredi matin, aura déjà cherché Jean-Michel. It’s Céline, says François, who Friday morning, will have looked for Jean-Michel. (36) C’est Manon, dit Vincent, qui samedi matin, aura déjà convaincu Jean-Bertrand. It’s Manon, says Vincent, who Saturday morning, will have convinced Jean-Bertrand. (37) C’est Lucie, dit Fabien, qui lundi matin, aura déjà amusé Pierre-Jean. It’s Lucie, says Fabien, who Monday morning, will have amused Pierre-Jean. (38) C’est Constance, dit Thomas, qui mercredi après-midi, aura déjà diverti Jean-Fabien. It’s Constance, says Thomas, that Wednesday morning, will have entertained Jean-Fabien. (39) C’est Madeleine, dit Thomas, qui jeudi après-midi, aura déjà reconnu Jean-Luc. It’s Madeleine, says Thomas, who Thursday afternoon, will have recognized Jean-Luc. (40) C’est Fanny, dit Robert, qui vendredi après-midi, aura déjà rassuré Marie-René. It’s Fanny, says Robert, who Friday afternoon, will have reassured Marie-René. (41) C’est Serge, dit Sarah, qui dimanche matin, aura déjà calmé Paul-Henri. It’s Serge, says Sarah, who Sunday morning, will have calmed Paul-Henri. (42) C’est Paul, dit Sandrine, qui lundi matin, aura déjà conseillé Marie-Madeleine. It’s Paul, says Sandrine, who Monday morning, will have counselled Marie-Madeleine. (43) C’est Vincent, dit Lydie, qui mercredi matin, aura déjà trouvé Anne-Laure. It’s Vincent, says Lydie, who Wednesday morning, will have found Anne-Laure.

On the Processing of Subjec t Clef t s in English-French Interl anguage

93

(44) C’est Romain, dit Sophie, qui dimanche matin, aura déjà agacé Jean-Baptiste. It’s Romain, says Sophie, who Sunday morning, will have bugged Jean-Baptiste. (45) C’est Luc, dit Aline, qui mardi après-midi, aura déjà consolé Marie-France. It’s Luc, says Aline, who Tuesday afternoon, will have consoled Marie-France. (46) C’est François, dit Sophie, qui jeudi matin, aura déjà blessé Anne-Josée. It’s François, says Sophie, who Thursday morning, will have hurt Anne-Josée. (47) C’est Céline, dit Lucas, qui vendredi matin, aura déjà soulagé Jean-Michel. It’s Céline, says Lucas, who Friday morning, will have comforted Jean-Michel. (48) C’est Fanny, dit Fabien, qui samedi matin, aura déjà sauvé Jean-Bertrand. It’s Fanny, says Fabien, who Saturday morning, will have saved Jean-Bertrand. (49) C’est Lucie, dit Vincent, qui lundi matin, aura déjà attrapé Jean-Pierre. It’s Lucie, says Vincent, who Monday morning, will have caught Jean-Pierre. (50) C’est Flore, dit Thomas, qui mercredi après-midi, aura déjà réprimandé Jean-Fabien. It’s Flore, says Thomas, who Wednesday afternoon, will have reprimanded Jean-Fabien. (51) C’est Océane, dit Kévin, qui jeudi après-midi, aura déjà grondé Jean-Luc. It’s Océane, says Kévin, who Thursday afternoon, will have yelled at Jean-Luc. (52) C’est Manon, dit Clément, qui vendredi après-midi, aura déjà réécouté Marie-René. It’s Manon, says Clément, who Friday afternoon, will have listened to Marie-René again.

4

Verbal Transitivity Development in First Language Acquisition: A Comparative Study of Russian, French and English Anna Frolova

Introduction In the early stages of linguistic development, children optionally omit sentence constituents, a phenomenon that is largely attested across different languages. This chapter examines direct object omissions in child Russian and compares Russian data to L1 French and English. In adult grammars, object omissions occur to a different degree across the languages of the world depending on the particular restrictions on object drop in a given language. Very little work has been done on object drop in Russian despite the fact that this language exhibits frequent object omissions and it does not seem to display visible morphology of verb-object agreement. Nevertheless, through a careful examination of a target language we can make testable predictions about transitivity development in first language acquisition. It is difficult to understand object omissions in child language without identifying the grammatical nature and contexts in which null objects can appear in adult speech. A comparison of developmental patterns in typologically different languages, such as French, English and Russian, can help us to hypothesize about universally represented structures as the starting point of language acquisition and about the grammatical elements that are specific to a particular language and that need to be acquired. The present study examines object omissions in adult and child Russian and compares new findings to previously reported results from French and English. With the exception of a single work (Gordishevsky & Avrutin, 94

Verbal Transit iv it y Development in First L anguage Acquisit ion

95

2003, 2004), object omissions in transitive contexts have not been studied before. Thus, the main purpose of this research is to see how children use direct objects in optional contexts in order to determine their patterns of use, and to identify contexts in Russian where objects cannot be omitted. First, we establish the contexts in adult Russian in which object omissions are allowed. By following a syntactic approach to verbal transitivity, we can better understand object omissions in adult Russian speech and make several predictions about child language development. For instance, we can predict that Russianspeaking children may have an object omission stage similar to their French and English peers. Omission rates in early stages will be higher in child production. Moreover, Russian-speaking children will retain this stage longer than English-speaking children because of the variability of null object contexts in the input. These predictions will be tested in an elicited production study targeting null object use in young Russian monolingual children. This chapter is organized as follows. In the second section, previous research on object omissions in adult grammars is presented. Furthermore, the syntactic transitivity approach (Cummins & Roberge, 2005) is applied to Russian data in order to elaborate further on object omission research by identifying contexts for object drop in adult Russian. The third section examines previous work on null objects in L1 acquisition. The fourth section presents a Russian replication of Pérez-Leroux et al.’s (2008) study on object omissions in French and English. The final section discusses the results and makes assumptions for future exploration of the null object phenomenon in Russian first language acquisition.

Null Objects in Adult Grammars Direct objects can be left unpronounced in many languages of the world. Some languages allow frequent object drop (Russian), while others are more restrictive (e.g. French, English). Object omission is allowed in languages with morphologically overt verb-object agreement, such as Pashto, Georgian, Swahili and Chichewa (Huang, 2000); however, this is not a necessary condition for object drop. Null objects also exist in languages which lack verbobject agreement, for example, in Chinese, German, Russian, and in many others (Huang, 2000; Sigurðsson, 2011). If, following Roberge (2007) and Cummins and Roberge (2005), we assume an obligatory representation of the direct object in the syntactic structure, we should be able to distinguish two contexts, referential and nonreferential, where null objects can be used, depending on grammatical restrictions in the particular language. According to Cummins and Roberge, non-referential null objects are bare cognate nouns by syntactic nature with generic or pragmatic interpretation. In English and French, this type of null object is often used with activity

96

The Acquisit ion of French in Mult ilingual Conte x t s

verbs, for example, to read, to eat, to draw. Non-referential null objects do not have an antecedent in the preceding context or a referent in the extra-linguistic discourse. The semantics of omitted objects in non-referential contexts are derived from the semantics of the verb. They can correspond to a prototypical object of that verb in the presence of contextual clues or they can receive a stereotypical interpretation if contextual clues are absent. The null object in (1) exhibits the maximum specific reading due to the contribution of pragmatic principles, whereas the omitted object in (2), whose semantics is derived from the verb, is interpreted as the least specific. (1) Je vais acheter un magazine au kiosque, et je lirai en t’attendant. I’ll buy a magazine at the stand, and I’ll read ØACC while I’m waiting for you. (Cummins & Roberge, 2005) (2) Pendant mon congé sabbatique j’ai toujours l’intention de lire. During my sabbatical I mainly intend to read ØACC. (Cummins & Roberge, 2005) Object omissions in non-referential contexts are the most common type of null objects allowed in many languages, for instance, in French (3), English (4) and Russian (5). (3) La magie des séries, c’est de surprendre, de dépayser. The magic of the playoffs is in surprising ØACC, disorienting ØACC. (Cummins & Roberge, 2004; cited from Larjavaara, 2000) (4) This is a lovely guitar, with an uncanny ability to impress ØACC and delight ØACC. (Cummins & Roberge, 2005) (5) U nego osobyj talant očarovyvat’. At himGEN particularADJ ACC talent ACC to charm ØACC. ‘He has a particular talent to charm Ø.’ (McShane, 2005) However, not all languages allow non-referential object drop. In the nonreferential context in Chinese, a bare generic noun must be used (6), (7), or can be left unpronounced only if the verb is followed by an adverb (8) (Tieu, 2008). (6) Lisi zai chang ge Lisi PROG sing song ‘Lisi is singing.’

Verbal Transit iv it y Development in First L anguage Acquisit ion

97

(7) ta zai pao bu he PROG run step ‘He is running.’ (8) ta pao (*bu) de hen kuai he run step DE very fast ‘He runs very fast.’ (Tieu, 2008) A different type of null object is used in referential contexts. Referential null objects can be clause-internally constrained by morphological agreement and they are clause-externally linked to an antecedent (Sigurðsson, 2011). Languages that require morphological agreement or accusative clitics to identify omitted objects are classified as pro drop languages. In topic drop languages, null objects are linked to an antecedent without identifying morphological agreement. Referential null objects are common in Romance languages, for example, in French where this type of object is allowed in the presence of the accusative clitic (9a). Unlike topic omissions, the referential null object in French is clause-internally constrained and is characterized as a pro. The semantics of referential pro ‘correspond to its φ-features (person, number, gender), which are recovered through feature-checking with the clitic’ (Cummins & Roberge, 2005: 52). On the other hand, clitic drop constructions are also attested in French (9b), although not all native speakers consider clitic omissions to be grammatical (Cummins & Roberge, 2005; Larjavaara, 2000). Both sentences in (9) share the same interpretation: (9) a. J’ai appelé Jeani mais je n’ai pas pu lei rejoindre Øi. (null object recovered by clitic) b. (*) J’ai appelé Jeani mais je n’ai pas pu rejoindre Øi. (clitic drop construction) ‘I called Jean, but I couldn’t reach him.’ (Cummins & Roberge, 2005) In the constructions where pro is recovered by an accusative clitic, null objects have a rigid interpretation due to the clitic. Semantic and pragmatic contributions are not significant for the empty pro interpretation as is the case for non-referential null objects. English does not allow referential null objects because this language ‘does not have an empty category pro or the means for its recovery by a clitic or other agreement marker. This type of referential null objects would be unrecoverable in English’ (Cummins & Roberge, 2005: 60, footnote). The pronoun ‘it’ must be used in this context as in (10).

98

The Acquisit ion of French in Mult ilingual Conte x t s

(10) What do you think of my cake? *I like Ø. (Cummins & Roberge, 2005) Russian grammar has no clause-internal restrictions (e.g. object clitic or verb-object agreement markers) on object omissions (Gordishevsky & Avrutin, 2003; McShane, 2005); nevertheless, object omissions seem to be frequent in adult speech if the referent was mentioned in the previous context as in (11) and (12) (see Miller & Weinert, 1998 for a statistical report on referential null subjects and objects in Russian). In referential contexts, Russian allows optionally omitted subjects and objects as is shown in the following examples. Following Sigurðsson’s (2011) approach, the omitted object in Russian could be considered a strong accusative pronoun. In (11b), both answers are possible and have the same meaning: (11)

a: b:

Čto ty delaeš’ s ètim rasteniem? what youNOM do1SG with this plant ACC ‘What are you doing with this plant?’ Polivaju./Ja polivaju ego. ØNOM water1SG ØACC/INOM water1SG it ACC ‘(I) am watering (it)./I am watering it.’ (Gordishevsky & Avrutin, 2004)

(12) Ja slučajno vypustila pticu, no brat pojmal ØACC. I accidentally let go3 SG PAST bird ACC but brotherNOM caught3d SG PAST ØACC. ‘I accidentally let the bird fly but my brother caught (it).’ (McShane, 2005) In referential contexts Russian can be compared to topic drop languages, such as Chinese (13) and German (14), because the referential object omissions are not preconditioned by verb-object agreement. The following example illustrates the omission of a referential object in Chinese: (13) Ta kanjian le. he see.Ø-AGR PERF.Ø-AGR ‘He saw [him/her, etc.].’ (Sugurðsson, 2011: 269) In German, dropped referential objects are usually first topicalized in the sentence initial position (14b). When the topic position is occupied by another element (e.g. the subject), the object cannot be omitted (14c). Only one argument can be dropped per sentence. (14) a. Ich hab’ ihn schon gesehen. I have him already seen. ‘I saw him already.’

Verbal Transit iv it y Development in First L anguage Acquisit ion

b. e Hab’ ich schon gesehen. e have I already seen ‘I saw (him/it/her/them) already.’ c. *Ich hab’ e schon gesehen. I have e already seen. ‘I saw *(him/it/her/them) already.’ (Huang, 1984)

99

e = empty

Unlike in German, in Russian the object can be omitted from any sentence position. Similarly to Chinese, several arguments can be omitted from the same sentence (11b). As is illustrated in (15), an omitted referential object in Russian must be linked to the antecedent or have a referent in the extralinguistic context (16)–(17). (15) a: Xočeš’ jabloko? ØNOM wantPRES 2d SG appleACC ‘Do you want an apple?’ b: Xoču. ØNOM want PRES IMPF 1st SG ØACC ‘Yes, (I) want (it).’ (Gordishevsky & Avrutin, 2003) (16) Idi suda, poceluju. ØNOM comeIMPER 2d SG hereADV, ØNOM kissFUTURE PF 1stPS SG ØACC ‘Come here, (I) will kiss (you).’ (Gordishevsky & Avrutin, 2003) (17) [Context: something falls, everybody looks.] Ne vstavaj, ja podnimu. ØNOM NEG get-upIMER 2d SG, INOM getFUTURE PF ØACC ‘Don’t get up. I’ll get (it).’ (Gordishevsky & Avrutin, 2003) According to Cummins and Roberge’s (2005) typology, (16) and (17) represent ‘deictic’ null objects. The semantics of deictic null objects comes from a salient referent in the extra-linguistic context. Deictic null objects are allowed in French, English and Russian, and are frequently used in imperative constructions. (18) (A hands a paper to B and says): Tiens, lis! ‘Here, read ØACC!’ (Pirvulescu & Roberge, 2004) To summarize, Russian allows frequent object omissions in referential and non-referential contexts, but there is no visible verb-object agreement

100

The Acquisit ion of French in Mult ilingual Conte x t s

morphology and no accusative clitic system. In referential contexts, Russian behaves similarly to topic drop languages, such as Chinese. In non-referential contexts, Russian allows object drop with activity verbs similarly to French and English. According to previous studies on object drop in Russian (Gordishevsky & Avrutin, 2003, 2004; McShane, 2005), null objects are optional in adult grammar. Are there any syntactic restrictions on object drop in adult Russian? I leave this question for future research, but will return to it in the discussion part of this chapter.

Object Omissions in First Language Acquisition Object omission in different child languages has been addressed by many studies. It was observed that in early stages of acquisition children optionally omit direct objects. Several studies explain argument omissions in child language (until age three) from a performance limitation perspective (Bloom, 1990) or base their analyses on the child’s pragmatic development (Gordishevsky & Avrutin, 2003, 2004; Schaeffer, 2000). Gordishevsky and Avrutin (2003, 2004), in the only study available on direct object omissions in monolingual acquisition of Russian, attribute null objects in child production data to the child’s pragmatic development. These authors conducted a longitudinal study on spontaneous productions of six Russian-speaking children. Results from two children (aged 1;09–2;06) were used to analyze direct object omissions. Gordishevsky and Avrutin (2003: 14) examined object omissions of obligatory transitive verbs and ‘optionally transitive verbs that were used in clearly transitive contexts’. According to their findings, ‘children violate the adult constraints on argument omissions in that they make no verbal reference to the omitted element; do not provide non-linguistic cues; and they use event time that is not close to the time of speech’. This observation is explained by the development of the pragmatic system; in particular, the authors hypothesize that young children have not yet acquired the pragmatic ‘concept of non-shared knowledge’ as formulated by Schaeffer (2000). Non-adult-like performance of object omissions decreases in finite contexts from 24% at Stage 1 (aged l;09–2;0) to 11% at Stage 2 (aged 2;0–2;06) and remains at a rate of around 40% in the non-finite contexts. The most common children’s non-target-like responses reported by Gordishevsky and Avrutin are illustrated in (19), where we can observe different types of omissions: (19) Non-adult like object omissions in L1 Russian a. CHILD: Mal’čiki boysNOM ‘Boys.’

