First Language Acquisition of Morphology and Syntax
Language Acquisition and Language Disorders (LALD) Volumes in this series provide a forum for research contributing to theories of language acquisition (first and second, child and adult), language learnability, language attrition and language disorders.
Editor Harald Clahsen
University of Essex
Lydia White
McGill University
Editorial Board Melissa F. Bowerman
Max Planck Institut für Psycholinguistik, Nijmegen
Katherine Demuth Brown University
Wolfgang U. Dressler Universität Wien
Nina Hyams
University of California at Los Angeles
Jürgen M. Meisel
Universität Hamburg
William O’Grady
University of Hawaii
Luigi Rizzi
University of Siena
Bonnie D. Schwartz
University of Hawaii at Manoa
Antonella Sorace
University of Edinburgh
Karin Stromswold Rutgers University
Jürgen Weissenborn Universität Potsdam
Frank Wijnen
Utrecht University
Mabel Rice
University of Kansas
Volume 45 First Language Acquisition of Morphology and Syntax. Perspectives across languages and learners Edited by Pedro Guijarro-Fuentes, María Pilar Larrañaga and John Clibbens
First Language Acquisition of Morphology and Syntax Perspectives across languages and learners
Edited by
Pedro Guijarro-Fuentes University of Plymouth
María Pilar Larrañaga John Clibbens University of Plymouth
John Benjamins Publishing Company Amsterdam / Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data First language acquisition of morphology and syntax : perspectives across languages and learners / edited by Pedro Guijarro-Fuentes, María Pilar Larrañaga and John Clibbens. p. cm. (Language Acquisition and Language Disorders, issn 0925-0123 ; v. 45) Includes bibliographical references and index. 1. Language acquisition. 2. Grammar, Comparative and general--Morphology. 3. Grammar, Comparative and general--Syntax. I. Guijarro Fuentes, Pedro. II. Larrañaga, María Pilar. III. Clibbens, John. IV. Title: 1st language acquisition of morphology and syntax. P118.F549 2008 401'.93--dc22 isbn 978 90 272 5306 4 (Hb; alk. paper)
2008004824
© 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
Table of contents
Introduction: First language, first language bilingualism and language impairment competence 1. Left-dislocated subjects: A construction typical of young French-speaking children? Christophe Parisse 2.
The development and interaction of Case and Number in early Russian Galina Gordishevsky & Jeannette Schaeffer
3. Verb movement and subject placement in the acquisition of word order: Pragmatics or structural economy? Marit R. Westergaard
1 13 31
61
4.
Three acquisition puzzles and the relation between input and output Manuela Schönenberger
87
5.
The acquisition of universal quantifiers in Spanish Linda Escobar & Vicenç Torrens
119
6. Subject-object asymmetry in children’s comprehension of sentences containing logical words Sharon Armon-Lotem
137
7. On the “vulnerability” of the left periphery in French/German balanced bilingual language acquisition Matthias Bonnesen
161
8.
The subjects of unaccusative verbs in bilingual Basque/Spanish children Pilar Larrañaga
9. Dominance, mixing and cross-linguistic influence: On their relation in bilingual development Tanja Kupisch 10. A cross-linguistic analysis of binding in Down syndrome Alexandra Perovic
183
209
235
First language acquisition of Morphology and Syntax
11. Balanced bilingual children with two weak languages: A French/German case study Natascha Müller & Antje Pillunat
269
Afterword Neil Smith
295
Index of subjects
299
Introduction First language, first language bilingualism and language impairment competence Language acquisition is a well established discipline in the linguistic field, with the beginnings of research in this area going back at least to Ronjat (1916). Research conducted since Ronjat’s seminal work on bilingualism has led to a high degree of specialisation and to the birth of different schools. The three main schools, i.e., generative, constructivist and functionalist, all collect empirical data, but often give diverging interpretations to one and the same database. The field is divided into three main domains according to type of participants studied: monolingual and bilingual (or multilingual) acquisition and impaired language (SLI, Down syndrome and others). Not surprisingly the first case studies were conducted on monolingual children speaking European languages such as English, French or German. More exotic languages are becoming increasingly popular subjects of study, however, and have a representation in this volume as well. It is in monolingual acquisition that the three theoretical approaches are most at odds. Whilst the interpretation of data in the constructivist approach relies very much on input, the generative approach bases its conclusions on theoretical constructs, as can be seen in two of the chapters in the present volume: Parisse and Müller & Pillunat, which focus on the topic of subject realization. Researchers became interested in the question as to whether monolingual and bilingual acquisition was qualitatively comparable in the nineteen-eighties. Much subsequent research that has compared bilingual and monolingual language acquisition has been able to show that both types of language acquisition are driven by the same principles. Only relatively recently have scholars working on theoretical approaches to language acquisition become interested in the field of language impairment. The general consensus amongst researchers is that children with SLI or Down syndrome go through the same stages as their monolingual or bilingual peers. If a difference is found, it is largely quantitative; the language of children with impairment is primarily delayed. This remains a debatable point, however, and some authors have claimed that the development of, for example, children with Down syndrome differs in some respects from that of typically developing children: particularly in the areas of morphology and syntax (Fowler 1990). This is the background against which the research described in this book is set. The book primarily seeks to address the generative approach to language acquisition in the three main domains: monolingual, bilingual acquisition and impaired language. As noted above, constructivists and generativists do not agree on the interpretation of data. This volume includes one chapter that is written in the
Introduction
constructivist tradition in order to suggest that a dialogue between these two approaches is worth pursuing. Such a dialogue will contribute to our better understanding of the fascinating world of language acquisition. The main purpose of this volume is to present original research from two different perspectives and to provide a basis for dialogue between researchers working in these diverging theoretical frameworks. The volume includes state-of-the-art research conducted mostly within the generative framework and reflects the diversity of methodological approaches and research interests that can be found in contemporary research work in the field. Many chapters in this volume address issues linked to the left periphery in various languages; some chapters are devoted to syntactic or morphological issues. We strongly believe that the result makes a contribution to our theoretical understanding of morpho-syntax, as well as contributing to the developing body of knowledge about language acquisition in different groups of learners.
1. Monolingual acquisition The four following chapters are devoted to monolingual first language acquisition in different languages (i.e., French, Norwegian, Spanish and Swiss German). The chapter by Parisse studies subject dislocations by young French-speaking children in the constructivist tradition, more specifically in the usage-based theory of language acquisition (UTB) by Tomasello (2000, 2003). In Parisse’s approach, French is a non pro-drop language where the use of a cliticized subject personal pronoun is mandatory if there is no lexical subject. In the presence of the subject in the form of a noun phrase or lexical subject, the cliticized personal pronoun is compulsory. This is what is called dislocation, which can be left or right dislocation. Hence, it is possible to move the dislocated subject anywhere, in front of or after the verb, and to insert elements between subject and verb. It is well known that the use of optional strong personal pronouns is often found in oral language and to a lesser extent in written texts. This is very common in children’s language and Parisse raises the following questions: (1) does children’s production merely reflect the distributional patterns of oral adult language?; (2) does children’s production reflect a specific developmental path in early French? In order to test hypothesis (1), a corpus of conversation transcripts between a child and an adult in a standardized situation was used by Parisse: the Bath Corpus. Parisse argues that children’s use of subject dislocations does not fully reflect their input. A second corpus, the House corpus, was used to test whether the use of subject dislocations follows a developmental pattern (hypothesis 2). The performance changes according to increasing age in that older children use more dislocated subjects. One possible explanation offered by Parisse is that this development pattern is the result of a growing systematic use of subject pronouns in front of a verb. In order to test this claim, the proportion of verbs preceded by a pronoun was computed for both corpora. This proportion tends to increase with age, and it correlates significantly with the use of constructions with
Introduction
pronouns in the children represented in both corpora. As opposed to this, there is no similar significant correlation in the adults’ production. Parisse argues that the production of optional pronouns is more likely to be a consequence of an automatic production of pronouns in front of a verb and that the optional character of the pronoun is learned at later age. He also claims that children build these constructions differently from adults. His interpretation of the data differs substantially from the approach used in the following chapters. The remaining chapters in this section are written in the generative tradition. In fact, the chapter by Gordishevski and Schaeffer reports the results of a longitudinal study with naturalistic settings. In particular, they investigated structural (NOM, ACC, GEN), inherent (DAT, INSTR), and lexical (ACC/DAT/GEN/INSTR/PREP assigned by prepositions) Case in the spontaneous speech of three Russian children raised monolingually. Inspired by Hoekstra & Hyams (1995), the researchers argue that their finding can be explained assuming the underspecification of the Number head. The results show that in the singular the children score very high, around 90% correct uses on NOM, on structural ACC, and on structural GEN as well as singular inherent DAT. Inherent INSTR was used only twice by two of the children. Similarly, the singular lexical Cases in total are produced correctly in 90% of the required contexts. In contrast, plural nouns show much lower success rates: 0% for structural ACC and for structural GEN but 14% for the lexical Cases. Plural inherent INSTR was used a few times by two children. Errors are found in around 40% of the cases and consist of substitution by NOM forms. The performance improves in all cases in the subsequent stage. Close examination of the children’s plural forms suggests that plural is not productive: (a) numbers of plural forms are extremely low; (b) most of the words used in plural represent pair-wise (‘hands’) or plural (‘fingers’) body parts, types of footwear and plural non-count nouns; (c) almost none of the plurals occur in their singular forms in the same transcripts. On the basis of these observations, Gordishevski and Schaeffer propose that the nominal Num head is initially underspecified, namely, it represents [-plural] only (cf. Müller 1994). They assume that in order to check Case features on N and D a “Case-chain” needs to be established. Following Hoekstra and Hyams (1995), Gordishevski and Schaeffer propose that in case of a plural N, an underspecified Num ([-plural]) breaks the Case-chain, because the N’s plural feature cannot be checked. This blocks Case feature checking, yielding default NOM forms in early Russian. When the noun is singular, underspecified Num does not break the Case-chain as it is [-plural], thus yielding forms correctly marked for Case. Chapter three by Westergaard argues for an analysis of verb-second (V2) word order which involves a split CP as suggested by (Rizzi 1997) on acquisition data from a dialect of Norwegian (Tromsø). Investigating three children, it is shown that, although verb movement (i.e., verb-second (V2) word order in non-subject-initial) is generally acquired early in declaratives, the children make certain errors in topicalisations and questions with respect to the position of pronominal subjects in negated utterances. More specifically, in the early files the children predominantly produce the Neg-S word
Introduction
order illustrated in (1), which is unusual in the target language unless the pronominal subject is stressed. There is a clear development to the more target-like word order in (2) in the files of all three children investigated. (1) (2)
nå skal ikke dem sove. now shall not they sleep ‘Now they shouldn’t sleep.’ og no kan æ ikke drikke det. and now can I not drink it ‘And now I can’t drink it.’
(Ole 2;3.15)
(Ole 2;10.0 )
The findings are accounted for within a split-CP model which departs from the existence of two subject positions in the IP domain: a high one for informationally given subjects and a lower one for subjects conveying new information. In addition a pragmatic account of the child data is also considered, but rejected in favor of an economy-based analysis which includes factors such as input frequency and complexity. Chapter four (Schönenberger) investigates word order acquisition as well. This chapter addresses three puzzles relating to the acquisition of verb-placement in Swiss German. Swiss German is a V2 language, which displays the verb-final pattern in embedded clauses. The children of Schönenberger’s study misplace the finite verb in embedded clauses: they generally move the finite verb to the second position in any type of embedded clause before they switch to the verb-final pattern around age 5;0. Since this finding is surprising, the author examined the input one of the children was exposed to. The examination of the input is based on two samples of speech in which primarily the child’s mother interacts with the child once at the beginning of the study. However, no significant differences in the input were found. Although Swiss German is verb-final, this pattern can sometimes be disguised, since Swiss German allows extra-position, which gives rise to word-orders in which the finite verb does not show up clause-finally (3). Such examples might be misanalysed by the child as involving verb-movement to the left rather than extra-position to the right.
(3) complementizer Subject . . . ti . . . Vfin XPi
Swiss German allows embedded V2–besides the verb-final pattern–in clauses introduced by wil “because”, which induces a difference in interpretation. The three puzzles are, therefore, as follows: (i) In the adult samples there is a large proportion of clear-cut verb-final sentences–over 65% that are unambiguously verb-final, but the child generalizes verb movement to all embedded clauses. (ii) In clauses introduced by wil V2 is allowed in the target grammar, but only in 15% of all wil-clauses do the adults use this option. (iii) Around age 6;0 the child mainly produces the verb-final pattern. There are a few V2 examples with wil, which in contrast to early wil-clauses with V2, occur with the appropriate pragmatic interpretation. However, even after age 6;0 the child still misplaces finite auxiliaries which select a participle. The dialect the child is exposed to does not allow the sequence Aux Participle but only Participle Aux in
Introduction
embedded clauses, which is attested in over 190 examples produced by the adults. In trying to solve these puzzles Schönenberger draws a comparison between the child data and adult matrix clauses, and also points out similarities between the child’s embedded clauses and matrix clauses in Old and Middle English. The next two chapters are devoted to children’s understanding of preverbal universal quantifiers in Spanish and logical words in Hebrew. Chapter five by Escobar and Torrens seeks to investigate clausal structures containing fronted phrases doubled by a clitic. The chapter highlightes a potentially difficult asymmetry: quantifiers in preverbal or postverbal position in contexts of Clitic Left Dislocation. Following previous work on the topic (e.g., Uribe-Etxebarria (1995) and Zubizarreta (1998)), the authors hypothesised that if Quantifier Raising is part of the early grammar of Spanish, the choice for a postverbal or preverbal quantifier subject would then be free. More crucially, the authors hypothesised that learners at all ages are expected to obtain the wide scope interpretation of postverbal quantifier subjects. The experimental design by Chien & Wexler (1991) also followed by Philip & Coopmans (1995) was chosen in order to collect data from children. According to this study’s findings, children do not show a special preference for a preverbal or postverbal quantifier subject in instances of Clitic Left Dislocation. However, Torrens and Escobar offer different accounts in order to explain these findings. What is clear then is that Spanish five-year-olds easily obtain the wide scope interpretation of postverbal quantifier subjects, unless they are presented with biased matching pictures (with an impaired object or an impaired agent). But adults also have problems with the mismatched pictures. In turn, this study supports the continuity hypothesis that children and adults share a common core of linguistic knowledge. Chapter six, on the other hand, by Armon-Lotem investigates the relative influence of syntax, semantics and pragmatics on children’s understanding of sentences containing logical words, such as, some and or, showing a subject-object asymmetry. Specifically, Armon-Loten’s chapter reports the results of a truth value judgment task administered to Hebrew speaking children, using Eyzeshehu yeled for some boy and o for or. Three hypotheses were tested: (1) Children derive the exclusive reading in the descriptive mode but accept an inclusive reading in the prediction mode. (2) Children judge the inclusive reading of or in the prediction mode as true more readily than the weaker reading of some in the prediction mode. (3) The specificity in subject position blocks the influence of non-specificity in the prediction mode. The group of children was split into two in order to be presented with sentences in the descriptive and sentences in the prediction mode respectively, where the implicature can be suspended in order to test hypothesis 1. For hypothesis 2, each child was presented with 5 sentences containing or, and 5 containing some. Hypothesis 3 was tested by a number of items in the description and in the prediction mode. The results of the Hebrew replication were similar for or, with children accepting the inclusive reading in the predicative mode but much less in the descriptive mode. The results for some, though similar too, showed a subject-object asymmetry, suggesting a possible syntactic interference.
Introduction
In the predictive mode, when the logical word was used with a direct object, the implicature was always erased, but when it was used with a subject the implicature was erased in less than third of the times. That is, while some NP was always accepted for every NP in the direct object position (4a), it was accepted for every NP only in third of times in subject position (4b): (4) a. b.
ani I ani I
menaxesh she guess that menaxesh she guess that
ha-yeled yoxal eyzeshehu tapuax the-boy will-eat some apple eyzeshehu yeled yoxal tapuax some boy will-eat apple
This indicates that, unlike adults, Hebrew speaking children are less likely to erase the scalar implicature in subject position than in object position. This subject-object asymmetry suggests that other linguistic factors, e.g., syntactic position, also contribute to children’s comprehension of sentences containing logical words.
2. Bilingual acquisition The next three chapters deal with the acquisition of syntax by bilingual children and the issue of vulnerability. Bonnesens’ chapter considers the question of the vulnerability of the CP or Left Periphery in Bilingual First Language Acquisition. Former studies on bilingual language acquisition (e.g., Meisel 1994; Meisel & Müller 1992) have revealed that children acquiring two languages from birth never go through a stage where the IP of one language is transferred into the other language. Numerous other studies have shown that the IP is not a vulnerable domain for cross-linguistic influence in bilingual first language acquisition. Recently Platzack (2001) has proposed that, in contrast to the IP, the CP is vulnerable even in monolingual acquisition, an approach which has been discussed by several authors in Müller (2003) from the perspective of bilingual language acquisition. In this chapter Bonnesen analyzes the left periphery of two French/German bilingual children, Annika and Pierre of the DuFDE-corpus (see Köppe 1994). Since the two languages differ clearly as far as the left periphery is concerned, the data are an excellent basis to verify the “vulnerability of the CP”- hypothesis. German is assumed to be a verb-second (V2) language (den Besten 1977; Platzack 1983). The verb-second structure is explained by the movement of the finite verb to C°. The initial phrase of a main clause occupies the Spec CP-position. In contrast to German, a preposed element in French (5a–b) results in a V3-struture. (5) a. Nun gehtV Peter ins Kino b. Maintenant Jean vaV au cinéma. Now John goes to the cinema.
The author suggests that the V2-structure of German can be transferred into French and/or that ungrammatical V3-structures are used in German, if the left periphery is
Introduction
indeed a vulnerable domain. In analyzing the French data, the status of the clitic subject pronouns is considered as well. According to Kaiser & Meisel (1991) and Kaiser (1992), the subject clitics are verbal affixes. Since Bonnesen adopts the prefix analysis of the subject clitics, a V3-structure on the surface may actually be a V2-structure if the subject pronoun is a clitic. In an utterance like “maintenant il part” (now he leaves) the verb and the subject clitic occupy the same position, I° (or T°/Agr°). If a child uses the German V2-structure in French, the Adverb “maintenant” would occupy SpecCP and the verb and the clitic would move to C°, though on the surface, the word order is the same as in a French structure. Hence utterances with only clitic subjects have been excluded as evidence for or against transfer of the V2-rule from German into French. If the subject is a DP, the movement of the verb to C° becomes visible resulting in “maintenant part Jean” (now leaves John) which is a German word order, but incorrect in French. The evaluation of the data presented by Bonnesen does not support the “vulnerability-hypothesis” of the CP. In German there is no evidence at all that the French grammar influences the German system. In Pierre’s French data there is no evidence of transfer either. Only in Annika’s French data, there are several utterances which can be explained by the German V2-structure. Nevertheless, correct V3-structures always occur. Hence, it cannot be argued that Annika transfers the general I-toC-movement-rule of German into French. Rather, her V2-structures during a certain period are probably only a problem of her performance. After the age of 3;08 V2structures show up (until the end of the recordings) exclusively with the adverbs “là” and “ici”. So the author argues that Annika has merely a lexical-driven V2-structure with these adverbs, which is much too weak to support the “vulnerability of the CP”hypothesis. Furthermore, he shows that in Annika’s data the position of the object is affected as well by the influence of German on the performance level, which indicates that her “vulnerability” is not at all linked only to the CP-level. In the same vein, Larrañaga’s chapter seeks to investigate the acquisition of word order within the generative perspective. This chapter examines the position of the subjects of unaccusative verbs and their pragmatic entailments in the speech of Basque/Spanish children raised bilingually. Basque and Spanish differ in the position of the head. Basque is head-final and Spanish head initial. Larrañaga adheres to the so-called Burzio’s (1984) generalization; subjects of unnacusative verbs are generated in the complement position. Hence, VS in Spanish and SV in Basque have the neutral reading. However, since both languages are free word order languages, both SV and VS are possible in both languages, but have different pragmatic readings. According to Larrañaga this is an ideal domain to test recent hypotheses. This chapter is based on Rizzi’s (1997) work on left periphery and on Müller & Hulk’s (2001) approach to cross-linguistic influence in bilingual first language acquisition. The data analysed with respect to the position of the subject and the information status show that there are two clear patterns of acquisition according to the language involved. The children studied by Larrañaga use SV in Basque from the outset of language acquisition and throughout
Introduction
the whole study where S has diverging target-like pragmatic readings. VS is almost non-existent in Basque. As opposed to Basque, both children show a strong preference for VS in Spanish until the age of 3;00. After this age, both SV and VS are used in Spanish. Hence, both children use the target-like word orders in both Basque and Spanish throughout the whole study. Larrañaga shows that there are no signs of cross-linguistic influence in the domain studied in Basque/Spanish bilinguals, although topic and contrastive subjects are possible in both languages and are represented in the C-domain if in preverbal position. This is the more surprising since topic subjects occur at a later stage in both languages. Larrañaga also shows that the preverbal subjects with a topic reading are used at a later stage because the C-layer is missing at early stages. The final chapter about bilingual first language acquisition is the one by Kupisch. In this chapter Kupisch seeks to investigate the relation between language dominance, mixed language utterances and cross-linguistic influence based on data from two German/French bilingual children who have been studied longitudinally. The children differ in terms of the extent to which one language can be considered dominant, but both children seem to produce fewer mixed utterances in their stronger language. The issue of cross-linguistic influence is then examined with respect to determiner acquisition. The languages under scrutiny – German and French – possess pronominal definite and indefinite determiners, but at the same time there are clear and evident inter-language differences regarding the frequency of occurrence, morphological load, and the prosodic structure of determiners. Monolingual German and French children acquire determiners at different rates, the process being faster in French. The bilinguals use more determiners in German than monolinguals of comparable ages, which is interpreted in favour of positive influence from French. According to Kupisch’s findings, language dominance does not cause any delay in the path of acquisition. On the contrary, children can differentiate between the two languages at an early stage. More importantly, even though the amount of mixing seems to be related to language dominance as one of the reasons, mixing does not seem to be unidirectional and does not consequently imply that the language development is slower. Hence, cross-linguistic influence does not seem to go together with the mixing rate and language dominance. In sum, the results from this chapter are inconsistent with mixing and dominance patterns, and the author argues that influence and mixing may represent different types of contact phenomena. Above all, the author commits to the proposal that cross-linguistic influence and mixing constitute in fact two different types of contact phenomena with the bilingual acquisition.
3. Impaired language The chapter by Perovic focuses on the acquisition of the binding principles in Down syndrome in L1 English and Serbo-Croatian. Down syndrome (DS) is a genetic disorder typically accompanied by moderate to severe learning disabilities. Language in
Introduction
this population is traditionally described as delayed, mirroring the pattern in typical language development, with no obvious signs of deficiency. In this study the author reports findings from an experimental investigation of the knowledge of binding, the module of grammar constraining the distribution of reflexives and pronouns, in two groups of young adults with DS. Individuals with DS, both English and Serbo-Croatian speaking, were found to have difficulties comprehending reflexives, but not pronouns. Since for the interpretation of reflexives the syntactic relation between the reflexive element and its antecedent is crucial, in contrast to pronouns that are interpreted by invoking extra-syntactic mechanisms, the results suggests a deficit in the DS population which is syntactic in nature. Such a pattern is exactly the opposite to that found in typically developing English children who obey the syntactic constraints on the distribution of reflexives early on, but have trouble applying co-reference rule, a constraint beyond syntax that rules the interpretation of pronouns. This chapter provides crucial evidence that language in DS speakers is not merely delayed but also deficient in important respects, with the deficit pertaining to an inability to establish the syntactic relation between the anaphor and its antecedent. Secondly, the findings further our knowledge of binding in non-impaired grammar, favouring one of the competing theories of binding phenomena: one that invokes both syntactic and extra-syntactic factors in the process of determining reference of a pronominal element (Reinhart & Reuland, 1993; Grodzinsky & Reinhart, 1993) as opposed to a theory that relies solely on syntactic constraints in the same process (standard Binding Theory of Chomsky, 1981, 1986). Chapter eleven by Müller & Pillunat reports on the language development by French/German bilingual children with more or less balanced bilingualism, amongst whom there is one child who has not been diagnosed with SLI but whose language very much matches with all the criteria proposed for children with SLI. The topic of the study is a rather robust grammatical phenomenon, namely pronoun use in the development of German and French. Research on monolingual children with SLI has revealed that some aspects of language are acquired in the same way as in typically developing monolingual children. The difference between children with SLI and typically developing children has been claimed to be of a quantitative nature (delay). Müller & Pillunat have studied a child with no diagnosed SLI who is raised with German and French from birth. The language development is measured, as in typically developing bilingual children, on the basis of MLU, absolute number of utterances per recording session, development of the nominal and the verbal lexicon, also in order to determine whether the two languages develop at equal pace. The acquisition process is studied from 2;8 – 3;8 and compared with that of typically developing German/French bilinguals (a) with a strong and a weak language or (b) with a balanced language development. It is shown that the pronominal system develops in this bilingual child as in typical monolingual and bilingual (balanced and unbalanced) children, i.e., the authors find a delay of object clitics in French and no delay of object pronouns in German, thus confirming results by researchers who have denied the existence of a difference
Introduction
between monolingual and bilingual children with SLI. A careful analysis of the contexts for the different kinds of pronouns reveals however that the bilingual child with SLI shows an overuse of strong pronouns. In this respect she differs from her typically developing monolingual and bilingual children. The authors argue that children with SLI face a problem with the interface between syntax and pragmatics. In sum, the present book includes a range of different languages and language combinations, and the effects of Specific Language Impairment and Down syndrome on acquisition, meaning that this collection gives a wide range of different perspectives on a number of topics of current importance in the theoretical study of morphology and syntax. The varied contributions illustrate the exciting range of empirical work that is currently under way in the field of language acquisition, and should stimulate further investigation and increased dialogue between researchers using different theoretical approaches.
Acknowledgments We would like to express our gratitude to the general editors of this series, Harald Clahsen and Lydia White: without their support and critical comments this project would not have got off the ground. We would also like to thank Kees Vaes and his editorial team for their help and support during the editorial process. We are also especially grateful to a large number of anonymous reviewers who read and commented on the chapters. This project started germinating during the Child Language Seminar which took place in Bristol in 2004. Financial support for the conference was provided by a British Academy Conference grant (BCG- 38722). Most of all, however we are very much indebted to all the authors of the chapters and Neil Smith for all the work they have put into this publication.
References den Besten, H. 1977. On the interaction of root transformations and lexical deletive rules. GAGL: Groninger Arbeiten zur Germanistischen Linguistik 20: 1–78. Burzio, L. 1984. Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel. Chien, Y.C. & Wexler, K. 1991. Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition 1: 225–295. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986. Knowledge of Language: Its Nature Origin and Use. New York NY: Praeger. Fowler, A.E 1990. Language abilities in children with Down syndrome: Evidence for a specific syntactic delay. In Children with Down Syndrome: A Developmental Perspective, D. Cicchetti & M. Beeghly (eds). Cambridge: Cambridge University Press.
Introduction Grodzinsky, Y. & Reinhart, T. 1993. The Innateness of binding and coreference. Linguistic Inquiry 24: 69–101. Hoekstra, T. & Hyams, N. 1995. The syntax and interpretation of dropped categories in child language: A unified account. In Proceedings of WCCFL XIV, Stanford University. Kaiser, G. 1992. Die klitischen Personalpronomina im Französischen und Portugiesischen. Eine synchronische und diachronische Analyse. Frankfurt: Vervuert. Kaiser, G. & Meisel, J.M. 1991. Subjekte und Null-Subjekte im Französischen. (Subjects and null subjects in French). In ‘Det, Comp und Infl’: Zur Syntax funktionaler Kategorien und grammatischer Funktionen [Linguistische Arbeiten 263], G. Fanselow & S. Olsen (eds), 110–136. Tübingen: Niemeyer. Köppe, R. 1994. The DuFDE Project. In Bilingual First Language Acquisition. French and German Development, J.M. Meisel (ed.), 15–27. Amsterdam: John Benjamins. Meisel, J.M. 1994. Getting FAT: Finiteness, agreement and tense in early grammars. In Bilingual First Language Acquisition: French and German Grammatical Development, J.M. Meisel (ed.), 89–129. Amsterdam: John Benjamins. Meisel, J.M. & Müller, N. 1992. Finiteness and verb placement in early child grammars: Evidence from simultaneous acquisition of French and German in bilinguals. In The Acquisition of Verb Movement, J.M. Meisel (ed.), 109–139. Dordrecht: Kluwer. Müller, N. 1994. Komplexe Sätze. Der Erwerb von COMP und von Wortstellungsmustern bei bilingualen Kindern (Französisch/Deutsch). Tübingen: Narr. Müller, N. 2003 (ed.). (In)Vulnerable Domains in Multilingualism [Hamburg Studies on Multilingualism 1]. Amsterdam: John Benjamins. Müller, N. & Hulk, A. 2001. Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition 1(4): 1–21. Philip, W. & Coopmans, P. 1995. Symmetrical interpretation and scope ambiguity in the acquisition of universal quantification in Dutch and English. Ms, Utrecht University. Platzack, C. 1983. Germanic word order and the COMP/INFL parameter. Working Papers in Scandinavian Syntax 2. Platzack, C. 2001. The vulnerable C-domain. Brain and Language 77: 364–377. Reinhart, T. & Reuland, E. 1993. Reflexivity. Linguistic Inquiry 24: 657–720. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer. Ronjat, J. 1916. Emprunts et faits de fonétique sintactique dans les parler de Labouheyre (Landes). Revue des Langues Romanes 59 : 38–43 Tomasello, M. 2000. Do young children have adult syntactic competence? Cognition 74: 209–253. Tomasello, M., 2003. Constructing a Language: A Usage-based Theory of Language Acquisition. Cambridge MA: Harvard University Press. Uribe-Etxebarria, M. 1995. On the Nature of SpecIP and its relevance for scope asymmetries in Spanish and English. In Contemporary research in romance linguistics [Current Issues in Linguistic Theory 123], J. Amastae, G. Goodall, M. Montalbetti & M. Phinney (eds). Amsterdam: John Benjamins. Zubizarreta, M.L. 1998. Prosody, Focus and Word Order. Cambridge MA: The MIT Press.
Left-dislocated subjects A construction typical of young French-speaking children? Christophe Parisse Few constructions in children’s production can be considered originals and not mere copies (partial or complete) of adult input. These rare constructions are precious to child language studies because they may reveal a developmental process at work. One of them is children’s left-dislocated subjects. In French, when there is no lexical subject, the use of a subject pronoun is mandatory. When there is a lexical subject, two constructions are possible: one with the use of a personal pronoun and one without. Children tend to produce a lot of pronouns in the context of lexical subjects, which makes it a good candidate for being a specific feature of child language. This study used two corpora of spontaneous language production, one with children aged three to four and one with children aged two to four. The data were analyzed using Tomasello’s framework (usage-based theory of language acquisition). In this theory, children’s language competence differs from the adults’ and develops with age. The first issue was to demonstrate the existence of this specific feature. Indeed, it appeared that children produced more leftdislocated subjects than adults and did not simply reproduce a feature of adults’ oral language. The presence of a developmental effect in the children’s production was also strongly suggested by the results. The second issue was to try to better understand what could be the reason for the difference in production of left-dislocated subjects between children and adults. Several explanations of the children’s results were put forward, in keeping with the predictions of the usage-based theory. Lexical and usage-based explanations of children’s behaviour were shown to be unlikely. More plausible explanations were performance limitation or a consequence of the development of the use of personal pronouns, even when they are not obligatory. It is proposed that children’s left-dislocated subjects should not be considered as a copy of adults’, but as a specific construction pattern that appears during the course of language development. This pattern could provide a path towards the development of more general linguistic abstractions, as proposed in the usage-based theory.