(1;11.19)

Verbal Transit iv it y Development in First L anguage Acquisit ion

101

ADULT:

A čto oni delajut? and what theyNOM are doingPRES IMPF ‘And what are they doing?’ CHILD: Stroiti _. ØNOM are buildingPRES IMPF 2d PL ØACC ‘(They) are building.’ (The child used 2nd person instead of 3rd person.) ADULT: Čto? ‘What?’ b. (Previously played with a stick, goes to mother) (1;10.28) CHILD: Mame dam _. ØNOM mommyDAT giveFUTURE PF 1SG ØACC ‘(I) give (it) to mommy.’ c. CHILD: Dostan’_. (1;10.11) getIMPERATIVE PF 2d SG ØACC ‘Get.’ ADULT: Čto tebe dostat’? whatACC youDAT getINFIN PF ‘What (do you want me to) get for you?’ CHILD: Dostan’ igrušku getIMPERATIVE PF 2dPS SG toyACC ‘Get (me) a toy.’ d. (Zlata plays with a dog. The dog runs away, and the child falls.) (1;9.24) CHILD: Sobaka. dogNOM ‘(The) dog.’ ADULT: Čto? ‘What?’ CHILD: Pobila. ØNOM beatPAST PF 3dPS SG ØACC ‘(It) beat.’ ADULT: Kogo sobaka pobila? whoACC dogNOM beatPAST PF ‘Who did the dog beat?’ CHILD: Zlatu. Zlata ACC ‘(The dog beat) Zlata.’ (Gordishevsky & Avrutin, 2003) While pragmatics is certainly developing in these early stages of language acquisition, Gordishevsky and Avrutin’s pragmatic approach cannot fully explain null objects in children’s production because it does not allow us to clearly identify non-target-like omissions. For instance, the responses in

102

The Acquisit ion of French in Mult ilingual Conte x t s

(19a, 19b and 19c) seem possible in adult language. By applying the syntactic transitivity account to adult Russian, we have a better method for identifying non-target-like omissions in child speech. From this perspective, the example in (19a) seems possible in adult speech, since the verb ‘to build’ can appear with a non-referential null object without a linguistic or extra-linguistic referent. Similar to verbs that are often called ‘optionally transitive’ in dictionaries, such as to draw, to write, to eat, the verb ‘to build’ can be used with a non-referential null object in French or English. Used without an overt object to emphasize the role of the subject in the event, the semantics of ‘to build’ provides the stereotypical interpretation for the null cognate object. The null objects in (19b) and in (19c) are also possible in adult speech, and they could be interpreted as deictic objects with a reference to the extra-linguistic context. In (19b) this interpretation is made available by the description of the context in which the object is mentioned. However, (19d) seems different from the other examples. The interpretation of the omitted object here is difficult because of the missing referent. While for the child this could be a deictic null object since she is talking about herself, an adult would presumably use a DP in this context. However, as was noted by an anonymous reviewer, the example in (19d) is produced by a child aged 1;9, and the omitted object can be attributed to performance limitations. This type of omission in older children may involve syntactic development which is described in the discussion section. Performance accounts and analyses making use of pragmatics examine omissions in child production until the age of three. Interestingly, according to many studies on Romance languages, illicit null objects are attested in child production well after the age of four (Grüter, 2006; Pérez-Leroux et al., 2008). Most of the literature on French and other Romance languages investigates object drop in constructions which contain an object clitic. According to several studies, object drop can be explained by difficulties in the acquisition of accusative clitics (see Hamann, 2003; Jakubowicz & Rigaut, 2000, among others). However, Wang et al. (1992) and Pérez-Leroux et al. (2008) report that non-target-like object omissions are also attested in child English, a non-clitic language. Moreover, the object drop stage in French has an optional character when children omit clitics who, at the same time, produce target-like clitic constructions. This finding seems to indicate that children know when to use an object clitic, but they choose to omit it optionally (Pérez-Leroux et al., 2008). Müller et al. (1996, 2006) propose that object omissions in early child grammar are due to the existence of universally available topic drop constructions attested in adult Asian languages. If children with a different L1 start with topic drop constructions, we would find similar omission rates and developmental patterns across languages. However, as it was shown in studies on typologically different languages, children with different L1s omit objects at different rates (Grüter, 2006; Pérez-Leroux et al. 2008; Wang et al., 1992).

Verbal Transit iv it y Development in First L anguage Acquisit ion

103

Wang et al. (1992) compare object drop in child Chinese and English in the production data of three- and four-year-olds. Their results show different omission rates at age 3;3 for Chinese and at age 3;6 for English. In child English omission rates are low (4.2%), and in child Chinese rates are higher (22.5%). Chinese children omit objects similar to the target grammar, where null objects are grammatical, and the rate of omissions is stable across age groups. English-speaking children have few omissions in the early stages and omissions disappear from child production by age four. Wang et al. (1992) suggest that object omissions can be related to target language possibilities because child grammars converge early to the adult target. The fact that omission rates are different across languages and across age groups supports the idea that they cannot be explained only by performance factors development. Grüter (2006) replicates Wang et al.’s (1992) study in order to compare object omissions in child French production to omissions in child Chinese and English. According to Grüter (2006), omissions in L1 French at 3;7 occur at a higher rate than in English but at a lower rate than in Chinese. These differences across languages indicate that different rates in object drop can be attributed to language-internal constraints, and not to language-external performance limitations. Pérez-Leroux et al. (2008) compare child-elicited production in French and English. Differently from previous studies, the authors investigated object drop in two contexts, referential versus non-referential, following Cummins and Roberge’s (2005) classification. This method enables us to compare omissions in obligatory versus optional contexts for object realization, as well as to see the differences in object use across typologically different languages, such as English, which licenses only non-referential omissions, and French, which also allows pro drop in referential contexts. Pérez-Leroux et al.’s (2008) study based on the syntactic approach to verbal transitivity (Cummins & Roberge, 2004, 2005; Pirvulescu & Roberge, 2004; Roberge, 2007) provides a better understanding of null objects in different languages and, in addition, allows us to make several predictions about object omissions in child language. According to this approach, verbal transitivity is provided by UG, and children must acquire the particular semantic relation that exists between the verb and its object: A child hearing a verb with no object needs to incorporate this fact into his or her grammar. For each verb the child will learn from the input the likelihood of it appearing without an object. The child must also decide, for verbs likely to have an object, whether to interpret the absence of object as reflecting the speaker’s intention to deemphasize the role of the object in the event (i.e. no intent to individuate), or to reflect a grammatical fact about the licensing of referential null objects (i.e. with an individuated interpretation). (Pérez-Leroux et al., 2008: 384)

104

The Acquisit ion of French in Mult ilingual Conte x t s

It is assumed that children start with a default instantiation of verbal transitivity, which implies a V-N structure. The null object in early child grammar is a null cognate noun used in referential and non-referential contexts. This null object type is represented in child grammar, and also in adult grammar, in the latter only in non-referential contexts. In the course of language acquisition, the referential use of the null cognate object becomes unavailable in child grammar: (20) a. Stage I:

VN V Ni → clitic or pronoun acquisition (early) b. Stage II: V N Cl V proi (in French) Contextual experience blocks N (Pérez-Leroux et al., 2008: 386)

Pérez-Leroux et al. (2008: 386) predict a relation between the variability of null object constructions in the target language and the timing of acquisition: ‘Children whose target language includes more variety in types of null objects, retain a referential null cognate object N longer than children whose target language only has a subset of null object constructions.’ The results reported by Pérez-Leroux et al. (2008) are discussed in more detail in the following section.

The Study This study is a replication of Pérez-Leroux et al. (2008). Our main purpose is to obtain preliminary results of object use in child Russian, a language that has not been sufficiently studied in the past, and to elaborate more rigorously on controlled studies on object omissions in child Russian. The results presented in this chapter show preferences and tendencies in child Russian production in optional contexts. The data from Russian are compared to Pérez-Leroux et al.’s (2008) findings on French and English.

Hypotheses Based on Pérez-Leroux et al.’s (2008) study, I can formulate the following hypotheses for Russian: •

Russian-speaking children will go through an object omission stage similar to children acquiring French and English when they overuse the default null objects in both referential and non-referential contexts. This will result in more frequent omissions than in adult speech.

Verbal Transit iv it y Development in First L anguage Acquisit ion

•

•

105

Russian-speaking children will retain this omission stage longer than English-speaking children but as long as French-speaking children, because of the availability of referential and non-referential null objects in the target grammar. In Russian, the omission rates across contexts should be similar for each age group because the two contexts are optional for object realization or omission.

Participants Thirty-nine monolingual Russian-speaking children participated in the study. In addition, 12 monolingual adult participants served as a control group. The recordings took place at daycare centres in St. Petersburg, Russia. The children were age-matched with French- and English-speaking children and were divided into three age groups as in Pérez-Leroux et al. (2008). The threeyear-old group (n = 16) was between 2;10 and 3;09 years of age (mean = 3;06; S.D. = 3.6 months). The four-year-old group (n = 10) ranged in age from 4;0 to 4;11 (mean = 4;05; S.D. = 4 months). The five-year-old group (n = 13) was between 5;02 and 6;01 years of age (mean = 5;07; S.D. = 3.3 months).

Methods This study replicates Pérez-Leroux et al.’s (2008) experimental design in order to obtain Russian data that are directly comparable to results for object omissions in French and in English. With this method we are able to control object omissions in two pragmatically different contexts for object realization. The experiment included two standard elicited production tasks in referential and non-referential contexts. Children listened to illustrated stories and were asked to answer simple questions about given scenes. The stories employed by Pérez-Leroux et al. (2008) were translated into Russian. In the preliminary study on Russian, the exact pictures from Pérez-Leroux et al. (2008) were used, but due to cultural differences some pictures were not easy to understand for Russian children. As a result, the original pictures were changed to pictures that were more familiar to Russian participants. For example, pictures in the Russian study were taken from famous children’s books in Russian (‘Doctor Ai-Bolit’; ‘Winnie the Pouh’, and others). A frog puppet named Croco, which is the name of the puppet in the original study, was used to facilitate communication with the children and always provided incorrect answers. The children were asked to help the puppet to understand the situation and to respond correctly. The experimenter interacted with each child individually, in a separate room at the child’s daycare centre. The procedure was audio-recorded. The following five transitive activity verbs were used once per context: risovat’ ‘to draw’, stroit’ ‘to build’, est’ ‘to eat’, pit’ ‘to drink’, rezat’ ‘to cut’. These verbs have similar characteristics in French, English and Russian in that they

106

The Acquisit ion of French in Mult ilingual Conte x t s

are frequently named ‘optionally transitive’ because they can appear with or without a direct object. Pérez-Leroux et al. (2008) also employed the verb ‘to hit’ which was excluded from Russian data because it typically requires a prepositional inanimate object and not an accusative object as in English or French: (21) udarit’ po mjaču to hit at ball DAT ‘to hit the ball ACC.’ In the referential context, the direct object was explicitly mentioned in the question in order to introduce it as a linguistic antecedent (22): ‘What is X doing with Y?’ (22) Referential context (Pérez-Leroux et al., 2008: 388): Story: The little girl wants to draw the prettiest thing in the room. Look! She takes a pencil and draws the flower. Experimenter: Hey, Croco! What’s the girl doing? Croco (the puppet): I know! The girl is smelling the flower! Child: [No] Croco: No, the little girl isn’t smelling the flower? Experimenter: Please tell Croco what the little girl is doing with the flower. Child: [She’s drawing it]. The exact stories and a very similar frog puppet were used in the Russian replication. The prompt question for the referential context and possible responses in Russian are illustrated in (23). In this context I expected answers without an object (23a) or with an accusative pronoun (23b): (23)

Russian prompt: Skaži požalujsta Kroko, čto devočka delaet s cvetkom. TellIMPERATIVE please CrocoDAT what girlNOM doesPRES IMP with flowerINSTR ‘Please tell Croco what the little girl is doing with the flower.’ Types of answer Null object: Ona risuet. sheNOM drawsPRES IMP ØACC ‘She is drawing.’ Pronoun: Ona risuet ego. sheNOM drawsPRES IMP it ACC ‘She is drawing it.’

In the non-referential context, the questions did not contain direct objects (24, 25): ‘What is X doing?’ I anticipated responses with a null object (25a) or with a lexical object (25b):

Verbal Transit iv it y Development in First L anguage Acquisit ion

107

(24) Non-referential context (Pérez-Leroux et al., 2008: 388): Story: Look at Clifford. He seems very happy. Experimenter: What is Clifford doing? Croco: I know! I know! He’s calling his friend. Child: [No] Croco: No? He isn’t calling his friend? Experimenter: Please tell Croco what Clifford’s doing. Child: [Clifford is eating (a bone)] (25) Russian Prompt: Skaži požalujsta Kroko, čto delaet Klifford. TellIMPERATIVE please CrocoDAT what doesPRES IMP CliffordNOM ‘Please tell Croco what Clifford is doing.’ Types of answer Null object: On est. heNOM eatsPRES IMP ØACC ‘He is eating.’ Lexical object: On est kostočku. heNOM eatsPRES IMP boneACC ‘He is eating a bone.’

Results The results for both contexts were classified as null (26), pronominal (27), and lexical (28): (26) Null object:

Ona risuet. sheNOM drawsPRES IMP ØACC ‘She is drawing.’

(27) Pronoun:

Ona risuet ego. sheNOM drawsPRES IMP it ACC ‘She draws it.’

(28) Lexical object:

Ona risuet cvetok. sheNOM drawsPRES IMP flowerACC ‘She is drawing the flower.’

Furthermore, responses that did not contain the target verbs (29) or contain morphologically complex (prefixed) verbal forms ((30), see the discussion section) were excluded from our analyses: (29) ‘Please tell Croco what the little girl is doing with the flower.’

108

The Acquisit ion of French in Mult ilingual Conte x t s

Cvetok. flowerACC ‘(The) flower.’ (30) ‘Please tell Croco what the little octopuses are doing.’ Oni razrezajut prostynu. They cutIMP SECONDARY sheetACC. ‘They are cutting a sheet.’

Results in the referential context As was noted earlier, in adult Russian direct objects can be omitted optionally in referential contexts. The expected results include the optional use of null objects or accusative pronouns. The results are shown in Table 4.1. According to a Shapiro–Wilk normality test, the dataset is not normally distributed (p < .05). A Kruskal–Wallis test shows statistical results for the proportion of null object responses across age groups (p = 0.014; df = 3). A Mann–Whitney test shows significant differences between the adult group and the three-year-olds (U = 33.5; p = 0.003; r = 0.56); adults and four-year-olds (U = 23.5; p = 0.015; r = 0.52); adults and five-year-olds (U = 40.5; p = 0.039; r = 0.53). The difference between other groups is not statistically significant. Pronominal responses were relatively rare in all groups (10–20%). The lexical DPs are unusual in the referential context in adult production (as in (28)). However, Pérez-Leroux et al. (2008) report similar results in adult control groups: 36% in English and 18% in French were DPs (cf. Tables 4.2 and 4.3). The authors suggest a possibility that adult participants prefer the more informative lexical response even after the object antecedent is explicitly introduced in the story (see also Müller et al., 2006). As can be seen from Tables 4.1, 4.2 and 4.3, null objects in Russian are considerably more frequent than in English and in French across all groups. It is important to note that in Russian the referential context used in this study is optional for pronominal or null objects, but that it is an obligatory context for pronominal use in English and in French. This fact can explain why French and English participants chose pronouns significantly more frequently than Russian participants. According to Pérez-Leroux et al. (2008), object omission Table 4.1 Distribution of responses by type of object in the referential context in Russian Group

Null

Pronoun

Lexical

Age 3 Age 4 Age 5 Adults

0.82 0.79 0.73 0.48

0.13 0.10 0.17 0.20

0.05 0.11 0.10 0.33

Verbal Transit iv it y Development in First L anguage Acquisit ion

109

Table 4.2 Distribution of responses by type of object in the referential context in English Group

Null

Age 2–3 Age 3–4 Age 4–5 Adults

0.350 0.083 0.017 0.000

Pronoun

Lexical

0.350 0.417 0.433 0.639

0.250 0.467 0.550 0.361

Source: Pérez-Leroux et al. (2008: 389).

Table 4.3 Distribution of responses by type of object in the referential context in French Group

Null

Age 3 Age 4 Age 5 Adults

0.345 0.257 0.115 0.014

Pronoun

Lexical

0.132 0.417 0.526 0.819

0.507 0.310 0.296 0.181

Source: Pérez-Leroux et al. (2008: 390).

in an obligatory context in French and English decreases over time. Consistent with the input of the target language, English children have a shorter omission stage than their French-speaking peers. Pérez-Leroux et al. (2008) report statistically significant differences between the younger group and all other groups in English. French children reach the adult-like performance later: a significant difference is observed between the three- and four-year-olds and adult groups. According to Table 4.1, Russian children produce null objects at a similar rate across all age groups. Wang et al. (1992) report similar results for Chinese children who omit objects more frequently than Chinese adults (22.5% versus 10%) and American children (22.5% versus 4.2%). Recall that, as in Russian, referential omissions in Chinese are grammatical in adult language. To summarize, the results from referential contexts suggest that Russian children drop objects more frequently than adults. Omissions in Russian are stable across age groups and occur at a higher rate than in languages where the referential object drop is not grammatical. However, the data do not allow us to reach a conclusion about illicit object omissions in Russian child production because referential object drop is a grammatical option in the target grammar.

Results in the non-referential context In the non-referential context, direct objects can be optionally omitted or can be realized as a lexical object (a DP) in Russian as well as in English

110

The Acquisit ion of French in Mult ilingual Conte x t s

and in French. Table 4.4 shows the results for non-referential object use in Russian. Tables 4.5 and 4.6 from Pérez-Leroux et al. (2008) are used to compare Russian with English and French. The data in non-referential contexts are not normally distributed for the group of age three (Shapiro–Wilk; p = 0.026). The results on proportions of null objects across age groups are significant (F3, 47 = 4.11; p = 0.011). Post hoc comparisons using an LCD test showed significant differences on the distribution of null object responses between the adult group and three-yearolds (mean difference = 0.37; 95% CI = −0.59, −0.15; p = 0.001; d = 1.4). The Table 4.4 Distribution of responses by type of object in the non-referential context in Russian Group

Null

Pronoun

Lexical

Age 3 Age 4 Age 5 Adults

0.76 0.64 0.54 0.39

0.00 0.00 0.02 0.02

0.24 0.37 0.44 0.60

Table 4.5 Distribution of responses by type of object in the non-referential context in English Group

Null

Age 2–3 Age 3–4 Age 4–5 Adults

0.400 0.250 0.317 0.264

Pronoun

Lexical

0.083 0.017 0.050 0.028

0.367 0.667 0.617 0.667

Source: Pérez-Leroux et al. (2008: 391).