Christophe Parisse
1. Introduction The theoretical approach to language acquisition presented by Tomasello (2000, 2003) argues that syntactic competence is built in a piecemeal fashion, starting with itemdependent knowledge and gradually generalizing towards the mastery of abstract syntactic structures. This approach considers that the grammars of the child and the adult are different and that the differences get smaller as the child develops her language competence. One of the consequences of considering language acquisition as a developmental process is that it is important to pay special attention to structures produced by a child that cannot be found in adult language. The difference between the child’s and the adult’s productions attests that these structures are constructed by the child and not merely reproductions of adult input. In other words, they represent the child’s grammar and not the adult’s grammar. Pinpointing such constructions opens a window into the underlying processes of language development and allows us to go further than simple linguistic descriptions. This technique is widely used to study the development of the lexicon and of morphosyntax, where it is easy to apply, as one needs only to find words that are not adult words to know that they correspond to original morphological or morphosyntactic constructions (setting aside the mispronunciation of existing words). It is unfortunately not so easy when one is interested in the study of syntactic constructions composed of more than one word (from two words up to a whole utterance). In this case, with the exception of obligatory markers and arguments that are never or seldom omitted by adults, it is not always easy to know if a construction is an original or a copy of an adult construction. Indeed, language flexibility and performance limitations are such that very few children’s first syntactic combinations cannot be found in adult language (see Parisse & Le Normand 2000). This is especially true when young children’s utterances are very short. The scarcity of constructions totally specific to children’s language can be compensated for by looking for constructions unusual in adult’s language instead of truly non-adult constructions. This can be done by comparing children’s and adults’ frequency of production. If there is a significant difference between children and adult productions for a given construction, then this construction becomes a good candidate for specific child construction and grammatical behaviour. This technique is applied in the study below which deals with the development of left-dislocations of the lexical subject in French children. The choice of this syntactic structure comes from previous work (Parisse & Le Normand 2001) in which it appeared to be relatively frequent in children, although it is not considered to be the most natural construction for adults. Thus, a difference in frequency between children’s and adults’ use could be expected and left-dislocated subjects could represent an example of grammatical behaviour specific to children.
Left-dislocated subjects
2. Usage-based theory of language acquisition Presentations of Tomasello’s usage-based theory (UBT) of language acquisition can be found in Tomasello (2000, 2003). UBT is rooted in a construction grammar framework (see Croft 2001; Goldberg 1995). Tomasello (2002) gives a short summary of the main steps in the development of productive language: 1. Children hear and attempt to learn utterances, i.e., phonological forms for expressing communicative intentions; in doing so they assign a communicative function both to the utterance as a whole and to certain of its constituents (sub-functions of the utterance’s overall function). 2. Children store comprehended and produced utterances and constituents, along with their functional analyses; token and type frequencies of stored pieces of language lead to the construction of (a) utterance schemas and (b) constituent schemas, both of which may contain relatively abstract ‘slots’ (and it is an empirical fact that children construct these abstractions only gradually and in piecemeal, item-based fashion). 3. Children produce utterances by ‘cutting and pasting’ from their inventory of stored utterance schemas and constituent schemas. (Tomasello 2002, 309, 310). A more detailed account of the theory from a syntactic point of view is found in Tomasello (2006). The issues of grammar learnability and the nature of the adult’s grammatical knowledge are discussed in Parisse (2005).
3. Left-dislocated subjects The analysis of left-dislocated subjects using Tomasello’s framework is rather infrequent, especially for French. So it seemed that the presentation of two recent works on the subject, Blasco-Dulbecco (1999) and De Cat (2002), although unrelated from a theoretical perspective, would provide more insightful analyses of the language data and would help the reader understand the language structures analysed in this chapter. French is a non pro-drop language, which means that subject pronouns are mandatory when there is no lexical subject. For example, in ‘il rit (he is laughing)’, the use of ‘il (he)’ is mandatory. When there is a lexical subject, it takes the place of the personal pronoun. However, it is possible to insert a personal pronoun between the lexical subject and the verb: the construction ‘le bébé rit (the baby is laughing)’, and the construction ‘le bébé il rit (the baby he is laughing)’ are both correct. Constructions of the second type are called left-dislocated subjects. They belong to the larger category of dislocations, a wide class of syntactic constructions that has already been studied, especially in French, where this grammatical construction is frequently used. Left-dislocated subjects represent the most frequent type of dislocation in French (Blasco-Dulbecco 1999).
Christophe Parisse
These constructions are often called double-markings of subject (i.e., the first marking is the lexical noun, the second marking is the subject pronoun). They are attested in children’s production as well as in adults’ production and can be used in affirmative utterances, for example, ‘papa il aime pas trop (dad he does not like [it] too much)’, as well as in interrogative utterances, for example, ‘et ton papa il est pas là? (and your dad he is not here?)’. There is some controversy as to whether the subject pronoun is really optional. For this to be true, the two constructions, ‘NP pro:subj V’ and ‘NP V’ should have the same meaning and function. This possibility is clearly presented in Grevisse & Goose (1993) : ‘the pronoun is redundant with the subject that is in its usual position – in oral language, the pronoun is considered as necessarily belonging to the verb . . . no pause is made between nominal subject and pronoun’. However, dislocations are useful as a way of introducing emphasis (to put extra-force on a element of the utterance in order to make it more important). In this case, there is usually a pause between the nominal subject and the pronoun, or at least a specific intonation contour (De Cat 2002). The use of dislocations also allows the NP to be moved around and material to be inserted between the NP and the verb. This also contributes to emphasis and focalisation effects (Wilmet 2003). Nonetheless, it remains true that many occurrences of dislocations are not emphatic. For Ashby (1977), less than half of all dislocations are clearly emphatic. Even more important is the fact that emphasis can be produced without the presence of a dislocation, so the existence of a personal pronoun between a lexical subject and a verb cannot be explained by emphasis only. Blasco-Dulbecco (1999) contains a historical presentation of dislocations. It appears that the phenomenon already existed in Latin and in Old French. Dislocations were limited to subjects and adjuncts without prepositions in that period. Prescription against its use is attested in Chifflet (1659), which shows that it was quite common at that time, and already considered as a marker of orality as it is today. Thus, the use of dislocations is not an evolutionary feature of French. Neither is it a phenomenon linked to a grammaticalisation process, contrary to some authors’ opinions, for example Ashby (1977), who suggested that the pronoun in front of a verb was becoming an obligatory construction in French. It remains true, however, that ancient and modern dislocations have different formats, and that double-marking did not then exist, so that a diachronic trend is not outright impossible. There is also no historical material available concerning oral language, so it is possible that constructions used in daily life were slightly different. What is known for certain is that it was often used in the 17th and 18th centuries with long subject noun phrases (Grevisse & Goose 1993). The explanations of the phenomenon given then were also quite similar to those given today, especially that of dislocations being a means of introducing or emphasising language material. Blasco-Dulbecco’s (1999) original work is based on a large set of attested examples (1500 oral language examples and 500 written language examples). Her first result is that dislocations account for 10.19% of all subjects oral French, and 2.83% of dislocated subjects in written French. Other results are that 69.25% of dislocations in oral language are left-dislocations, and that a majority of these in oral French (76.56%)
Left-dislocated subjects
concern subjects. The large difference between the figures for oral and written language shows the importance of using oral language data to assess the importance of a linguistic construction. Blasco-Dulbecco makes a thorough linguistic analysis of all dislocation types, using the ‘pronominal approach’ of Blanche-Benveniste et al. (1984). Dislocated elements may correspond to various syntactic components: subject, object, oblique. Blasco-Dulbecco starts her analysis with the oblique components. It appears that they are typical illustrations of dislocation usage, as they allow a lot of variation, some being used with either subject or object, some before the verb and some after the verb. When there is no preposition in the dislocated element, it is difficult to know exactly what is the relationship between the dislocated element and the verb, and the genuine cases of double-marking may be less frequent than they appear at first. This explains why the agreement in Gender or Number between the dislocated element and the verb (or the pronoun) is often missing. Her linguistic analyses led her to conclude that dislocations may cover a wide variety of situations. They may introduce new language material, but this is not always the case, especially when there is no exact reiteration of the dislocated element in the sentence. They are also a means of organising repetitions and creating enumeration effects. Finally, some elements are systematically used with a dislocated construction. There is a large difference in the nature of the function of dislocations before and after the verb. Before the verb, they have no syntactic function, but create a frame that helps organise the utterance that follows. After the verb, they represent a means of extending the verbal group, but have no pragmatic function. De Cat (2002) targets French child language as well as French adult language. Her work is based on the generativist framework, which makes it difficult to compare with Blaso-Dulbecco’s work and with the present work. However her description of linguistic data is very informative and some of her observations are certainly very useful for the analysis of the present work. De Cat’s work is based on a corpus of child-adult interactions, on a corpus of elicited judgements, and on data elicited for the study of intonation. The main argument of the author is that constructions such as ‘NP pro:subj V’ are different from constructions such as ‘NP V’ because the pronoun in ‘NP pro:subj V’ construction is not a simple inflectional morpheme but a syntactic argument that bears a theta-role (see also De Cat 2005). This means that the first construction is not a case of double-marking. The author demonstrates that it is possible to differentiate the prosody of a construction with dislocation and that of a construction without dislocation, and that ‘NP pro:subj V’ children’s constructions have a true dislocation prosody more frequently than a non-dislocation prosody. She finds that in declarative sentences, 31% of root clauses contain a dislocated element. In wh-questions, this figure rises to 46%, and falls to 18% in yes/no-questions. Although it is difficult to compare, the percentage of dislocations is certainly higher than in Blasco-Dulbecco’s study. Another interesting result is that when there is a nominal expressing the subject, 46% of subjects appear after the verb with subject clitic on the verb, 39% appear before the verb with subject clitic on the verb, and 16% appear before the verb without subject clitic on the verb. De Cat gives these percentages for three dialects of French. There are some
Christophe Parisse
differences between the dialects, the main one being that in French French, there are more dislocated subjects after the verb and less true lexical subject (heavy subjects) than in Belgian or Canadian French. Children use dislocations at a very early stage of language development, at the beginning of multi-word constructions, and demonstrate an early mastery of the pragmatic use of dislocations. Her main conclusion is that dislocations are discourse-motivated and constrained by the information structure; they are generated by adjunction, using minimal syntax, and they have no argumental status.
4. Goal of the study The goal of the present chapter is to find out whether the production of left-dislocated subjects is compatible with UBT. This implies obtaining some specific results. First and most important, the developmental hypothesis has to be confirmed. There should be differences between adults’ and children’s grammars and a developmental process in children’s grammar. These differences in grammars should manifest themselves in children’s and adults’ productions. Two consequent characteristics of children’s and adults’ productions will be checked below: 1. Whether a difference between adult and child production of left-dislocated subjects is confirmed or not; 2. Whether the children’s use of them has a developmental trend or not. Second, if the first test indicates that the existence of a usage-based account is possible, various hypotheses of UBT or related to UBT about the causes of a developmental trend can be tested. These hypotheses are: 1. Is there a lexical preference? Are there some specific verbs where dislocations are used by all children, whereas they are not for other verbs? 2. Could it be the result of item-based constructions? Are there some specific verbs where, on a child by child basis, dislocations are used, whereas they are not for other verbs? In this case, the lexical preference would change from one child to another. 3. Is there a relationship between the use of LS v. DS constructions and performance limitations? This could be checked by looking at the length of children’s utterances. If one or the other type of construction is used more characteristics of both types of constructions. 4. Is dislocation a consequence of the development in the use of personal pronouns? As language develops, production of obligatory subject pronoun becomes more and more systematic and easy for the child. If the development of dislocations is correlated with the development of subject pronouns, the production of dislocations could be a side effect of the development of pronouns before children learn not to use them when unnecessary.
Left-dislocated subjects
As numerous corpus analyses are presented in this chapter, it would be awkward to start by presenting all seven groups of results first and then a lengthy discussion about each group. So instead a short discussion will be presented with each series of results, which will also explain why the analyses were performed and organised as they were. After this, the totality of the results will be summarized, followed by a general discussion about what underlies the production of subject dislocation by children.
5. Methods The following work is based on analyses performed on two corpora of spontaneous language production. These corpora are described below and will be referred to as the ‘Bath corpus’ and the ‘House corpus’. The two corpora correspond to different children and were collected in different settings. They were not collected for the purpose of this study, but to serve as reference for spontaneous language production (Chevrie-Muller et al. 1997; Le Normand 1986).
5.1 Bath corpus The procedure used was the ‘Bain des poupées’ from a normalized language-testing tool (Chevrie-Muller et al. 1997). Children were recruited in kindergarten and recordings were made on site. All children were seen separately, in a quiet room within the school. During this procedure, a professional speech therapist plays with the child using a standardized set of toys in a standardized situation: giving a bath to two dolls. The speech therapist asks the children questions about the process involved (taking a bath) and about what the child does at home in a similar situation. The therapist interacts with the child until the question is answered and sufficient language material is produced. The interview goes on until the whole set of questions is answered. The corpus contains 108 recordings of children of three different age groups: 3;0, 3;6 and 4;0 for a total of 21,667 utterances and 121,042 words. The recording characteristics for each age group are summed up in Table 1.
Table 1. Characteristics of the ‘Bath corpus’ Age 3;0 3;6 4;0
Children N recordings N utterances 37 29 42
169 127 131
Adults N words MLU 502 439 477
2.96 3.38 3.63
N utterances N words MLU 229 166 177
1345 905 950
5.87 5.51 5.37
Note: All figures, except the number of recordings itself, correspond to average figures per recording.
Christophe Parisse
5.2 House corpus Children from this corpus were recruited in nurseries and kindergartens. Some recordings were made in a special room in the laboratory but most were made in a separate room in the child’s institution. All children were seen separately with a caretaker or a person familiar with the child. In this procedure (Le Normand 1986), children are in a play situation. They are not interrupted when they start speaking and may go on for as long as they wish. The familiar adult observer (usually one of the parents) plays with the child. The role of the adult observer is to ask questions if the child does not produce much language on her own. The length of the recording is standardized to 20 minutes. This procedure does not elicit a lot of language material from the observer and is not used to assess adults’ language. The corpus contains 316 recordings going from age 2;0 to age 4;0 by increments of 3 months. The average number of recordings per age is 35, the total number of utterances is 32,643 and the total number of words is 108,675. The recording characteristics for each age group are summed up in Table 2. Table 2. Characteristics of the ‘House corpus’ Age
N recordings
N utterances
2;0 2;3 2;6 2;9 3;0 3;3 3;6 3;9 4;0
40 31 36 36 40 34 34 33 32
66 83 89 110 113 130 118 117 109
N words 108 185 225 364 397 457 464 475 454
MLU 1.51 2.13 2.45 3.22 3.41 3.51 3.70 3.92 4.02
Note: Nb. utterances and nb. words correspond to average number of utterances and words per recording
6. Corpora analyses 6.1 Are children’s dislocations mere reproduction of adult language? Is it possible that the use of dislocations by children simply reproduces adults’ use? This would mean that dislocations are not a very interesting feature of children’s language to study, as they would then provide no particular insight into the development of children’s grammars. The Bath corpus can be used to test this hypothesis, as it contains both children and adults spontaneous productions. Left-dislocated subjects (DS) have the following structure: lexical noun group (noun group where the head is a lexical noun) followed by subject personal pronoun followed by verb (any tense or type). In all these constructions, the subject personal pro-
Left-dislocated subjects
noun is in the third person. Variations concerned Gender (feminine or masculine) and Number (singular or plural). Standard lexical subjects (LS) are constructions which are not subject dislocations, and consist of lexical noun group followed by verb. LS constructions are similar to DS constructions minus the personal subject pronoun. Table 3 gives the average number of such constructions found for the children and the adults in each age group in the Bath corpus. None of these two constructions was frequent, and some recordings had none at all. This confirmed previous observations made in English, German, and Russian that showed that lexical subjects are not frequent in oral language (Miller & Weinert 1998; Biber 2003). Table 3. Total number of lexical subjects and subject dislocations in the Bath corpus and ratio between DS and LS structures
Children
Adults
Age
LS
DS
DS/LS
LS
DS
DS/LS
3;0 3;6 4;0
13 7 1
17 12 9
1.31 1.71 10.50
23 9 7
16 5 12
0.69 0.55 1.71
Children’s LS constructions were less frequent than DS constructions for all age groups (see Table 3). For adults, this was not true, except for the 4;0 age group. The number of DS constructions produced by children did not appear to be much different from the number of DS constructions produced by adults t(107) = .51, p = .66, whereas the number of LS constructions produced by children was significantly lower than the number of LS constructions produced by adults t(107) = 2.04, p = .021. For each recording, the difference between the number of DS and LS constructions was computed. A statistically significant difference was found between children and adults, t(107) = 1.94, p = .027. The difference remained significant, t(17) = 2.44, p = .013 when the recordings with no DS nor LS construction (none for the child and none for the adult) were omitted from statistical computation. This suggests that the low frequency of DS and LS constructions did not influence the significance of the statistical results. It thus appears that children do not produce more dislocations than adults. The figures are roughly similar. Taking into account the fact that adults produced a slightly larger number of occurrences and of words, the children’s percentage of use of DS constructions was slightly higher than the adults’. However, it was the difference between the number of LS constructions and DS constructions that truly separated children from adults (see Table 3). In similar situations, when it was necessary to provide the lexical subject in an utterance, children tended to use a dislocated construction where adults often used an ‘a priori simpler’ lexical subject construction with no subject pronoun.
Christophe Parisse
6.2 Does the use of dislocations follow a developmental path? The use of dislocations appears to be a feature of children’s language, but this does not characterize or explain the use of dislocations by children. One important question that needs to be answered is whether the use of dislocations follows a developmental trend, or whether it is the same for all children and will change only through maturation at a later age. This can be measured with the House corpus, which follows children from age 2;0 to age 4;0. Results in Table 4 present the number of DS and LS constructions occurring in the Bath corpus. Contrary to what was observed in the Bath corpus, the number of DS constructions tended to increase with age to a much greater extent than the number of LS constructions. This is shown in the increased ratio of the number of DS constructions to the number of LS constructions (see Table 4, column 4). Table 4. Total number of lexical subjects and subject dislocations in the House corpus and ratio of DS to LS structures Age
LS
DS
2;0 2;3 2;6 2;9
21 22 33 38
1 10 23 54
Ratio DS/LS 0.06 0.45 0.70 1.42
The results obtained above could be confirmed using the Léveillé corpus (Suppes, Smith, and Léveillé 1973) from the CHILDES database. This corpus follows closely, two or three times a month, the development of language interaction between a child and his mother from age 2;2 to 3;3. The recordings were made at home, which contrasts with the corpus presented above. The same computations as above were done on the Léveillé corpus. The development in the production of LS and DS constructions by the child and his mother are presented in Figure 1. At first, the child produced many more LS than DS constructions. The number of LS constructions then tended to diminish whereas the number of DS constructions remained the same. Finally, the number of DS constructions increased similarly to the number of LS constructions. The most interesting result was that the development of the number of LS and DS constructions produced by the child did not in any way seem related to the use of these constructions by his mother. The mother used a much larger number of LS constructions than DS constructions for a long period (from age 2;2 to age 2;11) and, though she tended to use a reduced number of both constructions in the last recordings (from age 3;0 to age 3;3), this did not prevent a sudden growth of the use of such constructions by the child. It appeared that children frequently produce dislocated constructions. There was no obvious explanation as to why this construction would be chosen instead of the simple lexical subject construction, which children also used. The fact that they used both constructions prevents any explanation in terms of missing syntactic knowledge.
Left-dislocated subjects
14 12 10 LS-CHI DS-CHI LS-ADU DS-ADU
8 6 4 2
3 3;
9 2; 1 2; 0 11 3; 0 3; 2
2;
8 2;
7 2;
2;
6
3 2;
2;
2
0
Figure 1. Developmental trajectory for a child from age 2;2 to 3;3 (Leveille corpus).
The fact that they did not simply copy the adults’ production suggests that regulating mechanisms specific to the children were at work. There are more than one possible explanation to the results presented above. Some of the possible explanations can be tested using the material presented above (Bath and House corpora). The first explanation is that the choice of a given construction could be related to the verb used in the construction. If some constructions are verb specific for all children, then this would be a lexical choice. A second possibility is that some constructions are verb specific for specific children, which would mean they are usage-based constructions (Tomasello 2003). A third hypothesis is that the use of a pronoun allows for creating longer utterances. A fourth and final hypothesis is that the production of DS constructions could be a logical consequence in the development of the use of subject pronouns. All these four hypotheses are tested below.
6.3 Lexical choice The use of dislocations by children could by explained by many linguistic or psycholinguistic features. One of these is that this use is governed by a lexical choice. This means that some verbs will obligatorily use dislocations structure and some verbs will not, but that this is no a syntactic choice. To test the existence of a lexical obligatory choice, it is possible to use all the children’s corpora (Bath and House corpora). If there is a lexical tendency to use some types of verb with a type of construction, then there should be a partition between
Christophe Parisse
verbs that govern an LS construction and verbs that govern a DS construction. This did not seem to be the case. For all the children from the House and the Bath corpora, there were 64 verbs used in one or other construction. Twenty-eight of these verbs were used only once. Six verbs were produced more than once but used only in one context, either LS or DS. These six verbs represented only 18 of the total of 677 occurrences of verbs that occurred more than once in an LS or a DS construction. The 10 most frequently used verbs in such constructions are presented in Table 5 and no verb appears significantly more often in one type of construction than another. Table 5. Most frequent verbs used in LS or DS constructions (House and Bath corpora) Verb aller (to go) être (to be) avoir (to have) dormir (to sleep) manger (to eat) vouloir (to want) venir (to go) faire (to do) regarder (to look at) rentrer (to go into)
N of recordings 184 163 92 35 21 16 14 14 11 11
Lexical subject
Subject dislocation
117 109 57 9 4 5 9 5 7 3
200 116 55 45 17 13 11 12 5 8
Note: the first column gives the number of recordings containing at least one occurrence of the verb. More than one occurrence may occur per recording.
6.4 Item-based constructions There was no lexical tendency to use a verb in a specific construction when one considers the corpus of the children’s productions as a whole. However, it is possible that specific lexical constructions exist, but for specific children. Some children would use some verbs only in a certain type of construction, but other children would use the same verbs in different constructions, depending on how the verbs were modelled in their input. This corresponds to a usage-based hypothesis (Tomasello 2003). The existence of usage-based constructions should be tested. This will be done using the House corpus. The Bath corpus cannot be used here because it is too small to provide enough contexts to test this hypothesis. To test this, it is necessary to compute the individual use of LS and DS constructions for a given verb in the same recording session. This is presented in Table 6 for the House corpus. Once again, it appeared that LS constructions disappeared as children became older (verbs that were used only with LS constructions disappeared at age 4) and DS constructions became more and more frequent. It also appeared that, although some verbs were used in only one construction type and some children used only one construction type, most children used both types of construction and often the same verbs for both types (10 children
Left-dislocated subjects
Table 6. Use of specific constructions for the same child (House corpus) Age All LS All DS 2;0 2;3 2;6 2;9 3;0 3;3 3;6 3;9 4;0
12 8 9 8 7 4 4 2 0
1 2 4 4 8 14 10 6 6
Both LS & DS, different verbs
Both LS & DS, same verbs
0 2 6 4 5 3 7 7 9
0 3 3 8 12 6 4 14 10
out of 25 at age 4). Considering the relatively short length of the recordings, which did not provide the children with many opportunities to produce that many different constructions, it appeared unlikely that the choice between LS and DS constructions was usage-based, or that some children used the LS constructions more easily where other children seemed to prefer the DS construction, although this could still be the case for a few children.
6.5 Length effect It is well know that young children are sensitive to processing limitations (Valian, Hoeffner, and Aubry 1996). It is also possible that, during the process of language development, they tend to use simpler constructions before learning to use complex constructions (predicted by UBT). Both these reasons emphasize the fact that children will more readily use a less costly and simpler construction than a construction which is more costly and more complex. If processing limitations have an effect on children’s utterances, this would mean that the simplest structures will be produced more easily and could be used as a way to circumvent these limitations when producing long utterances. The measurement of utterance length could provide an explanation to children’s use of either LS constructions or DS constructions. As LS constructions are shorter than DS constructions, they should be easier to use and produce and their syntactic structure should also be simpler. This would make them useful to compensate for performance limitations. Concurrently, when children want to produce long utterances, they should automatically use the simpler construction to compensate for increased processing demands. This means that, in longer utterances, LS structures should be use more often than DS structures, even taking the supplementary element (the personal pronoun) present in DS structures into account. For all children of the House corpus, the average length of the DS utterances was 7.73 (6.73 minus the personal pronoun) whereas it was only 5.07 for LS utterances.
Christophe Parisse
This was also true when considering only the older children who often produced both types of constructions (DS: 8.10 –7.10 minus the pronoun; LS: 5.84). This is in clear contradiction with the idea that the shorter structure should help children to produce longer utterances, even taking the supplementary pronoun of DS structures into account. This would mean that either processing limitations bears no weight on the choice of one structure or another, or that LS structures, although they are shorter or correspond to a simpler tree structure, are not easier to process than DS structures. Further test should be done to choose between the two hypotheses. However, the difference between the length of the utterances that include LS or DS constructions needs to be explained. Whatever the results of further tests, there is indeed a possibility that a shorter structure (LS) is more complex than a longer one (DS). Two reasons may explain this: First, from a phonetic point of view, when a DS construction contains a pronunciation pause, it allows the child to produce a global utterance in two parts, each simpler than the counterpart LS construction. This pause can be considered as arising from limitations in performance. If the DS does not contain a pause but a rising pitch, then the phonological argument does not apply, but the second argument below does. Second, from a grammatical point of view, the use of two loosely related constructions (DS) – a lexical dislocated subject and a verb phrase that includes a pronoun – could be simpler for a child to compute than a unique construction (LS) that includes a lexical subject and a verb. This is because both grammatical structures in DS are simpler than the unique LS construction and because the operation of creating a DS – appending two syntactically independent constructs – is a very simple and very frequent operation in child language, and thus is very easy to produce. It would then be quite similar to the production of two successive but independent clauses.
If this explanation is verified, then it could be that performance limitations and construction simplicity are behind the use of DS constructions.
6.6 Consequence of the development of pronoun marking on verbs A last hypothesis is that DS constructions are produced because they contain verbs in their most usual form, with a personal subject pronoun before them. This would mean that personal pronouns are produced somewhat automatically, as if they were agreement markers, contrary to De Cat’s hypothesis (2002). The reason behind the increase in production of DS constructions would be that the use of personal pronouns becomes more and more automatic as children grow older. If pronouns are produced automatically before a verb in most contexts, the DS constructions would merely be LS constructions where the production of an optional pronoun is not blocked by the child. If this is true, there should exist a correlation between the development of
Left-dislocated subjects
personal pronoun marking and the development of DS constructions. This condition is not sufficient in itself to prove the hypothesis, but if not verified, would prove it wrong. To test this, the ratio of the number of subject pronouns produced before a verb to the number of verbs was computed for each recording. The correlation between this ratio and the number of DS constructions was then computed. The results showed a significant correlation value, df = 316, r = .428, z = 8.09, p <.0001, for the House corpus. But no significant correlation was found in the Bath corpus, either for the children, df = 108, r = .113, z = 1.159, p = .24, or for the adults, df = 108, r = .039, z = .396, p = .69. The difference in the House corpus and Bath corpus results may be the consequence of the low number of DS produced in the latter corpus or from the difference in the two situations. This discrepancy suggests that more data would be necessary to confirm the hypothesis of an automatic production of the subject pronoun.
7. Conclusion The results above should be handled with care because of the very small number of DS and LS constructions effectively produced by children and adults. It is quite possible that with larger numbers, the results would change. For instance, the differences in use of LS and DS structures for the adults showed that there were unknown factors involved in the production of these structures. It should be clear that the results need to be confirmed using more reliable methods or situations. For example, as DS and LS are indeed unusual constructions in oral language, the results obtained would be easier to confirm using induced language instead of spontaneous language. Further analysis of the contexts in which LS and DS structures are used is also necessary. It would be especially interesting to know whether children’s productions are similar analysis to adults’ productions in their context. Within the limits of the present results, the comparison between the number of subject dislocations produced by children and by adults showed that young children (especially children aged three to four) produced more dislocations than adults. There was a developmental trend in this use, as children aged two did not produce many dislocations and none were produced in large quantities before age three. In fact the surprising pattern that seemed to emerge was that children began by using more LS than DS constructions, then reversed the pattern when they were three to four years old, and would probably later go back to the first situation, which appears to be the adult norm (this last result was absent from the data as the children were not followed after age four, but they should develop towards the adult norm later). A series of statistical measures showed that it was unlikely that the reason behind children’s use of dislocations at age three and four was lexical or item-based in its nature, or the result of a specific developmental path for some children. A possible explanation for the use of dislocations was that they are a way of producing longer utterances by avoiding performance limitations or using simpler syntax (see part 6.5).
Christophe Parisse
This would mean that dislocations are constructed using mental operations such as concatenation which are less complex than operations involved in generating nondislocated constructions. When children will become older, processing limitations will diminish and the number of DS would also go down (as is attested in adults’ productions). Another explanation was that the production of subject pronouns before a verb had become sufficiently automatic that children could not stop their production, even when not in an obligatory context. This automatic behaviour will also change at a later age when the use of LS constructions will increase again. Even if the current results are tentative, they are useful in allowing hypotheses to be drawn for future work and in pointing out interesting phenomena in child language development. A proposal is that there are three different periods to be considered regarding the nature of LS and DS. At first, children start to produce what appears to be LS but is in fact the concatenation of a noun and a verb produced without morphosyntactic markers such as determiners or clitics. At this stage of development there is no difference between LS and DS from the children’s grammar point of view.1 Then, children start producing subject pronouns before all verbs, whether there is a lexical subject or not. In this second period, LS would mostly disappear and DS would become the norm. The process of concatenation that leads to DS is the similar with LS in the first period, but with elements that include a determiner or a pronoun. Finally, children learn to produce real LS constructions where lexical subject and verb are linked in a complex abstract structure and go on towards adult mastery. The previous production of DS structures would help them to develop more complex abstract categories, as hypothesized in UBT, and lead to the more complex LS construction. This development is helped by the development of written language. LS constructions are more frequent in written language that oral language (see Miller & Weinert 1998), to the point where the use of this construction is strongly advocated in the written language norm. This would lead children to integrate the LS construction in their oral language only after having started to learn the written language. A major consequence of this explanation is that LS structures would not be the prototype construction around which other constructions, such as DS constructions, are build, but only a type of construction among others, which are often easier to use, especially in oral language. The existence of a link between written language acquisition and the development of LS structures is made more plausible by the possibility of performance limitations in the production of LS structures. Written language is less prone than oral language to
1. This interpretation is much stronger than what De Cat (2002) proposes but it agrees with her basic interpretation that it is difficult to know what is a LS and a DS in young children: “Like adults, they [French children] frequently produce subject dislocations – and these are attested from the onset of word combination []. The difference between child and adult language in this context is restricted to the well-known but not fully understood null subject phenomenon in language acquisition.” (De Cat 2002, 246).
Left-dislocated subjects
performance limitations, as limits on working memory are removed by the presence of an external support that allows language users to analyse and reanalyse the language data at will.
8. Acknowledgements Grateful thanks to Sean Devitt for his help with the English version of this chapter.
References Ashby, W.J. 1977. Clitic Inflection in French. Amsterdam: Rodopi. Biber, D. 2003. Variation among university spoken and written registers: A new multi-dimensional analysis. In Corpus Analysis – Language Structure and LanguageUuse, P. Leistyna & C.F. Meyer (eds). Amsterdam: Rodopi. Blanche-Benveniste, C., Deulofeu, J., Stefanini, J. & van den Eynde, K. 1984. Pronom et Syntaxe. L’approche pronominale et son application au français, A. Selaf (ed.). Paris: CNRS. Blasco-Dulbecco, M. 1999. Les dislocations en français comtemporain. Paris: Honoré Champion. Chevrie-Muller, C., Simon, A.-M., Le Normand, M.-T. & Fournier, S. 1997. Batterie d’évaluation psycholinguistique (BEPL). Paris: ECPA. Chifflet, L. 1659. Essai d’une parfaite grammaire de la langue françoise. Genève: Slatkine. Reprints, 1973 ed. Anvers: Jacques Van Meurs. Croft, W. 2001. Radical Construction Grammar. Oxford: OUP. De Cat, C. 2002. French Dislocations. PhD dissertation, Language and Linguistic Science, University of York. De Cat, C. 2005. French subject clitics are not agreement markers. Lingua 115(9): 1195–1219. Goldberg, A.E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago IL: University of Chicago Press. Grevisse, M. & Goose, A. 1993. Le Bon Usage-Grammaire française. édition refondue par André Goosse ed. Paris - Louvain-la-Neuve: DeBoeck-Duculot. Le Normand, M.-T. 1986. A developmental exploration of language used to accompany symbolic play in young, normal children (2–4 years old). Child: Care, Health and Development 12:121–134. Miller, J. & Weinert, R. 1998. Spontaneous Spoken Language. Oxford: Clarendon Press. Parisse, C. 2005. New perspectives on language development and innateness of grammatical knowledge. Language Sciences 27(4): 383–401. Parisse, C. & Le Normand, M.-T. 2000. How children build their morphosyntax: The case of French. Journal Of Child Language 27(2): 267–292. Parisse, C. & Le Normand, M.-T. 2001. Local and global characteristics in the development of morphosyntax by French children. First Language 21:187–203. Suppes, P., Smith, R. & Léveillé, M. 1973. The French syntax of a child’s noun phrases. Archives de Psychologie 12(166): 207–269. Tomasello, M. 2000. Do young children have adult syntactic competence? Cognition 74: 209–253.
Christophe Parisse Tomasello, M. 2002. The emergence of grammar in early child language. In The Evolution of Language Out of Pre-language, T. Givon & B.F. Malle (eds). Amsterdam: John Benjamins. Tomasello, M. 2003. Constructing a language: A usage-based theory of language acquisition. Cambridge MA: Harvard University Press. Tomasello, M. 2006. Acquiring linguistic constructions. In Handbook of Child Psychology, D. Kuhn & R. Siegler (ed.). New York NY: Wiley. Valian, V., Hoeffner, J. & Aubry, S. 1996. Young children’s imitation of sentence subjects: Evidence of processing limitations. Developmental Psychology 32(1):153–164. Wilmet, M. 2003. Grammaire critique du français. Bruxelles: Duculot.
The development and interaction of Case and Number in early Russian Galina Gordishevsky & Jeannette Schaeffer This chapter reports the results of a spontaneous speech study on the interaction between Case and Number in child Russian. Following Hoekstra & Hyams (1995, 1996) and Müller (1994), we hypothesize that the Number head is initially underspecified in child grammars and represents [+singular] only. Our results support this hypothesis, exhibiting almost flawless Case licensing in the singular, but failure to use plural Cased forms correctly between the ages 1;8–2;0 and to a lesser extent between the ages 2;1–2;6. Our results also support the Full Clause or Full Competence Hypothesis (Hyams 1992; Wexler 1992; Poeppel & Wexler 1993), according to which functional categories are present in child grammar from the beginning. Keywords: Case, Number, underspecification, Russian child language
1. Introduction The goal of this study is two-fold. First and foremost, we investigate the acquisition of the relation between Case and Number by Russian-speaking children. Assuming that both Case and Number are generated by functional syntactic structure, our results contribute to the long-standing debate as to whether children’s grammars contain functional categories from the beginning or not. Second, we trace development of the two categories and the relation between them for an extended period of time. For this purpose, we examined the spontaneous speech of three young monolingual Russianacquiring children from the beginning of two-word combinations (around age 1;8) and through the production of multiple-word utterances (around age 2;6). Our results show that Case is produced correctly in the singular during the whole period, namely from the earliest stages of language acquisition, but not in the plural, at least before age 2. Inspired by Hoekstra & Hyams (1995, 1996), we argue that an underspecified nominal Number head blocks Case licensing in the plural in early Russian. After the age of two years, we observe some improvement in plural Case inflection. However, even at age 2;6, the children have not reached target like mastery yet. Our findings lend support to the Full Competence Model (Hyams 1992; Wexler 1992; Poeppel & Wexler 1993), which states that functional categories, including those
Galina Gordishevsky & Jeannette Schaeffer
responsible for Case licensing, are present from the beginning. Yet, certain functional categories, such as Number, may be initially underspecified (cf. Underspecification Hypothesis - Hoekstra and Hyams 1995, 1996). The structure of the chapter is as follows. In the next section (2), we provide a brief description of the Russian Case system, outline a theory of Case licensing and its relation to Number, and present a short review of previous studies on the acquisition of Case and Number cross-linguistically. Section 3 contains our hypotheses and predictions for Russian child language with respect to the investigated phenomena. After describing our methods in section 4, we show in section 5 how our results confirm our hypotheses. The results are further discussed in section 6. Finally, section 7 contains the conclusion.
2. Background 2.1 The Russian Case system Russian has an elaborate Case marking system, allowing for relatively free word order (cf. also Bailyn 1995), as exemplified in (1): (1) Russian free word order a. Miša podaril Maše knigu. Miša-NOM gave Maša-DAT book-ACC ‘Misha gave Masha a book.’ b. Miša podaril knigu Maše. Miša-NOM gave book-ACC Maša-DAT c. Miša Maše podaril knigu. Miša-NOM Maša-DAT gave book-ACC d. Maše podaril knigu Miša. Maša-DAT gave book-ACC Miša-NOM e. Maše knigu podaril Miša. Maša-DAT book-ACC gave Miša-NOM f. Knigu Maše podaril Miša. book-ACC Maša-DAT gave Miša-NOM g. Knigu Miša podaril Maše. book-ACC Miša-NOM gave Maša-DAT h. Maše Miša podaril knigu. Maša-DAT Miša-NOM gave book-ACC
All the combinations in (1) are instantiations of the same sentence with the meaning “Misha gave Masha a book”, and the list is far from exhaustive. As the reader can see, the subject can appear in sentence-initial position (as in the a, b, and c examples), in sentence-final position (as in d, e, and f), immediately preceding the verb or following the objects (as in g and h), and so on. The same variation is observed for the other
Case and Number in early Russian
arguments. In a language with such a free word order, overt Case marking is crucial in order to understand who did what to whom. 1 The Russian nominal paradigm includes six main Cases, namely nominative (NOM), accusative (ACC), genitive (GEN), dative (DAT), instrumental (INSTR), and prepositional (PREP), and four residual Cases, namely partitive, vocative, paucal, and locative. An elaborate discussion of the main Cases, including examples, follows in the next section (2.2). Here we include a brief discussion of the residual Cases. The partitive Case has distinct Case markings for a very limited number of lexical items, and therefore is not considered productive. For instance, partitive Case has a distinct marking with certain masculine mass nouns, as in (2) (from Brecht & Levine 1986: 19): (2) Partitive Case a. cena konjaka vs. b. rjumka konjaku price cognac-GEN glass cognac-PART ‘price of cognac’ ‘glass of cognac’
In most cases, however, the marking of the partitive Case is identical to genitive, and therefore it is usually not considered as a distinct Case. The partitive Case is observed only once in our spontaneous child speech data. Furthermore, the vocative Case, usually identical to nominative, is attested in colloquial speech. The vocative is normally used with [+human] nouns and names, and if the noun ends with the vowel [a], the ending is optionally omitted, as illustrated in (3): (3) Vocative Case a. Mam(a)! b. Tanj(a)! mommy Tanja ‘Mom!’ ‘Tanya!’
The paucal Case, assigned by numerals 2, 3 and 4 (see Franks 1995 and Rappaport 2002 for discussion), usually coincides with the genitive singular, as in the following examples: (4) Paucal Case a. dva studenta two student-GEN ‘two students’
b. tri kukly three doll-GEN ‘three dolls’
In our child data we observed only one use of (erroneous) paucal Case. Finally, the locative Case is a sub-Case of prepositional Case and seldom shows distinct Case marking, as in v sadu “in garden-LOC” versus pri sade “near garden-PREP” 1. We thank an anonymous reviewer for pointing out that whereas the sentences in (1) are truth-conditionally equivalent, they have different topic/focus configurations and therefore cannot be claimed to have exactly the same meaning. For a detailed discussion of the interaction of word order with information structure in Russian see e.g., King (1995) and Gordishevsky (2006/2007).
Galina Gordishevsky & Jeannette Schaeffer
(cf. Comrie 1986). This Case is very marginal in contemporary Russian, and the distinction is rarely made. We now turn to an elaborate discussion of the Russian Case and declension systems. Our analysis is largely based on Babyonyshev (1993), which is detailed enough to suffice for the purposes of the present study. For additional, and more detailed, accounts of the Russian Case system the reader is referred to Franks (1995); Bailyn (1995); and Halle (1994). Russian has four nominal declension classes, which contain nouns depending on a combination of the gender of the noun and its ending in the nominative singular form, as illustrated in (5): (5)
Russian Declension Classes (in the nominative singular form) Class I: ruka ‘hand’; papa ‘daddy’ Class II: dom ‘house’; okno ‘window’; pole ‘field’ Class III: kost’ ‘bone’; molodëž’ ‘youth’ Class IV: pal’to ‘coat’; kofe ‘coffee’
The first declension contains feminine and masculine nouns that end in the vowel [a] in the nominative singular form. The second declension includes masculine and neuter nouns that end in a consonant or in the vowel [o] or [e]. The third declension contains feminine nouns that end in a palatalized consonant, and the fourth declension consists of nouns of foreign origin of all Genders. While each of the first three declension classes has its own unique Case endings for both singular and plural forms (listed in Table 1), the nouns of the fourth declension class exhibit the same phonological form in all Cases in both singular and plural. This form is identical to the nominative singular form, and therefore the nouns that belong to this class are referred to as “declensionless”. They do not appear in the table. Table 1. The Russian Nominal Case paradigm Declension + Number 1stsg CASE
2nd sg
3rdsg
[+anim] [-anim]
1st pl
2nd pl
3rd pl
[+anim] [-anim] [+anim] [-anim] [+anim] [-anim]
NOM
a
Ø
Ø/o/e
Ø
y/i
y/i
y/i
y/i/a
i
i
ACC
u
a
Ø/o/e
Ø
Ø
y/i
ov/ev/ ej
y/i/a
ej
i
GEN
y/i
a
a
i
Ø
Ø
ov/ev/ ej
ov/ev/ ej/Ø
ej
ej
DAT
e
u
u
i
am
am
am
am
am
am
INSTR oj(u) om/em om/em
ju
ami
ami
ami
ami
mi
ami/mi
PREP
i
ax
ax
ax
ax
ax
ax
e
e
e
– shading shows homophonous (and therefore indistinguishable) forms within each declension – Ø marks the absence of a vocalic ending
Case and Number in early Russian
While examining the Russian Case paradigm in Table 1, note that sometimes the same Case marking morpheme is used for two or even more different Cases. For example, in second declension singular [+animate] nouns the [a] ending for accusative is homophonous with the ending for genitive Case, and in second declension singular [-animate] nouns the [Ø/o/e] endings for nominative are homophonous with the endings of accusative Case. This is marked by shading in the table. The same is observed in plural: the accusative endings of [-animate] nouns in all three declensions are homophonous with the nominative endings. In the analysis of our child data it will become clear that this observation is important: the homophonous endings present ambiguous data and can therefore not be taken as instances of flawless Case marking. The pronominal Case paradigm differs from the nominal one. We do not present this paradigm here, as it is irrelevant for our purpose, namely to investigate Case marking in nouns and its interaction with Number. Whereas nouns are argued to involve a Case-chain between N and D (cf. section 2.3), pronouns do not involve such a chain, as they are base generated under D (cf. Adger 2003: 256).
2.2 Environments of Case licensing In our analysis, we adopt the assumptions of the Minimalist framework by Chomsky (1998, 1999, 2001) with respect to Case licensing.2 We assume that Case features of DPs are valued by and then checked against the corresponding selectional features of (finite) verbs, prepositions, and nouns in three different ways: structurally, lexically, or inherently. In order not to commit ourselves to the terms “Case assignment” or “Case checking” we refer to this process as “Case licensing”. Structural Case is associated with particular structural positions and licensed under the relation of c-command. There are three structural Cases distinguished cross-linguistically and in Russian: nominative, accusative, and genitive. Examples of structural Cases in Russian appear in (6): (6) Structural Cases a. Mal’čik čitaet. boy-NOM reads ‘The boy is reading.’ b. Mal’čik čitaet knigu. boy-NOM reads book-ACC ‘The boy is reading a book.’ c. kniga mal’čika book-NOM boy-GEN ‘the book of the boy’ or ‘the boy’s book’
2. For a detailed description of the Case-licensing mechanism along the Minimalist lines see also Radford & Ramos (2001).
Galina Gordishevsky & Jeannette Schaeffer
Structural nominative Case is licensed in the specifier of little vP (the base subject position) by the [nominative] Case feature of (a finite) T under a c-command relation. The relevant structure is in (7). The structure is partial, showing only the projections and their features that are relevant for the present discussion. Structural accusative Case is licensed in the complement position of V (where the direct object is merged) by the [accusative] Case feature of little v under a c-command relation. This can also be seen in the structure in (7). (7)
T′ nP
T[nom]
n′
Subject[N,nom]
VP
n[acc] V
Object[N, acc]
Finally, structural genitive, also referred to as the “adnominal” genitive, is licensed in the specifier of an optional functional projection Possessor by the [genitive] Case feature of the D head under c-command, as shown schematically in (8). The irrelevant projections and features are omitted. (8)
DP spec
D′
D[gen]
PossP
Possessor[gen] Poss′ Poss
NP
Examples of structural (or “adnominal”) genitive are provided in (9): (9) Structural (Adnominal) Genitive Case a. Razrušenie goroda destruction city-GEN ‘destruction of the city’ b. Pomošč’ Peti help Petja-GEN ‘Petja’s help’
Case and Number in early Russian
Other instances of structural genitive Case are genitive of quantification and partitive genitive, illustrated in (10a) and (10b), respectively: (10) Other Structural Genitive Cases a. Genitive of Quantification pjat’ mal’čikov five boys-GEN ‘five boys’ b. Partitive Genitive Ja xoču čaju. I want tea-GEN(PART) ‘I want some tea.’
Another environment in which genitive functions as a structural Case is the genitive of negation construction. Thus, direct objects of transitive verbs can optionally appear in the genitive Case (instead of accusative) under sentential negation, resulting in somewhat different readings, as illustrated in (11a). Moreover, subjects of negated unaccusative existential verbs such as ‘be’, ‘exist’, etc. must appear in the genitive Case, as in (11b). (11) Structural Genitive of Negation a. Ja ne polučal pisem / pis’ma. I not received letters-GEN letters-ACC ‘I didn’t receive any/particular letters / the letters.’ b. U menja net(u) / ne bylo vremeni. at me no not have time-GEN ‘I have/had no time.’
The syntactic analysis of these constructions is irrelevant for the purposes of the present study and is therefore not presented here. The reason for this is that the children in our study used the genitive of negation exclusively with one word type, namely the negated form of verb ‘be’, which may be taken as evidence that they treat this instance of the genitive of negation as a lexical rather than structural Case (cf. section 2.4 for further explanation). Inherent Case is the Case carried by nominals bearing particular theta-roles, and is licensed by the Case feature of the theta-assigner, usually the verb. In Russian two inherent Cases are distinguished: dative and instrumental. Inherent dative Case is carried by nominals bearing the theta-roles of goal, beneficiary, or experiencer, as exemplified in (12); and inherent instrumental Case is carried by (nominals bearing the theta-roles of) instruments and agents of passive sentences, as illustrated in (13): (12) Inherent Dative Case a. Ja dala knigu Maše. – goal I gave book Maša-DAT ‘I gave the book to Masha.’
Galina Gordishevsky & Jeannette Schaeffer
b. c.
My kupili podarok Vanje. we bought present Vanja-DAT ‘We bought a present for Vanja.’ Mne xolodno. me-DAT cold ‘I am cold.’
(13) Inherent Instrumental Case a. On pišet ručkoj. he writes pen-INSTR ‘He writes with a pen.’ b. Pis’mo bylo napisano Valjej. letter was written Valja-INSTR ‘The letter was written by Valja.’
– beneficiary
– experiencer
– instrument
– passive agent
Finally, lexical (or “quirky”) Case is carried by nominal complements of particular lexical items such as verbs and prepositions regardless of the theta-role borne by the nominal carrying this Case or of its structural position. The lexical Case licensed by particular lexical items, i.e., some verbs and most prepositions, is specified in their lexical entry, hence the name “lexical” Case. In Russian, two types of lexical Case should be distinguished, namely lexical Cases licensed by verbs, and lexical Cases licensed by prepositions. The lexical Cases licensed by verbs include genitive, dative, and instrumental, as exemplified in (14), and the lexical Cases licensed by prepositions include accusative, genitive, dative, instrumental, and prepositional, as in (15): (14) Lexical Cases Licensed by Verbs a. Ona boitsja napadenija. she fears assault-GEN ‘She fears an assault.’ b. Ty mešaeš mne. you disturb me-DAT ‘You are disturbing me.’ c. Ona rukovodit gruppoj. she leads group-INSTR ‘She leads a group.’ (15) Lexical Cases Licensed by Prepositions a. Ja polozhila knigu na polku. I put book-ACC on shelf-ACC ‘I put the book on the shelf.’ b. On živët u roditelej. he lives at parents-GEN ‘He lives at his parents’ house.’ c. Busy rassypalis’ po polu. beads scattered on/over floor-DAT ‘The beads scattered on the floor.’
Case and Number in early Russian
d. e.
Derevo rastët pod oknom. tree grows under window-INSTR ‘The tree grows under the window.’ Rasskaži mne o Pete. tell me about Petja-PREP ‘Tell me about Petja.’
Following Babyonyshev (1993, 1997), we treat the dative Case borne by the nominal complement of the preposition k (‘to’) as inherent Case, rather than lexical Case. The reason is that complements of this preposition bear the theta-role of goal, which fits into the definition of inherent Case given above. An example of inherent dative licensed by the preposition k (‘to’) is given below: (16) Inherent Dative Case a. On podošel k oknu. – goal he approached to window-DAT ‘He came to the window.’
2.3 The interaction of Number and Case With regard to the internal structure of DPs, there are many competing analyses in the literature, none of which, to the best of our knowledge, can offer a satisfactory account of the DP-internal Case-licensing mechanisms. Therefore, we outline an account that incorporates some of the older ideas with the newest minimalist developments. In particular, we follow Ritter (1991, 1995) and Koopman (1999) in assuming that noun phrases contain a functional projection intermediate between DP and NP, namely NumP, whose head bears the Number feature. The structure of DP that we adopt is in (17): DP
(17) spec
D’ D
NumP spec
Num’ Num
NP spec
N’ N
Galina Gordishevsky & Jeannette Schaeffer
With regard to Case, we follow Hoekstra & Hyams (1995), assuming that in adult language a DP-internal Case-chain between N and D needs to be established. Along recent minimalist lines, we assume that in adult language the Num head has an unvalued uninterpretable Number feature that is valued and checked via an Agreement relation with the N head hosting a noun with an interpretable Number feature. This Number feature can be either singular or plural, depending on the reference of the noun carrying it. This is shown schematically in (18): DP
(18) spec
D’ D
NumP spec
Num’ Num
NP
[num] spec
N’ N
[sg/pl; Case]
As the target for Case feature licensing is the whole DP, the Case feature has to be in a position from which it can percolate up to the DP level. Such a position is D. In order to fulfill this requirement, a DP internal Case-chain is established. Part of the chain is formed by the Agreement relation between N and Num. The chain is extended to D to enable the Case feature to percolate up to DP. Thus, the Case-chain connects N, Num and D. By the Condition of Chain Uniformity (Chomsky & Lasnik 1993), each node involved in the chain carries the Case feature. The relevant structure is given in (19). The result of the DP-internal Case-chain is the Case feature on the DP. This Case feature on DP is subsequently licensed by the Case feature of the relevant head, such as T, v, or D, resulting in, for example, nominative, accusative or genitive Case.
2.4 Previous accounts of acquisition of Case and Number Let us now review briefly some previous studies on the acquisition of grammatical Case and Number in child language. Babyonyshev (1993) reports high percentages of correct structural and inherent Case inflection in early Russian, but slightly worse performance on lexical Case. In particular, in a spontaneous speech investigation of two monolingual Russian-speaking boys aged 1;6–2;0 and 2;1–2;7, she found 99.5%
Case and Number in early Russian
(19)
DP[Case] spec
D’ D [Case]
NumP
spec
Num’ Num [Case]
NP spec
N’ N [Case]
(596/599) correct usage of structural nominative Case and 90% (27/30) correct usage of structural accusative Case in appropriate positions. Concerning structural genitive, the children appear to be non-productive in its use, using it mainly with one construction, the genitive of negation, and only with one verb, the negated form of ‘be’. Babyonyshev suggests that this verb might have been analyzed by the children to assign lexical genitive Case, rather than structural genitive. Moreover, inherent dative Case was used appropriately in 87.5% (21/24) of the contexts, showing good mastery of the inherent Cases. More errors were found in Lexical Cases, however the percentages were above chance level. Lexical accusative was used appropriately in 62.5% (5/8) of the contexts, lexical genitive in 80% (8/10) of the contexts, and lexical prepositional in 75% (3/4) of the contexts. Babyonyshev’s findings are summarized below: (20) Findings of Babyonyshev (1993) (2 monolingual Russian-speaking boys aged 1;6–2;0 and 2;1–2;7) a. Structural Cases: NOM – 99.5% (596/599) ACC – 90% (27/30) b. Inherent Cases: DAT – 87.5% (21/24) INSTR – unattested c. Lexical Cases: ACC – 62.5% (5/8) GEN – 80% (8/10) PREP – 75% (3/4)
Babyonyshev’s findings present evidence against the lexical-thematic analysis of Radford (1986, 1990), outlined in (21a), which claims that child grammars “pass through a
Galina Gordishevsky & Jeannette Schaeffer
stage in which all functional categories are absent and only lexical categories that enter into thematic relations are present” (Babyonyshev 1997: 1). The finding that Case is checked correctly and early in child Russian presents evidence in favour of the Full Clause (or Full Competence) Hypothesis (Hyams 1992; Wexler 1992; Poeppel & Wexler 1993), outlined in (21b), according to which functional categories are present in the child’s grammar from the beginning. (21) a. b.
Lexical-Thematic Analysis (Radford 1986, 1990) Functional categories are absent from child grammar and only lexical categories that enter into thematic relations are present. Full Clause Hypothesis (Hyams 1992; Wexler 1992; Poeppel & Wexler 1993) Functional categories are present in child grammar from the beginning.
Although Babyonyshev demonstrates early mastery of Case distinctions by young Russian-speaking children, she does not distinguish between singular and plural. In a study on the acquisition of English plural, Zapf & Smith (2003) outline a developmental trend based on a summary of a number of previous studies on this topic. Thus, they distinguish four developmental stages, outlined in (22) below: (22) Stages of acquisition of plural in English (from Zapf and Smith 2003) Stage 1: 0–18 months No use of English plural Stage 2: 18–30 months Imitation of English plural Stage 3: 20–33 months Increased production, errors and over-generalizations Stage 4: 24–49 months English plural mastered
Zapf & Smith’s second and third stages are particularly interesting. During the second stage, children are said to “produce highly frequent regular and irregular plurals. For example, children use pluralized nouns such as ‘toys’, ‘shoes’, ‘animals’ or ‘keys’ ” (2003: 835). Zapf & Smith propose that this stage be characterized as the “rote stage”, because children use the plural forms that are frequent in the input and therefore might have been memorized. During the third stage, an increased and more successful production of the plural is observed, although errors are still attested. In section 5 we will show how our child data fit into these developmental stages. In a spontaneous speech study of two bilingual German/French children aged 1;6– 3;0 and 1;5–3;0, Müller (1994) reports a developmental stage at which the children do not mark Gender and Number distinctions, based on the finding that they do not make productive use of adult like determiners that mark these distinctions in either language. Müller suggests “that the absence of adult like determiners to encode grammatical Number and Gender distinctions is due to the unavailability of the corresponding grammatical features” (1994: 60). She further proposes that the relevant functional categories are present in the child grammar, but they are underspecified. In particular, she argues that initially [+singular] represents the unmarked feature, as opposed to [–singular].
Case and Number in early Russian
Hoekstra & Hyams (1995, 1996) put forward a similar proposal and argue that “in the early grammar the functional head Number may be left unspecified” (Hoekstra and Hyams 1995: 3). Since Number is part of both the verbal and the nominal domain, its underspecification has cross-categorial repercussions. In the verbal domain an underspecified Number head causes verbs to be non-finite, creating so-called “Root Infinitives”, while in the nominal domain underspecified Number results in determiner drop and lack of plural marking on nouns – all during the same developmental stages. The absence of a Number specification in the nominal domain breaks the D-N chain within a DP (cf. section 2.3), necessary for the realization of an overt determiner and marking of plural. This idea of ‘chain-breaking’ will be elaborated on in section 3, in which we formulate hypotheses and predictions regarding the relationship between Case and Number marking in child Russian based on Hoekstra & Hyams’ Number Underspecification Hypothesis. Despite the claim that functional categories may be underspecified in early child grammar, Hoekstra & Hyams (1995, 1996, 1997, 1998) maintain the Full Clause Hypothesis which states that all functional categories/projections in the syntactic tree are present from the beginning. In Hoekstra and Hyams’ (1998) own words: On this approach the child’s grammar has in principle the full clausal structure available. The Full Clause Hypothesis also underlies various analyses which postulate the underspecification of particular functional categories in the top structure of the clause. This has been the tack that we have taken (cf. Hoekstra & Hyams, 1995, 1996, 1997). (Hoekstra & Hyams 1998: 6)
3. Hypotheses and predictions Following Babyonyshev (1997) and Hoekstra & Hyams (1995, 1996, 1997, 1998), we adopt the Full Clause Hypothesis, i.e., we hypothesize that all functional categories are present in child grammar from the beginning. In particular, we hypothesize that the phrase structure positions responsible for Case licensing are available to the child from very early on.3 This first hypothesis is formulated in (23): (23) Hypothesis 1 All functional categories responsible for Case licensing are present in the child grammar from very early on.
3. Since the Case system does not have its own functional category/projection, we cannot hypothesize that it is Case itself that is underspecified. Rather, as outlined in section 2.3, Case results from the correct establishment of a chain between several heads within the DP, among which there are functional heads. Underspecification of one of these functional heads results in a DP without a Case feature, causing the DP derivation to crash.
Galina Gordishevsky & Jeannette Schaeffer
The specific prediction that follows from this hypothesis is that children do not make mistakes in Case licensing, that is, they are expected to use correct Case in obligatory environments. This first prediction is stated in (24): (24) Prediction 1 Young Russian-speaking children produce Case correctly.
However, we also follow Hoekstra & Hyams (1995, 1996, 1997, 1998) and Hyams (1996) in hypothesizing that the differences between child and adult grammars are due to the underspecification of functional categories. In particular, in the spirit of Hoekstra & Hyams (1995, 1996) and Müller (1994), we hypothesize that the Number head is initially underspecified in early grammars, namely it represents [+singular] only. This hypothesis is formulated in (25): (25) Hypothesis 2 The Number head is initially underspecified in child grammars and represents [+singular] only.
If this is true, this implies that plural nouns run into problems in early child grammar. A plural feature of the noun would not match the [+singular] feature on the underspecified Num. This is shown schematically in (26) below: (26)
DP spec
D’ D
NumP spec
Num’ Num
NP
[sg] spec
N’ N
[plural]
As a consequence of this feature mismatch, the Case chain cannot be established, resulting in a DP without a Case feature. Therefore, we predict that young Russian-speaking children fail to use Case correctly in the plural. This prediction is formulated in (27): (27) Prediction 2
Young Russian-speaking children fail to use Case correctly in the plural.
Case and Number in early Russian
This failure to use correct Case in the plural may manifest itself as the use of singular nouns in plural contexts (as also suggested by Hoekstra & Hyams (1995)). Another possible manifestation is the use of a default plural Case form for all plurals. When the noun is singular, Underspecified Number head does not break the Casechain as it is [+singular], thus yielding forms correctly marked for Case. This leads us to a reformulation of Prediction 1 to include only singular nouns. The revised Prediction 1a is in (28): (28) Prediction 1a Young Russian-speaking children produce Case correctly in the singular.
Furthermore, the Underspecification of Number may have consequences for the use of plural nouns in general. It has been observed cross-linguistically that young children do not use plural productively, i.e., they either do not produce plural forms or use them erroneously (cf. Pizzuto & Caselli (1992) for Italian, Grinstead (1994) for Spanish and Catalan, Ferdinand (1994) for French, and Clahsen et al. (1994) for German). At the same time, children as young as 1;6 exhibit productive control over singular nominal and verbal forms. Hoekstra & Hyams (1995, 1996) refer to this phenomenon as the “Avoid Plural Phenomenon” (APP) and claim that it is caused by the Underspecification of Number. We predict that the APP is also apparent in child Russian, and that the plural forms that are observed are frequently used plurals that represent mainly rotelearned forms (cf. Zapf & Smith 2003). This prediction is stated in (29): (29) Prediction 3 Initially plural is unproductive in child Russian, i.e., plural forms are rarely used compared to the singular and the attested plural forms are frequently used plurals that may be rote-learned.
The predicted rare usage of plural forms may be realized as avoidance of situations in which plurality must be expressed or as singular forms instead of plural ones. Now that we have laid out our hypotheses and predictions, we turn to the actual investigation of the Russian child data.