Table 4.6 Distribution of responses by type of object in the non-referential context in French Group

Null

Age 3 Age 4 Age 5 Adults

0.332 0.333 0.259 0.278

Source: Pérez-Leroux et al. (2008: 391).

Pronoun

Lexical

0 0 0.019 0

0.558 0.650 0.667 0.722

Verbal Transit iv it y Development in First L anguage Acquisit ion

111

following results approach significant levels: adults and four-year-olds (mean difference = 0.25; 95% CI = −0.49, −0.002; p = 0.048; d = 0.82) and three- and five-year-olds (mean difference = 0.22; 95% CI = 0.001, 0.43, p = 0.049, d = 0.79). By comparing data across all three languages in non-referential contexts (cf. Tables 4.4, 4.5 and 4.6), we observe a developmental pattern in Russian, but not in English or in French child data. While Russian children come gradually closer to achieving target performance, their French and English peers already seem to have adult-like production across all age groups. The comparison between two contexts in Russian with respect to the choice of the type of object suggests that even younger children show some sensitivity to contextual differences in object use. In the non-referential context, we observe a gradual decrease in object drop (from 76% to 54%) and an increase in lexical object use (from 24% to 44%) across all age groups, as well as a low production of pronouns (2%). In contrast, in the referential context, the rate of null objects is higher (from 82% to 73%); the lexical object was rarely used (5% to 11%), and the percentage of pronouns was higher (10% to 17%). On the other hand, in both contexts, the object was omitted more frequently by children than by adults. To test the hypothesis of whether the participants perform differently across contexts, a Friedman test, a non-parametric analogue of repeated-measures ANOVA, was conducted. The test shows significant differences between contexts on the distribution of null objects (n = 51; χ2 = 16.026; df = 1; p < .005). A Wicoxon Signed Ranks shows statistically significant results for all groups: age three (n = 16; p = 0.030; r = 0.62); age four (n = 10; p = 0.028; r = 0.79); age five (n = 13; p = 0.015; r = 0.74); adults (n = 12; p = 0.034; r = 0.76). According to these results, Russian-speaking children use null objects differently in referential and nonreferential contexts.

Discussion In the present study, Russian children’s elicited production of direct objects was tested in referential and non-referential contexts. With these results we can make the following conclusions. Referential and non-referential null objects are more frequent in child speech than in adult production. Russian-speaking children, similarly to their French and English peers, seem to undergo an omission stage where they use null objects more frequently than adults. In this experiment, we studied omissions only in optional contexts in Russian; in the future it would be important to establish if there are any obligatory contexts in which null objects are disallowed in Russian, and to investigate children’s production in these contexts (see below).

112

The Acquisit ion of French in Mult ilingual Conte x t s

As for the data from the non-referential context, Russian children’s production gradually becomes closer to adult production, with a higher percentage of lexical objects instead of null objects. These results are different from French and English data where children already behave in a target-like manner in the early stages. The results show that children use null, pronominal and lexical objects at different rates in pragmatically different contexts. The comparison of French, English and Russian suggests that the omission stage has a different duration across languages. In both referential and non-referential contexts in Russian, children omitted objects in a target-like manner but at a higher rate. However, as was suggested by Gordishevsky and Avrutin (2003), Russian children also produce non-target object omissions (19d). What kind of syntactic developmental mechanism in Russian is responsible for non-target omissions in child grammar? Interestingly, Forsyth (1970) observes the restriction on object omissions with perfective verbs in adult Russian: A potentially transitive verb, when used in its perfective forms, requires an object (expressed or implied). It makes sense to say včera ja pisalIMP ‘yesterday I wrote’, but in isolation the sentence včera ja napisalPERF is meaningless. The later demands an object such as pis’mo, sočinenie (‘a letter, a composition’) etc. In conversation, of course, it is common to answer such a question as vy napisaliPERF pis’mo? (‘did you write the/a letter’) by saying simply napisalPERF (‘ØNOM wrote ØACC’), implying the object already referred to. (Forsyth, 1970: 91) In sum, Russian perfective verbs cannot be used with non-referential null objects. Perfective verbs in the non-referential context can appear only with lexical nouns. There are no such restrictions on the use of morphologically simple imperfective verbs, e.g. the imperfective verbs can be used with overt or omitted objects in the non-referential context. Taking a closer look at the verb form used in (19d), we notice that the verb is used in its perfective aspectual form while in (19a) there is an imperfective verb. Yadroff (1997: 253) makes a similar observation and provides a syntacticsemantic account for it. He concludes that ‘Imperfective transitive [verbs] allow pro-arb objects [non-referential null object N] as well as pro-ref objects [referential null object pro], while perfective transitives only allow pro-ref objects’, as is illustrated in the following examples: (31) On pisal. heNOM wrotePAST IMPF.ØACC ‘He was writing.’ (pro-arb) / no particular context ‘He was writing (something).’ (pro-ref) / context specifies object

Verbal Transit iv it y Development in First L anguage Acquisit ion

113

(32) *On napisal. heNOM wrote PAST PF.ØACC ‘He wrote down.’ (pro-arb)/no particular context (33) a. b.

Miša napisal pis’mo materi? MishaNOM wrotePAST PF letterACC motherDAT? ‘Did Misha write a letter to his mother?’ Da, (on) napisal. yes, (heNOM) wrotePAST PF ØACC ‘Yes, he wrote (it).’ (pro-ref)/context specifies object. (Yadroff, 1997)

This asymmetry in the use of objects with imperfective and perfective verbs in Russian is explained by the difference in syntactic structures for these verbs and in the functioning of AspP. Consider imperfective (34) and perfective (35) structures: (34) Imperfective transitive structure On pisal (pis’mo). heNOM wrotePAST IMPF.ØACC (a/the letter). ‘He was writing (a/the letter).’

(35) Perfective transitive structure *On napisal. heNOM wrote PAST PF.ØACC ‘He wrote down.’ (pro-arb)/no particular context

114

The Acquisit ion of French in Mult ilingual Conte x t s

From (34) we can see that the imperfective verb pisat’ ‘to write IMP’ can be optionally used with an overt DP (e.g. pis’mo ‘letter’) or a null non-referential object N. However, when the perfective prefix ‘na-’ is added to the verb pisat’ ‘to write’, the use of an indefinite DP becomes obligatory. The example in (35) shows that the N cannot be used with the verb napisat’ ‘to write down PF’ in the non-referential context. The difference between imperfective and perfective structures is explained by the indexation mechanism described in detail in Yadroff (1997). Case marking plays an important role in null object distribution with imperfective and perfective verbs. All types of objects except the N are specified for accusative case. The direct object has to move to the SpecAspP for case verification. The N is the only object not marked for case, and as such it stays in situ. As is observed from the structure (35), the object of a perfective verb must be marked for accusative case, and for that reason, the N is not compatible with the perfective aspect. The experimental study described in this chapter was conducted using present imperfective contexts only. Perfective verbs are not used in present tense in Russian; however, morphologically complex imperfective forms derived from perfective verbs can be used in the present tense. These verbs are called ‘secondary imperfective’, and they have the same properties as perfective base verbs with respect to the object realization. In other words, if a perfective verb requires an object in a non-referential context, a derived secondary imperfective verb cannot be used without the object. In the present study very few responses contained secondary imperfective verbs (36). These responses were excluded from the general results. (36) Situation:

Two misbehaving little octopuses are trying to cut a sheet with a pair of scissors.

Verbal Transit iv it y Development in First L anguage Acquisit ion

Experimenter:

Adult: Child:

115

Skaži požalujsta Kroko, čto osminožki delaut. TellIMPERATIVE please CrocoDAT, what little octopusesNOM doIMP PRESENT ‘Please tell Croco what the little octopuses are doing.’ Oni razrezajut prostynu. They cutIMP SECONDARY sheetACC. ‘They are cutting a sheet.’ *Oni razrezajut. They cutIMP SECONDARY ØACC. ‘They are cutting.’

Our data show that at age three children mostly preferred to drop objects (five null objects out of six sentences), while at age four to five, participants used DPs (one null object out of seven sentences) with secondary imperfective verbs. Interestingly, several studies report object omissions with aspectually marked verbs (e.g. particle verbs) in child English and Dutch (Thrift & Hyams, 2001). The fact that object drop in child Russian can be related to the development of the aspectual system in this language may be an important finding in investigating the acquisition of transitivity. Based on our data, we suppose that Russian children may use null objects in strongly transitive contexts with perfective verbs. Following previous analyses on different child grammars, it is possible to assume that this null object in L1 would have different properties from null objects in adult language. For instance, both Müller et al. (2006) and Pérez-Leroux et al. (2012) suggest that in the early stages French-speaking children may use a referential null object not identified by agreement similar to topic drop in Chinese or Japanese. Let us assume that Russian children might use a null object of the topic drop type in all contexts and with all types of verbs, perfective and imperfective. Recall that in adult grammar most perfective verbs in non-referential contexts require an overt DP and object drop is not allowed (as is shown in (32) and (35)). Following Yadroff (1997), we suppose that perfective verbs might be used with null objects in child grammar because they are specified for accusative case and can move for case verification into SpecAspP: (37) Perfective structure in child Russian

116

The Acquisit ion of French in Mult ilingual Conte x t s

There is a possibility that Russian children may drop objects in perfective non-referential contexts until they acquire the perfective–imperfective distinction. The next step is to study children’s productions in perfective nonreferential contexts in order to examine if young Russian speakers are already aware of this syntactic restriction in null object use during early stages of linguistic development and if this mechanism is responsible for non-target object omissions in early Russian (Frolova, 2014).

Conclusions The main goal of this chapter was to extend research on null object phenomena in French and English to Russian first language acquisition. First, the syntactic approach to verbal transitivity has permitted us to establish contexts of null object use in adult grammar of Russian. It has been shown that Russian allows direct object drop in referential and non-referential contexts. In this respect, Russian behaves similarly to French, in which two types of null objects also exist. However, in French referential null objects are explained by the functioning of the accusative clitic system. From that point of view Russian is more similar to English because it does not have any verbobject agreement markers or clitics to license referential null objects. While English does not allow empty pro, the referential null objects are very common in Russian and the constructions in which they occur can be classified as topic drop. The replication of Pérez-Leroux et al.’s (2008) experimental study on null objects in child Russian suggested that children use null objects as adults do in optional contexts, but at higher rates. The results in the referential context illustrate that children omit objects more frequently than adults, and that rates of omission do not decrease across age groups. Null objects in the nonreferential context show some evidence for gradual change: children progressively approach target-like performance by replacing the null objects with lexical objects. A cross-linguistic comparison of children’s production in French, English and Russian indicates that children in all three languages use pronominal, null and lexical objects at different rates in referential and non-referential contexts. These results provide evidence against a purely pragmatic account for L1 acquisition by demonstrating that children at age three to five distinguish contextual pragmatic factors. The observation that the length of the omission stage is different for children in these three languages provides additional evidence against assumptions based on the child’s pragmatic development only. This study on object omissions in optional contexts has allowed us to evaluate several predictions and to identify non-target-like use of null objects in child Russian with morphologically complex verbs. According to several

Verbal Transit iv it y Development in First L anguage Acquisit ion

117

studies on aspect in Russian, the syntactic mechanisms that seem to be involved in the developmental process in child grammar are related to the aspectual system. The perfective aspect seems to impose restrictions on object omissions in the non-referential context. The next step in the research on null objects in Russian is to test these assumptions in an experimental study.

Acknowledgements The author wishes to thank two anonymous reviewers, Pedro GuijarroFuentes, Mellanie Elliott, Joanne Markle LaMontagne, Natascha Müller, Ana Theresa Pérez-Leroux, Mihaela Pirvulescu, Yves Roberge, Katrin Schmitz, Nelleke Strik and the audience at VII Congrès des Francoromanistes in Essen for their helpful comments and suggestions at different stages of this research. Travel to Germany was funded by the VII Congrès des Francoromanistes and by a graduate student conference grant from the French Department of the University of Toronto. An earlier version of this study was presented at AcquisiLyon09 Colloquium with a financial support from SSHRC (410-09-2026) and Di Sciullo 412-2003-1003. Any mistakes that remain are the author’s responsibility.

5

Static and Dynamic Location in French and German Child Language Anne-Katharina Harr and Maya Hickmann

Introduction Space is a basic domain of human cognition. It is therefore not surprising that all languages provide complex systems to express spatial information. Past linguistic research in this domain has typically aimed at showing the existence of universal spatial distinctions, for example, between static location (e.g. he’s in the kitchen), motion within a general location (e.g. he’s running in the house), and changes of location such as boundary crossing (e.g. he’s running into the house). Nevertheless, languages differ considerably in how they encode spatial information, for example, in terms of different frames of reference and/or by means of different lexical and grammatical devices for the expression of spatial relations or motion events (Levinson, 2006; Talmy, 2000a). Recent research has begun to examine the implications of this linguistic diversity for language acquisition (e.g. Allen et al., 2007; Bowerman & Choi, 2003; Gullberg et al., 2008; Hickmann et al., 2009; Slobin, 2004, see also Treffers-Daller & Tidball, Chapter 6, this volume). Despite recurrent acquisition sequences in children’s spatial language (spatial prepositions, motion expressions) that suggest the role of universal perceptual and cognitive determinants, results now indicate that language-specific factors also constrain the course of children’s language acquisition. The present chapter addresses this question by comparing how children express static and dynamic location in two typologically different languages, French and German. In this first section the theoretical background of our study as well as our predictions will be presented. After outlining our methodology in the second section, the results of our study will be shown in the third section. In the final section, our outcomes will be discussed in the context of previous research and some conclusions will be drawn. 118

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

119

Static and dynamic location in verb-framed versus satellite-framed languages Talmy (1985, 2000a) proposes a dichotomy between satellite- and verbframed language families, that express spatial information quite differently both in dynamic situations (voluntary and caused displacements in space) and in static situations (location and spatial relations among entities). Examples (1) and (2) first illustrate these differences with respect to expressions of voluntary motion events. As shown in (1), satellite-framed languages (e.g. Germanic languages) typically encode manner in the verb stem (klettern ‘to climb’, rennen ‘to run’) and path in verbal satellites such as particles (runter ‘down’, weg ‘away’). In contrast, as shown in (2), verb-framed languages (e.g. Romance languages) typically encode path in the verb stem (descendre ‘to descend’, partir ‘to leave’), while manner is expressed by peripheral constructions such as prepositional phrases (à quatre pattes ‘on all fours’) or gerunds (en courant ‘by running’). These two patterns correspond to the most prototypical ways of expressing these motion events, despite some exceptions in each language, such as French verbs that lexicalize manner and path simultaneously (e.g. grimper ‘to climb upwards’) or German verbs that encode path (e.g. überqueren ‘to cross’).1 (1) Der Junge rennt weg. The boy runs away. ‘The boy runs away.’ (2) Le garçon part en The boy leaves by ‘The boy runs away.’

courant. running.

The expression of caused motion events has been studied to a much lesser extent. In German, such complex events can be expressed by compact means, just like voluntary motion events. In accordance with Talmy’s typology, satellite-framed languages typically encode manner and cause in the (transitive) main verb. As illustrated in (3), such verbal phrases are easily combined with path particles or prepositional phrases. (3)

Johann rollt den Ball runter/ins Haus. John rolls the ball down/into the house. ‘John rolls the ball down/into the house.’

French seems to have three main construction types to express caused motion. These patterns differ in terms of their syntactic complexity and typically only encode two semantic components (see Harr, 2012: 129f; Hendriks et al., 2008: 18ff). First, like German, French provides speakers with

120

The Acquisit ion of French in Mult ilingual Conte x t s

a large set of intransitive motion verbs (encoding either manner or path) which can be used in transitive constructions to express caused motion events (e.g. monter quelque chose ‘to ascend something’). Secondly, like German, French causative verbs such as pousser (‘to push’) or tirer (‘to pull’) frequently encode the manner of the action causing the object’s displacement. Thirdly, unlike German, French has specific causative constructions combining the action verb faire (‘to make’) with a non-finite motion verb that encodes either manner or path (Example (4)). Since all three basic constructions only encode two semantic components, speakers have to add either a prepositional phrase or another kind of peripheral construction, such as a gerund, in order to convey all relevant information (in parentheses in Example (4)). (4)

Marie fait rouler la balle (en descendant). Mary makes roll the ball (by descending). ‘Mary rolls the ball down.’

With respect to the expression of location in static situations or in dynamic situations involving changes of location, French provides some neutral verbs2 that frequently express static location (être ‘to be’, se trouver ‘to be located’) and relatively neutral dynamic verbs expressing changes of location (mettre ‘to put’). Such verbs are semantically weaker than more specific verbs such as être assis (‘to be seated’) and a number of verbs expressing specific information, concerning, for example, manner of attachment (static: être cloué ‘to be fixed by a nail’; dynamic: coller ‘to stick’, accrocher ‘to hook’) or a general configuration (static: être bouché ‘to be corked’; dynamic: fermer ‘to close’). As for other devices, French has a rich set of prepositions that encode different spatial relations (e.g. sur ‘on’, au-dessus ‘above’, autour ‘around’), as well as some relatively neutral prepositions that are quite frequent, particularly à (‘at, to’) and de (‘from, of’).3 Examples (5) and (6) illustrate how these two semantically weak prepositions can be used alone in spatial contexts, and Examples (7) and (8) show how they can be combined with more specific prepositions. (5) Marie est à la maison. Mary is at home. ‘Mary is at home.’ (6) Jean part de Munich. John leaves from Munich. ‘John leaves Munich.’ (7) Henri va jusqu’à la Tour Eiffel. Henry goes [all the way] to the Eiffel Tower. ‘Henry goes all the way to the Eiffel Tower.’

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

(8)

121

Il met l’écharpe autour de son cou. He puts the scarf around of his neck. ‘He puts the scarf around his neck.’

German follows the typical satellite-framed lexicalization pattern described above for motion events. Just like French, German also provides speakers with some neutral verbs to express static location (sein ‘to be’) and dynamic changes of location (machen ‘to make’, tun ‘to do’). Apart from these semantically weak verbs, German verbs typically encode posture in descriptions of static locations (hängen ‘to hang’, sitzen ‘to sit’). As for changes of location, speakers can either express a change of posture (setzen ‘to put sth in a sitting position’, legen ‘to put sth in a lying position’), a specific manner of attachment (kleben ‘to stick’, stecken ‘to put in a specific manner’) or information about the manner of the action resulting in the location change (ziehen ‘to pull’). In contrast to French, German provides speakers with a large set of particles that typically express path. Examples (9) and (10) illustrate two changes of location combining a neutral and a specific verb with path-particles. (9)

(10)

Er tut den Deckel He takes the lid ‘He takes the lid away.’

weg. away.

Sie legt den Stift drauf. She lays the pencil on it. ‘She puts the pencil on it.’

Unlike French, most German prepositional phrases require case marking on the nominal element which gives unambiguous information about whether the prepositional phrase is to be interpreted as denoting a static location (dative) or a directional component in a dynamic situation involving a change of location (accusative and in rare cases dative). Examples (11) and (12) illustrate both situations. (11)

Das Glas steht auf dem The glass stands on the ‘The glass stands on the table.’

Tisch. table (dative).

(12)

Ich stelle das Glas auf I stand the glass on ‘I put the glass on the table.’

den the

Tisch. table (accusative).

Recent research suggests that, when verbalizing motion and location, speakers preferentially choose those means of expression that are most

122

The Acquisit ion of French in Mult ilingual Conte x t s

typical of their native language, thereby partially affecting their spatial conceptualization (Bowerman, 1996; Bowerman & Choi, 2003; Levinson, 2006). According to Slobin (1996, 2006), these language-specific patterns make some aspects of reality more salient than others, thereby inviting speakers to direct their attention to different types of information. From a developmental point of view, children must identify the lexicalization patterns most typical of their language, along with their meanings, and on this basis direct their attention to the types of information that are most relevant for this particular language.

Implications for child language Although most researchers agree that language acquisition is influenced by both general cognitive and language-specific factors, there is no consensus as to the relative weight that must be attributed to each of these two types of determinants. General cognitive factors include purely cognitive capacities (e.g. working memory) and perceptual constraints, such as the meanings that can be expressed by young children (Johnston, 1985), the formal complexity of the linguistic devices to be acquired (Johnston, 1985; Slobin, 1973), and of course cognitive processing mechanisms (Cromer, 1974). According to Piaget and other researchers in related cognitivist approaches, such general cognitive factors underlie children’s linguistic development in the domain of spatial expressions accounting for recurrent language-independent sequences of acquisition (Johnston, 1985; Piaget, 1964). On the basis of a review of several studies concerning the acquisition of prepositions across languages, Johnston (1988) concludes that all children follow a similar developmental sequence (containment < support < occlusion < proximity < sagittal axis). This recurrent order seems to reflect the relative complexity of spatial concepts and suggests that universal cognitive factors determine the acquisition process. Whereas Johnston (1988) only examined the use of concrete spatial prepositions and/or the expression of concrete spatial meanings, other studies investigated the acquisition of functional uses of prepositions in varied constructions across different languages (Alexaki et al., 2009; Kochan et al., 2007). The results show that in English and Greek, prepositions appear earlier in concrete spatial contexts than in purely grammatical constructions in which they often assign a specific case. In contrast, French children seem to start out by using prepositions to mark pragmatic and argumentative functions to justify actions, disambiguate intentions or interpersonal positioning (Kochan et al., 2007: 147). Thus, the acquisition of concrete uses of prepositions before functional ones cannot be generalized to all languages, because language-specific factors also play a role. Slobin further argues that the order in which linguistic devices are acquired depends not only on their meaning but also on their formal complexity (Slobin, 1973). Therefore, a child may have acquired a specific

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

123

conceptual representation while still lacking the linguistic (morphologicosyntactic) means to express it. The third type of cognitive constraints concerns limitations that are related to information processing. It is obvious, for example, that speakers are confronted with cognitive constraints in their working memory capacities or in the speed with which they can perform various cognitive operations (Johnston, 1985). For children, this implies primarily limitations on the complexity and length of their productions as well as on their comprehension of different types of utterances (Cromer, 1974). Apart from these general developmental factors, language-specific properties also seem to influence the ways in which children acquire specific linguistic devices. In order to acquire adult-like linguistic knowledge during first language acquisition, children have to learn to map linguistic forms onto semantic concepts while learning to construct grammatical utterances. In other words, children’s acquisition of target-like ways to express spatial information is partially input driven: they acquire those lexicalization patterns that are most frequent in their input and which correspond to the most typical patterns specific to their language. Thus, depending on the properties of their language, children learn to verbalize situations in ways that are most typical of their mother tongue. Several studies (Bowerman, 1996; Bowerman & Choi, 2003; Harr, 2012; Hickmann, 2007; Hickmann et al., 2009; Ji et al., 2011) have begun to show that, from the youngest age onwards, children’s utterances resemble those of adults speaking the same language more than those of children of the same age speaking a typologically different language. Thus, studies concerning early child Korean (Bowerman & Choi, 2001; Choi, 2011; Choi & Bowerman, 1991; Choi et al., 1999) show that young Korean children produce utterances that resemble much more those of adults speaking their language than those of same-aged children learning English, furthermore showing that even their early comprehension during the prelinguistic phase may be influenced by the specific properties of their language. For example, as compared to early productions in English, Korean learners regularly express different aspects of motion events (e.g. systematically distinguishing voluntary and caused motion) and they encode different dimensions of spatial relation (e.g. ‘tight fit’, expressed in Korean verbs). A study (Hickmann & Hendriks, 2006) comparing the expression of location and changes of location in English and French also shows striking crosslinguistic differences. Whereas English adults produced the typical pattern relying most frequently on satellites (e.g. down, up) and focusing on posture (e.g. to hang) and manner of action (e.g. to push), French speakers concentrated more on specific manners of attachment (e.g. accrocher ‘to hook’), which they expressed in verbs. French children produced this adult-like pattern from three years onwards, although they showed two developmental progressions with age: (1) they used an increasing number of specific verbs; and (2) in their descriptions of static location, they produced a number of

124

The Acquisit ion of French in Mult ilingual Conte x t s

overgeneralizations with several prepositions, including sur (‘on’), which was used with non-canonical situations of contact/support (for which adults used the neutral preposition à (‘at/to’) together with semantically rich verbs), as well as with situations without contact/support (for which adults used au-dessus ‘above’). Only one study (Harr, 2012) has directly compared spatial language in child German and French, focusing on how children (aged four to 10) express voluntary and caused motion. The results show striking cross-linguistic differences. From a very early age on, German-speaking children use the compact structures available in their language to express simultaneously some kind of manner and some aspect of path, as well as cause in the case of caused motion, all within a single clause. As illustrated in (13), they typically encode manner and cause in the main verb (schieben ‘to push’), expressing path in particles (rauf ‘up’) or prepositional phrases (auf den Hügel ‘on the hill’). (13) Popi schiebt die Kiste (auf den Hügel) Popi pushes the box (on the hill) ‘Popi pushes the box up the hill.’

rauf. up.

(6 years)

In contrast, given its typological properties, French expressions of caused motion require the use of more complex forms to encode all three semantic components simultaneously, as illustrated by the adult utterance in Example (14). Since path is typically encoded in the main verb (monter ‘to ascend’, traverser ‘to cross’), manner and often also cause have to be expressed by other linguistic means, especially subordinate constructions such as gerunds (en faisant tourner le ballon ‘by making roll the ball’). As a result, French speakers at all ages frequently omit some semantic components, producing utterances of lesser semantic density as compared to German. In particular, young French children rarely express all three components (cause, path and manner) together within one clause (Example (15)) or they distribute different types of information across clauses (Example (16)). (14) Popi descend la colline en faisant Popi descends the hill by making ‘Popi rolls the ball down the hill.’

tourner le ballon. (adult) roll the ball.

(15) Il traverse la route. He crosses the street. ‘He crosses the street.’ (16) Il

a

poussé

la roue

et

(8 years)

il

est rentré

dans le garage. (4 years) He has pushed the tyre and he is BACK-entered in the garage. ‘He pushed the tyre and he entered into the garage.’

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

125

Thus, typological factors seem to have an impact on the types of semantic information that are expressed by children and on the semantic density of their utterances. From very early on, children seem to express those types of information that are most salient to them as a function of the most prototypical and accessible linguistic means available in their native language. Extending a previous study focusing on English and French (Hickmann & Hendriks, 2006), the present paper compares the relative impact of such language-specific determinants versus general developmental factors in French and German children (6 to 10 years of age) and adults in two controlled situations: one in which they localized static entities (static task) and another in which they described actions that consisted of displacing objects (dynamic task).

Predictions On the basis of previous results (Bowerman, 1996, 2007; Harr, 2012; Hickmann, 2007; Hickmann & Hendriks, 2006) and following Talmy’s typology, the present study aimed to determine the role of both languagespecific properties and an increasing cognitive maturity with age on two aspects of speakers’ descriptions: the types of information they express (e.g. manner, path, cause) and the linguistic means they use to express this information (e.g. verb and other linguistic devices outside of the verb). Responses were elicited from children and a control group of adults in two tasks that required either verbally locating one entity in relation to another (static task) or describing the displacement of an entity from one location to another (dynamic task). Two types of predictions were made. First, the study tests the general hypothesis that descriptions produced by children and adults in both static and dynamic tasks should follow the typological properties of their language. In this respect, the prediction was that children’s responses should be more similar to the descriptions produced by adults of their language group than to those of children of the same age but belonging to another typologically distinct language group. More specifically, French speakers were expected to rely predominantly on verbs in order to express specific information relevant to the tasks, such as manner of attachment (e.g. accrocher ‘to hook’) and general spatial and/or functional configurations (e.g. fermer ‘to close’). When used, prepositional phrases should frequently be semantically neutral and/or mostly provide information about location or ground entities in both static and dynamic situations, and they should rarely focus on path in the dynamic task. In contrast, German speakers of all ages were expected to produce frequent path particles in the dynamic task as well as to use case marking unambiguously indicating whether they encode path (dynamic location) or a general location (static location). They should express some type of manner in main verb roots, including position or posture and manner of attachment in both tasks, as well as manner of causing motion in the dynamic task.

126

The Acquisit ion of French in Mult ilingual Conte x t s

Secondly, it was also expected that general cognitive factors should influence children’s responses regardless of their language. In particular, increasing cognitive maturity should partially determine the ability to process and to express multiple semantic components simultaneously (e.g. memory of what to express, access to an increasing repertoire of semantically differentiated lexical units, use of complex linguistic structures). As a result, young children should express less specific information than adults and their responses should become increasingly specific and semantically dense with age. In particular, in comparison to adults, young children should produce more general-purpose verbs and other devices that can be used in a wide range of contexts, while more informative linguistic devices should increase with age. The same prediction can be made for both languages, despite crosslinguistic differences that may result from typological properties.

Methodology Participants The participants were monolingual French and German children and adults. Two age groups of 12 children each were tested in kindergartens and schools in France (Paris) and Germany (Munich). Their ages were approximately four years (mean 4;3, range, 3;11–4;10 hereafter four-year-olds) and six years (mean 6;1, range 5;8–6;9, hereafter six-year-olds). These ages were selected in light of previous studies (Harr, 2012; Hickmann, 2007) showing that developmental progressions are most striking between four and six years, during the transition from pre-school to primary school. Teachers were asked to exclude those children who were below or above the average. Each group contained an equal numbers of boys and girls, and all were of similar socio-economic status (middle class). The control group comprised 12 French and German native adults, university students in Paris and Munich.

Materials Based on previous results (Bowerman, 1996; Hickmann & Hendriks, 2006), two sets of stimuli were used, one for a static task (localizing entities) and one for a dynamic task (describing actions).4 The stimuli in the static location task consisted of a set of 20 pictures (Appendix 1) that were drawn in black and white, except for one entity that was coloured in yellow on each picture (the figure) and that had to be located in relation to another entity (the ground). The pictures showed spatial configurations that involved nonprototypical contact/support situations (e.g. a handle on a door, a spider on a ceiling) in order to elicit responses containing varied prepositions and verbs, since previous results concerning other languages showed that

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

127

cross-linguistic differences were more striking with this type of item than with prototypical situations (e.g. an apple in a bowl, a cup on a table). German speakers were expected to predominantly use the preposition an (‘on’) or auf (‘in’) together with a verb expressing some other semantic component (such as posture). Although French speakers could use the related prepositions sur (‘on’), other prepositions were also expected, particularly neutral prepositions such as à (‘at/to’) or de (‘from’), together with specific verbs expressing varied components (such as manner of attachment). In the dynamic location task the participants had to describe a set of 20 actions that were mimed by the experimenter (Appendix 2) and that involved putting something into or onto something else (e.g. putting a shoe on a doll, a lid onto a pan, a toy into a bag, two Lego pieces onto each other) (for more details, see Bowerman, 1996).

Procedure Participants were seen individually in their school or university setting. Younger children were introduced to a doll and were asked to blindfold her as part of a game in which they would be telling her secrets. This procedure ensured that subjects produced full descriptions. Older children and adults were asked to describe the pictures/actions for a future naïve addressee who would not have seen them. Tasks and items were presented in different orders. Half of the subjects described first static location, then dynamic location, while the other half followed the reversed order. Within each task order, subjects were also randomly assigned to one of two item orders. In the static task, pictures were randomized and presented within two booklets (Booklet 1 and 2 in Order 1, the reverse in Order 2). Subjects flipped through the pages and had to describe on each page ‘where the yellow thing was’. In the dynamic task, actions were presented in four successive blocks (Blocks 1–4 in Order 1, the reverse in Order 2). The experimenter acted out an action and then asked the subject to describe what she had done. The entire session was audiotaped and transcribed.

Coding The analyses below examine all utterances that described location and changes of location. These utterances were coded with respect to two dimensions: the types of semantic information that were encoded (Semantic components expressed) and the linguistic means whereby this information was expressed in the response (Information locus).

Semantic components expressed Expressed information fell into the types of semantic components shown in Table 5.1. Responses that expressed none of these semantic components and/or were entirely vague were also coded (e.g. es ist da ‘it’s there’; tu le mets

128

The Acquisit ion of French in Mult ilingual Conte x t s

Table 5.1 Semantic components expressed in descriptions of static and dynamic location Specific semantic component French

German

Manner of attachment of action of position (posture) Path Cause Figure or ground properties Location

stecken (‘to stick’) schieben (‘to push’) hängen (‘to hang’) aus + accusative (‘from’) tun (‘to do’) eintüten (‘to put into a bag’) auf + dative (‘on’)

accrocher (‘to hook’) tirer (‘to pull’) être assis (‘to be sitting’) sortir (‘to exit’) mettre (‘to put’) chausser (‘to put shoes on’) sur (‘on’)

‘you put it’). In this respect, a further distinction was made between linguistic devices that expressed relatively specific information (e.g. verbs such as coller ‘to glue’) or other devices (such as um ‘around’) versus entirely or relatively neutral information (e.g. verbs such as être ‘to be’, mettre ‘to put’, tun ‘to do’, geben ‘to give’ or other devices such as à ‘at/to’, da ‘there’).

Information locus Analyses of information locus examined how speakers encoded relevant semantic information with particular attention to a major distinction between two loci in the responses: main verb roots (e.g. hängen ‘to hang’, accrocher ‘to hook’) and all other devices that could express relevant information outside of the verb root, which were subdivided into four types: (a) German particles expressing path (e.g. weg ‘away’, an ‘to’); (b) prepositional phrases providing information about path or location (e.g. auf dem Kopf ‘on the [dative] head’, auf den Haken ‘on the [accusative] hook’, sur la bouteille ‘on the bottle’); (c) locative adverbs (e.g. da ‘there’, ici ‘here’); and (d) adverbial phrases that expressed some kind of manner (e.g. fest ‘tightly’). As illustrated below, responses were then classified according to whether speakers encoded specific semantic information only in verb roots, only in other devices, in both, or not at all (indicated in the graphs below as Verb only, Other only, Verb & Other, None, respectively): (1) Verb only: da hängen (‘to hang there’); accrocher à/être accroché à (‘to hook at/to be hooked at’). (2) Other devices only: auf dem Baum sein (dative) (‘to be on the tree’); auf den Topf tun (accusative) (‘to put on the pot’); mettre dans la casserole (‘to put in[to] the pot’); être sur la tête (‘to be on the head’). (3) Verbs and other devices: auf den Haken hängen (accusative) (‘to hang on the hook’); zusammenstecken (‘to fit together’); coller sur le mur (‘to stick on the wall’); emboîter dans le carton (‘to in-fit in the box’). (4) None: da sein (‘to be there’); mettre là (‘to put there’).

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

129

Results Information locus Static task Figure 5.1 shows the distribution of semantic information in French speakers’ responses. Among the responses of French adults, 32% contain only a specific verb and 16% express specific information both in the verb and in other devices. Examples (17) and (18) illustrate both response types. (17) C’est un manteau qui est accroché à un cintre. That’s a coat which is hooked at/to a hanger. ‘That’s a coat which is on a hanger.’

(adult)

(18) C’est un panneau cloué sur un arbre. That’s a board nailed on a tree. ‘That’s a board nailed on a tree.’