4. Methods 4.1 Subjects We investigated the spontaneous speech of three monolingual Russian-acquiring girls, between the ages of 1;8–2;6. Their speech was recorded in 30–60-minute sessions with 3-week intervals by one of the authors (Galina Gordishevsky) in their home settings in the presence of at least one of the caregivers. All the girls are monolingual native speakers of Russian, born in Israel to families of new immigrants and raised in strictly Russianspeaking environments by native Russian-speaking parents (who acquired Hebrew as a
Galina Gordishevsky & Jeannette Schaeffer
second language) and monolingual grandparents and/or elderly nannies. Exclusively Russian picture books and cartoons were read/shown to all the children both during the sessions and on a regular basis. Two of the girls (namely, ZLA and KAT) started going to Russian-speaking daycare centres around the age of two years.4 The recordings yielded spontaneous speech transcripts of around 200 utterances each (on average). The transcripts were made by native Russian-speaking research assistants in the CHILDES (Child Language Data Exchange System) format (see MacWhinney 2000) with some adaptations for Russian. Each transcript was crosschecked by an additional research assistant and finally reviewed by the first author. The analyzed transcripts included both the children’s and the adults’ utterances and detailed descriptions of contexts in which these utterances were produced (such as accompanying gestures, looking in particular direction, pointing to objects etc.). Overall, 34 transcripts were analyzed. Being interested in possible development, we divided the files into two stages, namely Stage 1, including 18 files and covering the ages 1;8–2;0, and Stage 2, which included 16 files and covered the ages 2;1–2;6. Details regarding the subjects’ age, MLU, number of transcripts and utterances analyzed in both developmental stages, are provided in Tables 2 and 3. Note that the ages of the three children are directly comparable due to the high similarity of their MLU counts and due to the fact Table 2. Subject information – Stage 1 child’s MLUwords No of files name age range range analyzed ZLA 1;8.10–2;0.9 1.07–2.66 MIC 1;8.13–2;0.9 1.48–1.92 KAT 1;9–2;0.14 1.54–1.99 overall
7 6 5 18
average No of utterances per file
overall No of utterances
205 260 115 200
1437 1559 577 3573
Table 3. Subject information – Stage 2 child’s MLUwords No of files name age range range analyzed ZLA 2;1.1-2;5.12 2.49-3.08 MIC 2;1.0-2;3.24 2.19-2.54 KAT 2;2.17-2;6.2 2.33-2.98 overall
6 5 5 16
average No of utterances per file 244 235 182 222
overall No of utterances 1462 1174 908 3544
4. The resemblance of our findings on Case use in the three children to the findings of Babyonyshev (1993, 1997) strengthens our observation of no influence of a second language on the children’s productions whatsoever.
Case and Number in early Russian
that all of them entered the two-word stage at the same age, i.e., around the beginning of data collection (cf. Gordishevsky (2006/2007) for more information).
4.2 Analysis We analyzed all nouns bearing overt Case marking. As mentioned in section 2.1, pronouns are base-generated in D and do not involve a Case-chain between N, Num and D. Being interested in the interaction between Case and Number, we excluded pronouns from our analysis. A number of additional elements bare Case marking in Russian. These include various DP modifiers that agree with the head noun in Number, Gender and Case, such as a) adjectives, which exhibit Case marking paradigm different from the nominal one. Adjectives are assumed to receive their Case features via the process of noun-adjective agreement within the morphological component of the grammar, but not bear Case features in syntax (Babyonyshev 1997); b) nominal and pronominal possessors (= possessive forms of nouns and pronouns) that exhibit an adjectival Case marking paradigm different from the nominal one and are therefore often treated as adjectives rather than nominals. These elements are argued to be merged as specifiers of an optional functional category “Possessor” (cf. Alexiadou & Wilder 1998); and c) demonstratives such as tot/etot ‘this/that’ that also agree with the head noun in Number, Gender and Case, and exhibit an adjectival Case marking paradigm. Initially, we coded and analyzed all these types of elements associated with the DP projection as bearing overt Case. However, since there is no consensus in the literature as to the proper analysis of these elements, we decided to be conservative and included only nouns in our final analysis, as stated in (30): (30) Elements Bearing Case in Russian i. Included in analysis (bearing structural, inherent, or lexical Case and involving a Case-chain between N and D via Num): – nouns; ii. Excluded from analysis (bearing structural, inherent, or lexical Case, but not involving a Case-chain between N and D): – pronouns; iii. Excluded from analysis (bearing Case by agreement): a. adjectives; b. nominal and pronominal possessors; c. demonstratives
We also excluded from our analysis four groups of uninflected nouns, described in (31), three of which are non-existent in adult Russian. These include onomatopoeic nouns, incomplete nouns, non-declinable child-invented forms which cannot be classified within any of the declensions described earlier, and adult like nouns of foreign origin that belong to the fourth declension class and are also non-declinable.
Galina Gordishevsky & Jeannette Schaeffer
(31) Uninflected Nouns (excluded from analysis) a. onomatopoeic forms referring to nouns: mjau (‘cat’); kva (‘frog’); xrju-xrju (‘pig’); igogo (‘horse’); bi(-bi) (‘car’); aljo (‘telephone’); njam-njam (‘food/eat’) b. incomplete nouns, lacking inflectional ending: goljo for golovka (‘head’); tiljat’ for ventiljator (‘fan’); pizi for puzyr (‘bubble’); dok for doctor (‘doctor’) c. non-declinable child-invented nominal forms: bababu for buterbrod (‘sandwich’); mimi for Michelle (child’s name); soso for soska (‘pacifier’) d. non-declinable adult nouns (declension class IV): kofe (‘coffee’); kino (‘cinema’); pjero (name of book character); tabis (‘Teletubbies’); barbi (‘Barbie’)
Moreover, we made the distinction between singular and plural nouns, as in (32). The physical situation surrounding the conversation as well as the utterances preceding and following the target utterance were taken into account in determining the plurality or singularity of the target form. The verb form was also taken into consideration where relevant. Furthermore, distinction was made between structural (namely, nominative, accusative, and genitive), inherent (namely, dative and instrumental), and lexical (accusative, genitive, dative, instrumental, and prepositional) Cases. (32) Number and Case Distinctions for analysis i. Number: singular vs. plural; ii. Case: (a) structural (NOM, ACC, GEN); (b) inherent (DAT, INSTR); (c) lexical (ACC, GEN, DAT, INSTR, PREP)
All instances of vocative Case were included in the nominative group, since vocative Case endings are identical to nominative ones, as explained in section 2.1. We treated the dative Case licensed by the preposition k (‘to’) as inherent Case, since complements of this preposition bear the theta-role of goal (see section 2.1). Consequently, all the instances of dative Case licensed by this preposition were included in the inherent Case group. We excluded from our counts homophonous Case forms (see Table 1), as they could not unambiguously point to correct Case licensing. Of such forms, our child data contained only the forms that were ambiguous between nominative and accusative Case. Finally, we did not include the genitive of negation in our analysis, since all the 63 instances of the genitive of negation we found at both developmental stages appeared only in one construction, namely as subjects of the negated form of verb ‘be’. As hinted at in section 2.4, this construction might be analyzed by children as lexical genitive Case licenser rather than structural Case licenser. Due to lack of consensus on the status of these forms, they were excluded from our analysis.
Case and Number in early Russian
5. Results As a first step, we counted all singular and all plural noun forms in the two developmental stages outlined in section 4, regardless of their obligatory contexts.5In Stage 1, we obtained 1978 (=91%) singular forms and 183 (=9%) plural forms, and in Stage 2 – 1703 (=87%) singular forms and 261 (=13%) plural forms. Furthermore, out of the 54 different word types used in plural in Stage 1, 35 (=65%) consist of frequently used plurals that might be rote-learned by the children. The rest 19 (=35%) word types represent productive plurals that might not be rote-learned. For a more elaborate description of these forms the reader is referred to the discussion section. In Stage 2, out of the 85 word types used in plural, 38 (=45%) represent frequently used rote-learned plurals, and 47 (=55%) – non-rote learned productive forms. Next, we analyzed Case forms and their relation to Number in the two developmental stages. In both stages, we found very high percentages of correct performance on all Cases in the singular. The results are summarized in Tables 4–9, including the individual data and the total numbers and percentages for the three children. Table 4 shows that in developmental Stage 1, the children score 99% (1287/1303) correct on NOM, 88% (278/315) correct on structural ACC, and 88% (51/58) Table 4. Structural Cases – Stage 1 – singular
Case
Child
NOM
ACC*
GEN
total
ZLA MIC KAT Total
100% (534/534) 97% (515/529) 99% (238/240) 99% (1287/1303)
86% (114/133) 90% (138/154) 93% (26/28) 88% (278/315)
90% (27/30) 84% (21/25) 100% (3/3) 88% (51/58)
97% (675/697) 95% (674/708) 99% (267/271) 96% (1616/1676)
*in ACC - excluded: 51 ambiguous NOM/ACC forms
Table 5. Structural Cases – Stage 2 – singular
Case
Child
NOM
ACC*
GEN
total
ZLA MIC KAT Total
100% (502/503) 100% (272/272) 100% (188/188) 100% (962/963)
91% (99/109) 95% (72/76) 91% (71/78) 92% (242/263)
0% (0/1) 100% (2/2) 100% (2/2) 80% (4/5)
98% (601/613) 99% (346/350) 97% (261/268) 98% (1208/1231)
*in ACC - excluded: 106 ambiguous NOM/ACC forms
5. Most singular forms were used correctly in singular contexts: Out of the 1978 singular forms in Stage 1, 17 were erroneously produced in plural contexts, while in Stage 2, 7 out of 1703 singular forms were used instead of a plural form. See Table 14 and surrounding text for a discussion of errors in the plural. As for the plural forms: none of them was produced in a singular context.
Galina Gordishevsky & Jeannette Schaeffer
Table 6. Inherent Cases – Stage 1 – singular
Case
Child
DAT
INSTR
ZLA MIC KAT Total
97% (30/31) 81% (25/31) 97% (36/37) 92% (91/99)
used once used twice –
Table 7. Inherent Cases – Stage 2 – singular
Case
Child
DAT
INSTR
total
ZLA MIC KAT Total
86% (30/35) 100% (10/10) 94% (15/16) 90% (55/61)
– 100% (5/5) 100% (5/5) 100% (10/10)
86% (30/35) 100% (15/15) 95% (20/21) 92% (65/71)
Table 8. Lexical Cases – Stage 1 – singular
Case
Child ACC* ZLA MIC KAT Total
GEN
DAT
INSTR
PREP
100% 100% (7/7) 100% (2/2) 100% (2/2) 80% (16/20) (8/8) 80% 89% (16/18) – 82% (9/11) 96% (22/23) (8/10) 100% 100% (7/7) – 100% (2/2) 100% (3/3) (2/2) 90% 94% 100% 87% 89% (18/20) (30/32) (2/2) (13/15) (41/46)
total 90% (35/39) 89% (55/62) 100% (14/14) 90% (104/115)
*in ACC – excluded: 17 ambiguous NOM/ACC forms
correct on structural GEN in the singular. Unsurprisingly, in Stage 2, similarly high results were obtained in the singular, namely 100% (962/963) correct on NOM, 92% (242/263) correct on structural ACC, and 80% (4/5) correct on structural GEN, as shown in Table 5. Furthermore, as shown in Table 6, in developmental Stage 1, singular inherent DAT is 92% (91/99) adult like. Inherent INSTR was used only three times, with two word types: once by one child and twice by another child. We do not draw conclusions concerning the acquisition of this Case based on these rare uses. Table 7 shows similarly high success rates on inherent Cases in the singular in developmental Stage 2: 90% (55/61) correct usage of DAT and 100% (10/10) correct usage of INSTR.
Case and Number in early Russian
Table 9. Lexical Cases – Stage 2 – singular
Case
Child
ACC*
GEN
DAT
ZLA MIC KAT Total
96% (23/24) 96% (22/23) 100% (9/9) 96% (54/56)
94% 100% (32/34) (2/2) 93% 100% (14/15) (4/4) 100% – (7/7) 95% 100% (53/56) (6/6)
INSTR
PREP
total
83% (19/23)** 90% (9/10) 100% (13/13) 89% (41/46)
92% (33/36) 96% (26/27) 100% (11/11) 95% (70/74)
92% (109/119) 95% (75/79) 100% (40/40) 94% (224/238)
*in ACC – excluded: 50 ambiguous NOM/ACC forms **all 4 errors appeared in one file with one word type
Finally, Tables 8 and 9 show that the singular lexical Cases are produced correctly 90% (104/115) and 94% (224/238) of the time in developmental Stages 1 and 2, respectively. These results include only lexical Cases licensed by prepositions, while lexical Cases licensed by verbs were not attested. In contrast, plural nouns show much lower success rates, at least in ACC and GEN. The results for plural are summarized in Tables 10-13. Table 10 shows that Russian children between 1;8 and 2;0 never produce plurals with correct structural ACC or GEN. In contrast, structural NOM is 92% (92/100) correct. This paradox is explained in the discussion section. As is displayed in Table 11, slightly older Russian speaking children (between age 2;1 and 2;6) do somewhat better, namely 20% (1/5) correct on structural ACC (we excluded from our counts 107 forms that were ambiguous between NOM and ACC), and 17% (1/6) correct on structural GEN. Again, structural NOM is mostly correct (97.5% (117/120)). Tables 12 and 13 contain the results of the correct use of inherent and lexical Cases in the plural. Thus, in Stage 1 (Table 12), plural inherent INSTR was used four times by one child, but only with one word type, and 33% (1/3) correct by another child. Plural inherent DAT is absent in our data. Lexical Cases also show 33% (3/9) correct rate on the overall. We also attested 4 nouns, whose target forms were plural Cased non-NOM forms (but unclear which Case exactly – the context did not provide enough clues), which surfaced as NOM plural or singular forms. These forms were not included in the table, but are included in the total number of plural nominative forms. In Stage 2 (Table 13), plural inherent INSTR was 100% (2/2) correct, while plural inherent DAT – 67% (2/3) correct. Lexical Cases show a similar success rate, namely 68% (13/19) on the overall. In total, we obtained 29 Case (and Number) errors in plural in Stage 1, and 19 such errors in Stage 2. The errors mainly consisted of substitutions with both NOM plural
Galina Gordishevsky & Jeannette Schaeffer
Table 10. Structural Cases – Stage 1 – plural
Case
NOM
ACC*
GEN
Total
92% (92/100)
0% (0/7)
0% (0/2)
*in ACC – excluded: 50 ambiguous NOM/ACC forms
Table 11. Structural Cases – Stage 2 – plural
Case
NOM
ACC*
GEN
Total
97.5% (117/120)
20% (1/5)
17% (1/6)
*in ACC - excluded: 107 ambiguous NOM/ACC forms
Table 12. Inherent and Lexical Cases – Stage 1 – plural
Case
inherent INSTR
inherent DAT
lexical Cases*
Total
33% (1/3) + used four times by one
–
33% (3/9)
comments
child with only one word type
*in ACC – excluded: 21 ambiguous NOM/ACC forms
and NOM singular forms, and to a much lesser degree of substitutions with non-NOM singular forms and non-NOM plural forms (only in Stage 2), as indicated in Table 14. As table 14 shows, the children in the younger age group substituted NOM plural and NOM singular forms for the target non-NOM plural forms in 42% (12/29) and 48% (14/29) of the contexts, respectively. The children in the older age group made such substitutions in 47% (9/19) and 16% (3/19) of the cases, respectively. A small proportion of substitutions consisted of non-NOM singular and plural forms, namely 10% (3/29) non-NOM singular in the younger group, and 21% (4/19) non-NOM singular and 16% (3/19) non-NOM plural in the older group. These misuses of singulars in place of plurals, as well as of nominative plural in place of other plurals, present the most relevant statistics. These are the errors that show children actually lack some knowledge, rather than just avoid the use of plurals.6
6. We thank an anonymous reviewer for directing our attention to this observation.
Case and Number in early Russian
Table 13. Inherent and Lexical Cases – Stage 2 – plural
Case
inherent INSTR
inherent DAT
lexical Cases*
Total
100% (2/2)
67% (2/3)
68% (13/19)
total inherent Cases: 80% (4/5)
*in ACC – excluded: 6 ambiguous NOM/ACC forms
Table 14. Types of errors (substitutions) in plural
NOM Case
non-NOM Case
plural
singular
plural
singular
Stage 1 Stage 2
42% (12/29) 47% (9/19)
48% (14/29) 16% (3/19)
– 16% (3/19)
10% (3/29) 21% (4/19)
6. Discussion Our results are summarized in (33): (33) Summary of results Russian-speaking children aged 1;8–2;0 (Stage 1) a. show very high percentages of correct performance on all Cases in the singular, namely 96% on structural Cases, 92% on inherent Cases, and 90% on lexical Cases; b. show much lower success rates with plural nouns, namely 0% on structural ACC and GEN, 33% on inherent Cases and lexical Cases, replacing the Case-marked forms with default NOM forms. c. produce few productive plural nouns (35% out of all plurals used). d. substitute plural Cased non-NOM nouns with singular forms 58% of the time and with plural forms in the default NOM Case in 42% of the cases. Russian-speaking children aged 2;1-2;6 (Stage 2) a. show very high percentages of correct performance on all Cases in the singular, namely 98% on structural Cases, 92% on inherent Cases, and 94% on lexical Cases. b. show higher success rates with plural nouns compared to the younger children, but are far from adult like (20% on structural ACC and 17% on structural GEN, 80% on inherent Cases and 68% on lexical Cases). c. produce slightly more productive plural nouns than the younger children (55% out of all plurals used). d. substitute plural Cased non-NOM nouns with singular forms in 37% of the cases and with default NOM plural forms 47% of the time.
Galina Gordishevsky & Jeannette Schaeffer
First of all, the high percentages of correct Case licensing in the singular in both Stage 1 and Stage 2 confirm our Prediction 1a, repeated in (34) for convenience: (34) Prediction 1a Young Russian-speaking children produce Case correctly in the singular. This supports our Hypothesis 1, as repeated in (35): (35) Hypothesis 1 All functional categories responsible for Case licensing are present in the child grammar from very early on.
Our findings resemble the results obtained by Babyonyshev (1993) in that we found high percentages of correct Case inflection in the structural and inherent Cases. However, we did not find any difference between lexical Cases on the one hand and the structural and inherent Cases on the other, as Babyonyshev did. Nevertheless, the high percentages of correct performance on all Cases in the singular obtained in this study present additional counter-evidence against the lexical-thematic analysis of Radford (1986; 1990), according to which functional categories are absent from early child grammar and only lexical categories are present. Similar to Babyonyshev’s findings, our results lend support to the Full Clause Hypothesis (Hyams 1992; Wexler 1992; Poeppel & Wexler 1993), according to which functional categories are present in the child’s grammar from the beginning. Second, the finding that Case licensing in plural is poor, especially in Stage 1, lends support to our Prediction 2, repeated in (36): (36) Prediction 2 Young Russian-speaking children fail to use Case correctly in the plural.
As shown in the results section, the observed errors mainly consisted of substitutions with both NOM singular and NOM plural forms, and hardly of substitutions with non-NOM forms (only a few in Stage 2). As outlined in section 2.3, in order for Case licensing to take place, a DP internal chain between N and D has to be established. This chain is mediated by the Num head. If the Num head is specified as [+singular] only, it breaks the chain in the case of a plural noun. Consequently, the Case feature cannot reach the DP level, and therefore cannot be licensed. We propose that, in order to avoid crashing of the DP, young children’s grammars assign default Case in such situations, which is nominative in Russian. This is illustrated by the so-called ‘Who wants…’ test, as illustrated in (37). As opposed to English, which renders an answer with ACC Case (‘me’, ‘us’), the correct Case in Russian is NOM:7 (37) Nominative – default Case in Russian Q: Kto xochet moroženoe? who wants ice-cream ‘Who wants an ice-cream?’
7. Although we decided to leave pronouns out of our analysis for theoretical reasons (they do not involve an N-D Case-chain, because they are base-generated in D - cf. Adger 2003: 256), we
Case and Number in early Russian
A: a. Ja. b. My. I-NOM we-NOM ‘I.’ ‘We.’
Our developmental Stage 1, and to a slightly lesser extent Stage 2, correspond to the developmental stage reported by Müller (1994), at which children do not mark Number distinctions in German and French (the children in her study were 1;6–3;0 and 1;5–3;0). Our findings support Müller’s (1994) and Hoekstra and Hyams’ (1995) Underspecification Hypothesis, according to which the nominal Number head is initially underspecified and represents the unmarked [+singular] feature only. This was our Hypothesis 2, repeated in (38): (38) Hypothesis 2 The Number head is initially underspecified in child grammars and represents [+singular] only.
Finally, close examination of the children’s plural forms suggests that plural is initially not productive, strengthening Hoekstra & Hyams’ (1995; 1996) Avoid Plural Phenomenon, discussed in section 3. As mentioned in the results section, in Stage 1, the numbers of plural forms are extremely low: 183 plural forms vs. 1978 singular forms. Moreover, out of the 54 different word types used in plural, 35 (65%) represent pairwise (‘hands’, ‘knees’, ‘lips’, ‘eyebrows’) or plural (‘fingers’, ‘teeth’) body parts, types of footwear (‘slippers’, ‘boots’, ‘sneakers’, ‘socks’), pluralized nouns (‘pants’, ‘glasses’), plural non-count nouns (‘beads’, ‘money’, ‘hair’, ‘cereal’, which are all plural in Russian), and entities that usually occur in groups (‘toys’, ‘blocks’, ‘colours’, ‘tears’). These forms are frequently used plurals and might be rote-learned by the child. Only 19 (35%) word types used in plural represent forms that could not possibly have been memorized, such as ‘plates’, ‘houses’, ‘presents’, ‘boys’, ‘dogs’, ‘tigers’. In addition, almost none of the plurals occur in their singular form in the same transcripts. All these observations suggest that plural is unproductive at this stage (Prediction 3, repeated in (39)), and provides additional evidence for the hypothesis that the Number feature is initially underspecified.
did code for them. Unlike their English speaking peers, who produce phrases such as ‘me have it’, or ‘her sleeping’ (Rispoli 1994; Vainikka 1994; Schütze 1997), Russian speaking children’s subject pronouns always correctly bear NOM Case. This follows from the hypothesis that if for some reason NOM Case cannot be checked, the element appears with default Case. As the default Case is ACC in English, but NOM in Russian, early English subject pronouns sometimes occur with ACC Case (and occasionally with GEN), but early Russian pronouns always appear as NOM. Interestingly, the majority of non-NOM Case subject pronouns in English child language occur with so-called ‘Optional Infinitives’, which is why they are argued to result from underspecified Agreement and/or Tense (Schütze 1997).
Galina Gordishevsky & Jeannette Schaeffer
(39) Prediction 3 Initially plural is unproductive in child Russian, i.e., plural forms are rarely used compared to the singular and the attested plural forms are frequently used plurals that may be rote-learned.
This finding places the children of the younger age group within the second developmental stage outlined in Zapf & Smith (2003) (see section 2.4). A detailed analysis of the children’s plural forms in Stage 2 suggests that plural gradually becomes productive. Although the children in Stage 2 are still far from adult like mastery of the plural, development towards the target grammar is observed. Thus, while the numbers of plural forms used are still low compared to singular ones, with only a slight increase in the numbers (261 plural forms vs. 1703 singular forms), the proportion of productive use of plural is much higher compared to Stage 1. Recall that in Stage 1 the so-called rote-learned nominal forms comprised 65% (35/54) of all plural nouns, with only 35% (19/54) word types representing non-rote learned forms. In Stage 2, on the other hand, the children used 85 word types in plural, out of which 38 (45%) belonged to allegedly rote-learned word groups described above, and 47 (55%) represented non-rote learned forms, such as ‘soldiers’, ‘circles’, ‘pockets’, ‘sticks’, ‘pictures’, ‘pigeons’, ‘dolphins’, ‘bunnies’, ‘daughters’, to mention just a few. Finally, increased numbers of nouns started to occur in their singular form in the same transcripts where they appeared in plural. All these factors unequivocally point to gradual development towards target like mastery of plural forms, i.e., acquisition of the specification of Number. The stage at which children exhibit increased production with errors still present (in particular, overgeneralization errors in English) has been described by Zapf & Smith (2003) as a stage intermediate between the rote-learned stage (at which the children aged 1;8–2;0 in our study are) and the stage at which (the English) plural is mastered. The results we obtained for the older group of children (aged 2;1–2;6) place them within this particular developmental stage, at which they are already beyond the rote-learned stage, but have not reached adult like mastery of the plural yet. One might wonder whether the children’s non-adult like behaviour with respect to plural Case results from a phonological deficit, namely an incapability to produce codas, i.e., syllable-final consonants.8 An examination of the Russian Case paradigm in Table 1 reveals that some, but not all, plural endings contain a coda. For example, the plural PREP ending is -ax in all declensions, while the plural INSTR ending of all declensions is -(a)mi. Our findings show that the children do not succeed even at those non-NOM plural forms that do not require a coda, such as INSTR -(a)mi ending, ACC and GEN -ej ending. Moreover, children do produce
8. Thanks to Outi Bat-El who raised this issue at IATL20.
Case and Number in early Russian
codas (e.g., on nouns in NOM, GEN and INSTR) from the earliest ages observed, as the following examples show: (40) Syllables containing a coda a. acik dusim boy-NOM shower-INSTR ‘The boy is under the shower.’ b. asos’kav socks-dimin-GEN c. masil marker-NOM
(ZLA 1, age 1;8)
(ZLA 3, age 1;10) (MIC 4, age 1;11)
Thus, the evidence just presented refutes a phonological explanation of erroneous Case marking in the plural.
7. Conclusion To conclude, in this study we have shown that while young Russian-speaking children’s performance is adult like on Case marking on singular nouns, they have difficulty with Case marking in the plural. We argued that this difficulty stems from the underspecification of the Number feature in early grammar. Underspecified Number, i.e., a default specification of [+singular], breaks the Case-chain in case of plurals, preventing the Case feature from getting to the DP level where it can be licensed. The result is two-fold: a) failure to use correctly Cased plural forms and their replacement mainly with default nominative forms, and b) avoidance of productive plural forms. From age 2 on, Russian-speaking children start to use plural forms more productively, suggesting that the functional feature Number is beginning to get specified for plural as well. Consequently, the number of plurals with correct Case is higher. Our findings lend support to the Full Competence Model, namely that functional categories, including those responsible for Case licensing, such as Number, are present from the beginning, as evidenced by the high percentages of correct Case in the singular. Our results regarding plural nouns provide evidence in favour of the Underspecification Hypothesis which states that certain functional categories, such as Number, although present, are initially not fully specified. We are left with a number of open research questions. The first one is how the children would behave with respect to plurals in a structured experimental task that elicits plurals. In a controlled task a higher number of tokens can be elicited than in a free-speech sample, thus yielding more reliable results. The second question concerns further development of the plural feature and its interaction with Case after the age of 2;6 and the point at which the full convergence on the target grammar is achieved. We leave these issues open for future research.
Galina Gordishevsky & Jeannette Schaeffer
References Adger, D. 2003. Core Syntax: A Minimalist Approach. New York NY: Oxford University Press. Alexiadou, A. & Wilder, C. (eds). 1998. Possessors, Predicates and Movement in the DP. Amsterdam: John Benjamins. Babyonyshev, M. 1993. Acquisition of the Russian Case system. In MIT Working Papers in Linguistics 19 [Papers on Case & Agreement 2], C. Phillips (ed.), 1–43. Cambridge MA: MIT. Babyonyshev, M. 1997. Case in Early Russian. Ms. Bailyn, J.F. 1995. A Configurational Approach to Russian ‘Free’ Word Order. PhD dissertation, Cornell University. Brecht, R.D. & Levine, J.S. 1986. Case and meaning. In Case in Slavic, R.D. Brecht & J.S. Levine (eds), 17–34. Columbus OH: Slavica. Chomsky, N. 1998. Minimalist Inquiries. MIT Working Papers in Linguistics 15. Chomsky, N. 1999. Derivation by Phase. MIT Working Papers in Linguistics 18. Chomsky, N. 2001. Beyond Explanatory Adequacy. MIT Working Papers in Linguistics 20. Chomsky, N. & Lasnik, H. 1993. The theory of principles and parameters. In An International Handbook of Contemporary Research, J. Jacobs, A. von Stechow, W. Sternefeld & T. Vennemann (eds.), 506-569. Berlin: Walter de Gruyter. Clahsen, H., Eisenbeiss, S. & Penke, M. 1994. Underspecification and lexical learning in early child grammars. Essex Research Reports in Linguistics 4. Comrie, B. 1986. On delimiting cases. In Case in Slavic, R.D. Brecht & J.S. Levine (eds.), 86-106. Columbus OH: Slavica. Ferdinand, A. 1994. On the development of the verbal system in French. Paper presented at the 18th Annual Boston University Conference on Language Development. Franks, S. 1995. Parameters of Slavic Morphosyntax. Oxford: Oxford University Press. Gordishevsky, G. 2006/2007. Subject Omission in Adult and Child Russian. PhD dissertation, Ben-Gurion University of the Negev. Grinstead, J. 1994. Consequences of the maturation of number morphology in Spanish and Catalan. MA thesis, UCLA. Halle, M. 1994. The Russian declension: An illustration of the theory of distributed morphology. In Perspectives in Phonology [Lecture Notes 51], J. Cole & C. Kisseberth (eds), 29-61. Stanford CA: CSLI. Hoekstra, T. & Hyams, N. 1995. The syntax and interpretation of dropped categories in child language: A unified account. In Proceedings of WCCFL XIV, Stanford University. Hoekstra, T. & Hyams, N. 1996. Missing heads in child language. In Proceedings of GALA 1995, C. Koster & F. Wijnen (eds.), 251-260. Groningen: Centre for Language and Cognition. Hoekstra, T. & Hyams, N. 1997. Agreement and the finiteness of V2: Evidence from child language. In Proceedings of BUCLD 22, A. Greenhill, M. Hughes, H. Littlefield & H. Walsh (eds.), 360-374. Somerville MA: Cascadilla. Hoekstra, T. & Hyams, N. 1998. Aspects of root infinitives. Lingua 106: 81-112. Hyams, N. 1992. The genesis of clausal structure. In The Acquisition of Verb Placement, J. Meisel (ed.), 371-400. Dordrecht: Kluwer. Hyams, N. 1996. The underspecification of functional categories in early grammar. In Generative Perspectives on Language Acquisition, H. Clahsen (ed.), 91-127. Amsterdam: John Benjamins. King, T.H. 1995. Configuring Topic and Focus in Russian. Stanford CA: CSLI. Koopman, H. 1999. The internal and external distribution of pronominal DPs. In Beyond Prin-
Case and Number in early Russian ciples and Parameters, K. Johnson & I. Roberts (eds), 91-132. Dordrecht: Kluwer. MacWhinney, B. 2000. The CHILDES project: Tools for analyzing talk. 3rd edn. Mahwah NJ: Lawrence Erlbaum Associates. Müller, N. 1994. Gender and number agreement within DP. In Bilingual First Language Acquisition, J. Meisel (ed.), 53–88. Amsterdam: John Benjamins. Pizzuto, E. & Caselli, M.C. 1992. The acquisition of Italian morphology: Implications for models of language development. Journal of Child Language 19(3): 491–557. Poeppel, D. & Wexler, K. 1993. The full competence hypothesis of clause structure in early German. Language 69(1): 1–33. Radford, A. 1986. Small children’s small clauses. In Research Papers in Linguistics 1: 1–44. University College of North Wales, Bangor. Radford, A. 1990. Syntactic Theory and the Acquisition of English Syntax: The Nature of Early Child Grammars of English. Oxford: Blackwell. Radford, A. & Ramos, E. 2001. Case, Agreement and EPP: Evidence from an English-speaking child with SLI. Essex Research Reports in Linguistics 36: 42–81. Rappaport, G.C. 2002. Numeral phrases in Russian: A minimalist approach. In Russian Morphosyntax: A Festschrift for Leonard H. Babby, G. Greenberg & J.S. Levine (eds.), Journal of Slavic Linguistics 10(1/2): 329–42. Rispoli, M. 1994. Pronoun case overextensions and paradigm building. Journal of Child Language 21: 157–172. Ritter, E. 1991. Two functional categories in noun phrases: Evidence from Modern Hebrew. In Perspectives on Phrase Structure: Heads and Licensing [Syntax and Semantics 25], S. Rothstein (ed.), 37–62. San Diego CA: Academic Press. Ritter, E. 1995. On the syntactic category of pronouns and agreement. Natural Language and Linguistic Theory 13: 405–443. Schütze, C. 1997. INFL in child and adult language: Agreement, Case and Licensing. PhD dissertation, MIT. Vainikka, A. 1994. Case in the development of English syntax. Language Acquisition 3: 257–325. Wexler, K. 1992. Optional Infinitives, head movement and the economy of derivation in child grammar. In Occasional paper 45. Cambridge MA: Center for Cognitive Science, MIT. Zapf, J.A., & Smith, L.B. 2003. The protracted course of the acquisition of the plural. In BUCLD 27 Proceedings, B. Beachley, A. Brown & F. Conlin (eds), 834–845. Somerville MA: Cascadilla.