(adult)

French children at age four and six predominantly express semantic information only with prepositional phrases outside of the main verb (four years 74%, six years 72%), as illustrated in Example (19). Between four and six years specific verbs tend to increase slightly either as the only means of

Figure 5.1 Information locus in the static task: French

130

The Acquisit ion of French in Mult ilingual Conte x t s

expressing specific information (four years 6%, six years 16%) or in combination with other devices (four years 2%, six years 6%). However, such uses remain relatively low as compared to adults. (19) Le bandeau il est sur la The headband it is on the ‘The headband is on the head.’

tête. head.

(4 years)

Figure 5.2 shows information locus in German descriptions of static locations. As is typical for satellite-framed languages and as illustrated in Examples (20) and (21), German speakers at all ages most frequently express relevant semantic information in linguistic devices outside of the main verb root (four years 73%, six years 68%, adults 66%). (20) Das Telefon ist an der The telephone is on the ‘The telephone is on the wall.’ (21) Das Loch ist im The hole is in + the ‘The hole is in the towel.’

Wand. wall.

Handtuch. towel.

Figure 5.2 Information locus in the static task: German

(4 years)

(adult)

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

131

A developmental progression can also be observed in German with respect to responses that express specific information both in verb roots and in other linguistic devices. This response type increases with age and reaches the adult target-like pattern by six years (four years 13%, six years 30%, and adults 33%). Example (22) illustrates such a typical German response. (22) Das Stirnband sitzt aufm The headband sits on + the ‘The headband is on the head.’

Kopf. head.

(6 years)

Dynamic task As shown in Figure 5.3, responses produced by French speakers in the dynamic task are rather heterogeneous. First, a substantial proportion of these responses do not encode any specific semantic information. About 30% of children’s responses are of this type at both ages and 10% still occur among French adults. As illustrated in (23) and (24), such cases most often consist of using the verb mettre (‘to put’), sometimes in combination with a neutral preposition such as à (‘to’). (23) Tu mets les jouets. You put the toys. ‘You put the toys.’

Figure 5.3 Information locus in the dynamic task: French

(6 years)

132

The Acquisit ion of French in Mult ilingual Conte x t s

(24) Tu te mets un pansement You [to]you put a band aid ‘You put a band aid on your hand.’

à la main. at/to the hand.

(adult)

When responses express specific semantic information, French children encode it either in the verb root (four years 31%, six years 27%) or in other devices (four years 37%, six years 35%). Example (25) illustrates a typical response with a specific verb and Example (26) a response with a specific preposition. (25) Tu l’ accroches. You it hook. ‘You hook it.’ (26) Tu lui mets autour du You him/her put around the ‘You put [it] around his/her neck.’

(4 years)

cou. neck.

(6 years)

Utterances with specific information both in the verb and in other linguistic devices increase with age, at first only slightly (four years 3%, six years 10%), reaching their peak at adult age (32%) where they represent the most important response pattern. Response (27) illustrates this response type. (27) Vous les accrochez les légos l’un dans l’ You them hook the legos the one into the ‘You hook the legos into each other.’

autre. (adult) other.

As shown in Figure 5.4, response patterns are less heterogeneous in German than in French. Although German children produce a large number of utterances in which specific semantic information is expressed in satellites (four years 53%, six years 45%), this response pattern decreases for adults (10%). In this response type, speakers most often use the neutral verbs tun (‘to do’) or machen (‘to make’) in combination with satellites or similar linguistic devices which express path information (Example (28)) or reciprocity (Example (29)). (28) Das reintun so. This in-do so. ‘To put it so into.’

(4 years)

(29) Die zusammentun. Them together-put. ‘To put them together.’

(6 years)

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

133

Figure 5.4 Information locus in the dynamic task: German

Responses with both specific verbs and specific satellites strongly increase with age and become the most frequent pattern among adults (four years 43%, six years 54%, adults 89%). Examples (30) and (31) illustrate this response type. (30) So aneinander stecken. So to one another put tightly. ‘To put them tightly together.’ (31) Du setzt der Lisa eine Kappe You dress the Lisa a cap ‘You dress Lisa with a cap.’

(4 years)

auf. on.

(adult)

Semantic components expressed Static task Figures 5.5 and 5.6 show which types of information were expressed in the static task by French speakers in verbs and in other linguistic devices, respectively. As shown in both of these figures, there is practically no developmental change with respect to semantic components expressed. French speakers of all ages most frequently encode information about manner in the main verb, typically encoding a specific manner of attachment (e.g. être collé

134

The Acquisit ion of French in Mult ilingual Conte x t s

Figure 5.5 Static task: Semantic components expressed in French verbs

‘to be stuck’). This response pattern increases slightly between age six and adulthood (four years 49%, six years 48%, adults 61%). Often such responses also include information about the figure entity (e.g. être accroché ‘to be hooked’), which decreases with age to some extent (four years 41%, six years 40%, adults 32%). In nearly all of their responses, speakers of all three age

Figure 5.6 Static task: Semantic components expressed in other French linguistic devices

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

135

groups express information about the location of the figure. French speakers most often used prepositional phrases (e.g. à ‘at/to’, sur ‘on’ and dans ‘in’) and less often locative adverbs such as ici (‘here’) to express information outside of the main verb. Examples (32) and (33) illustrate this typical French pattern in static location descriptions. (32)

Le pansement est collé sur The band aid is stuck on ‘The band aid is stuck on the foot.’

(33) Le pendentif est accroché The pendant is hooked ‘The pendant is on the chain.’

le the

au to + the

pied. foot.

collier. chain.

(4 years)

(adult)

Figures 5.7 and 5.8 show the semantic components expressed in the static task for German. As in French, there is hardly any developmental change in German. Speakers of all age groups most frequently express some kind of manner in the main verb (four years 86%, six years 97%, adults 88%) and provide information about the location of the figure in other linguistic devices (four years 86%, six years 95%, adults 90%). As is typical for German, verbs often encode some posture (e.g. hängen ‘to hang’, sitzen ‘to sit’) that applies to different contexts and figures (see Example (34)). In some contexts speakers use verbs that express specific information about manner of

Figure 5.7 Static task: Semantic components expressed in German verbs

136

The Acquisit ion of French in Mult ilingual Conte x t s

Figure 5.8 Static task: Semantic components expressed in other German linguistic devices

attachment (see Example (35)). Such manner verbs are combined with prepositional phrases and a dative case marking on the nominal determiner that unambiguously expresses static location (e.g. an + Dat. ‘on + Dat.’, auf + Dat. ‘on + Dat.’). Speakers rarely used locative adverbs such as draußen (‘outside’). (34) Die Lampe hängt an der The lamp hangs on the (dative) ‘The lamp is on the ceiling.’ (35) Die Flasche, in der (dative) The bottle, in it (dative) The bottle, in it is a cork.’

Decke. ceiling.

steckt tightly-fits

ein a

(6 years)

Korken. cork.

(adult)

Dynamic task Figures 5.9 and 5.10 show the semantic components expressed in the dynamic task for French. French speakers typically express cause in the main verb (four years 66%, six years 62%, adults 51%) and location in other linguistic devices (four years 94%, six years 89%, adults 79%). There are few developmental changes although manner information in the main verb increases with age (four years 23%, six years 24%, adults 37%). Whereas French children often use the neutral verb mettre (‘to put’; see Example (36)), adults make

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

137

Figure 5.9 Dynamic task: Semantic components expressed in French verbs

use of more specific verbs such as nouer (‘to knot’) or enfiler (‘to pull on’; see Example (37)). (36) T’ as mis les You have put the ‘You have put the legos.’

légos. legos.

(4 years)

Figure 5.10 Dynamic task: Semantic components expressed in other French linguistic devices

138

The Acquisit ion of French in Mult ilingual Conte x t s

(37) Vous enfilez le peignoir You pull-on the robe ‘You pull the robe on the doll.’

à to

la the

poupée. doll.

(adult)

Verbs coded as ‘Residual’ (see Figure 5.9) mainly comprise information about the figure (e.g. emboîter ‘to put into a box’, boucher ‘to cork’). As in the static location task, French speakers most frequently mention location in other linguistic devices (four years 94%, six years 89%, adults 79%), using the same prepositions as mentioned above (Example (37)). Recall that our conservative strategy coded as location all prepositional phrases that did not unambiguously express path. As shown in Figures 5.11 and 5.12, lexicalization patterns differ considerably in French and in German, particularly with respect to linguistic devices outside of the verb root. Like French speakers, German speakers of all age groups express most frequently cause in the main verb using neutral verbs such as machen (‘to make’) or tun (‘to do’) (four years 67%, six years 65%, adults 52%). With age, information about manner increases steadily (four years 32%, six years 35%, adults 46%). The most frequent manner verbs in German express posture (e.g. setzen ‘to put into a vertical position’, legen ‘to put into a horizontal position’, hängen ‘to hang’; see Example (38)), manner of action (e.g. werfen ‘to throw’) and specific manners of attachment (e.g. binden ‘to tie’, kleben ‘to stick’). Path is most frequently expressed unambiguously either in verbal particles or in prepositional phrases (four years 64%, six years 67%,

Figure 5.11 Dynamic task: Semantic components expressed in German verbs

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

139

Figure 5.12 Dynamic task: Semantic components expressed in other German linguistic devices

adults 69%). As illustrated in (38), adults frequently use prepositional phrases governing the accusative case (e.g. in + Acc ‘in-path’, an + Acc. ‘on-path’). (38) Spielzeug wird in eine Schüssel (accusative) reingelegt. (adult) Toys will in a bowl (accusative) in-put in a horizontal position. ‘Toys are put into a bowl.’ When describing dressing actions, German speakers across all ages use two extremely frequent particle verbs (aufsetzen ‘to put on’ and anziehen ‘to dress’). As shown in (39), the verb stem encodes manner and the particle encodes path, which corresponds to the typical satellite-framed pattern. (39) Der Lisa den Schal anziehen. The Lisa the scarf on-pull. ‘To dress Lisa with the scarf.

(6 years)

German four-year-olds most frequently produce locative adverbs that only provided vague information about the location of objects (e.g. da ‘there’, daran ‘on-it’). In about 20% of the cases, German speakers use adverbs expressing information about the reciprocity of the dynamic action (coded in the category ‘Residual’), e.g. zusammen (‘together’) or ineinander (‘in each other’), which are very typical of German and very frequent in everyday language. Example (40) illustrates an utterance produced by a fouryear-old who uses a neutral causative verb in combination with such a reciprocal adverb.

140

The Acquisit ion of French in Mult ilingual Conte x t s

(40) Du tust sie aneinander. You put them to each other. ‘You put them together.’

(4 years)

Discussion This study aimed to examine lexicalization patterns in two typologically different languages from a developmental perspective, with particular attention to the impact of language-specific and general developmental factors during first language acquisition. Verbalizations of static and dynamic location in French and German were compared in light of Talmy’s dichotomy between satellite- and verb-framed languages. Several striking results are discussed below as well as their implications for further research directions. French speakers follow the typical verb-framed pattern and primarily rely on verbs to express specific information, sometimes in combination with other linguistic devices, such as prepositional phrases or adverbs. As predicted on the basis of typological properties, they mainly encoded a specific manner of attachment in the verb which in some cases also provided information about properties of the figure or ground entities (e.g. accrocher ‘to hook’, emboîter ‘to put into a box’). Other verbs expressed particular spatial and/or functional dispositions such as fermer (‘to close’). When French speakers combined such verbs with other linguistic devices, they most frequently used prepositional phrases to specify the location of the figure (static task) or the endpoint of its change of location (dynamic task). Since most French prepositional phrases do not unambiguously encode path, they nearly all encoded location. In descriptions of location, German speakers of all age groups predominantly rely on satellites which are very typical of a satellite-framed language. Although German children use a large number of neutral verbs, specific verbs increase strikingly with age. Note that German speakers never rely on verbs alone to describe location. In line with our predictions, they mostly encode information about manner in the verb (e.g. posture, manner of action) and express path and location by means of other linguistic devices. As is typical in German, speakers frequently use prepositional phrases that unambiguously encode location (static task) or path (dynamic task). In the dynamic location task, other linguistic devices also frequently provided information about reciprocity (e.g. ineinander ‘in each other’, zusammen ‘together’). Our results clearly show that children’s descriptions of static and dynamic location are much more similar to descriptions produced by adults of the same language group than to those of same-aged children belonging to a typologically distinct language. Thus, children tend to express the same types of information as adults in their language. Nevertheless, general developmental factors also have an impact on children’s verbalizations. In both languages,

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

141

children express gradually more specific information and this developmental progression concerns the verb as well as other linguistic devices. Whereas French and German children at four years of age produce a large number of neutral verbs (e.g. mettre ‘to put’, tun ‘to do’, être ‘to be’), specific verbs expressing a particular manner of attachment or a particular manner of doing an action to cause a displacement increase until adult age. This increase of specific verbs is more striking in German than in French where children increasingly rely on specific verbs alone, particularly in descriptions of static location. With respect to other linguistic devices, it has been observed that in descriptions of static location French and German children predominantly rely on other linguistic devices such as prepositional phrases (in French) and particles (in German). One reason for this result might be that these devices are easier to acquire in that they do not require verbal inflexion and are highly frequent in children’s input. In line with this outcome, only minimal developmental changes were observed in linguistic devices outside of the main verb. With respect to the semantic information expressed in such linguistic devices, children of both languages differ only minimally from adults. These results are only partially in line with previous studies on the acquisition of spatial language by French and German children of the same ages (Harr, 2012). Thus, the present results concerning how children express location (localizing static entities) and changes of location (object displacements mimed by the experimenter) only show few developmental changes with age, whereas such changes were observed in both languages (but especially in French) in children’s descriptions of two types of motion events: spontaneous motion carried out by an agent (voluntary motion) or object displacements caused by an agent. This observation can perhaps be explained by the fact that responses in the present study implied fewer semantic components because they only involved static configurations (static task) or very simple events that had just occurred in the immediate situation (dynamic task). In contrast, previous experimental situations were more complex and required the expression of more semantic components. In particular, in descriptions of caused motion, most relating to the present study, many types of information were built into the stimuli (cause, manner causing displacements, manner of motion of the agent and displaced entity, path of the agent and displaced entity). In order to provide complete descriptions in these situations, French adults frequently produced rather complex subordinate constructions, such as gerunds or relative clauses, whereas French children often omitted some of the relevant information in their responses. German children also expressed increasing semantic information with age, but developmental changes were less striking, as was also observed for English-speaking children with both types of motion events (Hickmann & Hendriks, 2010; Hickmann et al., 2009). In the present study, then, children’s responses were very similar to the adult target pattern from four years on in both languages, including in

142

The Acquisit ion of French in Mult ilingual Conte x t s

French. This result suggests that children’s cognitive and linguistic system at four years of age can process spatial information in simple static and dynamic tasks where all relevant semantic components can be expressed in compact syntactic constructions without the need to use more complex structures (e.g. subordination). However, they seem to encounter difficulties when confronted with more complex tasks that lead children to select only those dimensions that are most salient to them. In both simple and complex tasks, however, the results show the strong impact of language-specific properties that influence children’s productions from an early age onwards. In conclusion, from very early on, children’s lexicalization patterns follow the target adult system. However, before they reach this target, they still undergo various adjustments that are characterized by a constant interaction between language acquisition and cognitive development. Although the results of our study are in line with previous studies which provide support for Talmy’s typological distinction between verb- and satellite-framed languages, much more research is still needed in this domain. First, further research is necessary to examine cross-linguistic variations within the same language family. In comparison to previous findings concerning English (Hickmann & Hendriks, 2006), German speakers rely more on specific verbs than English speakers in similar tasks. This is especially true in the dynamic task, where German speakers of all age groups show a greater lexical diversity in the (manner) verb lexicon than English speakers. Similarly, given the properties of Korean grammar, particularly the existence of compact resultative verb constructions, Korean children produce motion expressions of higher semantic density than French children, even though these two languages are presumably of the same typological family (Choi, 2011). Such within-family variations must be further investigated on the basis of more languages. As proposed by Slobin (2004), comparisons of adult verbalizations across different languages may lead to a more complex typological continuum ranging from highly V- to highly S-languages, rather than a mere dichotomy. A second question concerns the cognitive implications of our findings. Specifically, are the cross-linguistic differences observed on the linguistic level reflected in deeper differences on the conceptual level? If we assume that different lexicalization patterns imply different ways of ‘thinking for speaking’ (Slobin, 1996), future psycholinguistic research must address this challenging question by means of a variety of methodologies, such as non-verbal tasks and eye-tracking, that will allow us to gain more direct insight into the cognitive processes that underlie language production and acquisition.

Notes (1) Although French provides verbs that lexicalize path and manner in literary and/or higher registers, common verbs rarely combine both information components. Linguistic analyses (Kopecka, 2006) have also shown that some French motion verbs contain a prefix expressing spatial and aspectual components, for instance the verb

St at ic and Dynamic Locat ion in French and Ger man Child L anguage

143

accourir (‘to run quickly to’) which is a combination of the allomorph ac- of the Latin prefix ad- (‘to’) and the verb courir (‘to run’). However, in contrast to similar devices in satellite-framed languages, this phenomenon is not productive and is quite marginal. With respect to French grimper (‘to climb upward’), combinations with path markers as in *?Ils grimpent à travers la falaise (‘They climb upward across the rock.’) are at best odd, still intrinsically implying an upward direction, and can only be interpreted (through additional inferences) as involving motion in an upward direction but dispersed at different places and/or over a large distance. Any other combination of grimper with path markers that do not indicate upward direction as in **Ils grimpent à travers le champ (‘They climb upward across the field’) is impossible. (2) By semantic neutral we understand those linguistic elements which can be used in many different contexts and therefore do not encode a specific semantic meaning. Semantic neutral verbs have also often been called light verbs or general-purpose verbs. (3) Neutral prepositions are often highly polysemous and can also express temporoaspectual marking (Vandeloise, 1991), for instance in Marie arrive à quatre heures (‘Mary arrives at four o’clock.’) or in Arrête de pleurer! (Lit.: ‘Stop from cry!’ ‘Stop crying!’). More generally, the prepositions à (‘at, to’) and de (‘from, of’) have become grammaticalized in many syntactic contexts where they do not have any spatial or temporal meaning, e.g. Je t’invite à dîner. (Lit. ‘I you invite at/to dinner.’ ‘I invite you for dinner.’) or Je me souviens de mon enfance. (‘I remember at/to my childhood.’ ‘I remember my childhood.’). (4) We thank Melissa Bowerman for lending us some of her stimuli (see Bowerman, 1996) to which new stimuli were added for the purposes of this study. These two tasks also included other items not discussed here (see Hickmann & Hendriks, 2006). In the static task other items involved prototypical contact/support situation items as well as distractors showing various spatial relations (such as containment). In the dynamic task each item actually consisted of two successive actions: after placing an object somewhere (doing actions, which are analyzed here), the experimenter then removed it from its new location (undoing actions, not analyzed here).

Appendix 1: Pictures in the Static Task (1) a spider [figure] on a ceiling [ground] (2) a ring on a finger (3) a coat on a hook (4) a hole in/on a towel (5) a balloon on a stick (6) a shoe on a foot (7) a telephone on a wall (8) a panel on a tree (9) a crack in/on a cup (10) a band-aid on a leg (11) a picture on a wall (12) rain drops on a window (13) a hook on a wall (14) a bandana on someone’s head

144

The Acquisit ion of French in Mult ilingual Conte x t s

(15) insects on a wall (16) a pendant on a chain (17) a handle on a door (18) a cork on/in a bottle (19) a lamp on a ceiling (20) an earring on an ear

Appendix 2: Actions in the Dynamic Task (1) Putting a lid onto a pan (2) Putting a cap onto a pen (3) Putting a band aid onto a hand (4) Putting a hook onto a wall (5) Putting a jacket onto a hook (6) Putting a straw hat onto a doll (7) Putting a beret onto a doll (8) Putting a scarf onto a doll (9) Putting a shoe onto a doll (10) Putting a bathrobe onto a doll (11) Putting one pop bead into another (12) Putting one Lego piece into another (13) Putting one Clippo piece into another (14) Putting two pop beads to join them together (15) Putting two Lego pieces to join them together (16) Putting several Lego pieces into a bag (17) Putting toys into a box (18) Putting toys into a bowl (19) Putting a cassette into its case (20) Putting a puzzle piece into a puzzle

6

Can L2 Learners Learn New Ways to Conceptualize Events? A New Approach to Restructuring in Motion Event Construal Jeanine Treffers-Daller and Françoise Tidball

Introduction This chapter discusses the ways in which English-speaking learners of French talk about motion events in their second language (L2). Speakers can choose how to construe an event, as they take different perspectives on what is happening in a story and explain the story differently. If a speaker chooses to say that a figure arrives at a particular location, while another says that s/he runs towards or into this location, both descriptions are appropriate verbalizations of the event. The key point is, however, that the choices speakers make are not idiosyncratic: they reflect language-specific perspectives on motion events. Speakers of English opt to choose a manner verb such as to run much more often than speakers of French, who often focus on path alone: they use path verbs such as descendre ‘to descend’ or traverser ‘to cross’ to describe the movement of the figure, and only add manner in an adjunct if there is a special reason for highlighting this, as in il descend/traverse en courant ‘he descends/crosses while running’ (Hickmann, 2006: 13). In other words, speakers of English choose to verbalize manner much more often than speakers of French, who generally express manner only when it needs to be foregrounded for some reason (see also Harr & Hickmann, this volume, Chapter 5). As Slobin (1987: 435) puts it, in preparing how to verbalize a particular motion event, speakers select those 145

146

The Acquisit ion of French in Mult ilingual Conte x t s

characteristics that (a) fit some conceptualization of the event, and (b) are readily encodable in the language. According to Slobin, language exerts an influence on thought when speakers prepare to speak, and this is what has become known as ‘thinking-for-speaking’. This can happen because speakers are used to particular ways of encoding their thoughts, for example about motion, which are being used time and again. From now on these habitual ways of verbalizing thought are referred to here as conceptualization patterns. Von Stutterheim and Nüse (2003) provide important empirical evidence for thinking-for-speaking: they show that language-specific principles of information organization are available at the moment of utterance planning. It is these principles that guide language users in the selection and structuring of knowledge while speaking. While many authors agree that these language-specific principles are learnt during first language (L1) acquisition, it is unclear to what extent L2 learners are capable of conceptual restructuring in the L2. According to Slobin (1996: 89), L2 learners find this difficult, because the training one receives in childhood is ‘exceptionally resistant to restructuring in adult second-language acquisition’. Therefore learners may continue to rely on the conceptualization patterns from their L1 in, for example, construing motion events. This reliance on L1-mediated concepts when speaking in the L2 is referred to as conceptual transfer (Jarvis & Pavlenko, 2008). In their studies of the cognitive processing of motion events, Carroll and von Stutterheim (2003: 398) show that even advanced L2 learners ‘remain rooted in at least some of the principles of conceptual organization as constituted in the course of L1 acquisition’. They show that L2 learners continue to base the production of L2 speech on conceptualization patterns they have acquired for their L1, and this reveals itself in the information they select for verbalization (e.g. whether or not they pay attention to manner) and in the segmentation, structuring and linearization of the information, as well as in the perspectives they take on the event (see also Daller et al., 2011 for further discussion of conceptual transfer in motion event construal among TurkishGerman bilinguals). The debate regarding learners’ ability to restructure their conceptual system is, however, far from settled. According to Hendriks et al. (2008), even advanced English-speaking learners of French continue to rely on thinking-for-speaking patterns from the L1. Similar results were obtained by Larrañaga et al. (2012) and Negueruela et al. (2004), but Cadierno (2004), Cadierno and Ruiz (2006), Navarro and Nicoladis (2005) and von Stutterheim (2003) claim that the evidence for L1 transfer of thinking-for-speaking patterns is limited. In recent work, however, Cadierno (2010) finds more evidence for L1 transfer of conceptualization patterns than in her earlier work. Because of the mixed evidence in this domain, Schmiedtova et al. (2011) call for more studies to gain a better understanding of the role of different factors in the domain of event conceptualization in L2.

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

147

As Pavlenko (2005, 2011) and Athanasopoulos (2011) have demonstrated, a range of factors affect bilingual cognition: factors they mention are age of acquisition, context of acquisition (whether learners live in the L1 context or the L2 context), length of stay in the target language culture, and amount of language use as well as general and specific language competence. The former refers to the general level of competence in the two languages, and the latter to knowledge of the specific domain of investigation (e.g. motion or colour). In addition, the effects may not be the same in different cognitive domains. Predicting whether or not conceptual restructuring will take place remains therefore very difficult. According to Hendriks et al. (2008: 21) studies of students of different proficiency levels can provide more insight into whether or not learners are able to reconceptualize spatial information in L2 acquisition. The current study aims to shed light on this issue by comparing motion event construal among L1 English adult learners of French at two different levels (intermediate and advanced). The discussion in this field is also relevant for the wider discussion about the role of transfer in L2 acquisition. It is interesting in this context that Ringbom (2007: 1) hypothesizes that the primary concern of learners is to discover similarities between the L1 and the L2 or other languages they already know: this view is clearly in line with Slobin’s view that L2 learners are reluctant to restructure their L1 conceptualization patterns. Kellerman (1995: 141) goes even further and suggests that learners may actively look for ‘the linguistic tools which permit them to maintain their L1 perspectives’. In the domain of motion, Latinate path verbs enter, arrive and descend may well be examples of such tools for English L1 learners of French. Larrañaga et al. (2012) consider the existence of such cognates to be a double-edged sword: on the one hand, they facilitate learning of motion expressions by L2 learners but, on the other hand, they may trigger fossilization, as learners could be led to believe that motion events are based on the same pattern in both languages, and fail to notice the differences. Support for Slobin’s approach can also be found in the cognitive linguistic notion of entrenchment (Langacker, 1987). As motion expressions are so widely used in language, they become cognitively routinized or entrenched (Langacker, 1987) in the speakers’ minds. For this reason, L2 learners will experience difficulties in learning new patterns of motion event construal which differ from the entrenched patterns. While most authors recognize that in L2 acquisition the first language influences the second language (and vice versa), researchers do not agree on exactly what features can be transferred, what the constraints on transfer are, whether transfer is important right from the start (Montrul, 2006; Schwartz & Sprouse, 1996) or only at a later stage when the processor is ready (Pienemann, 1998; Pienemann et al., 2005), and whether it continues to play a role in the advanced stages of L2 acquisition, as Lefebvre et al. (2006: 10)

148

The Acquisit ion of French in Mult ilingual Conte x t s

claim is the case, if learners do not have access to sufficient positive evidence. We hope the current study will shed some light on the issue of the role of language competence in L2 in the transfer of conceptualization patterns from L1. In this context, Athanasopoulos’ (2006, 2011) finding that intermediatelevel learners were clearly influenced by L1-based cognitive processing while more advanced learners had shifted towards L2-based cognitive patterns is particularly interesting for the current study, even though the study of Athanasopoulos focused on a different domain, namely the consequences of the L2 acquisition of a grammatical number system on categorization. In the current study we test the hypothesis that English-speaking learners of French will struggle to acquire the target-like expression of motion in their L2, and will transfer L1 conceptualization patterns to their L2. Evidence for conceptual transfer will be sought, for example, in the amount of attention paid to manner in describing motion events in the L2. We expect English-speaking learners to select manner more often for verbalization than native speakers of French, and to use manner of motion verbs in situations where the moving figure crosses a boundary, which is not allowed in French because of the boundary-crossing constraint (Slobin & Hoiting, 1994: see the section on ‘Transfer and Simplification’). Finally, we predict that language competence in the L2 is an important mediating factor: learners who have not had extensive exposure to everyday spoken and written French will rely to a larger extent on thinking-for-speaking patterns from the L1 than learners who had been on a year abroad prior to data collection. As Jarvis (2000) and Jarvis and Pavlenko (2008) have shown, some of the confusion in the literature about the importance of transfer in L2 acquisition is probably due to the lack of methodological rigour in studying transfer. Researchers have far too often considered transfer as a ‘you-know-it-whenyou-see-it phenomenon’, and have assumed that particular interlanguage features were the result of transfer simply if these features occurred in the speech or writings of learners and native speakers of a particular language. Jarvis (2000) and Jarvis and Pavlenko (2008) therefore propose to compare the frequency of particular features in the L2 of learners with different L1s who study the same target language (intergroup heterogeneity) in addition to comparisons between the features in the interlanguage of learners and features in their L1s (cross-linguistic performance congruity). If a second group of L2 learners with a different L1 is not available, every effort should be made to compare the data with results from the available literature on the acquisition of the same feature by other groups of L2 learners. The approach we try out in this chapter builds on the work of Jarvis (2000) and Jarvis and Pavlenko (2008) but differs from it in that we explore what constitutes relevant evidence for transfer as well as for restructuring. Evidence for the former is sought in the differences between the overall distribution of relevant features in the speech of L2 users and monolingual users of the L1 (the source language). This comparison provides information about

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

149

the distance between the speech of L2-users and monolingual speakers of the source language. We will assume there is evidence for transfer if the following situation occurs: with respect to the differences in the overall distribution of features between and L2-users and monolingual uses of the L1: • •

If the differences are not significant, it is likely that the L2-users have transferred features from the L1. If the differences are significant, there is no evidence that the L2-users have transferred features from the L1.

Of course additional evidence from L2-users with a different L1 will need to be provided too (intergroup heterogeneity). In addition, we propose to make analyses of the differences in the distribution of features in the speech of L2-users and monolingual users of the L2 (the target language). This comparison provides information about the distance between the speech of the L2-users and monolingual users of the target language and constitutes crucial evidence for (the lack of) restructuring. With respect to the differences in the overall distribution of features between L2-users and monolingual users of the L2, we assume that: • •

If the differences are significant, there is no evidence that the L2-users have restructured their interlanguages towards the L2 norm. If the differences are not significant, it is likely that the L2-users have (to a certain extent) restructured their interlanguages.

Thus, this second source of information provides evidence about a different kind of intergroup heterogeneity, which can be used to argue the case for or against restructuring. Our approach is partly based on Grosjean’s (2001) language mode continuum. According to Grosjean, in their everyday lives bilinguals and L2-users find themselves in various language modes that correspond to points on a monolingual–bilingual mode continuum (see Grosjean, 2008, for a thorough discussion). At one end of the continuum, they find themselves in a completely monolingual mode (e.g. when speaking to monolingual users) and at the other end they find themselves in a bilingual mode (e.g. when speaking to friends or relatives who are bilingual too). Grosjean (2001) has also pointed out that L2-users travel along the language mode continuum, so their speech could contain more or less transfer or evidence for restructuring depending on the situation in which they were recorded. Thus, any data collection from bilinguals (L2-users) necessarily constitutes a snapshot of their language use only. We have adapted Grosjean’s (2001) model to the one presented in Figures 6.1–6.4, which illustrate the fact that L2-users find themselves on the interlanguage continuum between native speakers of their L1 and native speakers

150

The Acquisit ion of French in Mult ilingual Conte x t s

of their L2. As Cook (1992, 2008) has pointed out, both languages coexist in the mind of the L2-user, and therefore the speech patterns of speakers who know more than one language do not correspond fully to those of monolingual users of the L1 or the L2 (see Cook, 1992, 2008, for further details on multicompetence). Contrary to other studies which focus exclusively on a single category of motion, we have chosen to make a comprehensive analysis of a range of motion categories, i.e. manner, path, deixis and cause, as this enables us to see in which proportion each type is being used. Although motion events can often be verbalized in different ways, for example by choosing either a path verb or a manner verb or a deictic verb, some choices are more common than others in each language. It is by studying the overall distribution of different types of motion expressions that language-specific preferences for one or the other verbalization are most clearly revealed, and not by focusing on just one aspect of motion. The current study focuses on the characteristics of L2 speech only, and does not look at possible consequences of language learning for the speakers’ L1. For a comprehensive review of different processes of cognitive restructuring that can take place in the minds of L2-users with respect to both L1 and L2, the reader is referred to Pavlenko (2011). With respect to the distribution of a range of features in the speech of L2-users and monolinguals, the following four possible scenarios might occur: (A) The L2-users are significantly different from monolingual users of L2 but not from monolingual users of L1: this is the transfer scenario. (B) The L2-users are significantly different from monolingual users of L1 but not from monolingual users of L2: this is the restructuring scenario. (C) The L2-users are significantly different from both groups of monolinguals: this is the creative or hybrid scenario. (D) The L2-users are not significantly different from either group of monolinguals: this is the convergence scenario. It is important to note at this point that the absence of significant differences between the L2-users and monolinguals does not mean that the speech patterns of both groups are the same: we agree with Cook (1992, 2008) that the speech patterns of L2-users are unlikely to be completely the same as those of monolinguals. Each possible scenario will now be presented briefly. Under Scenario A (Figure 6.1), the learners have not yet moved very far on the interlanguage continuum: they find themselves closer to the source language than to the L2 target. This is the transfer scenario: the learners are not statistically significantly different from monolingual users of the L1 but they are significantly different from monolingual users of the L2 with respect to a particular feature or the distribution of a range of features.

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

151

L2 -users

Differences n.s.

Differences significant

Figure 6.1 The transfer scenario (A)

Thus, for example, if there is no statistically significant difference between Level 1 learners and native speakers of English in the frequency with which they conflate manner with motion in the main verb, this will constitute evidence that the learners still follow the English patterns of verbalizing manner. Under Scenario B, the learners have moved far enough on the interlanguage continuum towards the target language that they are significantly different from monolingual users of the L1 but no longer significantly different from monolingual users of the L2. This is the restructuring scenario in Figure 6.2. Scenario B is one in which learners have started restructuring their interlanguage grammars to such an extent that the distribution of features in their L2 no longer reflects the distribution of features in the L1. Such a scenario might occur if English-speaking L2-users of French discover that native speakers of French use verbs to express path and start using these with a frequency that approaches that of monolingual users of French. Scenario C occurs when the learners produce hybrid constructions that are found in neither their L1 nor their L2 or when the frequency distribution of a particular feature in the learner data is significantly different from the

L2 -users

Differences significant

Figure 6.2 The restructuring scenario (B)

Differences n.s.

152

The Acquisit ion of French in Mult ilingual Conte x t s

L2-users

Differences significant

Differences significant

Figure 6.3 The creative or hybrid scenario (C)

frequency distributions in both the L1 and the L2. This is the creative or hybrid scenario (see Figure 6.3). An example of this could be if L2-users of French start using path verbs significantly more often than native speakers of English, but the native speakers of French use these significantly more often in French than the L2-users. Another example of such creative constructions could be the use of verbless static expressions to describe motion events, as described by Giacobbe (1992), which are not common among adult native speakers of either English or French. Further details of these are given in the section on ‘Transfer and Simplification’. Finally, Scenario D, depicted in Figure 6.4, is the convergence scenario, where L2-users’ productions are not significantly different from either the source or the target language. While this seems hardly possible if the target languages are very different from each other, it might occur in situations where languages are similar with respect to a particular phenomenon, or where convergence between two languages has given rise to constructions or distributions that are intermediate between both languages. The existence of cognates or (perceived) syntactic similarities between the source and the target languages could trigger such a scenario. In the domain of motion, for

L2-users

Differences n.s.

Figure 6.4 The convergence scenario (D)

Differences n.s.