Verb movement and subject placement in the acquisition of word order Pragmatics or structural economy? Marit R. Westergaard This chapter reports on a study investigating three children acquiring a dialect of Norwegian (Tromsø). It is shown that, although verb-second (V2) word order is attested early in non-subject-initial declaratives and questions, the children produce non-target-consistent word order in these constructions at an early stage, failing to move pronominal subjects past negation or an adverb. This is analyzed within a split-CP model which also assumes the existence of two subject positions in the IP domain, a high one for informationally given subjects and a lower one for subjects conveying new information. A pragmatic account of the child data is considered, but rejected in favour of an economy-based analysis which includes factors such as input frequency and complexity.
1. Introduction In the literature on the acquisition of verb movement and verb-second (V2) word order in Germanic languages, sentences such as (1) and (2) have been used to argue for different theories of language development. Clahsen, Penke & Parodi (1993/94) used the verb-initial topicalization construction in (1) to argue for a stage in the language acquisition process of German where certain functional projections are missing. More specifically, within a weak continuity structure building approach, they claimed that children go through a stage where their clause structure includes an IP, but no CP projection. Under the assumption that negation is attached to the VP, the finite verb in (1) has moved to the head position of the only functional projection available. Consequently, there is no specifier position for the subject to move to between the finite verb and negation, and the result is therefore a word order where negation precedes the subject (Neg-S).1
1. It should be noted that while the English translation of example (1) is ambiguous between an epistemic and a deontic reading, the German modal verb dürfen only allows a deontic interpretation.
Marit R. Westergaard
(1) darf nich Julia haben. (Mathias, stage II) ((X)-V-Neg-S) may not Julia have ‘Julia may not have that.’ Target form: Das darf Julia nicht haben. (From Clahsen, Penke & Parodi 1993/94: 423)
However, sentences with the opposite word order (S-Neg) have also been attested in non-subject-initial declaratives in early child German. Poeppel & Wexler (1993) found eight such examples (and no counterexamples) in the data of the child Andreas that they investigated, one of which is illustrated in (2).2 Poeppel and Wexler take these examples to be evidence of the continuity approach to language acquisition, where all functional projections (basically IP and CP) are available to the child from the outset of language acquisition. More specifically, they argue that the S-Neg word order of sentence (2) shows that the subject cannot be inside the VP, and thus that there must be two functional specifier positions above VP, one for the subject (SpecIP) and one for the topicalized element (SpecCP).3 (2)
den tiegt a nich wieda.
(Andreas, age 2;1) (X-V-S-Neg)
that gets he not again ‘He won’t get that back.’ Target form: Den kriegt er nicht wieder. (From Poeppel & Wexler 1993:23)
In this chapter I present similar data from Norwegian child language which I analyze within an economy approach to language acquisition that retains aspects of both continuity and structure building. More specifically I argue that young children will move elements only as high up in the structure as there is clear evidence for in the input. In Norwegian, the two word orders in (1) and (2) are both grammatical in the adult language, indicating that there are two subject positions available in such constructions. The choice between the two is related to information structure, the higher one being reserved for informationally given subjects and the lower one for subjects conveying new information. This means that the subject position below negation will be preferred with full DPs, while pronominal subjects normally occur above negation. Since the behaviour of pronominal vs. DP elements in these constructions is similar to that of object shift in Scandinavian, the construction in (2) could be referred to as
2. Poeppel & Wexler (1993) investigated one transcript of Andreas at age 2;1, altogether 282 complete declarative sentences. 3. Poeppel & Wexler (1993) do not discuss the possibility that the pronominal subject in (2) could be a clitic, nor do they provide information about whether the subject in the seven other examples are pronouns or full DPs.
Verb movement and subject placement
‘subject shift’. I will show that children initially prefer the word order in (1) with full DP as well as pronominal subjects, which means that any type of subject may appear in the low subject position at an early stage. The Norwegian child data investigated in this chapter (see section 3.1 below) indicate that a change occurs around the age of 2;6–3;0, leading to a preference for the more target-consistent word order found in (2) with pronominal subjects. This is illustrated by the sentences in (3) and (4) from the Norwegian corpus. (3) nå skal ikke dem sove. now shall not they sleep ‘Now they shouldn’t sleep.’
(Ole.09, age 2;3.15) (X-V-Neg-S)
(4) og no kan æ ikke drikke det. and now can I not drink it ‘And now I can’t drink it.’
(Ole.19, age 2;10.0 ) (X-V-S-Neg)
The chapter is organized at follows: In the next section I give a brief overview of word order in Norwegian and outline the syntactic model that I adopt. Then, in section 3, I present the Norwegian child data, first some general observations about the acquisition of V2, and then an investigation of the children’s production in constructions such as (3) and (4). In section 4, I present two possible accounts of the child data; a pragmatic analysis in terms of a lack of the concept of non-shared knowledge, and a structure building continuity approach which includes a principle of economy of movement. In section 5, I provide several arguments against the pragmatic analysis and discuss some further aspects of the economy-based approach, taking into account factors such as input frequency and complexity. Section 6 provides a summary and conclusion.
2. The word order of Norwegian Like German, Norwegian is a V2 language with the finite verb appearing in second position in all main clauses. This is standardly assumed to be the result of verb movement to the C position of the clause (see e.g., Vikner 1995). Unlike German, however, Norwegian is an SVO language, which means that in subject-initial declaratives, the V2 requirement is only visible if there is a sentence adverb or negation present. This is illustrated in (5), where the finite verb is assumed to have moved across negation. In non-subject-initial declaratives we see verb movement across the subject, as shown in example (6). The V2 requirement generally also holds in main clause questions, which is illustrated by the yes/no-question in (7) and the wh-question in (8).
Marit R. Westergaard
(5) Jeg liker ikke jordbær / *Jeg ikke liker jordbær. (Standard Norwegian) I like not strawberries / I not like strawberries ‘I don’t like strawberries.’ (6) Jordbær liker jeg faktisk /*Jordbær jeg liker faktisk. Strawberries like I actually / Strawberries I like actually ‘Strawberries I actually like.’ (7) Liker du jordbær? like you strawberries ‘Do you like strawberries?’ (8) Hvilken bok likte du best? which book liked you best ‘Which book did you like best?’
However, there are certain clause types in Norwegian which do not display V2, as discussed in Westergaard (2006a). These include exclamatives and embedded questions, where there is no verb movement across the subject, as well as all embedded clauses, where there is no verb movement across adverbs or negation, as illustrated in (9).4 (9) Jeg vet [at han ikke liker jordbær]. I know that he not likes strawberries ‘I know that he doesn’t like strawberries.’
Furthermore, many dialects of Norwegian do not have a strict V2 requirement in whquestions, see e.g., Vangsnes (2004, 2006). In the dialect spoken by the three children in this study (Tromsø), there is a word order distinction based on the length of the whconstituent. As shown in (10) and (11), the finite verb is obligatorily in second position if the question is introduced by full DP wh-constituents or one of the long (disyllabic) question words, korfor, korsen or katti (‘why’, ‘how’ or ‘when’). In contrast, V2 is not required after the monosyllabic wh-words ka, kem and kor (‘what’, ‘who’ and ‘where’), as illustrated by the two possible word orders in (12)–(14). (10) Ka slags bok leste du?/ *Ka slags bok du leste? what kind book read you/ what kind book you read ‘What (kind of) book did you read?’
(Tromsø)
(11) Korfor gikk ho?/ *Korfor ho gikk? why went she / why she went ‘Why did she go?’ (12) Ka legen sa?/ Ka sa legen? what doctor.def said/ what said doctor.def ‘What did the doctor say?’ 4. Embedded V2 is possible, but not preferred, in certain embedded clauses in Norwegian, notably in complements to so-called bridge verbs, see Vikner (1995) and Bentzen (2003, 2005).
Verb movement and subject placement
(13) Kor du bor?/ Kor bor du? where you live/ where live you ‘Where do you live?’ (14) Kem den nye læreren er?/ Kem er den nye læreren? who the new teacher.def is/ who is the new teacher.def ‘Who is the new teacher?’
There does not seem to be any significant difference in meaning between the two word orders in (12)–(14) when the sentences are uttered in isolation. However, in actual discourse, the choice of word order is not random. In Westergaard (2003), a sample of spontaneous speech from an adult in a corpus of child language was investigated, and it was argued that the choice of the two word orders is dependent on the information structure of the sentence, more specifically on whether the subject conveys given or new information in the discourse. This was reflected in certain statistically significant patterns for subject and verb types in the two constructions, V2 being preferred with full DP subjects and the verb være ‘be’, while non-V2 is chosen with pronominal subjects and any other verb. Some typical examples from the corpus are provided in (15) and (16). In the terminology used in Westergaard (2005a), V2 word order is preferred when the subject is characterized as [+focus] (informationally new) while non-V2 is used when the subject is [-focus] (i.e., given information). (15) kor er mitt fly? where is my plane ‘Where is my plane?’
(INV, file Ole.17)
(16) kor vi lande henne? where we land loc ‘Where should we land?’
(INV, file Ole.17)
In Westergaard (2005a), the two word orders of wh-questions in the Tromsø dialect are accounted for within a type of Split-CP approach to clause structure. This is inspired by the Split-CP model of Rizzi (1997, 2001), according to which the CP domain consists of a high ForceP(hrase) and a low Fin(iteness)P(hrase) with several functional heads in between expressing aspects such as topic and focus. The present approach is in many ways different from this; most notably, the ForceP of Rizzi’s system is divided into different functional heads for individual clause types; e.g., an Int(errogative) head is present in wh-questions, a Pol(arity) head is present in yes/no-questions, a Top(ic) head is present in main clause declaratives, while embedded declaratives are bare FinPs. Thus the illocutionary force of a clause is reflected in the topmost head. Some of the heads in this model that are relevant to the data discussed in this chapter are provided in (17): (17)
CP[ ForceP[(Int°
. . . Top° . . . Pol° . . .) . . . Foc° Top° . . . Fin° IP[ . . .
Marit R. Westergaard
V2 word order in this model is the result of an EPP head feature on syntactic heads in the CP domain, which attracts the verb.5 This means that there may be several sources for V2 word order, and this accounts for different V2 grammars. In Westergaard (2006a) I show that V2 grammars may differ with respect to the clause types which require V2. For example, unlike Standard Norwegian, Danish has V2 in exclamatives, Belfast English allows V2 in embedded questions, and Icelandic displays verb movement in embedded clauses. In order to explain the two possible word orders in wh-questions with monosyllabic question words, Westergaard (2005a, b) refers to the Head Preference Principle of van Gelderen (2004) and argues that in the Tromsø dialect, the monosyllabic wh-words are heads and may therefore in themselves fulfill the EPP head requirement on Int°, which then renders non-V2 word order possible. When V2 does occur in wh-questions, this is argued to be the result of the requirement of a lower head in the CP domain, e.g., the lower Top° head of Rizzi’s (1997) system, see Westergaard (2006a).6 This projection is argued to attract elements with low information value, [-foc], which means that a subject conveying given information may be attracted to the specifier position of this head, and when the subject is [+foc], it is the verb which moves, yielding V2 word order.7 This then ensures that elements that convey new information stay in lower positions when other elements move up, thus serving as a syntactic spell-out of the well-known pragmatic principle of end focus (see e.g., Firbas 1992). Finally in this section, we consider the clause types that were discussed in the Introduction, the subject shift constructions. In questions and non-subject-initial declaratives with V2, where there is verb movement across the subject (to the CP domain), two different subject positions become visible if the sentences also contain negation or a sentence adverb. That is, the subject may appear either preceding or following the adverb or negation. Examples illustrating this in non-subject-initial declaratives are given in (18), while similar examples are provided for yes/no-questions in (19). (18) a. b.
Dette kan ikke Peter gjøre. this can not Peter do Dette kan Peter ikke gjøre. this can Peter not do ‘This Peter can’t do.’
5. This abbreviation refers to the Extended Projection Principle (EPP) of earlier versions of generative theory (originally from Chomsky 1982), which ensured that all clauses have a subject. Within the Minimalist model, e.g., Chomsky (1995), an EPP feature on a syntactic head will require that this head projects a specifier in order for the uninterpretable EPP feature to be deleted. 6. In Westergaard & Vangsnes (2005) and Westergaard (2005) the head responsible for the word order in wh-questions was argued to be the Foc° head. 7. Recall that the verb in these V2 constructions is often være ‘be’, which is inherently informationally light.
Verb movement and subject placement
(19) a. b.
Kan ikke Peter gjøre dette? can not Peter do this Kan Peter ikke gjøre dette? can Peter not do this ‘Can’t Peter do this?’
As mentioned in the Introduction, this means that the two word orders attested in German child language in (1) and (2) are both grammatical in Norwegian. In Westergaard & Vangsnes (2005), the two subject positions are argued to be distinguished by the information value of the subject; given subjects occurring before and new or focused subjects following negation or a sentence adverb. In accordance with this, full DP subjects will tend to follow adverbs/negation in these constructions, although they may precede if they are already given in the discourse. When the subject is a pronoun, on the other hand, the position preceding negation will be preferred, as shown in (20). In fact, the Neg-S word order is extremely dispreferred for pronominal subjects unless they are stressed, as illustrated by (21). (20) a. b.
Dette kan du ikke gjøre. this can you not do Kan du ikke gjøre det? can you not do this
(21) a. *?Dette kan ikke du gjøre. b. Dette kan ikke DU gjøre.
Syntactically, the two subject positions are in the IP domain of the clause, and in Westergaard (2005) it is argued that the low position, for informationally new or [+foc] subjects, is the specifier of the T(ense)P(hrase). Informationally given or [-foc] subjects, on the other hand, must move to the specifier of a higher projection in the IP domain, the In(ner)Top(ic)P(hrase). Negation and sentence adverbs are adjoined to TP and thus appear between the two subject positions. This is illustrated in (22). It could be noted that according to Frey (2004), there are two subject positions in the IP domain also in German (the higher one called a medial topic position), but these are only visible in embedded clauses. Likewise, van Kemenade & Los (2006) show that certain discourse adverbs distinguish between given and new subjects in embedded clauses in Old and Middle English. (22)
InTopP[
DP[-foc] TP[ Sentence Adverbs/Negation TP[ DP[+foc] . . .
The word order facts surveyed in this section show that children who are exposed to the Tromsø dialect of Norwegian have to acquire a V2 grammar with certain exceptions, most notably the ‘optional’ V2/non-V2 in wh-questions, where the variation is dependent on the information value of the subject as [+/-foc]. Likewise, the choice between the two subject positions in V2 clauses is dependent on the same distinction, although syntactically the subject positions are in the IP domain.
Marit R. Westergaard
In the next section, we turn to the child data on word order, focusing on verb movement and V2, as well as the two different subject positions in V2 clauses.
3. The child data 3.1 The corpus The corpus used for this study was collected in Tromsø in 1997/98. Three children, Ina, Ann, and Ole, who were around 1 year 9 months at the start of data collection, were recorded approximately every two weeks until the age of three. The corpus consists of altogether 70 recorded sessions, most of them lasting about an hour. Table 1 gives an overview of the corpus, specifying the age of the three children and the total number of files and child utterances.8 Table 1. Overview of the Norwegian corpus of child language, Tromsø dialect Name of child
Age
Files
Ina 1;8.20–3;3.18 Ina.01–27 Ann 1;8.20–3;0.1 Ann.01–21 Ole 1;9.10–2;11.23 Ole.01–22 Total
No of child utterances 20,071 13,129 13,485 46,685
3.2 The acquisition of verb movement and V2 In other acquisition studies of word order in V2 languages, e.g., Jordens (1990) for Dutch, Poeppel & Wexler (1993) for German, Santelmann (1995) and Platzack (1996) for Swedish, V2 is attested from the earliest occurrences of multi-word utterances. This is also the case in the data from the three children in the Norwegian corpus – as illustrated in (23)–(26), verb movement is attested across negation in subject-initial declaratives, and across the subject in wh-questions, yes/no-questions and non-subject-initial declaratives. (23) æ vet ikkje. I know.pres not ‘I don’t know.’ (24) kor e babyen? where be pres baby.def ‘Where is the baby?’
(Ann.02, age 1;9.18)
(Ina.06, age 2;1.0)
8. Apart from 10 files that have been collected and transcribed by the author, the corpus has been collected by Merete Anderssen. (see Anderssen 2006)
Verb movement and subject placement
(25) er det båt? be.pres it boat ‘Is that (a) boat?’
(Ina.03, age 1;10.23)
(26) der er stor stor Ole Brumm. there be.pres big big Ole Brumm ‘There is (a) big big Winnie the Pooh.’
(Ann.01, age 1;8.20)
Furthermore, the children were found to produce wh-questions with both V2 and nonV2 from a relatively early stage. These questions also appear with the same preference patterns for subject and verb types as in the adult grammar. That is, while V2 word order tends to occur with the verb være ‘be’ and a full DP subject, non-V2 constructions are preferred with all other verbs and pronominal subjects, as illustrated in (27) and (28), cp the examples in (15) and (16) which were produced by an adult. As in the adult data, there is no one-to-one correspondence between subject type (pronoun vs. full DP) and word order, indicating that children are not simply making a distinction between pronouns and full DPs. In Westergaard (2003, 2005) it was argued that this suggests that children are sensitive to patterns of information structure from a very early age and that they make a syntactic distinction between informationally given subjects and subjects conveying new information. (27) kor er Ann sin dukke hen? where is Ann poss doll loc ‘Where is Ann’s doll?’ (28) ka du skal finne? what you shall find ‘What do you want to find?’
(Ann.04, age 1;11.0)
(Ina.5, age 2;0.5)
As we saw in section 2, there is no V2 in embedded clauses in Norwegian. Although there are very few embedded clauses in the corpus, due to the young age of the children, certain cases of overgeneralization of V2 word order are attested, as illustrated in example (29), where the verb precedes the adverb også ‘also’. This word order is also found in other non-V2 constructions, as in the wh-question in (30).9 Similar examples are attested in the production of somewhat older Norwegian children in Bentzen (2003).10 It is notable that this type of overgeneralization only affects verb movement across negation or an adverb, never across the subject (e.g., in embedded questions), see Westergaard & Bentzen (2007). 9. This is a subject question, which in the target grammar (the Tromsø dialect) requires the presence of the relative complementizer som in second position. Thus, there is no verb movement to the CP domain in these cases. 10. Even more examples of this kind are elicited in a production experiment described in Westergaard & Bentzen (2007), where it is suggested that children possibly do not “unlearn” this word order until after the age of 6.
Marit R. Westergaard
(29)
det er ho mamma som har også tegna. (Ina.26, age 3;2.05) it be.pres det mommie who have.pres also draw.part ‘It is mommie who has also drawn.’ Target form: Det er ho mamma som også har tegna.
(30)
kem som vil ikkje være ilag med han? who that will not be together with him ‘Who doesn’t want to be with him?’ Target form: Kem som ikkje vil være i lag med han?
(Ina.25, age 3;1.8)
In the next section I turn to the clause types where there is verb movement across the subject (to the CP domain), viz. non-subject-initial declaratives and questions. When these clauses also contain negation or an adverb, the two subject positions in Norwegian become visible, as was illustrated in examples (18)–(21) in section 2. The child data shows that children initially have certain word order problems in these subject shift constructions.
3.3 Non-subject-initial declaratives and questions with negation/adverbs In this section, I present the Norwegian child data on non-subject-initial declaratives and questions containing negation or an adverb – i.e., sentences of the type illustrated in (1) and (2) from German child language. There are a total of 213 examples of these clause types attested in the Norwegian corpus as a whole: Adding up both declaratives and the two question types, the child Ann produces 43, Ole 69, and Ina as many as 101 examples of such sentences.11 The vast majority of the examples contain the negative element ikkje/ikke ‘not’, while sentence adverbs are only occasionally attested. Most of these sentences occur with pronominal subjects. In fact, only 7 of Ina’s, 9 of Ann’s and 6 of Ole’s sentences of this type have a full DP subject, and all of these 22 examples occur with the expected word order, with the subject following negation. This is illustrated in (31)–(36), examples which have been selected from relatively early as well as late files in the corpus. (31) der snakke ikkje mannen. there speak.pres not man.def ‘There the man doesn’t speak.’
(Ina.09, age 2;2.12)
11. The distribution of each construction is the following: In Ina’s data there are 4 wh-questions, 19 yes/no-questions and 78 non-subject-initial declaratives, in Ann’s data 2 wh-questions, 3 yes/ no-questions and 38 declaratives, and in Ole’s data 6 wh-questions, 7 yes/no-questions and 56 declaratives. In other words, the three children are similar with respect to the relative frequency of each construction type.
Verb movement and subject placement
(32)
komte ikke reven med mæ # i senga mi? come.past not fox.def with me in bed.def mine ‘Didn’t the fox come with me in my bed?’ Target form: Kom ikke reven med mæ i senga mi?
(33) no har ikke Ann mat til han. now have.pres not Ann food for him ‘Now Ann doesn’t have food for him.’
(Ina.18, age2;8.12)
(Ann.09, age 2;2.19)
(34) der kom ikke pappaen heller. there come.past not daddy.def either ‘There the daddy didn’t come either.’
(Ann.14, age 2;6.0)
(35) der står ikke alle folkan. there stand.pres not all people.def/pl ‘There all the people are not standing.’
(Ole.12, age 2;5.18)
(36) korfor kommer ikke mummien sæ laus? why come.pres not mummi.def refl loose ‘Why is the Mummy troll stuck?’
(Ole.17, age 2;8.24)
When these constructions occur with pronominal subjects, on the other hand, the picture is quite different. There is a mixture of the two word orders (Neg-S and S-Neg) attested in the data of all three children, as illustrated by the examples in (37)–(39) with Neg-S and examples (40)–(42) with the opposite word order. (37) a. b.
nei det må ikkje du gjøre. no that must not you do ‘That you mustn’t do.’ har ikkje han fota her? have.pres not he foot.pl here ‘Doesn’t he have feet here?’
(Ina.17, age 2;7.22)
(38) a. b.
no kan ikke han sove mer. now can not he sleep more ‘Now he can’t sleep any more,’ den like ikke æ å se. that like.pres not I to see ‘That I don’t like to look at.’
(Ann.10, age 2;3.9)
(39) a. b.
nå skal ikke dem sove. now shall not they sleep ‘Now they shouldn’t sleep.’ det får ikke æ lov til. that get.pres not I allowed to ‘That I am not allowed to do.’
(Ole.09, age 2;3.15)
(Ina.13, age 2;5.25)
(Ann.11, age 2;4.0)
(Ole.12, age 2;5.18)
Marit R. Westergaard
c.
går ikke det an å kjøre rundt? go.pres not it particle to drive around ‘Isn’t it possible to drive around?’
(Ole.11, age 2;4.21)
(40) a. b.
nei, nå må han ikke røre. no now must he not touch ‘No, now he mustn’t touch.’ korfor gjør han ikke det? why do.pres he not that ‘Why doesn’t he do that?’
(Ina.21, age 2;9.18)
(41) a. b.
nei det kan dem ikke. no that can they not ‘No, that they can’t do.’ men no var det ikke mer. but now be.past it not more ‘But now it was all gone.’
(Ann.17, age 2;8.4)
(42) a. b.
og no kan æ ikke drikke det. and now can I not drink it ‘And now I can’t drink it.’ korfor ser æ ikke skoan? why see.pres I not shoe.def/pl ‘Why don’t I see the shoes?’
(Ina.25, age 3;1.8)
(Ann.19, age 2;9.17)
(Ole.19, age 2;10.0)
(Ole.17, age 2;8.24)
Strictly speaking, all the sentences in (37)–(39) could be grammatical, in case the pronominal subjects following negation are stressed. Since intonation and stress is not marked in the corpus, this information is not available. However, most of these sentences with Neg-S word order and pronominal subjects seem somewhat odd in the context they occur, indicating that they do represent a non-target word order. The following two examples from Ina’s data, recorded around age 2;7, may also express some uncertainty in the child grammar about word order. In (43) the pronominal subject appears twice, both preceding and following negation, while in (44) there is an instance of the negation ikke ‘not’ on both sides of the pronominal subject. (43) . . . har han ikke han shorts på sæ? . . . have.pres he not he shorts on self ‘Doesn’t he have shorts on?’
(Ina.16, age 2;7.8)
(44) æ [//] kan ikke du ikke hjelpe mæ da? I . . . . . . . can not you not help me then ‘Can’t you help me then?’
(Ina.16, age 2;7.8)
More importantly, there is a clear development across the files of the three children with respect to the order of pronominal subjects and negation in these constructions: While the non-subject-initial declaratives and questions with Neg-S word order tend
Verb movement and subject placement
to occur at a relatively early age, the sentences with the more target-consistent S-Neg word order mostly occur in the later files of the three children. This is illustrated in Table 2, where the child data in the corpus are divided into five age periods. Table 2. The number of non-subject-initial declaratives and questions with Neg-Spro and Spro-Neg word orders in the Norwegian child data Period 1 2 3 4 5
Ina
Ann
Ole
Age
Neg-S
S-Neg
Neg-S
S-Neg
Neg-S
S-Neg
1;9–2;3 2;3–2;6 2;6–2;8 2;8–3;0 3;0–3;4
0 7 7 10 5
0 4 4 36 21
0 9 1 0
2 7 3 12
1 21 0 6
1 3 13 18
As we see from the table, the children produce hardly any examples of nonsubject-initial declaratives or questions with negation in period 1, before the age of approximately 2;3. In period 2, up until approximately age 2;6, all three children produce more of the non-target-consistent Neg-S than the S-Neg word order. This situation changes for the children Ole and Ann in period 3, when they start producing more S-Neg word order, while Ina at this stage is still producing more non-target forms. Some development is possibly taking place in her grammar at this stage, indicated by the uncertainty expressed by the two examples in (43) and (44) above. Soon after, Ina catches up with the other two children with respect to the production of target-consistent forms, as there is a major change in the relative frequency of Neg-S vs. S-Neg word order in period 4 (age approximately 2;8 to 3;0), and this situation is stable into period 5 (up to age 3;4), for which the corpus only contains data from Ina. This development is illustrated in Figure 1 for all three children, where it is clearly displayed that the Neg-S word order (represented by solid lines) is more frequent in the early files of all three children, while the target-consistent S-Neg word order (represented by broken lines) gradually takes over. The two lines for Ann and Ole cross between Periods 2 and 3, suggesting that a certain change has occurred in their systems, while the lines for Ina’s development indicate that the change takes place somewhat later in her grammar. Since the two subject positions involve a distinction between informationally given and new subjects, a possible account of the developmental pattern in the Norwegian child data is that the children at the early stages have not yet mastered the pragmatic issues related to information structure. Another possible analysis of these data is that there is an economy principle at play in early child systems, ensuring that syntactic movement will only take place if there is clear evidence in the input. These two analyses are outlined in the next section.
Marit R. Westergaard 40 36
35 30 25 21
20
18
15
13
10 5 0
21
2 1 0 Period 1
9 7 4 3 Period 2
7 4 3 1 0 Period 3
INA, Neg-S INA, S-Neg ANN, Neg-S ANN, S-Neg OLE, Neg-S OLE, S-Neg
12 10 6 0 Period 4
5 Period 5
Figure 1. The development of Neg-Spro word order (solid lines) and Spro-Neg word order (broken lines) in non-subject-initial declaratives and questions with negation in the Norwegian child data.
4. Two possible analyses: pragmatics vs. economy 4.1 Delay in pragmatic competence It has often been argued in the language acquisition literature that syntax is early, but pragmatics is late. This is presumably because much of syntactic structure is assumed to be given by UG, while pragmatic principles need to be learned by language in use and will therefore take some time. Thus, aspects of child language that remain unexplained by syntactic principles are often referred to as children’s problems with pragmatic issues, e.g., in the Optional Specificity analysis of Hyams (1996). Other examples of this are Chien & Wexler (1990), who, reporting on experiments done with English-speaking children on the binding of reflexives and pronouns, argue that children’s early grammars have all the relevant syntactic principles. To explain younger children’s non-target-consistent performance on the binding of pronouns in their experiments, they suggest that children lack a pragmatic principle, thus arguing for a dissociation between syntax and pragmatics. In similar work on the acquisition of binding principles in Russian, Avrutin & Wexler (1992) identify a pragmatic principle (Principle P), which is responsible for the contra-indexation of pronouns with potential antecedents. They show that children’s behaviour with respect to the syntactic elements of the Binding Theory is adult-like, while young children have major problems with the interpretation of sentences where Principle P is involved. Again, this is taken as evidence for the modular structure of the language faculty, and they claim that
Verb movement and subject placement
“the syntactic submodule is far ahead of its pragmatic counterpart” (p. 303). Finally, Batman-Ratyosyan & Stromswold (2002), in experimentation studies on word order in Turkish, find that while the inclusion of discourse context in their examples improves the performance of older children in the study (3-, 4- and 5-year-olds), it has the opposite effect on younger children (2-year-olds). On the other hand, the 2-year-olds benefit from morphosyntactic cues (Case marking). Referring to the concept of Theory of Mind, the authors suggest that “acquiring discourse/pragmatics requires that one be aware of the intentions and knowledge states of others” (p. 803), and conclude that learning morphosyntax is easy, while discourse/pragmatics is hard. This means that pragmatic issues such as information structure are often also assumed to be acquired late, as children have to reach a certain age before they are able to judge what is given or new information for their interlocutors. Schaeffer (1995), working on the acquisition of scrambling in Dutch, finds that young children sometimes fail to scramble (especially pronominal) objects when it is required in the target grammar. In Schaeffer (2000), scrambling is argued to be driven by referentiality, which can only be grammatically marked if a distinction is made between discourse-related and non-discourse-related objects. In order to make this distinction, it is necessary to take the knowledge of the interlocutor into account, which requires what Schaeffer calls the “Concept of Non-shared Knowledge”. This is a pragmatic principle which young children (2-year-olds) are argued to lack, and this accounts for their failure to produce the target-consistent word order in scrambling contexts. However, studying the acquisition of Russian, where the distribution of null elements is dependent on contextual requirements, Gordishevsky & Avrutin (2004) find that information structure is generally available to children at a very early age. Nevertheless, they find a certain overuse of null subjects in early child language, and in order to explain this, they hypothesize that young children lack the concept of non-shared knowledge mentioned above. For the data they discuss, this means that young children tend to treat arguments as given information, even when that is inappropriate in the discourse situation – that is, the children are assuming that these are given information also for their interlocutors. Finally, in a discussion of null subjects in various child languages, Rizzi (2005) evaluates both grammatically based (parameter missetting) and extra-grammatical explanations and claims that there need not be a contradiction between the two. More specifically, he argues that early subject drop is a genuinely grammatical option in child languages which may be used to alleviate an immaturity in the production system, and concludes that “language development is grammatically based, but performance driven” (p. 102). Based on previous research, which supports pragmatic analyses of certain delays in children’s acquisition, the Norwegian child data presented in section 3 could be given a similar analysis: That is, if young Norwegian children lack the concept of non-shared knowledge, this should make them unable to distinguish between informationally given and new subjects in discourse. This would then cause them to treat all subjects as identical in information value and place them in the low subject posi-
Marit R. Westergaard
tion at an early stage. Only after the children have acquired the distinction between informationally given and new subjects would they be able to move the former to the higher subject position.
4.2 Economy of movement In Westergaard (2005), Norwegian children’s acquisition of V2 word order was analyzed within a continuity model which includes structural economy. The continuity aspect of this approach is taken care of by a universal pool of possible functional categories, where rules for their relative order are given by UG. In the process of language acquisition, children select categories from this universal set, based on UG principles and cues in the input, as stated in (45). (45) Children build clause structure based on – UG (universal pool of categories+various rules/constraints) – input cues
The cues in this model are assumed to be pieces of I-language structure which are expressed in the primary linguistic data that children are exposed to, as in Lightfoot’s (1999, 2006) cue-based approach to language acquisition and change. As an example, the cue for V2 syntax, according to Lightfoot (2006: 86) is formulated as in (46), which is a piece of structure “where a phrasal category occurs in a Specifier of a CP whose head is occupied by a verb.” In a similar fashion, the cue for subject shift could be formulated as in (47), which would be a piece of structure where a [-foc] subject occurs in the specifier of the highest projection in the IP domain, i.e., the InTopP (above negation). (46) Cue for V2 syntax:CP[XP CV. . .]
(from Lightfoot, 2006: 86)
(47) Cue for subject shift: InTopP[ DP [-foc] . . .]
Within this model it is assumed that UG provides children with the information that all clauses have a CP domain, minimally a FinP, which is necessary for clause typing. Furthermore, UG will match a syntactic head with the illocutionary force of the sentence, and children therefore need no input information to know that e.g., a wh-question is an IntP and a non-subject-initial declarative a TopP.12 However, the presence of certain other functional projections must be triggered by primary linguistic data. For example, Norwegian children need input in order to project a TopP in subject-initial declaratives, as it is not the case that all languages universally have this structure.
12. Within the Split-CP model sketched in section 2, Westergaard (2006a) argues that cues may be linked to particular clause types, so that there are separate cues for V2 depending on the functional head involved (Int°, Top°, Pol°, etc.).
Verb movement and subject placement
This means that children take an economic approach to structure building and that they will not build more structure than is triggered by input. This idea is similar to the principle of structural economy in the Lexical Learning Hypothesis of Clahsen, Eisenbeiss & Penke (1996), originally from Safir (1993). Related to this, it is also argued that children’s movement operations target the lowest possible position in the structure that is compatible with the input, as stated in (48).13 (48) Structural economy – children only build as much structure as there is evidence for in the input – children only move elements as far as there is evidence for in the input
An economy approach to movement is in line with much work within the minimalist program of Chomsky (1995), and has been used extensively by Platzack (1996) in his argumentation for the “Initial Hypothesis of Syntax”, which basically argues that children will start out assuming that all features in the language they are acquiring are weak (while only strong features trigger movement). It also corresponds to what is often seen in child language data, viz. that when children produce non-target-consistent word order, this virtually always involves lack of movement or less movement than in the adult language, not an overgeneralization of movement, see e.g., Barbier (2000) on the lack of scrambling and object shift in Dutch and German, Josefsson (1996) on the lack of object shift in Swedish, or Radford (1994) on the lack of inversion in some English-speaking children’s wh-questions. Economy of movement has also been attested in the production of the three Norwegian children in this corpus. In Westergaard (2004) it was shown that in addition to the non-subject-initial declaratives with target-consistent V2 (e.g., example (26) above), the children produce occasional examples without verb movement, as illustrated in (49) and (50). Similar examples are also attested in Swedish child language by Santelmann (1995) and Platzack (1996). (49)
der Ann har et. there Ann have.pres one ‘There Ann has one.’ Target form: Der har Ann et.
(Ann.08, age 2;1.28)
(50)
på øyan æ har solbrilla. on eyes.def I have.pres sunglasses ‘On my eyes I have sunglasses.’ Target form: På øyan har æ solbrilla.’
(Ole.02, age 1;10.22)
13. If one assumes the extension condition of the minimalist program (Chomsky 1995), where movement is always argued to expand the tree, then the lack of movement may be subsumed under the more general principle of structural economy.
Marit R. Westergaard
Furthermore, the Norwegian children in the corpus also occasionally fail to move a pronoun in object shift constructions. These are very similar to the subject shift constructions in that pronominal objects are required to move across negation or an adverb in the adult grammar unless they are stressed. The example from Ina’s data in (51) indicates that the children have similar problems with object shift constructions in that there is lack of movement. (51)
eg finn ikkje han. I find.pres not him ‘I don’t (can’t) find it.’ Target form: Eg finn han ikkje.
(Ina.13, age 2;5.25)
But if children’s production of non-target word order is the result of economy of movement, we may then ask why the children seem to overgeneralize verb movement across adverbs and negation in the embedded contexts that were illustrated in (29)–(30) in section 3. In Westergaard (2005) and Westergaard & Bentzen (2007) this is in fact argued also to be due to economy of movement, in the following way: As mentioned above, in the syntactic model assumed here, UG does not provide Norwegian children with the information that subject-initial main clauses are TopPs with verb movement to Top° and subsequent subject movement to SpecTopP. That means that this must be learned from input. Unfortunately, there are no clear cues in subject-initial declarative main clauses that the verb (and accordingly also the subject) moves all the way to the TopP. Nevertheless, there should be ample evidence in the input that there is verb movement across negation and adverbs in these sentences.14 Therefore, children apparently realize very early that finite verbs move across negation and adverbs in main clauses, and they produce target-consistent word order from the first appearance of relevant constructions, such as sentence (23) from section 3, repeated here. (23’) æ vet ikkje. I know.pres not ‘I don’t know.’
(Ann.02, age 1;9.18)
Given children’s economy approach to language acquisition, they will only move the verb as high up in the clause structure as there are clear cues for in the input. Since sentence adverbs and negation are adjoined to TP, as indicated by the structure in (22) in section 2, the lowest possible projection in the clause structure that will ensure that the verb occurs to the left of negation is the InTopP. Verb movement to the InTop° head is in accordance with the children’s economy approach to structure building, it will
14. In an adult sample investigated in Westergaard (2006a), there are 43 such examples, which makes up 6.4% of the total number of complete sentences. However, in the extended cue-based model it is assumed that children only focus on the relevant clause type for word order cues, in this case subject-initial declaratives, which means that the evidence for verb movement is attested in as much as 35% of the relevant triggering contexts.
Verb movement and subject placement
result in V-Neg word order, and it will thus correspond to what is found in the adult language. But the syntactic representation of a sentence like (23) in the child’s grammar will differ from that of the target grammar, in that the child version of the sentence is a simple FinP with verb movement to InTop° (i.e., V-to-I movement in traditional terminology), while the corresponding adult structure is a full TopP. If this analysis is correct, we would expect to see verb movement to the InTopP also in embedded clauses and other non-V2 contexts. Recall that in the syntactic model adopted here, embedded clauses are bare FinPs, which means that they have an InTopP, but embedded declaratives do not normally have a TopP.15 Consequently there is no verb movement in Norwegian embedded clauses in the adult grammar. If the verb moves to the InTop° head in the child grammar, the children should also produce a word order with the verb preceding negation and other adverbs in constructions which do not have verb movement to the CP domain. And this is exactly what we saw in section 3: an overgeneralization of V2 word order from main clause declaratives to embedded clauses and non-V2 wh-questions, as was illustrated in (29) and (30). The children’s overgeneralization patterns in embedded clauses provide evidence for verb movement to a lower head in main clause declaratives than in the target grammar and can therefore be considered to be the result of an economy principle of movement . Based on the syntactic model presented here, which combines continuity and structure building, as well as the children’s non-adult-like production in other clause types, it could be argued that Norwegian children’s delay in target-consistent word order in the subject shift construction is due to economy of movement. In the next section I discuss the two analyses outlined in this section, and I will argue in favour of this latter approach, rejecting the pragmatic account of the child data as the children’s lack of the concept of non-shared knowledge.
5. Discussion In this section I first present three sets of arguments against a pragmatic account of the delay in subject shift constructions. Then I discuss some further factors supporting an economy approach to the child data, focusing on aspects such as input frequency, complexity and non-saliency of the landing site of the subject pronoun.
15. The TopP is assumed to be present in embedded clauses only in cases of embedded topicalization, as in sentence (i). In accordance with Vikner (1995), I consider this some kind of CP recursion, with the difference that the topmost CP is a bare FinP (like other embedded declaratives). (i)
Hun sa FinP [at TopP[denne boken hadde hun aldri lest]]. she said that this book.def had she never read ‘She said that she had never read this book.’
Marit R. Westergaard
The first set of arguments against a pragmatic account concerns the fact that the two subject positions seem not only to be both present (cf. Table 2), but also to be treated differently from early on. It is not clear that the non-target-consistent forms presented in section 3 indicate that the children’s grammars do not distinguish between given and new information at an early stage of development. If that were the case, one would expect to see pronouns and full DPs occurring in both subject positions, without a clear pattern. As we saw in section 3, however, full DP subjects are always placed in the position for informationally new subjects, after negation. The developmental data suggest that as soon as children realize that certain subjects are required to move to the high subject position, only informationally given subjects undergo this movement. The following example from Ole, where he produces both a pronoun as well as a full DP subject, illustrates how the two are placed in the appropriate positions with respect to the adverb: (52) nå vil han også bamsen sove. now will he also teddy.def sleep ‘Now he the teddy will also sleep.’
(Ole.09, age 2;3.15)
Furthermore, it does not seem to be relevant to explain the data by referring to the children’s lack of a pragmatic principle such as the concept of non-shared knowledge. The way this is used by Gordishevsky & Avrutin (2003), children are argued to overuse null subjects because they erroneously treat them as given information in the discourse situation, i.e., as knowledge that is shared between speaker and hearer. In the Norwegian data presented here, on the other hand, the situation is that the given information seems to be treated as if it were new, i.e., pronouns occur in a position that would normally be reserved for new information. The second set of arguments is related to the non-simultaneity of constructions which could be argued to be due to the lack of the principle of non-shared knowledge. Since subject shift has been shown to be very similar to object shift constructions, the two should be accounted for by the same principle, and the prediction would be that both constructions should fall into place at the same time. In Westergaard (2006b), however, I show that object shift seems to be a considerably later acquisition. According to economy of movement, non-target production in the two constructions may initially be due to the same economy principle, but one may be more persistent than the other for various other reasons. In Westergaard (2006b) it is suggested that an extreme lack of object shift constructions, both in child-directed speech and in the children’s own production, could be a contributing factor here. Likewise, a similar delay in subject movement is attested in Schönenberger (2002), where two girls acquiring Swiss German are found to overgeneralize V2 word order in embedded contexts until the age of approximately five. At the earliest stage of the recordings (approximately age 3;10–4;03) all subjects appear below the verb in these embedded clauses, while around age 4;04 the children start distinguishing between pronouns and full DPs, moving pronominal subjects to a higher position above the verb.
Verb movement and subject placement
It could of course be argued that Schönenberger’s data indicate that the two children discover the concept of non-shared knowledge around 4;04 and that this accounts for the change in word order (note that input could not account for this change, as the children are still producing non-adult-like word order). However, in my opinion it seems odd that such a general principle should be acquired so much later by the Swiss German children than the Norwegian ones, and I think that an account that takes syntactic complexity and possibly also frequency into account is more likely. Thus, the fact that the Swiss German data are attested in embedded clauses, which are arguably both more complex and less frequent than the main clauses in the Norwegian child data, could be one alternative explanation. The third set of arguments concerns children’s early acquisition of information structure in other constructions. Some recent studies come to the conclusion that information structure is in place early. One example is De Cat’s (2003) study of French acquisition data, where word order in the target grammar is dependent on the topicality of an element. She shows that French children abide by pragmatic requirements on syntactic structure, and argues that the competence to encode topics is available to them from the earliest stages of language acquisition. The children in the present study are also found to distinguish between informationally given and new subjects at an early age in the two types of wh-questions allowed in their dialect of Norwegian (see Westergaard 2003). Recall that the choice of word order in these questions in the adult grammar is dependent on information structure, more specifically on whether the subject is [+/-focus], as was illustrated in examples (15)–(16) in section 2. If it were the case that pragmatic principles are acquired late, we should expect children to experience certain problems in the acquisition of the subtle word order distinctions in whquestions. However, as mentioned in section 3.2, the children in this study very early produce both V2 and non-V2 word orders in a systematic way, i.e., only after those whwords which allow it, and with the same preference patterns for subject and verb types as adults. These patterns were illustrated in (27) and (28) above and repeated here as (27’) and (28’). The fact that children so early distinguish between given and new subjects with respect to verb movement indicates that young children are extremely sensitive to patterns of information structure. (27’) ka du skal finne? what you shall find ‘What do you want to find?’ (28’) kor er Ann sin dukke hen? where is Ann poss doll loc ‘Where is Ann’s doll?’
(Ina.5, age 2;0.5)
(Ann.04, age 1;11.0)
Moreover, the non-subject-initial declaratives without verb movement that were illustrated in (49)–(50) in the previous section are similar to the non-V2 wh-questions with respect to subject and verb types, while the early target-consistent forms with V2 normally appear with be and full DP subjects (see Westergaard 2004). Thus, the
Marit R. Westergaard
children seem to be distinguishing between informationally given and new subjects even in the absence of input for such a distinction. Based on the argumentation presented so far, I would like to discard the pragmatic account of the Norwegian child data on subject shift. But then one might ask how a principle of economy of movement fares any better? Why is subject movement delayed when other types of movement, e.g., V2, seem to fall into place much earlier? And why is target-consistent word order in object shift constructions a later acquisition than subject shift? If all movement operations are affected by the same economy principle, one would also expect simultaneous acquisition. However, the economy principle of movement must obviously be overridden by evidence in the input. Thus, a delay in acquisition should take the following factors into account when answering these questions, which I discuss in turn below: the input frequency of the sentence types that express the cue, the general complexity of the construction, and finally, the saliency of the landing site. The cue for subject shift was formulated in (47) in the previous section, and this should be expressed in all questions and non-subject-initial declaratives which also contain negation or an adverb. In a small sample of adult speech investigated in Westergaard (2005) (from a one-hour recording in the Norwegian child language corpus), there were 42 relevant examples, comprising 6.3% of all complete clauses in the sample (see also Westergaard, 2006a). The examples in the input sample also show that the distinction is not simply a matter of pronoun vs. full DP, as there is no one-to-one correspondence between subject type and word order. Out of the 42 relevant examples, the majority of them being non-subject-initial declaratives, 28 occur with the subject preceding negation, which means that the cue for subject shift is attested in 4.2% of all complete clauses in the sample, while 14 have the opposite word order. While all the 28 subjects preceding negation are pronominal (i.e., typical examples of given information), the 14 subjects following negation are divided equally between pronouns (presumably stressed) and full DPs. This means that, although the cue for subject shift has a relatively low frequency in the input, it is not completely inconsiderable. There are even sentence pairs in the input sample that should clearly reveal the nature of these subject types to the children, thus being responsible for the eventual acquisition of these constructions. Examples are provided in the two yes/no-questions containing negation in (53) and (54), which occur adjacent to each other in the corpus. (53) har han ikkje et dyr? has he not an animal ‘Doesn’t he have an animal?’
(INV, file Ole.14)
(54) har ikkje han Postman Pat en kattepus? has not det Postman Pat a kitty ‘Doesn’t Postman Pat have a kitty?’
(INV, file Ole.14)
In comparison, the evidence for verb movement in non-subject-initial declaratives is attested as much as 19.6% in the input sample, considerably more than the evidence
Verb movement and subject placement
for subject shift. When considering object shift, which seems to fall into place somewhat later than subject shift, the input sample contains only 3 examples, which makes up as little as 0.4%. Thus it could be said that there is massive (and consistent) evidence for V2 in the input, and this could to a certain extent account for the earlier acquisition of target-consistent word order in these constructions. Likewise, the even later acquisition of object shift could be related to the extremely low frequency of this construction in the input. However, input frequency alone presumably cannot explain the different types of delay related to economy of movement. As argued in Westergaard & Bentzen (2007), input frequency may only have an effect in connection with other factors, e.g., complexity. I would like to argue that questions and non-subject-initial declaratives are quite difficult for young children, involving not only verb movement to the CP domain (resulting in V2), but also movement of a non-subject element or a wh-constituent to initial position. The third movement operation required in these constructions, movement of an informationally given ([-foc]) subject across negation, then adds a further complexity to the construction. Another important factor is that there are other cues for subject placement in the language which are competing with the cue for subject shift. Finally, I would like to point out that this type of subject movement does not involve a particularly salient landing site, compared to e.g., subject-initial declaratives, where the subject is attested in initial position as soon as multi-word utterances appear in the data. This is the case regardless of e.g., the presence of negation or the finiteness of the verb, as illustrated in (55). This, I would argue, is due to the general saliency of the initial position in the input. There is also a relatively high frequency of subjectinitial declaratives in the primary linguistic data that children are exposed to. In the adult sample studied in Westergaard (2005) this construction makes up 21.9% of all complete sentences, while in a much larger study of child-directed speech in Swedish, Josefsson (2004) finds that subject-initial declaratives represent on average 27% of all main clauses containing a verb. (55) æ ikke låne. I not borrow.inf ‘I will not borrow (it).’
(Ole.03, age 1;10.22)
Based on the discussion provided in this section, I argue that the brief delay in the children’s movement of a [-foc] subject to the InTopP in subject shift constructions is simply due to a general economy principle of movement, and not linked to children’s failure to distinguish between informationally given and new subjects. That is, the original cause of the non-adult-like word order is that children will not move subjects to the higher position unless there is clear evidence for this in the primary linguistic data. The persistence of this non-target behaviour in the child data is then related to factors such as input frequency, complexity, and the non-saliency of the landing site.
Marit R. Westergaard
6. Summary and Conclusion In this chapter I have investigated the acquisition of word order in so-called subject shift constructions in Norwegian child language and attested a certain delay in the children’s movement of informationally given subjects ([-foc]) to a position preceding negation in V2 constructions, i.e., questions and non-subject-initial declaratives. Within a Split-CP approach to clause structure which also assumes two subject positions in the IP domain, the higher subject position is argued to be in the specifier of the In(ner) Top(ic)P, while the low position is SpecTP. A pragmatic account of this delay in terms of a lack of the concept of non-shared knowledge was considered, but argued not to be relevant in this case, mainly because the children are found to distinguish between informationally given and new subjects very early in wh-questions. Instead, I outlined a model of structural economy in language acquisition, where children are assumed to build clause structure based on a universal set of functional categories and cues in the input. Within this approach, various kinds of non-target-consistent word order in the children’s production were argued to be due to economy of movement, e.g., object shift constructions, certain cases of lack of V2 in non-subject-initial declaratives, as well as overgeneralizations of V2 in embedded contexts. This economy principle is argued also to be responsible for the delay in movement of pronominal subjects to the higher position above negation and adverbs in subject shift constructions. Finally, the persistence of this delay is related to the input frequency of the clause types where the cue for subject shift is expressed, the general complexity of the construction, and the non-saliency of the landing site of pronominal subjects.
References Anderssen, M. 2006. The Acquisition of Compositional Definiteness in Norwegian. Doctoral Dissertation, University of Tromsø. Avrutin, S. & Wexler, K. 1992. Development of Principle B in Russian. Language Acquisition 2(4): 259–306. Barbier, I. 2000. An experimental study of scrambling and object shift in the acquisition of Dutch. In The Acquisition of Scrambling and Cliticization, S.M. Powers & C. Hamann (eds), 41–69. Dordrecht: Kluwer. Batman-Ratyosyan, N. & Stromswold, K. 2002. Morphosyntax is easy, discourse/pragmatics is hard. In Proceedings of the 26th Annual Boston University Conference on Language Development 2, B. Skarabela, S. Fish & A.H.-J. Do (eds), 793–804. Somerville MA: Cascadilla. Bentzen, K. 2003. V-to-I Movement in the absence of morphological cues: Evidence from Northern Norwegian. In P. Svenonius, A. Dahl and M. R. Westergaard (eds), Nordlyd 31.3, Tromsø University Working Papers on Language and Linguistics: Proceedings from the 19th Scandinavian Conference of Lingustics, 573-588. University of Tromsø. Bentzen, K. 2005. What’s the better move? On verb placement in Standard and Northern Norwegian. Nordic Journal of Linguistics 28.2, 153-188.
Verb movement and subject placement Chien, Y-C. & Wexler, K. 1990. Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition 1(3): 225–295. Chomsky, N. 1982. Some Concepts and Consequences of the Theory of Government and Binding. Cambridge MA: The MIT Press. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Clahsen, H., Penke, M. & Parodi, T. 1993/94. Functional categories in early child German. Language Acquisition 3, 395–429. Clahsen, H., Eisenbeiss, S. & Penke, M. 1996. Lexical learning in early syntactic development. In Generative Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations and Crosslinguistic Comparison [Language Acquisition and Language Disorders 14], H. Clahsen (ed), 129–159, Amsterdam: John Benjamins. De Cat, S. 2003. Syntactic manifestations of very early pragmatic competence. In Proceedings of the 27th Annual Boston University Conference on Language Development, B. Beachley, A. Brown & F. Conlin (eds), 209–219. Somerville MA: Cascadilla. Firbas, J. 1992. Functional Sentence Perspective in Written and Spoken Discourse. Cambridge: Cambridge University Press. Frey, W. 2004. A medial topic position for German. Linguistische Berichte 198: 153–190. Gelderen, E. van. 2004. Grammaticalization as Economy. Amsterdam: John Benjamins. Gordishevsky, G. & Avrutin, S. 2004. Optional omissions in an optionally null subject language. In Proceedings of GALA 2003, Vol. 1, LOT Occasional series 3, University of Utrecht, J. van Kampen & S. Baauw (eds), 187–198. Available online at http://lotos.library. uu.nl/index.html. Hyams, N. 1996. On the underspecification of functional categories. In Generative Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations and Crosslinguistic Comparison. [Language Acquisition and Language Disorders 14], H. Clahsen (ed.), 91–128. Amsterdam: John Benjamins. Jordens, P. 1990. The acquisition of verb placement in Dutch and German. Linguistics 28: 1407–1448. Josefsson, G. 1996. The acquisition of object shift in Swedish child language. In Children’s Language 9, C.E. Johnson & J.H.V. Gilbert (eds), 153–165. Mahwah NJ: Lawrence Erlbaum. Josefsson, G. 2004. Input and output: Sentence patterns in child and adult Swedish. In The Acquisition of Swedish Grammar [Language Acquisition and Language Disorders 33], G. Josefsson, C. Platzack & G. Håkansson (eds), 95–133. Amsterdam: John Benjamins. Kemenade, A. van & Los, B. 2006. Discourse adverbs and clausal syntax in Old and Middle English. In The Handbook of the History of English, A. van Kemenade & B. Los (eds), 224–248. Oxford: Blackwell. Lightfoot, D. 1999. The Development of Language: Acquisition, Change and Evolution. Oxford: Blackwell. Lightfoot, D. 2006. How New Languages Emerge. Cambridge: Cambridge University Press. Platzack, C. 1996. The initial hypothesis of syntax: A minimalist perspective on language acquisition and attrition. In Generative Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations, Crosslinguistic Comparisons [Language Acquisition and Language Disorders 14], H. Clahsen (ed), 369–414. Amsterdam: John Benjamins. Poeppel, D. Wexler, K. 1993. The full competence hypothesis of clause structure in Early German. Language 69, 1–33. Radford, A. 1994. The syntax of questions in child English. Journal of Child Language 21, 211–236. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar: Handbook of Generative Syntax, L. Haegeman (ed), 281–337. Dordrecht: Kluwer.
Marit R. Westergaard Rizzi, L. 2001. On the position ‘Int(errogative)’ in the left periphery of the clause. In Current Studies in Italian Syntax, G. Cinque & G. Salvi (eds), 287–296. Amsterdam: Elsevier. Rizzi, L. 2005. On the grammatical basis of language development: A case study. In The Oxford Handbook of Comparative Syntax, G. Cinque & R. Kayne (eds), Oxford: Oxford University Press. Safir, K. 1993. Perception, selection, and structural economy. Natural Language Semantics 2, 47–70. Santelmann, L. 1995. The Acquisition of Verb Second Grammar in Child Swedish: Continuity of Universal Grammar in Wh-questions, Topicalizations and Verb Raising. PhD dissertation, Cornell University. Schaeffer, J. 1995. On the acquisition of scrambling in Dutch. In Proceedings of the 19th Annual Boston University Conference on Language Development 2, D. MacLaughlin & S. McEwan (eds), 521–532. Somerville MA: Cascadilla. Schaeffer, J. 2000. The Acquisition of Direct Object Scrambling and Clitic Placement: Syntax and Pragmatics. Amsterdam: John Benjamins. Schönenberger, M. 2002. Embedded V-to-I in Child Grammar: The Acquisition of Verb Placement in Swiss German. Dordrecht: Kluwer. Vangsnes, Ø. A. 2004. On wh-questions and V2 across Norwegian dialects: A survey and some speculations. Working Papers in Scandinavian Syntax 73, 1–59. Vangsnes, Ø. A. 2006. Microparameters for Norwegian wh-grammars. Linguistic Variation Yearbook 5, 187–226. Vikner, S. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: OUP. Westergaard, M.R. 2003. Word order in wh-questions in a North Norwegian dialect: Some evidence from an acquisition study. Nordic Journal of Linguistics 26(1), 81–109. Westergaard. M.R. 2004. The interaction of input and UG in the acquisition of verb movement in a dialect of Norwegian. Nordlyd: Tromsø Working Papers in Language Acquisition 32(1), 110–134. Westergaard, M.R. 2005. The Development of Word Order in Norwegian Child Language: The Interaction of Input and Economy Principles in the Acquisition of V2. Dr. philos. dissertation, University of Tromsø. Westergaard, M.R. 2006a. Triggering V2: The amount of input needed for parameter setting in a Split-CP model of word order. In Language Acquisition and Development: Proceedings of GALA 2005, A. Belletti, E. Bennati, C. Chesi, E. DiDomenico & I. Ferrari (eds), 564–577. Cambridge: CUP. Westergaard, M. 2006b. Non-target word order in early child language: Pragmatics vs. economy. Paper given at CASTL/CLS workshop on Variation and Stability in Grammar, University of Nijmegen, 25 April. Westergaard, M. & Bentzen, K. 2007. The (Non-) Effect of Input Frequency on the Acquisition of Word Order in Norwegian Embedded Clauses. In Frequency Effects in Language Acquisition: Defining the Limits of Frequency as an Explanatory Concept, [Studies on Language Acquisition], I. Gülzow & N. Gagarina (eds), 271–306. Berlin/New York: Mouton de Gruyter. Westergaard, M.R. & Vangsnes, Ø.A. 2005. Wh-questions, V2, and the left periphery of three Norwegian dialect types. Journal of Comparative Germanic Syntax 8: 117–158.
Three acquisition puzzles and the relation between input and output* Manuela Schönenberger
I discuss three puzzles arising from the study of two children acquiring Swiss German, a verb-second (V2) language which displays the verb-final pattern in embedded clauses. Before switching to the verb-final pattern the children use verb movement in embedded clauses which presents the following puzzles bearing on the relation between input and output. (i) The children generalize verb movement to all embedded clauses although a large proportion in the input unambiguously show the verb-final pattern. (ii) The children produce V2 in wil ‘because’-clauses, although the adults usually produced the verb-final pattern. (iii) One of the children still misplaces finite auxiliaries at age 8;0. In trying to solve these puzzles I draw a comparison between the child data and adult matrix clauses, and also note similarities between these acquisition data and diachronic English data.
1. Introduction Two children acquiring a variant of Lucernese Swiss German were observed longitudinally – age 3;10–8;01–to study the acquisition of verb placement in embedded clauses. Adult Swiss German displays verb movement in matrix clauses, resulting in verb-second (V2), and the verb-final pattern in embedded clauses. While the children’s verb placement in matrix clauses is target-consistent, their verb placement in embedded clauses is not. The children produce a large number of examples with verb movement, often resulting in verb-placement errors. The verb-final pattern first appears at age 4;11 and gradually supplants verb movement. These data pose a puzzle for language acquisition. Children are assumed to set parameters governing word order very early according to Wexler (1999), hence the production of a systematic error in this domain should not occur. It also appears from the literature that children acquiring German or other dialects of Swiss German than Lucernese do not behave like this. To investigate this issue I examine the role of the input to the Lucernese children. I will show that the input clearly flags the verb-final pattern in embedded clauses * I am grateful to two anonymous reviewers for very detailed and helpful comments.