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

153

example, French L2-users of English could make use of path verbs such as to enter to describe a movement into a closed space, as in to enter a room, which is possible but not the preferred expression in English (see the next section for more details). A slight, but non-significant increase in the use of such cognates could be seen as an example of Scenario D. In this chapter we hope to show that this model provides us with a rigorous method for identifying conceptual transfer as well as restructuring in L2 acquisition and for measuring the distance between the productions of learners and those of monolingual speakers of either the source or the target language. Approaches such as those of Jarvis (2000) and Jarvis and Pavlenko (2008) offer a rigorous method for identifying the role of the L1 in L2 learning, and are therefore very helpful in providing evidence about the distance between the productions of learners and those of speakers of the source language. For studies of restructuring in L2 acquisition which focus on the learners’ success in acquiring target-like patterns, it is crucially important to study the distance between the learners’ productions and those of speakers of the target language. We hope to show with the current chapter how this can be done. The structure of the chapter is as follows. Below we first present a summary of the main differences between motion event construal in French and English. The next section focuses on the existing literature on the L2 acquisition of motion in French and other Romance languages. After that the hypotheses are formulated, and the methods for the current study are explained. Then we present the results of our study, and in in the final section these are discussed in the light of Slobin’s thinking-for-speaking theory, and some thoughts for further research are offered.

Motion Event Construal in English and French Since the groundbreaking work of Talmy (1985, 2000a, 2000b) on the typology of lexicalization patterns, which formed the basis for Slobin’s (1996) highly influential thinking-for-speaking framework, many researchers have become interested in the ways in which speakers map conceptual structure onto surface structures, and in cross-linguistic influence in the conceptualization of motion in L2 learners and bilinguals. Before discussing the issue of transfer, we will look at Talmy’s and Slobin’s framework and how this applies to differences between English and French. Speakers of English and French differ in the way they construe motion events. According to Talmy’s (1985, 2000b) typology, English and French belong to two different types of languages, satellite-framed and verb-framed languages, respectively, because of the different ways in which they encode path and manner in motion events. In French, a verb-framed language

154

The Acquisit ion of French in Mult ilingual Conte x t s

(V-language), path is typically encoded in the verb (entrer ‘to enter’, sortir ‘to exit’), whereas in English, a satellite-framed language (S-language), the path component is typically expressed in a particle associated to the verb (in, out), as in (1a, 1b). (1) a. b.

The man goes into the bank L’ homme entre dans The man enters in ‘The man goes into the bank.’

la the

banque bank

Deictic motion verbs express motion either towards the speaker (venir) or away from the speaker (aller),1 but do not express the vector, i.e. the arrival, traversal or departure that the figure can execute with respect to a particular ground, nor do these verbs specify any details regarding the conformation component of the path, i.e. the spatial relation of the path to the ground.2 The path verb entrer, for example, specifies that the motion is an arrival and that it is a movement into an enclosure, whereas the verbs venir and aller do not provide any such information. Venir and aller can be seen as verbs that conflate deixis and motion in the main verb. In English, the verbs take and bring are also deictic motion verbs, but they express caused motion rather than voluntary motion. In French this distinction is expressed in prefixes, emporter/emmener ‘take away from the speaker’ versus apporter/amener ‘bring towards the speaker’. English and French differ from each other with regard to the encoding of manner. In English, manner is characteristically conflated with manner in the verb as in (2a); in French, it is expressed by an adjunct to the verb, as in (2b). (2) a. b.

The man runs into the bank L’ homme entre dans The man enters in ‘The man runs into the bank.’

la banque the bank

en courant running

In other words, as Hickmann et al. (2009) put it, English and French native speakers differ from each other with respect to the semantic density of their motion expressions. They show that descriptions of motion are denser in English in that manner and path are often packed into the same clause in English, as in (2a), while the same information is spread over different clauses in French, as in (2b). Whereas English speakers habitually express manner, it is often omitted by French speakers. When it is the habitual way of moving, such as flying for a bird, as in (3), manner is not mentioned unless it needs to be foregrounded for a purpose, as in (4).

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

155

(3) a. b.

The bird flew out of the hole L’ oiseau est sorti The bird is exited ‘The bird left the hole.’

(4) a.

Paul a monté la côte à pied, Sophie a pris le bus. Paul has ascended the hill on foot, Sophie has taken the bus ‘Paul went up the hill on foot; Sophie took the bus.’

du from.the

trou hole

The ease with which information can be processed is a major factor in the lexical and grammatical choices made by speakers (Slobin, 2004). As path is encoded in the satellite in English (in, out, up, down), manner and motion can be neatly conflated in the verb, the obligatory constituent of the sentence, thus making the expression of manner easy to process. As Slobin (2003: 4) puts it, English speakers get manner ‘for free’, whereas for French speakers elaborating on manner is ‘more of a luxury’. Slobin (2004: 232) also notes that in V-languages manner adjuncts are generally added to path verbs, and not to manner verbs, whereas in S-languages, manner adverbs qualify manner verbs, thereby augmenting attention to manner. English has developed a rich lexicon of manner verbs and English speakers learn to use a wide variety of such verbs from an early age. As Slobin (2003) has demonstrated in his study, English-speaking preschool children used 34 types of manner-of-motion verbs while French-speaking preschool children used far fewer types. Thus, manner of motion is a salient lexical domain for English speakers whereas it is not part of the habitual expression of motion in French. However, patterns that are different from the overall typology can be found in English as well as French. As Talmy (1985: 64) has shown, English has a system of lexicalization doublets, i.e. the same verb can be used without an incorporated idea of motion, as shown in (5a) or in constructions which do imply movement, as in (5b). (5) a. b.

The craft floated/was afloat on a cushion of air The craft moved into the hangar, floating on a cushion of air.

In (5a) the verb float does not imply movement, but in (9b) it does. The expression of motion in (5b) is similar to the French expression in (2b), in that manner is expressed in a subordinate clause. In her study of the semantic structure of motion verbs in French, Kopecka (2006) shows that French also has satellite-framed patterns, but these have generally been neglected in studies of the expression of motion in French. According to Kopecka, 22 verbal prefixes, such as a(d) ‘to, toward’ as in accourir ‘run to’ and em/en ‘away, off’ as in s’envoler ‘fly away’, make it possible to express elements of path in the main verb.

156

The Acquisit ion of French in Mult ilingual Conte x t s

Although French does have common verbs which conflate manner and motion (marcher ‘to walk’, courir ‘to run’, ramper ‘to crawl’, glisser ‘to slip/slide’), most authors agree that their use is restricted by the boundary-crossing constraint (Slobin & Hoiting, 1994): they are not used when the path involves crossing a boundary. Thus, l’homme court dans la maison can only depict a man running around inside a house and not a man running into a house. A path verb is used in French to express the boundary crossing il entre dans la maison ‘he enters the house’, leaving the manner to be added (en courant ‘by running’) or inferred. An exception to this rule is the use of manner verbs which denote instantaneous actions such as to throw oneself or plunge (Slobin, 2004: 226). These can be used in French in a boundary-crossing situation, e.g. plonger dans la piscine ‘to plunge into the swimming pool’. Cummins (1996) proposes that the difference between French and English in the expression of motion does not reside in the lexicalization by the verb but in the semantic content of the prepositions. French has no equivalent to the English preposition into to express transition from an initial location along a path to a final location: into can encode both the path and the attainment of final location, whereas dans can only locate an entity or an activity at a final location. Thus, Il arrive dans la maison ‘he arrives in(side) the bank’ and Il court dans la maison ‘he runs in(side) the house’ both depict the man at the same location inside the house. Similarly, the preposition to has a different semantic content from its French equivalent à, as it can express path and end location – compare ‘he wobbled at me’ with ‘he wobbled to me’ (examples from Cummins) – whereas à can only locate the entity at its end point. This often causes difficulties for English learners of French who transfer the notion of path contained in ‘to’ to à and translate ‘the bus to the station’ by ‘le bus à la gare’ (‘the bus at the station’). Interestingly, Pourcel (2004) reports some unexpected lexicalization patterns by French native speakers who conflate verb and manner and express path in the satellite, as in (6–7). (6) Un homme court en traversant la rue A man runs crossing the road ‘A man is running across the road.’ (Pourcel, 2004: 353) (7) Un homme pédale à vélo en montant la A man pedals on his bike going up the ‘A man is cycling up/upwards.’ (Pourcel, 2004: 353)

rue street

Although this lexicalization pattern is unusual as it does not fit the typological properties of French as a V-language, it does occur and illustrates the fact that some satellite-framed constructions can be found in French.

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

157

Cummins (1996: 48) also gives an example of two motion events where the prepositions ‘locate the theme at the final location’ (8). (8)

Max a couru dans sa chambre / au Max has run in his bedroom / to.the ‘Max ran into his bedroom/to the store.’

magasin store

While this seems to point to a weakening of the boundary-crossing constraint in Canadian French and a shift towards the S-language pattern of English, Cummins (p.c.) does not believe that there is a difference between Canadian French and metropolitan French in this regard. The existence of examples such as (9), and Stringer’s (2010) Example (10) from a five-year-old, which appear to involve a boundary crossing as well, support Cummins’ point of view. (9) Je me suis levée et j’ai couru dans le salon pour rejoindre ma soeur et ma mère était là, allongée sur le canapé!! (http://entite.over-blog.com/article-2788877.html, accessed 9 February 2015) (10) Il est en train de grimper dans sa maison (5-year-old) He is in process of climb in his house ‘Now he’s climbing into his house.’ [context: tree house] (Stringer, 2010: 21) Thus, a more detailed analysis of the ways in which motion is expressed in French and English shows that there is variation in both languages. The existence of variation clearly complicates the task of learners of French, who need to discover which expressions represent the habitual ways of describing motion, and which ones are more marginal patterns (see also Hendriks et al., 2008). Finally it should be pointed out that there is no complete overview of all motion verbs in either English or French. The most complete list of English motion verbs can probably be found in Levin (1993), who provides a typology of English verbs. For French there is Krassin’s (1984) study of the semantic field of motion in French. This study is one of the rare works which provides not only a semantic classification but also a list of 207 motion verbs. Other studies which deal with the syntactic structures associated with French motion verbs and provide a contrastive analysis of the expression of motion in French and English are Braun (1976) and Vinay and Darbelnet (1958), but these do not offer a complete list. In most studies a distinction is made between motion involving no change of location, as in gigoter ‘to wiggle’, trembler ‘to shake’, s’accroupir ‘to crouch’ or se pencher ‘to lean over’ and, on the other hand, motion involving movement from one location to another, as in culbuter ‘to tumble, fall’ or se précipiter ‘to throw oneself’, which are examples of translational motion or translocation (Zlatev et al., 2006). Krassin’s (1984) list consists of verbs of translational

158

The Acquisit ion of French in Mult ilingual Conte x t s

motion, although she provides examples of other kinds of motion verbs too, for example vaciller ‘to sway’, which does not necessarily imply movement across space. In the current study both types of verbs are being analyzed.

Transfer and Simplification in the Expression of Path and Motion in L2 Learners of French There are only a few studies into transfer in the L2 acquisition of French motion verbs and even less on transfer in the conceptualization of events, as most research focuses on transfer in the acquisition of syntax (e.g. White, 1991). The first study into this topic is that of Schlyter (1984), who investigated Swedish learners of French, but this does not use Talmy’s framework which was published a year later. She found some evidence for transfer among learners who learn French in classroom settings, in that they used manner of motion verbs in combination with a path satellite, 3 as in (11a): (11) a.

On a fait la bicyclette à des endroits We have done the bicycle to ART places ‘We cycled to different places.’ (Schlyter, 1984: 13)

différents different

In Standard French the following expression is normally used: (11) b.

On est allé en bicyclette à We are gone on bike to ‘We went on bike to different places.’ (Schlyter, 1984: 13)

des endroits différents ART places different

Schlyter also found that learners overuse venir ‘to come’ in the initial stages of language acquisition, where native speakers prefer path verbs such as sortir ‘to go out’. According to Schlyter, informants who acquire French in naturalistic settings are less likely to transfer Swedish ways of expressing motion into French. Transfer in structures such as those in (12) and (13) would only be common among learners in classroom settings. (12) Le fleuve coule dans un tunnel [for: le fleuve entre dans un tunnel] ‘The river flows into a tunnel.’ (Swedish learner of French) (Schlyter, 1984: 36) (13) Je peux pagayer à l’école [for: Je peux aller à l’école en pagayant] ‘I can paddle to school.’ (Swedish learner of French) (Schlyter, 1984: 36) Harley and King (1989) find similar results in the interlanguage data from Anglophone learners of French, which they interpret as representing transfer of the English way of expressing motion into French. Harley and

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

159

King (1989) show that Anglophone learners overuse venir and aller, as in (14) and (15), whereas native speakers of French would prefer to use verbs which conflate motion and path, such as sortir ‘go out’ or entrer ‘go in’.4 The authors attribute this to transfer from English. (14) Après un heure Madame Dupont a venu au balcon After one hour Mrs Dupont has come to.the balcony ‘After one hour Mrs Dupont came to the balcony.’ (Canadian immersion student, 6th grade) (Harley & King, 1989: 427) (15) Elle est allé dans la maison She is gone in the house ‘She went into the house.’ (Canadian immersion student, 6th grade) (Harley & King, 1989: 427) The authors observe that immersion students make substantially less use of the following verbs: arriver ‘arrive’, descendre ‘go down’, monter ‘go up’, partir ‘leave’, passer ‘go along’, redescendre ‘go down again’, rentrer ‘go home’ and sortir ‘leave, go out’, and make more use of the following verbs which the authors consider to be easily translatable in that the syntactic and semantic structures associated with these correspond to those found in English: aller, courir, entrer, grimper ‘to climb’, marcher ‘to walk’, retourner ‘to return’, sauter ‘to jump’ and venir. Despite the apparent ease with which they select these verbs, students in Harley and King’s study were found to transfer aspects of the syntactic information from their L1: perhaps even the entire subcategorization frame which is associated with the verb, as can be seen in sentences such as (16), where the immersion students use entrer with a direct object (as is common in English) instead of with a prepositional phrase beginning with dans ‘in’. (16)

Trois bandits entre Ø Three bandits enter Ø (Harley & King, 1989: 426)

le the

banque bank

The use of English subcategorization information on French verbs may be interpreted as evidence that learners transfer not only phonetic forms but also some semantic and syntactic information attached to lexical items into the target language. Transfer need, however, not necessarily be involved in the production of (14) and (15), as it is not impossible that they are the result of simplification. Students in the study may have chosen to simplify the constructions by splitting motion and path and opting for a deictic motion verb (venir or aller) followed by a path particle. In fact, learners do not always resort to deictic

160

The Acquisit ion of French in Mult ilingual Conte x t s

motion verbs as a simplification strategy. Giacobbe (1992), for example, shows that an adult learner with Spanish as her L1 starts from a path-less, static conceptualization of space. The learner does not use aller, despite the fact that Spanish ir and French aller are so similar in their conjugation (cf. Sp. vas, va, van versus Fr. vas, va and vont). Instead she uses the similarity between the French and Spanish prepositions a, de and en to construe the event, as in (17) and (18). (17) Y après la femme al camion de la police And afterwards the woman into.the van of the police ‘And afterwards the woman goes into the police van.’ (Giacobbe, 1992) (18) Après de la prison Afterwards from the prison ‘Afterwards s/he returns from prison.’ (Giacobbe, 1992) Giacobbe’s study could be interpreted as evidence for Pienemann’s (1998) claim that at the early stages of naturalistic L2 acquisition wholesale transfer may not be an option because of processing problems. Simplifications in motion event construal can sometimes also be found in child language. According to Bowerman’s (1982; in Allen et al., 2007) universal hypothesis children have an initial strategy of isolation or differentiation of semantic elements. Thus, they would have an early preference for mapping each semantic element onto a different lexical item or syntactic structure. Some evidence for this hypothesis can be found in Stringer (2006: 139), who shows that young French children often split complex trajectories into subevents. Predicates such as traverser ‘to cross’ imply that the figure goes into a particular reference element (ground) and comes out on the other end, which is more complex than predicates which express only one of these two subevents. As (19) shows, young children do not use traverser, but split the event into two or three different subevents, whereas adults express the whole trajectory using the verb traverser; see (20). (19)

Il va dans la rivière, il nage, il ressort de la rivière He goes in the river, he swims, he comes.out of the river ‘He goes in the river, he swims, he gets out of the river again.’ (French NS; 3;6) (Stringer, 2006: 140) (20) Il traverse la rivière en nageant He crosses the river by swimming. ‘He swims across the river.’ (French NS, adult) (Stringer, 2006: 140)

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

161

Stringer also found examples of children and adults who split motion and path, as in (21), although children as young as three were found to use verbs such as entrer in these contexts, as (22) shows. (21)

(22)

Il va dans la caverne He goes in the cavern ‘He goes in the cavern.’ (French NS 7;6) (Stringer, 2005: 185) Il rentre dans la caverne He enters in the cavern ‘He enters the cavern.’ (French NS 3;2) (Stringer, 2005: 174)

It does not appear to be the case that the acquisition of path and motion necessarily involves a developmental stage in which these two are split. In their study of the acquisition of path and manner in Japanese, Turkish and English, Allen et al. (2007) tested the universal hypothesis, but found little evidence for it. Instead of spreading path and manner over different clauses, children prefer synthetic expressions of path and manner in one clause, and they acquire the language-specific ways of expressing path and manner by the age of three. Further evidence for French was obtained by Hickmann (2003) and Hickmann et al. (2009), who observed that French children often start out by verbalizing just one component of a motion event, and generally acquire path before manner. In a recent study of motion event construal among English-speaking learners of French, Hendriks and Hickmann (2011) investigated voluntary motion and Hendriks et al. (2008) studied caused motion among adult English-speaking learners of French. Both studies show that both intermediate and advanced learners struggled with the target-like expression of motion in French. Many learners of French expressed manner and cause of motion in the main clause, but in order to express path the intermediate-level learners often used expressions which did not imply a change of location, as in (23), or expressed in a separate clause, as in (24). (23) Papy pousser/poussé une roue dans une caverne dans les bois. (LOW-INT_01) ‘Popi to push/pushed a tire in a cave in the woods.’ (Hendriks et al., 2008) (24) Popi tirer/tiré une sac et # et ascende le # le toit. (LOW-INT_01) ‘He pull/pulled a bag and # and ascends the # the roof.’ (Hendriks et al., 2008) Advanced learners sometimes transformed French path verbs into satellite-like devices, as in (25).

162

The Acquisit ion of French in Mult ilingual Conte x t s

(25) Il rouler/roulé le roue dans # entre le ferme. (ADV_06) *‘He to roll/rolled the tire in # between/enter the shed’ Importantly, Hendriks et al. (2008) point out that native speakers of French produce a great variety of structures to express manner, cause and path of motion. It is this variation which makes learning the target system opaque to English learners of French and explains the fact that learners continue to rely on the L1 in construing motion events.

Hypotheses of the Current Study We assume that English-speaking L2 learners of French will find acquiring new ways of thinking-for-speaking patterns difficult, because of the cognitive entrenchment of event construal patterns from the L1. With respect to the acquisition of path, we hypothesize that learners of French at lower intermediate levels may go through a phase where path and motion are split, and use deictic verbs such as aller, combined with a preposition such as à, which functions as a path satellite, or even opt for static expressions such as those illustrated by Giacobbe (1992). The static expressions and the overuse of deictic motion verbs are predicted to disappear in later stages of L2 acquisition, if the learners are exposed to sufficient input. The overuse of deictic motion verbs can either be interpreted as transfer from English (as deictic verbs in combination with path satellites are common in English) or as simplification. A comparison with native speakers of English may shed some light on which interpretation is more likely to be correct. An additional reason why learners may prefer aller and venir is that these are highly frequent verbs: they are among the first 25 verbs in the Corpaix word list (Véronis, 2000). The only other motion verbs in the first 25 in this list are arriver ‘to arrive’ and partir ‘to leave’ (see also Treffers-Daller & Calude, 2015). With respect to the acquisition of manner, we hypothesize that the learners: (a) will select manner more often for verbalization than native speakers of French; (b) will conflate manner in the main verb more frequently than native speakers of French; and (c) will use manner verbs in expressions in which a boundary crossing is predicated. With respect to caused motion, we expect learners to find it difficult to restructure their grammars and learn the fine-grained vocabulary of French. Intermediate-level learners will be expected to use simple verbs of caused motion, such as jeter ‘to throw’ or to avoid complicated constructions involving caused motion, but it is unlikely that this is related to transfer from English. We also hypothesize that language competence in the L2 is an important mediating factor and predict that transfer of conceptualization patterns is more prevalent among intermediate-level learners, who have had little contact with everyday spoken French, than among advanced learners who had been

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

163

on a year abroad prior to data collection. As is always the case in L2 acquisition, learners will differ from one another as to how successful they are in the long run and we can therefore expect variation in the performance of L2 learners, related to factors such as motivation and other individual characteristics of learners, but these are beyond the scope of the current paper.

Methods Our informants were 94 students at the University of the West of England (UWE), Bristol, divided into four groups (see also Table 6.1), all of whom were around 20 years old. There were two groups of learners of French, namely 21 first year students (all with an A level in French) and 20 final year students. All students had received some instruction regarding the differences between French and English in the construction of motion during their second year, but not much time is spent on this issue. The final year students had all undertaken either a one-semester or one-year placement in France the previous year. French native speaker data were obtained from a group of 10 Erasmus students from France who spent one academic year at UWE and 13 students in their first year at a French Business School following a three-month course in English as a foreign language. Native speaker data for English were gathered from a group of 30 monolingual Level 1 linguistics students at UWE Bristol. In the current study, a comparison with learners with a different L2 background can only be made on the basis of the available literature as we do not have access to another group of L2 learners of French. Therefore Schlyter’s (1984) study of Swedish learners and Giacobbe’s (1992) study of a Spanish learner of French will be used to provide the necessary information about intergroup heterogeneity (see the introductory section). All students undertook the same task under the same conditions: in the booths of the interpreting lab they individually recorded their description of two picture stories presented as cartoon strips of six pictures each (Plauen, 1996 [1952]). These were entitled Erfolglose Anbiederung ‘Unsuccessful Ingratiation’ (Appendix 1) and Unbeabsichtigte Helden ‘Unintentional Heroes’ (Appendix 2). Students were asked to tell two stories in order to ensure there were enough data to analyze for each student. The main protagonists in the Table 6.1 Overview of informants in the current study Groups N Mean age

Learners of French, Level 1

Learners of French, Level 3

Native speakers of French

Native speakers of English

21 19.3

20 22.4

23 20.3

30 19.7

164

The Acquisit ion of French in Mult ilingual Conte x t s

stories are a father and a son. In the first story they go for a walk along the beach with their dog and they play fetching the stick from the water. Another man comes along and wants to play with them but the dog is not interested. From now on this story will be referred to as the Lake Story. In the second story, the pair are witnesses to a bank robbery and kidnap in a bank. The father beats up the robber and frees the employees. This story will be referred to as the Bank Story. The students could use as much preparation time as they liked, they could tell the stories in any order they wished, and there was no time limit. In order to familiarize themselves with the data collection procedure, and to prepare for speaking in their L2, they were asked to explain on tape in French why they learnt French and to count to 30 in the L2 before telling the stories. We assumed this would also help trigger an L2-speaking mode (Grosjean, 2001), although it is unlikely that speakers were in a completely monolingual French mode while telling the stories, because the data collection took place in the UK and, prior to the recordings, the students were with their peers with whom they normally spoke English. To what extent this has impacted the students’ way of telling the stories is unclear and further research will need to show whether collecting data from the same informants in a French-speaking environment (e.g. at the end of the year abroad) would lead to significantly different results. Each participant also completed a C-test, which provided us with an external criterion for the linguistic competence of our participants (see Daller & Xu, 2009; Eckes & Grotjahn, 2006, for a discussion about the validity of this test as a measure of general language ability). This test consisted of a series of six short texts on a variety of topics in which, from the second sentence onwards in each text, the second half of every second word was deleted and students had to provide the missing half. In total, the students had to complete 120 words. The French C-test we used proved highly reliable (Cronbach’s Alpha 0.963, six items). As can be seen in Table 6.2, native speakers achieved the highest scores, and Level 3 learners outperformed Level 1 learners. A one-way ANOVA and a Tukey post hoc test showed that there are significant differences between all three groups in their performance on the C-test (F(2,61) = 105.371, p < .001; see Tidball & Treffers-Daller, 2007 for more details). In terms of the Common European Framework of Reference for Languages, the students on the course varied from A2/B1 at Level 1 to Table 6.2 Language proficiency of the learners of French, in comparison with native speakers Oral exam Level 1 Level 3 Native speakers

53 55

Written exam

C-test

53 61

51% 77% 92%

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

165

B2/C1 at Level 3. Native speakers of French who were enrolled on the course were excluded from the learner groups. Prior to the analysis, all data were transcribed in CHAT format (MacWhinney, 2000). In order to find all motion verbs in the French dataset, we created an include file with all 207 motion verbs listed in Krassin (1984). As the latter only lists verbs which can be used with human protagonists, we checked manually whether any motion verbs not listed in Krassin were used in those parts of the Lake Story where the movements of the dog are described. This turned out not to be the case. Upon scrutiny of the stories told by the students it did, however, become clear that 12 motion verbs which were repeatedly used by informants when telling the two stories were missing from this list. As giving an overview of the motion events in the stories was only possible with a list that was as complete as possible, we added those 12 verbs to the file with all the French motion verbs.5 With the help of wildcards we ensured that all inflected forms of the verbs could be traced in the data. We then ran the FREQ command in CLAN with the list of motion verbs as an include file, which allowed us to find all motion verbs in the two stories of all informants ([email protected]@). This made it possible to establish which groups used most motion verbs and which percentage of these verbs were manner verbs, deictic verbs, path verbs or verbs of caused motion. A small number of verbs belong in more than one category (e.g. échapper ‘escape’ encodes path as well as manner), and these were therefore counted as belonging in both categories. Although emporter ‘take away’, apporter ‘bring’, amener ‘take away’, and emmener ‘bring’ encode deixis as well as caused motion, these were classified as verbs of caused motion for the purposes of this chapter because we were interested in the frequency of usage of the basic deictic verbs aller ‘go’ and venir ‘come’ for reasons explained in the earlier section on ‘Transfer and Simplification’, and these therefore needed to be counted separately. As aller and venir are both used for a variety of functions in French, for this chapter only those instances in which they were employed as main verbs which expressed a motion event were analyzed. Thus we excluded utterances in which aller was used as an auxiliary, as in le chien va le chercher ‘the dog goes and fetches it’, or venir was employed to express the recent past as in il vient de sauver une banque d’ un holdup ‘he just saved the bank from a robbery’. In the current study we did not include periphrastic expressions of motion such as mettre à terre ‘to put down’ or faire tomber ‘to make fall’ or collocations such as prendre une plongée ‘take a dip’, but we did analyze phrasal verbs such as knock down, as studying motion in English is hardly possible without including phrasal verbs. In order to ensure that other verbalizations of manner were not ignored, we also explored the use of adverbial expressions of manner such as très vite ‘very fast’ in the data. Finally, we wanted to find out whether the learners of different levels were able to construe motion events in the context of a

166

The Acquisit ion of French in Mult ilingual Conte x t s

boundary crossing, as these are assumed to be particularly challenging for English learners of French. We therefore studied the construal of a single motion event, namely the action of one of the main protagonists in the bank story, the bank robber, who runs into the bank (see picture in the Appendix). While the stories contain many different kinds of motion events, this event was described by almost all speakers in each group, and there was little avoidance.

Results In this section we first present the overall results of our analysis of different types of motion verbs in the two stories of the learners and the native speakers, then we briefly sketch the use of manner adverbials, and we finish with a comparison of the construal of one motion event which constitutes a boundary crossing. The final section offers a summary of the findings.

Overall use of motion verbs in the two stories First of all we wanted to know whether native speakers of English and French produced similar numbers of motion verbs in telling the stories, and this was indeed the case: members of both groups produced approximately 15 tokens each. Thus, we can conclude that for native speakers of both languages the stories were comparable in that they generated similar numbers of motion events in each language. As one might expect, the average number of motion verbs that are being used by the learners increases from Level 1 (200 tokens; 21 types) to Level 3 (314; 27 types); both learner groups produce fewer motion verbs than the native speakers of French (368 tokens, 33 types), and these differences are significant (χ2 (2) = 30.75, p < .001). These results are entirely predictable, given the obvious differences in language competence among the three groups. Tables 6.3 and 6.4 offer a perspective on the diversity of manner verbs and the verbs of caused motion used by the different groups. Table 6.3 shows that the native speakers of both languages employ an equal number of types of manner verbs, but almost half of the types used by the native speakers of French occur only once in the data. In English most types are used very frequently, which clearly reveals the different perspectives on the events in the stories: while manner verbs are available in both languages, the French choose to use these much less frequently than the English. On average, when verbs from both stories are counted together, each native speaker of French uses 1.8 manner verbs, and the learners use a similar number of manner verbs on average, while the mean figure for native speakers of English is 6.4.

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

167

Table 6.3 Manner verbs (types and tokens) used by learners and native speakers of French found in both stories Native speakers English (n = 30)

Level 1 (n = 21)

Level 3 (n = 20)

Native speakers French (n = 23)

walk (53) run (53) swim (13) fly (12) rush (9) jump (7) hurry (4) float (2) climb (2) stride (1) march (1) Total 157 (m = 6.43)

courir ‘run’ (20) marcher ‘march’ (10) nager ‘swim’ (6) sauter ‘jump’ (1) plonger ‘plunge’ (1)

courir ‘run’ (20) nager ‘swim’ (12) marcher ‘swim’ (4) voler ‘fly’ (1) flotter ‘float’ (1)

Total 38 (m = 1.8)

Total 38 (m = 1.9)

courir ‘run’ (19) nager ‘swim’ (8) marcher ‘march’ (3) plonger ‘plunge’ (2) se précipiter ‘ rush’ (2) se baigner ‘bathe’ (2) pénétrer ‘penetrate’ (1) s’enfuir ‘flee’ (1) enjamber ‘stride’ (1) sauter ‘jump’ (1) s’envoler ‘fly away’ (1) Total 41 (m = 1.8)

Notes: aVoler has two different meanings: to steal and to fly. Both meanings occur in the story, but only occurrences of the meaning fly have been counted, as steal is not a motion verb.

The results displayed in Table 6.4 reveal that verbs of caused motion are used to an equal extent by both native speaker groups: on average, the French use 6.1 verbs of caused motion and for the English this figure is 6.4. Thus, both groups make extensive use of this type of motion verb. The averages for the learners are understandably much lower. The picture obtained by studying manner verbs or verbs of caused motion in isolation from other types of motion is, however, not complete: we argue that all types of motion verbs should be studied together to obtain a comprehensive account of motion event construal. As we will see below, the similarity between the groups disappears when manner verbs or verbs of caused motion are studied as a proportion of all motion verbs. As explained above, much more revealing information can be obtained from looking at the different categories of motion verbs expressed as a percentage of the total number of motion verbs, as this gives a comprehensive picture of motion event construal in all groups. Figure 6.5 offers an overview of the result of these calculations for all four groups of motion verbs. It shows, for example, that the percentage of manner verb tokens is highest among the native speakers of English and lowest among the French native speaker group, while the Level 1 and the Level 3 students occupy the middle position. The differences between the four groups in their use of manner verbs are statistically significant (χ2 (3) = 152.7, p < .001), but this could be due to the

168

The Acquisit ion of French in Mult ilingual Conte x t s

Table 6.4 Verbs of caused motion (types and tokens) used by learners and native speakers of French Native speakers English (n = 30)

Level 1 (n = 21)

Level 3 (n = 20)

Native speakers French (n = 23)

throw (53) take (48)

jeter ‘throw’ (15) lancer ‘throw’ (10)

jeter ‘throw’ (36) lancer ‘throw’ (12)

knock down (44)

amener ‘bring’ (1)

ramener ‘bring back’ (5) enlever ‘remove’ (3)

lancer ‘throw’ (39) bousculer ‘shove’ (21) jeter ‘throw’ (20)

push (11)

bring (9) turn (7) drop (5) carry (5) chuck (4) pull (2) lift (2) lead away (1) cart off (1) send (1) Total 193 (m = 6.4) Total 26 (m = 1.2)

rattraper ‘catch up’ (3) emmener ‘take away’ (1) emporter ‘take away’ (1)

Total 61 (m = 3.1)

renverser ‘overthrow, make fall’ (19) récupérer ‘retrieve’ (11) emmener/amener ‘take away’ (10) enlever ‘remove’ (8) envoyer ‘send’ (8) lâcher ‘release’ (3) apporter ‘bring’ (1) tourner ‘turn’ (1)

Total 141 (m = 6.1)

large differences between the English native speakers on the one hand and all French-speaking groups on the other hand. The results are, however, still significant when the English native speaker group is excluded (χ2 (2) = 7.52, p < .05). A full overview of all intergroup comparisons is offered in Table 6.5. When separate χ2 tests are carried out to establish which of the groups are significantly different from each other, it becomes clear that the Level 1 students are significantly different from the native speakers of French (χ2(1) = 7.5, p < .01) for all types of verbs except path verbs, as well as from the native speakers of English (χ2 (1) = 36.0, p < .001), but they are not significantly different from the Level 3 students (χ2 (1) = 3.08, p = .101). The Level 3 students are also significantly different from the native speakers of English (χ2 (1) = 73.2, p < .001) but they are not significantly different from the native speakers of French for manner and path. In other words, the Level 1 students are still far away from the target-like expression of manner in the

Can L2 Lear ners Lear n New Ways to Conceptualize Event s?

169

Table 6.5 Overview of intergroup comparisons for all motion verbs in the two stories

Manner Path Deixis Caused motion

1–3

1-NS (Eng)

1-NS (Fr)

3-NS (Eng)

3-NS (Fr)

ns ns ns ns

** ** ** **

** ns ** **

** ** ** **

ns ns ** **

Notes: ** = p < .01; * = p < .05.

L2, but the Level 3 students are more similar to native speakers of French in this respect. Figure 6.5 also reveals that the Level 1 learners make most use of deictic motion verbs. The overall differences between the four groups are significant (χ2 (3) = 71.54, p < .0001), but individual comparisons between groups do not always lead to significant results. A comparison between the Level 1 learners and the native speakers of French shows that these groups are significantly different from each other (χ2(1) = 25.46, p < .001), but the differences between the Level 1 and the Level 3 learners are not significant. The Level 1 learners are also significantly different from native speakers of English (χ2 (1) = 6.0, p < .05), and there are significant differences between the Level 3 students and the native speakers of French (χ2 (1) = 17.2, p < .001). 120 100 80 manner verbs caused motion verbs

60

deictic verbs 40

path verbs

20 0 NS English

Level 1

Level 3

NS French

Figure 6.5 Path verbs, deictic verbs, verbs of caused motion and manner verbs (tokens) as a percentage of all motion verb tokens in the speech of learners and native speakers

170

The Acquisit ion of French in Mult ilingual Conte x t s

Verbs of caused motion are used more by the advanced learners than by the intermediate learners, but native speakers also make extensive use of these. The overall differences between the four groups in their use of this type of motion verbs are significant (χ2 (2) = 68.0, p < .001). A comparison of the individual groups reveals that the Level 1 learners are significantly different from the native speakers of French verbs (χ2 (1) = 38.04, p