Manuela Schönenberger
and thus does not seem to be the source of the children’s verb-movement grammar. Matrix clauses, on the other hand, always show verb movement and occur much more frequently in the input than embedded clauses. Thus structures with verb movement are predominant in the input. The children’s verb placement in embedded clauses, however, does not show ‘general V2’, which is characteristic of matrix clauses in Swiss German, but only ‘partial V2’, which closely resembles that found in matrix clauses in diachronic data in English. Thus the children have not simply mapped the structure of matrix clauses in Swiss German onto embedded clauses. I propose instead that they have set the verb-placement parameter to [+movement], which gives rise to general V2 in matrix clauses and partial V2 in embedded clauses. Another feature of the children’s verb placement – the unusual distribution of finite auxiliaries – is seen as yet another misset parameter, which is unrelated to the verb-placement parameter. Since the children eventually show target-consistent verb placement in embedded clauses it must be possible to reset parameters. The children’s predominant use of V2 in a context where it is allowed, namely in clauses introduced by wil ‘because’, may be due to not yet fully specified pragmatic knowledge. This paper is organized as follows. Section 2 discusses verb placement in matrix and embedded clauses in adult Swiss German. Section 3 gives an overview of the acquisition literature on verb placement in embedded clauses. Section 4 focuses on the relation between input – what the child hears – and output – what the child produces with respect to verb placement in embedded contexts. Section 4.1 describes how the Lucernese data were collected. Section 4.2 examines the extent to which the verb-final pattern is present in the embedded clauses produced by the adults and compares it to the verb placement used by the children. Section 4.3 looks at verb placement in clauses introduced by wil in the adult samples and the child data. In contrast to clauses introduced by a complementizer, clauses introduced by wil allow for the verb-final pattern as well as V2. Section 4.4 studies the frequency of sequences of complex verbal clusters containing a finite auxiliary in the adult samples and the child data. Section 5 addresses the question of parameter setting. Section 5.1 considers the role verb placement in matrix clauses in the input might have played in encouraging the children to set the parameter to [+movement] in embedded clauses. Section 5.2 highlights similarities between diachronic English data and the Swiss-German acquisition data. Section 5.3 shows how the children’s verb placement can be derived. Verb movement in the children’s embedded clauses is analysed as targeting different functional heads in the C-domain, an option available in the Split-CP framework proposed by Rizzi (1997). Section 6 contains my conclusions.
2. Verb placement in Swiss German Swiss German is a verb-second (V2) language, which displays the verb-final pattern in embedded clauses. A typical feature of V2 is the fact that one and only one maximal projection can precede the finite verb. This maximal projection is not restricted to the
Three acquisition puzzles
subject, although subject-initial V2 clauses are by far the most frequent.1 The phenomenon of V2 is illustrated in (1). (1) a. b. c.
En Frosch lacht ‘Prinzessin aa. a frog smiles the-princess at ‘A frog smiles at the princess.’ Noch erem Kuss verwandlet-er sich in-en Prinz. after her kiss transforms-he self into-a prince ‘After her kiss it turns into a prince.’ En Prinz hät ‘Prinzessin immer wöle hürote. a prince has the-princess always want(ed) marry ‘A prince the princess has always wanted to marry.’
The apostrophe ‘ in (1a) and (1c) stands for the definite article which has been totally assimilated to the initial plosive in the nominal Prinzessin. Embedded clauses usually show the verb-final pattern, as in (2). Relative clauses (2a) and clauses introduced by a complementizer (2b, 2c) only allow the verb-final pattern.2 Wh-complements (2d) and clauses introduced by wil ‘because’ (2e) sometimes allow for verb movement, which influences the interpretation of the sentence. (2) a. b. c. d. e.
De Frosch [wo ufeme Seeroseblatt hocket] isch glücklech. the frog which on-a waterlily leaf sits is happy ‘The frog, which sits on the leaf of a waterlily, is happy.’ Er freut sich [dass-ere de Ball in ‘Teich keit isch]. he pleases self that-her the ball in the-pond fallen is ‘It is glad that her ball fell into the pond.’ [Wenn-er si lieb frooget] küsst si en villecht. if-he her dearly asks kisses she him maybe ‘If it asks her in a friendly manner she might kiss it.’ Eus ninnt’s wunder [wenn si’s entlech wooget]. us takes-it wonder when she-it finally dares ‘We wonder when she’ll finally dare do it.’ [Wil de Frosch eigentlich en Prinz isch] sött si pressiere. because the frog actually a prince is should she hurry up ‘She should hurry up because the frog is actually a prince.’
1. In a sample containing the matrix clauses produced by the mother of one of the children 335 of 564 (59%) of declaratives are subject-initial (see Schönenberger 2001). 2. There is a certain type of relative clause which is not introduced by wo, but by der, die, das and which is only construable with V2, as shown in (i). These types of relative clauses are rare (cf. fn 3). (i)
Es git Lüt, die findet de Grendel doof. there gives people who find the Grendel stupid ‘There are people who find Grendel stupid.’
Manuela Schönenberger
The subtle impact verb placement can have on the interpretation of a sentence is illustrated in (3). In (3a) the verb-final pattern is used. The clause introduced by wil mentions the reason or cause for the frog’s happiness. The frog’s happiness is due to its having been recently kissed. In (3b) the V2 pattern is used. In this case, wil is often followed by an intonational break. The V2-clause introduced by wil gives the reason why the speaker thinks that the frog is happy. In other words, based on the frog’s pleased look the speaker concludes that the frog is happy. Thus the wil-clause in (3b) does not give the reason why the frog is happy; the frog’s happiness is not due to his looking pleased. The option of using wil+V2 is not restricted to Swiss German, but is also attested in other varieties of German, and alternative readings to that discussed above are possible (see e.g., Eppler (1999); Günthner (1996); Uhmann (1998)). (3) a. b.
De Frosch isch glücklech [wil-er en Kuss öbercho hät]. the frog is happy because-he a kiss received has ‘The frog is happy because it has received a kiss.’ Am Frosch goot’s wohl guet [wil [er luegt so to-the frog goes-it surely good because he looks so zfride drii]] . contented verbal particle ‘The frog must be happy because it looks so pleased.’
The verb-final pattern can sometimes be masked due to ‘extraposition’ – a type of rightward movement, provided the underlying order of Swiss German is assumed to be SOV – which is shown in (4) and (5). Because something occurs to the right of the finite verb in these embedded clauses, it appears that the finite verb has undergone movement to the left. However, this is not the case. In (4a) a temporal adjunct (am Sunntig), which usually linearly precedes the finite verb, has been extraposed to the right and hence follows the finite verb. Extraposition of Prepositional Phrases (PPs) is possible in both Swiss German and German. Sentential complements (4b) – in contrast to nominal complements – generally follow the predicate which selects them, in both Swiss German and German. (4) a. b.
I weiss [dass p’Fee gern lang pfuuset am Sunntig]. I know that the-fairy gladly long sleeps on Sunday ‘I know that the fairy likes to sleep for a long time on Sundays.’ I weiss [dass de Oberon nöd weiss dass-er i Gfoor isch]. I know that the Oberon not knows that-he in danger is ‘I know that Oberon doesn’t know that he’s in danger.’
The examples in (5) involve Verb Projection Raising (VPR), which is a typical feature of Swiss German, but also occurs in varieties of German and Dutch, and in Old English. In essence, VPR is the phenomenon in which the non-finite complement of a modal (or auxiliary) can appear to the right of the modal (or auxiliary). In (5a) the infinitive dichte follows the finite modal wöt, as does the infinitive läse in (5b). In (5b’)
Three acquisition puzzles
both the infinitive läse and its direct object churzi Gedicht follow the modal wöt. The string to the right of wöt in (5b’) is referred to as the raised cluster. Sometimes the word order of an embedded clause superficially looks like subject-initial V2, as in (5b’). (5) a. b. b’.
Si säget [dass de Caliban wider wöt dichte]. they say that the Caliban again wants poems write ‘They say that Caliban wants to write poems again.’ Wer weiss [öb de Prospero churzi Gedicht wöt läse]. who knows if the Prospero short poems wants read ‘Who knows if Prospero wants to read short poems.’ Wer weiss [öb de Prospero wöt churzi Gedicht läse]. who knows if the Prospero wants short poems read ‘who knows if Prospero wants to read short poems.’
VPR is a complex issue (see Haeberli & Haegeman (1995); Haegeman & van Riemsdijk (1986); Haegeman (1998); Koopman & Szabolcsi (2000); Schmid & Vogel (2004); Schönenberger & Penner (1995) a.o.), but the focus for this discussion is that the type of verb which can function as a V(erb)-R(aising) verb can differ between languages and even between dialects. In Lucernese a modal is an optional VR-verb, while an auxiliary selecting a participle is not a VR-verb. In Bernese a modal is an obligatory VR-verb, while an auxiliary selecting a participle is an optional VR-verb. Thus both orders in (6a) and (6b) are possible in Lucernese, while only the order in (6b) is possible in Bernese. Both orders in (7a) and (7b) are acceptable in Bernese, while only the order in (7a) is acceptable in Lucernese. (6) a. . . .dass de Grendel tanze cha that the Grendel dance can ‘that Grendel can dance’ b. . . .dass de Grendel cha tanze that the Grendel can dance (7) a. . . .dass de Grendel gfrässe hät that the Grendel eaten has ‘that Grendel has eaten’ b. . . . .dass de Grendel hät gfrässe that the Grendel has eaten
Moreover, the type of constituents which are allowed to occur in a raised cluster is not necessarily the same across languages which allow VPR. In Swiss German, nonpronominal complements and adjuncts of the non-finite verb selected by a VR-verb can optionally occur in the raised cluster, while adjuncts modifying the VR-verb and unstressed pronouns are excluded from the raised cluster. In (8) the non-pronominal object t’Nationalhymne can optionally occur in the raised cluster, while the pronoun si is excluded from it (8b). (8) a.
. . .dass de Grendel t’Nationalhymne /si cha singe that the Grendel the-national anthem her can sing ‘that Grendel can sing the national anthem/it’
Manuela Schönenberger
b. . . .dass de Grendel cha t’Nationalhymne /*si singe that the Grendel can the-national anthem her sing
In Section 3 a brief overview of the acquisition literature relating to verb placement in embedded clauses is given. In Section 4 verb placement in adult Lucernese is compared to verb placement in child Lucernese.
3. Overview of acquisition literature on embedded verb placement The general conclusion that emerges in the acquisition literature on German is that German-speaking children do not experience any serious difficulties in acquiring the verb-final pattern typical of embedded clauses. Clahsen & Smolka (1985:148) state that “as soon as the first embedded clauses are used, the finite verb appears in sentencefinal position.” (For a detailed study based on a corpus of about 800 embedded sentences produced by seven German-speaking children see Rothweiler 1993; for a similar observation in Dutch see Krikhaar 1992.) Gawlitzek-Maiwald, Tracy & Fritzenschaft (1992) provide the first significant counter-evidence to the general claim that children place the verb correctly in embedded clauses in German. Their study is based on two monolingual children, one of whom sometimes places the finite verb incorrectly. Brandt & Diessel (2004) report on two children who produce verb-placement errors in relative clauses. Penner (1990) studies the verb placement of four 3 year-old children acquiring Bernese. One of these children makes verb-placement errors in embedded clauses, which seem to be restricted to relative clauses. Penner’s (1996) study is based on diary data and elicitation data from a Bernese-speaking child, age 1;11–4;11. The corpus comprises over 1100 utterances; most but not all of these are subordinate clauses. All embedded clauses show the verb-final pattern until around age 3;02, when the child starts to vacillate between moving the finite verb and leaving it in clause-final position. This stage lasts for several months before the child acquires the verb-final pattern. Swedish and Norwegian are Germanic V2 languages with SVO. Håkansson (1989), and Håkansson & Dooley Collberg (1994) show that Swedish children go through a developmental stage during which they place modals and auxiliary verbs, but not lexical verbs, incorrectly with respect to negation in embedded clauses. This developmental stage lasts from the onset of subordinate clauses around age 2;0 until about 3;06. Westergaard & Bentzen (2007) found errors in the production data of Norwegian children, who place finite verbs in embedded clauses incorrectly with respect to negation and adverbs. They note that incorrect verb movement across the subject is not attested in their data on embedded questions. Stromswold (1990), on the other hand, observes that English-speaking children occasionally produce ungrammatical SubjectAux inversion in embedded questions of English–the only modern Germanic non-V2 language.
Three acquisition puzzles
The Lucernese-speaking children of my study behave unlike any other children. They produce a large number of verb-placement errors, which are not restricted to a specific type of embedded clause, and they do so over an extended period of time. Moreover, any type of finite verb – be it lexical, an auxiliary or a modal – is affected.
4. Comparing input and output In this section the children’s verb placement in embedded clauses is compared to the adult’s. After a brief note on data collection I shall concentrate on the following three questions: (i) The child moves the finite verb to the left in embedded clauses, although Swiss-German generally uses the verb-final pattern in this context. How clearly does the input flag the verb-final pattern? (ii) V2 is an option in clauses introduced by wil ‘because’. The child exclusively uses this option before age 4;11. Does the input contain many examples with V2 introduced by wil? (iii) Even after the child starts to use the verb-final pattern productively, auxiliary verbs are still often misplaced: they precede the participle, resulting in the sequence Auxiliary Participle, which is not allowed in the target grammar. Are there any structures in the input containing a finite auxiliary in which the auxiliary precedes a non-finite verb?
4.1 A note on data collection The data of the Lucernese children, called Moira and Eliza, were extracted from audio recordings made by Moira’s mother. A tape recorder was used to record the children while they were playing or when Moira interacted with other members of her family, mainly her mother. In total there are 100 tapes of 90 minutes each, amounting to 150 hours of recorded data. Most of these data pertain to Moira. Her corpus consists of over 5500 embedded clauses produced spontaneously and several hundred elicited embedded clauses. Eliza’s corpus is much smaller, but still contains over 600 spontaneous embedded clauses and several hundred elicited embedded clauses. The elicited data come from elicitation sessions starting at age 4;11 and were carried out over several months. During these elicitation sessions the child was asked to repeat a sentence uttered by the mother. These sentences contained a matrix clause and an embedded clause with the verb-final pattern introduced by different subordinators. To analyse the acquisition of verb placement, all embedded clauses produced by the children were transcribed, and spontaneous production data were classified separately from elicited data. To increase accuracy in transcription, each tape was listened to at least twice and inconsistencies in transcription were corrected. All the transcriptions were prepared by the author. To examine whether Moira’s verb placement in matrix clauses is target-consistent, three samples of her natural production data were transcribed and compared to her mother’s.
Manuela Schönenberger
The children appear to behave alike in all relevant aspects so I focus on Moira’s data, of which considerably more is available than for Eliza. My examination of the input is based on two samples of speech in which adults (primarily Moira’s mother) interact with Moira, once at the beginning of the study, when she is 3;10 and once at the end, when she is 8;01. Each input sample contains over 400 embedded sentences. I chose the two ages in order to determine whether the adults used less complex sentences when the child was young; the earliest recordings of Moira’s speech were made at 3;10. I found no significant differences in the input, as noted in the discussion of Tables 1.1 and 1.2 below. A detailed discussion of the spontaneous and elicited data and classification of embedded clauses into monthly bins between age 4;11–6;01 can be found in Schönenberger (2001). The input sample studied in Schönenberger (2001), which was based on 8 tapes selected at random, only partially overlaps the input samples discussed here.
4.2 The occurrence of the verb-final pattern in the input The verb-final pattern is clearly predominant in the adult’s verb placement in embedded clauses. Only a few of the examples produced by the adults show verb movement, i.e., embedded V2, as in (9). Note in passing that example (9a) does not involve direct speech, since the finite verb sig in the bracketed constituent is in the subjunctive not in the indicative. However, even when adults use the verb-final pattern, it is not always clearly marked as such, which I consider a potential source of confusion for the child. It is unclear what input patterns might be confusing to the child. I hypothesize that there are two potential confusing factors: (i) the pattern superficially looks like embedded V2, i.e., the subject (SU) is the only constituent which occurs between the subordinating element (sub.) and the finite verb (Vfin), and (ii) one or more constituents occur to the right of the finite verb, as a result of ‘extraposition’ rather than verb movement to the left. This is discussed below. (9) embedded V2: a. Si seit [si sig ä Prinzessin]. she says she is.subj a princess b. Ich ha ned gwüsst [was lauft zersch]. I have not known what runs first ‘I didn’t know what is showing first (at the cinema).’ c. . . . [wil du wötsch-en jo vermischt] because you want-it jo mixed ‘because you prefer it mixed (with water)’
(mother) (father)
(mother)
The examples in (10) do not provide the child with a clear cue about the verb placement used in embedded clauses, since they do not contain enough information. They consist of a subordinator, a subject, and a finite (intransitive) verb, but no other constituent, e.g., an adverbial, to the left of the finite verb which could have given an unambiguous cue to the child about the verb placement used in the sentence.
Three acquisition puzzles
(10) sub. SU Vfin: a. Du chasch au anesitze und verzelle [wenn’t wotsch]. you can also down-sit and talk if-you want ‘You can also sit down and chat if you want to.’ b. I weiss ned [wo si wohnet]. I know not where they live ‘I don’t know where they live.’
(mother)
(mother)
The examples in (11) might confuse the child, because the finite verb does not occupy the clause-final position. The word order in these examples might be particularly confusing, since only one constituent, namely the subject, occurs between the subordinator and the finite verb, which looks like embedded subject-initial V2. In contrast to the examples in (9), which show embedded subject-initial V2, the examples in (11) are really verb-final: the fact that a string of words – abbreviated as ‘XP’ – ends up to the right of the finite verb is not the result of verb movement to the left, as is the case in the examples in (9), but due to the extraposition of a PP in (11a) and VPR in (11b). (11) sub. SU Vfin XP: a. [wenn iich schaffe im Huus] when I work at home b. [wo-n-er sind go Velo fahre] when-you.pl are go bicycle ride ‘when you went cycling’
(mother) (mother)
Some examples which unambiguously mark the verb-final pattern are given in (12). In (12a) the finite verb is transitive and is preceded by its non-pronominal object. In (12b) the finite verb is a particle verb. Moreover, the finite verb in both examples occupies the ‘absolute’ clause-final position. (12) sub. . . . Vfin: a. Was mue-mer säge [wemmer ka Linse wett]? what must-one say if-one no lentils wants ‘What should one say if one doesn’t want any lentils?’ b. [Wenn du ufschtoosch] gömmer voruse. when you up-stand go-we out ‘When you get up we go outside.’
(mother)
(mother)
In contrast to the examples in (12), those in (13) are less clearly marked as being verbfinal, because a string of words occurs to the right of the finite verb. As opposed to the examples in (11), however, the word order in (13) does not superficially look like embedded subject-initial V2, since another constituent besides the subject occurs between the subordinator and the finite verb. In (13a) the perfect participle gseh intervenes between the subject and the finite auxiliary händ, and in (13b) the sentential negation ned follows the subject and precedes the finite modal wett.
Manuela Schönenberger
(13) sub. SU . . . Vfin XP: a. ä Fee weisch [wommer gseh händ i de Muus] a fairy know.2sg that-we seen have in the mouse ‘a fairy–you know–that we saw in [the TV programme] “[Die Sendung mit] der Maus”’ b. [wil si ned wett nass werde] because she not wants wet get ‘because she doesn’t want to get wet’
(mother)
(mother)
The different word order patterns discussed above are listed separately in Tables 1.1 and 1.2 below, which summarize the results of an analysis of two of Moira’s input samples, one early and one late.3,4 Even without quantitative statistical comparison it is clear from the tables that there is little difference in the rate at which these patterns occur in the two input samples. The majority of the embedded sentences produced by the adults in the early input sample display the verb-final pattern ‘Sub. . . . Vfin’, which occurs in 301 of 460 examples, i.e., 65%. The neutral pattern ‘Sub. SU Vfin’ is attested in 52 of 460 examples, i.e., 11%. Examples in which the finite verb does not occupy the clause-final position, either due to extraposition of an XP – ‘Sub. SU Vfin XP’ (43 examples) and ‘Sub. . . . Vfin XP’ (53 examples) – or because of actual verb movement resulting in V2 (11 examples) amount to (43 + 53 + 11)/460, i.e., 23%. In the late input sample the patterns are distributed as follows: the verb-final pattern is 69% (269/391), the neutral pattern is 12% (47/391), and the patterns which might be confusing for the child is 19% (28 + 38 + 9/391). However, Moira presumably based her hypotheses of verb placement on earlier input than that contained on the first recordings, i.e., my early input sample. Of course the earlier input is unknown, but given that there are no significant differences between the ‘early’ and the ‘late’ recorded input samples, and that the main person interacting with Moira before age 3;10 in any case is her Mother, it seems likely that the prior input was not substantially different with respect to word order in embedded clauses. Indeed, if the ‘very early’ input did flag the verb-final pattern less clearly, this would imply that (i) the mother used many more intransitive verbs than in these samples, (ii) that if she used modifiers she would predominantly put them in the postverbal position, and (iii) that she produced many more examples with verb movement in clauses with wil and wh-complements. This seems unlikely.
3. For reasons of space, V2-complements of bridge verbs (9a) are omitted from these tables. There are 19 in the ‘early’ input sample and 18 in the ‘late’ input sample. V2-complements of bridge verbs are never introduced by a complementizer, while the verb-final complements of bridge verbs are. Moreover, there are two relative clauses in the early input sample-none in the late input sample-which are introduced by die and show the V2 pattern. These are also omitted from Table 1.1. 4. The abbreviation ‘compl.’ stands for complementizer. Only the complementizers wenn ‘when, if ’, dass ‘that’, and öb ‘whether, if ’ are taken into account, as other complementizers are rarely used.
Three acquisition puzzles
Table 1.1. Early Input (to Moira): adult-child (3;10–4;02) interactions V-position V2 Sub. SU Vfin Sub. SU Vfin XP Sub. . . . Vfin Sub. . . . Vfin XP Total
compl.
wil ‘because’
0 19 17 143 31 210
7 0 3 23 4 37
relative wo 0 7 14 103 13 137
wh
Total
4 26 9 32 5 76
11 52 43 301 53 460
Table 1.2. Late Input (to Moira): adult-child (7;08–8;01) interactions V-position V2 Sub. SU Vfin Sub. SU Vfin XP Sub. . . . Vfin Sub. . . . Vfin XP Total
compl.
wil ‘because’
relative wo
wh
Total
0 14 11 137 22 184
7 1 0 20 3 31
0 8 8 72 10 98
2 24 9 40 3 78
9 47 28 269 38 391
Table 2. Adult-Adult interactions in a related dialect V-position
compl.
V2 Sub. SU Vfin Sub. SU Vfin XP Sub. . . . Vfin Sub. . . . Vfin XP Total
0 13 9 117 18 157
wil ‘because’
relative wo
wh
Total
8 0 0 13 0 21
0 2 5 58 13 78
2 8 1 15 2 28
10 23 15 203 33 284
I also recorded two adult speakers of a related Swiss German dialect interacting with each other and analysed the data to see how the patterns are distributed. As shown in Table 2 the patterns used by these two speakers are the same, which is expected, but they also occur with roughly the same frequency.5 The rate at which clear-cut verb-final examples are found is 72% (=203/284). The neutral pattern occurs in 8% (=23/284) of the data, and examples in which the finite verb does not occur clause-finally comprise 20% (=(15 + 33 + 10)/284) of the data. These are clearly consistent with the 69%, 12%, and 19% proportions respectively in Moira’s late input and hence validate the claim that this input is indeed adult-like. The same is true of the early input sample.
5. There are 57 examples with bridge verbs selecting a V2-complement, which are not listed in the table.
Manuela Schönenberger
4.2.1 The occurrence of verb movement in the output Given what we know about the input, we would expect Moira to produce mainly the verbfinal pattern. Even if we sum all examples which do not clearly mark the verb-final pattern, i.e., all examples in which the finite verb does not occupy the clause-final position and all neutral examples, the clear-cut verb-final pattern still outweighs these by a factor of 2. However, our expectation is not borne out: Moira moves the finite verb in any type of embedded clause, often giving rise to patterns which are not attested in the target grammar. In fact, both Moira and Eliza move the finite verb in clauses introduced by a complementizer, as in (14) and (15), and in relative clauses (16), resulting in verb-placement errors, since these contexts are incompatible with verb movement. (Child examples in which the verb placement is inconsistent with the target grammar are preceded by the diacritic %.) (14) a. %[Wenn wot-r bisle] denn muesch au mitgo. if wants-he pee then must.2sg also with-go ‘if he wants to pee then you have to go with him’ b. %Da isch de Sattel [dass wird si do ned dreckig]. this is the saddle that gets she here not dirty ‘This is the saddle so that she doesn’t get dirty here.’ c. %Ich luege immer [öb chunt öper oder ned]. I look always if comes somebody or not ‘I always look whether somebody is coming or not.’ (15) a. %[Wenn du trinksch öppis]. . . when you drink something b. %Weisch du [dass du häsch es Halstuech aa]? know you that you have a shawl on ‘Do you know that you are wearing a shawl?’ c. %Mues halt luege [öb ich find nomel öpis]. must halt look if I find again something ‘I have to see whether I can find something else.’ (16) a. b. c. d.
%Isch das dine Fuess [wo chützelet mine]? is this your foot which tickles mine %mit Schmirfarbe [wo chammer wider usmache] with smear colours which can-one again out-make ‘with colours which one can rub out again’ %einisch [wo hät t’Eliza öpis wele lose] once when has the-Eliza something want(ed) listen ‘once when Eliza wanted to listen to something’ %de Güselchübel [wo chammer alles dristecke] the dustbin where can-one everything in-put ‘the dustbin where one can put in everything.’
(E:4;02)
(M:4;03) (M:4;05)
(M:4;01) (E:5;00)
(E:4;10)
(M:4;03)
(E:3;11)
(M:4;04)
(M:4;00)
Both children also generally apply verb movement in a context in which it is an option in the target grammar, namely in clauses introduced by wil ‘because’ (17) and in
Three acquisition puzzles
wh-complements (18). However, verb movement in these contexts does not always sound acceptable, as it might violate constraints on discourse. When Moira answers her mother’s question (17b) she provides the reason why Tanja receives a present (17b’), and hence she should have used the verb-final pattern. Verb movement in the wh-complement in (18a) sounds acceptable, because the matrix predicate can license an interrogative complement. In (18b), on the other hand, the wh-complement cannot be interpreted as an indirect question, and verb movement is excluded. (17) a. aber violetti ned [wil [die hämmer scho]] but purple not because those have-we already ‘but not purple ones because those we have already’ b. Wiso chunt das ‘Tanja öber? ‘Why does Tanja receive this?’ b’. %[Wil [si hät zum Geburtstag]] because she has to-the birthday ‘because it’s her birthday’ (18) a. Etz wömmer luege [wär hät recht]. now want-we look who has right ‘Now we shall see who is right.’ b. %Los [wie tue-n-ich rede]! listen how do-I talk ‘Listen to how I talk!’
(M:4;06)
(mother) (M:3;11)
(M:4;07)
(E:3;10)
Besides examples in which it can clearly be determined whether verb movement has applied or not, there are also examples where this is not possible. These examples are referred to as ‘ambiguous’: they are structurally ambiguous between a derivation with verb movement and a derivation with the verb-final pattern. Note that many of these examples are only ambiguous because Moira (or Eliza) produced them rather than an adult, and the child uses verb movement, while the adult does not. This is the case for the ambiguous examples with a complementizer (19a/20a) and the relative clauses (19b/20b). All of these structurally ambiguous examples sound grammatical. (19) Sub. SU Vfin: a. [öb si chunt] if she comes b. Mir gfallet alli [wo-n-ich mache]. me.dat please all which-I make ‘I like all the ones which I make’ (20) Sub. SU Vfin XP: a. [wenn ich wör mit eme lange Schtegge ässe] if I would with a long stick eat ‘if I ate with a long stick’ b. di einzig [wo-n-ich ha mit gäl] the only one which-I have with yellow
(M:4;07) (M:4;10)
(M:4;05)
(M:4;06)
Manuela Schönenberger
Tables 3.1, 3.2, and 3.3 quantify the verb placement used by Moira in spontaneous speech in clauses introduced by a complementizer, in relative clauses, and in whcomplements, respectively (see appendix for more detailed tables). As can be deduced from these tables, Moira hardly produces the verb-final pattern before age 4;11. Discarding the ambiguous cases (191 examples), the verb-final pattern is found in 22 of 1051 (2%). Table 3.1. Moira’s verb placement with complementizers Age 3;10–4;11 5;00–8;01
V-movement
no V-movement
ambiguous
Total
84.9% (541) 9.4% (144)
2.7% (17) 71.6% (1093)
12.4% (79) 19.0% (290)
637 1527
ambiguous
Total
21.5% (77) 24.8% (207)
358 835
ambiguous
Total
Table 3.2. Moira’s verb placement in relative clauses Age 3;10–4;11 5;00–8;01
V-movement 77.4% (277) 14.5% (121)
no V-movement 1.1% (4) 60.7% (507)
Table 3.3. Moira’s verb placement in wh-complements Age 3;10–4;11 5;00–8;01
V-movement
no V-movement
85.4% (211) 12.4% (71)
0.4% (1) 46.6% (266)
14.2% (35) 41.0% (234)
247 571
The patterns used by both children before age 4;11 are listed in Figure 1. The diacritic % precedes patterns which are not found in the target grammar, and the diacritic # precedes patterns which are possible in the target grammar but impose a restriction on the interpretation of the sentence. The two verb-placement patterns in clauses introduced by a complementizer, which are non-target consistent, rely on the occurrence of an overt subject. Thus only examples with overt subjects can be used to determine how often a given pattern occurs. In Swiss German a 2nd sg pronoun can be dropped in matrix clauses, but not in embedded clauses. When a 2nd sg subject pronoun is not used contrastively, it can surface as an enclitic on the subordinator. When it is clearly audible, it is t. But often the enclitic undergoes assimilation, which is difficult to hear in the recordings. Other subject pronouns are clearly audible, even when they are unstressed and surface as enclitics on the subordinator. Moreover, many relative clauses are subject relative clauses and hence do not contain an overt subject. The same holds for whcomplements. Tables 4.1–4.3 contain the figures only for examples in which verb movement has applied and in which the subject (SU) is clearly audible. In clauses introduced by a complementizer, the subject can either precede or follow the finite verb. The former order is very rare in relative clauses and wh-complements before
Three acquisition puzzles Clauses introduced by a complementizer a.
% complementizer Vfin SU …
b.
% complementizer pronominal SU Vfin …
Relative clauses a.
% wo Vfin (SU) …
b.
% wo pronominal SU Vfin … (very rare)
Wh-complements a.
# Wh-constituent Vfin (SU) …
b.
% Wh-constituent pronominal SU Vfin … (very rare)
Complements of bridge verbs (without dass ‘that’) and clauses with wil ‘because’ a.
bridge verb … [XP Vfin … ]
b.
# wil [XP Vfin … ]
Figure 1. Moira and Eliza’s verb-placement patterns before age 4;11.
Table 4.1. Verb movement in Moira’s comp clauses with overt SU Age 3;10–4;11 5;00–8;01
comp V SU . . .
comp SU V . . .
Total
74.7% (357) 26.2% (34)
25.3% (121) 73.8% (96)
478 130
Table 4.2. Verb movement in Moira’s relative clauses with overt SU Age
wo V SU . . .
wo SU V . . .
Total
3;10–4;11 5;00–8;01
98.3% (175) 81.8% (54)
1.7% (3) 18.2% (12)
178 66
Table 4.3. Verb movement in Moira’s wh-complements with overt SU Age
wh V SU . . .
wh SU V . . .
Total
3;10–4;11 5;00–8;01
96.8% (184) 88.3% (53)
3.2% (6) 11.7% (7)
190 60
Manuela Schönenberger
age 4;11. In all of the examples in which the subject precedes the finite verb it is pronominal. After age 4;11 there is a sharp increase in the use of this pattern in clauses introduced by a complementizer. The subject preceding the finite verb can now also be non-pronominal. In summary, Moira and Eliza produce many examples with verb movement, although the verb-final pattern is clearly attested in the input. In clauses introduced by a complementizer two different verb-placement patterns obtain, depending on whether the subject is pronominal or not. The subject type seems to be irrelevant in the context of relative clauses and wh-complements; the finite verb immediately follows the relative complementizer wo or the wh-constituent.
4.3 Verb placement in clauses introduced by wil ‘because’ Clauses introduced by wil ‘because’ can be construed with the V2 pattern as well as the verb-final pattern. The difference in verb placement is paired with a difference in meaning, as mentioned in Section 2. The V2 pattern with wil, shown in (21a), is found in 21% (7 of 34) of cases in the early input sample, and in 23% (7 of 30) of cases in the late input sample. Thus in the majority of clauses with wil the adults use the verb-final pattern (21b). (21) a. b.
Es isch en Hund gsi [wil ich ha gseh] dass-er it is a dog been because I have seen that-he es Halsband aahät. a collar on-had ‘It was a dog because I saw that it was wearing a collar.’ De Drache hät müese brüele [wil s Ei verbroche the dragon has have cry because the egg broken isch]. is ‘The dragon had to cry because the egg was broken.’
(mother)
(mother)
Most of the wil-clauses produced by Moira show the V2 pattern before age 4;11. The use of V2 is possible in (22a’): Moira provides the reason why she thinks that the child (in the picture book) is a girl. In many cases, however, her use of the V2 pattern sounds odd, as in (22b’), where she provides the reason why she is eating a specific type of food. (22) a. A was a gsehsch dass’ es Meitli isch? ‘How can you recognize that it’s a girl?’ a’. [wil s hät en lange Rossschwanz] because she has a long pony-tail b. Worum issisch du da? ‘Why are you eating this?’ b’. %[wil ich ha s gern]. ‘Because I like it.’
(mother) (M:3;10) (mother) (M:4;01)
Three acquisition puzzles
The first constituent in Moira’s V2 sentences can also be a non-subject, shown in (23): (23) a. a’. b. b’.
“Chasch jo min Fründ sii.” reading from a book: “You can be my friend.” [wil dem isch’s jo au langwilig, oder] because him is-it jo also boring or ‘because he is bored too, right’. Worum weisch das? ‘How come you know this?’ [wil do isch au es Meitli]. because here is also a girl ‘Because here there’s a girl as well.’
(mother) (M:4;05)
(mother) (M:3;10)
Moira’s data are summarized in Table 5. When clauses introduced by wil start to appear in Eliza’s data, Eliza first shows a clear preference for V2 over the verb-final pattern as does Moira: Eliza uses V2 15 times before she produces the first verb-final example with wil at age 5;04. Table 5. Moira’s verb placement in clauses introduced by wil ‘because’ Age 3;10–4;11 5;00–8;01
V2
wil … Vfin
ambiguous
Total
55.4% (188) 33.8% (230)
1.2% (4) 43.0% (293)
43.4% (147) 23.2% (158)
339 681
In summary, as in all the other embedded contexts, Moira generally moves the verb in clauses introduced by wil before age 4;11, but the pattern which results is general V2: the first constituent is not restricted to being a subject. In the input samples the V2 pattern can also be found, but the use of the verb-final pattern seems much more common. A gradual preference for the verb-final pattern can be observed in Moira’s data after age 5;05 (see Table 8 in the appendix). The frequency of use of V2 vs. V-final in corpora of German-speaking adults can vary quite substantially, see e.g., Eppler (1999); Uhmann (1998). However, a speaker’s use of a given pattern depends on the interpretation he wishes the sentence to convey. Moira appears to use the V2 pattern indiscriminately before age 4;11. Interestingly, the German-speaking children of Rothweiler’s (1993) study clearly prefer the verb-final pattern to the V2 pattern: out of 77 examples the verb-final pattern occurs in 68 examples, the V2 pattern in 9 examples. In all other embedded contexts these German-speaking children use the verb-final pattern, in stark contrast to Moira and Eliza.
4.4 The sequence ‘Aux non-finite verb’ vs. ‘non-finite verb Aux’ Swiss German has a number of auxiliary-like elements which can trigger Verb Projection Raising (VPR). When VPR applies, the non-finite complement occurs to the right of the finite auxiliary-like element:
Manuela Schönenberger
(24) a. b.
[dass t’Cheshire Chatz wöt gschnell verschwinde] that the-Cheshire cat wants quickly vanish ‘that the Cheshire cat wants to vanish quickly’ [dass-mer kum cha sis Lache vergässe] that-one hardly can his smile forget ‘that one can hardly forget his smile’
A finite auxiliary selecting a participle can optionally trigger VPR in some dialects, e.g., Bernese. Thus the order ‘Participle Aux’ as well as the order ‘Aux Participle’ is acceptable in Bernese. In the variant of Lucernese the children are acquiring a finite auxiliary selecting a participle cannot trigger VPR and therefore it always follows the participle, i.e., ‘Participle Aux’. However, in verbal clusters with more than two verbs the finite auxiliary usually precedes the infinitival complement in both Bernese and Lucernese, resulting in the sequence ‘Aux Infinitive Infinitive’.6 In the early input sample there are 118 cases with ‘Participle Aux’ order, shown in (25). There are 14 cases which superficially look as if the finite auxiliary were a VRverb – the non-finite complement is ‘extraposed’, i.e., occurs to the right of the finite auxiliary – as in (26). These non-finite verbs are not participles, though, but infinitives. Thus 11% (14 of 132) of all the strings with a finite auxiliary co-occurring with a nonfinite verb might encourage Moira to analyse auxiliaries as VR-verbs, independent of whether they select a participle or an infinitival complement. The proportion is the same in the late input sample, i.e., 11% (12 of 113). (25) a. b.
[wenn s ned probiert häsch] if it not tried have.2sg ‘if you haven’t tried it’ [dass de Chnopf ufgange isch] that the bud opened is ‘that the bud opened’
(mother)
(26) a. b.
[dass’ nüm hät chöne laufe] that-it no more has be able run ‘that it was no longer able to run’ [dass’ geschter sind go schwimme] that-they yesterday are go swim ‘that they went swimming yesterday’
(mother)
(mother)
(mother)
Moira seems to treat finite auxiliaries selecting a participle as optional VR-verbs. In (27) the finite auxiliary occurs to the left of the participle (and optionally of its complement and modifier), which is excluded in the target grammar.
6. The Swiss-German dialects display the IPP (infinitivus pro participio) effect, which means that in a configuration where a participial form is expected an infinitival form appears instead. This is usually the case when a verb embedded under an auxiliary itself selects an infinitival form.
Three acquisition puzzles
(27) a. %s Chind . . . [wo Geburtstag hät gha] the child who birthday has had ‘the child who had his/her birthday’ a’. %[Wo s mit sine Auge hät die schöne Fraue gseh] when it with its eyes has the beautiful women seen ‘when it saw the beautiful women with its own eyes’ b. %[Wil ich nonig uf de Welt bi gsi] bisch du ano because I not yet on the world am been are you also äleige gsi. alone been ‘Because I hadn’t been born you were alone too.’ b’. %S erschte mol [wo-n-ich no bin mit em Mami gange] the first time when-I still am with the mum gone ‘the first time when I still went with my mum’
(M:7;06)
(M:7;11)
(M:7;08)
(M:7;09)
This finding emerged only after Moira started to produce the verb-final pattern. Before age 4;11 any type of finite verb could be affected by verb movement. Thus finite auxiliaries, modals, and lexical verbs often surface in positions in the child data where they are never found in the target grammar. The fact that Moira treats finite auxiliaries as VR-verbs makes an analysis of her verb placement more complex. Since VPR exists in the target grammar, finite modals can occur in non-clause-final position, not because they have undergone movement to the left, but because their complement has undergone extraposition to the right. The same applies to finite auxiliaries selecting an infinitival complement. Given that Moira also treats finite auxiliaries selecting a participle as VR-verbs, examples such as (27), which are ruled out in Lucernese, but possible in Bernese, must be classified as ‘ambiguous’ rather than as cases of verb movement. Moira’s data are summarized in Table 6. The embedded clauses with the order ‘Aux SU PP’ are clearly the result of verb movement, abbreviated as ‘V-mvt’ in the table, because the finite auxiliary precedes the subject, which is generally excluded from the raised cluster. The order ‘SU Aux PP’ is also likely to involve verb movement, since the constituent to the right of the finite auxiliary contains at least one constituent
Table 6. Moira’s sequences of Aux PP and PP Aux (in embedded contexts with overt SU) Age 3;10–4;11 5;00–8;01 Total
Aux SU PP (v-mvt) 160 34 194
SU Aux PP (V-mvt) ambiguous 9 13 22
22 83 105
SU…Aux PP (verb-final)
SU…PP Aux (verb-final)
0 70 70
1 384 385
Manuela Schönenberger
besides the participle, which is generally excluded from the raised cluster, as in (28): (28) a. %[dass si hät’s gseh] that she has-it seen ‘that she saw it’ b. %[wemmer hät sich als Maa ine Frau verchlaidet] when-one has self as man in-a woman up-dressed ‘when one, as a man, dresses up as a woman’
(M:4;08)
(M:4;11)
The label ‘ambiguous’ refers to examples with the word order ‘SU Aux (XP) PP’, which would all be acceptable in Bernese: (29) a. %[wo-n-ich ha pauet] which-I have built b. %[wo du häsch Tomate gern gha] when you have tomatoes dear had ‘when you liked tomatoes’
(M:4;10) (M:5;05)
The examples of the type in (27), which are again permitted in Bernese, contain at least one constituent other than the subject between the subordinator and the finite auxiliary. I regard these as involving the verb-final pattern, i.e., the finite auxiliary occurs to the left of the participle (and its complement/s and modifier) as a result of VPR. Sequences in which the participle precedes the finite auxiliary clearly display the verb-final pattern. It is less clear whether Eliza treats the auxiliaries ha ‘have’ and si ‘be’ as optional VRverbs. In her spontaneous data there are only 4 cases which support this assumption.
5. Intermediate grammar and parameter setting Moira and Eliza’s verb placement in embedded clauses differs from the target grammar but is nevertheless highly regular. This is evidence that they have set the parameter governing verb placement, but to an incorrect value, i.e., [+movement]. This setting is correct for matrix clauses, but not for embedded clauses. Around age 4;11 the verbfinal pattern finally emerges in Moira’s data and gradually replaces verb movement. For several months she wavers between moving the finite verb and leaving it in clausefinal position. In Eliza’s data the verb-final pattern emerges a couple of months later. My final recordings at age 8;01 show that both children’s verb placement is fully targetconsistent. The implication is that they have reset the verb-placement parameter to the correct value, i.e., [–movement] for embedded clauses. Given the unexpected data on verb placement in Swiss German the question arises whether there are other cases discussed in the acquisition literature in which children
Three acquisition puzzles
produce patterns inconsistent with the target grammar. There are. I shall only mention two. Thornton (1990) and (2004) discusses two such cases in English. One concerns the production of medial wh-questions (30) by some of the English-speaking children she studied in an experimental setting. The other involves the non-inversion in whyquestions (31) in the spontaneous production of one English-speaking child. (30) a. %What do you think what Cookie Monster eats? b. %Who do you think who Grover wants to hug? (31) a. %Why Sophie has a magic bed? b. %Why Daddy’s never coming to watch?
(A.L.:4;05) (A.L.:4;03)
Crain & Thornton (1998:38) note that “[t]he appearance of the medial-wh in the language of children learning English cannot be explained as children’s response to the input. Although these constructions are not grammatically well formed in English, structures like [30] are attested in dialects of German [32] and other languages.” They conclude that “English-speaking children’s non-adult productions therefore support the Continuity Hypothesis” and that “the medial-wh phenomenon constitutes a serious challenge to the Input Matching Model.” (32) Wer glaubst du wer nach Hause geht? who believe you who to home goes ‘Who do you believe is going home?’
Thornton (2004) observes that non-inversion with why closely resembles the behaviour of perché ‘why’ in Italian and concludes that the child’s non-inversion with why is also in agreement with the continuity hypothesis. Before turning to a language whose patterns resemble those found in the child data I will show that except for verb placement the word order used by Moira is entirely consistent with the target grammar.
5.1 Essential similarities with word order in the target grammar Moira’s intermediate grammar of course shows many similarities to the target grammar. The order of the constituents she produces in embedded clauses is the order adopted by adults except for verb movement, e.g., most of the embedded clauses in the target grammar are also subject-initial. Only in 2% of all verb-final clauses in the input samples, 20 of ((460 – 11) + (391 – 9)), does the subject not immediately follow the subordinator. Moreover, her non-target-consistent verb placement in clauses introduced by a complementizer shows two different patterns, depending on whether the subject is pronominal or not. In the target grammar, unstressed subject pronouns must be adjacent to the finite verb in non-subject-initial clauses, while non-pronominal subjects can be separated from it, as shown in (33b). Unstressed subject pronouns must also be adjacent to the subordinator in embedded clauses, while non-pronominal subjects can be non-adjacent,
Manuela Schönenberger
as in (34b). In Moira’s clauses introduced by a complementizer an unstressed subject pronoun either immediately follows the complementizer, as in the target grammar (34a), or immediately follows the moved finite verb, as in matrix clauses of the target grammar (33a). Thus the distribution of pronominal subjects in Moira’s data is target-consistent. (33) a. b.
Geschter hät-er /de Fido wohrschinlech wider yesterday has-he the Fido probably again Geschter hät wohrschinlech *er/de Fido wider yesterday has probably he the Fido again ‘Probably he/Fido ate again yesterday.’
(34) a. b.
[dass-er /de Fido döt schlooft] that-he the Fido there sleeps [dass döt *er /de Fido schlooft] that there he the Fido sleeps ‘that he/Fido sleeps there’
gfrässe. eaten gfrässe. eaten
Despite these similarities between Moira’s grammar and the adult’s, there is one crucial difference: Moira moves the finite verb in both root and embedded contexts, while the adult does so in root contexts only. Although there is verb movement in matrix clauses, Moira’s verb movement in embedded clauses does not coincide with the order used in matrix clauses, namely V2. There are two main differences between matrix V2 and embedded verb movement in Moira’s clauses introduced by a complementizer. (i) In her embedded clauses the finite verb is often immediately adjacent to the complementizer, and (ii) if the finite verb has moved but is not adjacent to the complementizer the only constituent that can ever intervene is a pronominal subject (and, later – after age 4;11–also a non-pronominal subject). In a root context, on the other hand, Moira produces many non-subject-initial V2 sentences (128 of the 296 matrix clauses at age 3;10 are nonsubject-initial). Thus Moira’s verb placement in embedded clauses, although V2-like, should not be confused with genuine V2. In other words, Moira does not produce the order: complementizer XP Vfin . . ., which is typical of V2 languages like Icelandic and Yiddish (cf. Vikner 1995). Let us now consider clauses introduced by wil ‘because’, wh-complements and relative clauses. Moira’s wil-clauses with verb movement show genuine V2. Moira uses topicalization (as well as drop, i.e., subject topic drop and expletive drop) in wil-clauses, which is possible in matrix clauses as well as wil-clauses with V2 in the adult grammar. What she is slow to discover is the fact that the V2 pattern and the verb-final pattern in wil-clauses are usually construed with different readings. This is presumably pragmatics not syntax. Pragmatic development can be delayed, see e.g., Chien & Wexler (1990) and Thornton & Wexler (1999) for the Principle B lag, Schulz (2003) for presupposition failure, Wexler (to appear) for the a –> the error. Westergaard (2003), however, shows that children are sensitive to at least certain aspects of pragmatics–the distinction
Three acquisition puzzles
between known vs. new information – very early in their language development. Before age 4;11, Moira’s wh-complements all involve V2. Again she seems to use V2 indiscriminately in this context, regardless of whether the matrix predicate can license an interrogative complement. In this respect her grammar resembles Belfast English. Henry (1995) shows that wh-complements in this variety of English are generally compatible with I-to-C movement, resulting in V2. Moira’s relative clauses introduced by wo, which is a head in the target grammar, do not pattern with clauses introduced by a complementizer but with wh-complements, which suggests that she treats wo as a wh-constituent, i.e., a maximal projection. In contrast to the cases discussed above, Moira’s very late misplacement of finite auxiliaries is not linked to verb movement, but to her analysis of finite auxiliaries as optional VR-verbs, a possible setting for Bernese. To reset the parameter to the Lucernese value appears to be a rather lengthy process, given that she produces the Bernese order Auxiliary Participle until the age of 8;0.
5.1.1 The presence of verb movement in the input Since both Moira and Eliza appear to have set the verb-placement parameter to [+movement] before age 4;11, I have attempted to determine to what extent verb movement is present in the input. In the transcription of the first recording, when Moira was age 3;10, the adults interacting with her produced 411 matrix clauses and 104 embedded clauses. All the matrix clauses show verb movement, while only 12 of the embedded clauses do so: there are 11 V2 complements of bridge verbs and 1 relative clause, introduced by the relative operator die (see footnote 2). Therefore, in the input, structures with verb movement amount to 82% (423 of 515) of the total and outweigh those without verb movement by a factor of 4. Moreover, of the 92 embedded clauses with the verb-final pattern only 62 unambiguously flag the verb-final pattern, thus occurring only in 12% (62 of 515) of the input clauses. However, in Moira’s embedded examples with complementizers only those in which a constituent is interspersed between the complementizer and the moved finite verb result in a V2 pattern. Since the intervening constituent is always a subject, this is subject-initial V2 rather than general V2. As discussed above, the overwhelming number of sentences show verb movement in the input, but in how many of these is the constituent preceding the finite verb a subject? Of the 411 matrix clauses 249 are subject-initial, and of the embedded clauses with verb movement all 12 are subjectinitial. Of the remaining 92 embedded clauses with the verb-final pattern 20 superficially look like subject-initial V2 (cf. examples (10) and (11)). Thus 55% (281 of 515) of the sentences show the word order (sub) SU Vfin (XP), which coincides with that used by Moira in many of her embedded clauses introduced by a complementizer. One might hypothesize that Moira produces subject-initial V2 in this context because subject-initial V2 is more frequent in the input than non-subject-initial V2, or alternatively because verb movement in clauses introduced by a complementizer cannot give rise to non-subject-initial V2 in her grammar. I assume the latter, because it is a
Manuela Schönenberger
natural choice, and suggest that the complementizer occupies the position the finite verb moves to in topicalized constructions. Even so, I can conclude that the fact that verb movement is present in the majority of sentences, due to the predominance of matrix clauses, can mislead a child into setting the verb-placement parameter to [+movement]. However, it is unclear why children other than Moira and Eliza in the literature are not similarly misled.
5.2 Partial V2 in Old and Middle English matrix clauses The children’s V2-like patterns in embedded clauses are reminiscent of those found in Old and Middle English matrix clauses. (For discussions of diachronic English data see van Kemenade (1987); Pintzuk (1991); Kroch & Taylor (1997); Roberts (1997); Haeberli (2002); among many others.) Old and Middle English matrix clauses with an operator, e.g., wh-questions, usually show the V2 pattern, i.e., the finite verb immediately follows the operator. In Moira’s wh-complements with verb movement the finite verb also immediately follows the wh-constituent. Her relative clauses pattern alike, i.e., the finite verb immediately follows wo, which suggests that Moira treats the relative complementizer wo as the relative operator itself. The fact that relative wo is homophonous with the wh-constituent wo ‘where’ may have encouraged her to do this. Moreover, the relative operator is usually null. What is more striking is the fact that clauses introduced by a topic generally show V2 order in Old English, provided the subject is nonpronominal. If the subject is pronominal, it usually precedes the finite verb. This closely resembles what we find in Moira’s clauses introduced by a complementizer before age 4;11. In Middle English, non-pronominal subjects could also precede the finite verb in clauses with topicalization. Similarly, around age 4;11, non-pronominal subjects start to appear between a complementizer and the moved finite verb in Moira’s data.
5.3 Derivation of verb movement in the child grammar My analysis of the children’s verb movement is very similar to that developed in Fischer et al. (2000) for Old and Middle English, but was developed independently (Schönenberger 2001). Based on Rizzi (1997) I adopt a minimally Split-CP system and propose that verb movement in clauses introduced by a complementizer targets a lower head in the C-domain than in relative clauses and wh-complements. Fischer et al. also adopt a Split-CP and suggest that verb movement in topic-initial matrix clauses targets a lower head than in clauses with an operator, i.e., wh-questions, negative-initial sentences, etc. Before age 4;11 the finite verb in clauses introduced by a complementizer moves to the same head in the C-domain, i.e., Fin°. Between age 3;10 and 4;04 pronominal and nonpronominal subjects generally follow the finite verb, as shown in (35) and (36). Around age 4;04 pronominal subjects start to occur in between the complementizerand the moved finite verb. I suggest that pronominal subjects can now optionally move into [Spec,FinP],
Three acquisition puzzles
(35)
ForceP Force
FinP
dass Spec
Fin' Fin
TP
keiV Spec
…
ichSU
nüme tSU um tV
%dass kei ich nüme um that fall I no longer over ‘that I no longer fall over’
(36)
(M:4;02)
ForceP Force
FinP
wenn Spec
Fin' Fin
TP
hätV Spec de TigiSU
%wenn hät de Tigi Hunger when has the Tigi hunger ‘when Tigi is hungry’
… tSU Hunger tV
(M:4;00)
as in (37), or remain in [Spec,TP] (38), while non-pronominal subjects stay in [Spec,TP], as in (36). Following Roberts (1995/6) I assume that [Spec,FinP] can only host light elements such as pronouns. In Fischer et al.’s analysis verb movement in topic-initial clauses targets the same head as in the children’s clauses introduced by a complementizer. However, in my analysis movement into the highest functional head in the C-domain is blocked because of the complementizer, while in Fischer et al.’s analysis it is empty, because topics are maximal projections, not heads, which have moved into the highest Spec
Manuela Schönenberger
position.7 Moroever, I assume movement of pronouns into [Spec,FinP] to be optional, whereas for Fischer et al. this movement is obligatory in Old and Middle English. ForceP
(37) Force
FinP
öb Spec
Fin'
erSU Fin
TP
gsehtV Spec
… tSU ä Tür tV
t'SU
%öb er gseht ä Tür if he sees a door ‘if he sees a door’
(M:4;08)
ForceP
(38) Force
FinP
dass Spec
Fin' Fin
TP
haV
Spec
…
ichSU
tSU kei Auge tV
%dass ha-n-ich kei Auge that have I no eyes ‘that I don’t have eyes’
(M:4;06)
7. Since clauses introduced by wil allow two different patterns in the child as well as the adult grammar, I assume that wil occupies the complementizer position in verb-final wil-clauses, whereas in wil-clauses with V2 it is a subordinating conjunction rather than a complementizer which occupies the head of Conj(unction)P and selects a ForceP complement. Inside the ForceP complement a topic moves into [Spec,ForceP] and the finite verb moves into Force, resulting in non-subject-initial V2. Whether subject-initial clauses are FinPs or TPs, as proposed by e.g., Zwart (1997) for Dutch, requires further investigation.
Three acquisition puzzles
Verb movement in relative clauses and wh-complements results in just one single word order: ‘wo Vfin . . .’ (39) and ‘wh Vfin . . .’ (40). Since nothing generally intervenes between the subordinator and the finite verb – not even a pronoun – verb movement must have targeted a higher head in the C-domain, namely Force°. This results in strict V2, characteristic of operator contexts in the diachronic data in English, for which Fischer et al. also suggest that the finite verb moves to the highest head in the C-domain. ForceP
(39) Spec
Force'
woOBJ Force
FinP
machtV Spec
Fin' Fin
TP
t'V
Spec
…
de AladinSU
tSU tOBJ tV
%das, wo macht de Aladin that which makes the Aladdin ‘that which Aladdin makes’ (40)
(M:3;10)
ForceP Spec
Force'
wivilOBJ Force
FinP
chuntV Spec
Fin'
erSU Fin
TP
t'V
Spec
…
t'SU
tSU tOBJ öber tV
%wivil chunt er öber how much gets he verbal particle ‘how much he gets’
(M:3;10)
Manuela Schönenberger
Around age 4;11 non-pronominal subjects start to appear between the complementizer and the moved finite verb, which I take to show that the finite verb targets a lower functional head than Fin°, shown in (41). I assume that the complementizer is now base-generated in Fin° and then moves up to Force° as in the target grammar; dass ‘that’ contains the features [+declarative], [+finite]. The trace in Fin° blocks verb movement to that position. It is also only at this age that Moira starts to produce examples with the verb-final pattern, which indicates that the verb has not undergone any movement. This is of course consistent with the target grammar. ForceP
(41) Spec
Force' Force dassk
FinP Spec
Fin' Fin
TP
tk DPh
T'
t‘Rahel T
VP
hätV th
V' DP drü Chind
%Weisch du [dass t’Rahel hät drü Chind]? know you that the-Rahel has three children ‘Did you know that Rahel has three children?’
tV
(M:4;11)
Fischer et al. also link the appearance of preverbal non-pronominal subjects in Early Middle English to a change in verb movement properties. Ultimately lexical verbs will cease to move in English, as is the case in Modern English, where only modal verbs, auxiliaries, and periphrastic do can undergo movement. In summary there are three close parallels between the Swiss-German child data and the diachronic data in English: (i) In operator-contexts both show strict V2, i.e., both pronominal and non-pronominal subjects follow the finite verb. Wh-questions
Three acquisition puzzles
and relative clauses contain operators, and verb movement targets the highest head in the C-domain. (ii) In non-operator contexts V2 does not obtain, and pronominal subjects distribute differently from non-pronominal subjects. Sentences with topicalization in the diachronic data in English and clauses introduced by a complementizer in the child data are non-operator contexts. Verb movement targets a lower head in the C-domain. (iii) Some time during the Middle English period non-pronominal subjects start to occur in preverbal position in non-operator contexts. This is also true of the child data after age 4;11. Verb movement no longer targets a head in the C-domain.
6. Concluding remarks Moira and Eliza behave unlike any other children discussed in the acquisition literature: they use verb movement in a context where it is usually banned. By concentrating on Moira’s data, which far exceeds that of Eliza, I have tried to show that the verb placement she uses in embedded clauses is the result of misset parameters. She sets the verb-placement parameter for both matrix and embedded clauses to [+movement]. This setting gives rise to target-consistent V2 in matrix clauses, and to partial V2 in embedded clauses. Partial V2 is not found in the target grammar, but is attested in matrix clauses of Old and Middle English. By examining two input samples I have tried to show that although the verb-final pattern is sometimes masked in embedded clauses, the majority of embedded clauses unambiguously show the verb-final pattern. It is impossible to determine the precise input for any given child, but one is left with the distinct impression that Moira partially ignores cues in the input. Her intermediate grammar, however, is not the result of simply mapping the word order found in matrix clauses to embedded clauses, which would have resulted in embedded general V2 or embedded subject-initial V2. This can be taken as evidence against an input matching model. My analysis in terms of parameter missetting is compatible with the continuity hypothesis, although it conflicts with the assumption that children set ‘basic’ wordorder parameters very early. That Moira takes so long to abandon her verb-movement grammar suggests that it is difficult to recover from a misset parameter. Even after she has switched to the verb-final pattern, finite auxiliaries continue to surface in unusual positions until age 8;0. This can be seen as another misset parameter, albeit non-basic, pertaining to VR-verbs, where the initial value is set to Bernese rather than Lucernese. Before age 5;06 Moira predominantly uses V2 in clauses introduced by wil ‘because’, even in contexts where the verb-final pattern would have been more appropriate. This suggests that Moira is not yet aware that the difference in verb placement is coupled with a difference in interpretation, which is consistent with current understanding that certain aspects of pragmatic knowledge are slow to mature.
Manuela Schönenberger
References Brandt, S. & Diessel, H. 2004. The development of relative clauses in German. Poster presented at CLS, University of Bristol. Chien, Y.-C. & Wexler, K. 1990. Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition 1: 225–295. Clahsen, H. & Smolka, K.-D. 1985. Psycholinguistic evidence and the description of V2 phenomena in German. In Verb Second Phenomena in Germanic Languages, H. Haider & M. Prinzhorn (eds), 137–167. Dordrecht: Foris. Crain, S. & Thornton, R. 1998. Investigations in Universal Grammar: A Guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge M: The MIT Press. Eppler, E. 1999. Word order in German-English mixed discourse. http://www.phon.ucl.ac.uk/ publications/WPL/99papers/eppler.pdf. Fischer, O., van Kemenade, A, Koopman, W. & van der Wurff, W. 2000. The Syntax of Early English. Cambridge: Cambridge University Press. Gawlitzek-Maiwald, I., Tracy, R. & Fritzenschaft, A. 1992. Language acquisition and competing linguistic representations: the child as arbiter. In The Acquisition of Verb Placement. Functional Categories and V2 Phenomena in Language Acquisition, J. Meisel (ed.), 139–180. Dordrecht: Kluwer. Günthner, S. 1996. From subordination to coordination? Verb-Second position in German causal and concessive constructions. Pragmatics 6(3): 323–356. Haeberli, E. 2002. Inflectional morphology and the loss of V2 in English. In Syntactic Effects of Morphological Change, D. Lightfoot (ed.), 88–106. Oxford: OUP. Haeberli, E. & Haegeman, L. 1995. Clause structure in Old English: Evidence from negative concord. Journal of Linguistics 31: 81–108. Haegeman, L. 1998. Verb movement in embedded clauses of West Flemish. Linguistic Inquiry 29: 631–656. Haegeman, L. & van Riemsdijk, H. 1986. Verb projection raising, scope and the typology of verb movement rules. Linguistic Inquiry 17: 417–466. Håkansson, G. 1989. The acquisition of negative placement in Swedish. Studia Linguistica 43(1): 47–58. Håkansson, G. & Dooley Collberg, S. 1994. The preference for modal + neg. An L2 perspective applied to L1 acquisition. Second Language Research 10(2): 95–124. Henry, A. 1995. Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: Oxford University Press. Kemenade, A. van. 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Koopman, H. & Szabolcsi, A. 2000. Verbal Complexes. Cambridge MA: The MIT Press. Krikhaar, E. 1992. Voeg woordloze bijzinnen in kindertaal. MA thesis, University of Utrecht. Kroch, A. & Taylor, A. 1997. Verb movement in Old and Middle English: Dialect variation and language contact. In Parameters of Morphosyntactic Change, A. van Kemenade & N. Vincent (eds), 297–325. Cambridge: Cambridge University Press. Penner, Z. 1990. On the acquisition of verb placement and verb projection raising in Bernese Swiss German. In Spracherwerb und Grammatik. Linguistische Untersuchungen zum Erwerb von Syntax und Morphologie, M. Rothweiler (ed.), Linguistische Berichte Sonderheft 3: 166–189. Penner, Z. 1996. From empty to doubly-filled complementizers: A case study in the acquisition of subordination in Bernese Swiss German. Working Papers of the University of Konstanz 77.
Three acquisition puzzles Pintzuk, S. 1991. Phrase Structures in Competition: Variation and Change in Old English Word Order. PhD dissertation, University of Pennsylvania. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar: A Handbook of Generative Syntax, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer. Roberts, I. 1995/6. Remarks on the Old English C-system and the diachrony of V2. In Language Change and Generative Grammar, E. Brandner & G. Ferraresi (eds), Special Issue of Linguistische Berichte 7: 154–168. Roberts, I. 1997. Directionality and word order change in the history of English. In Parameters of Morphosyntactic Change, A. van Kemenade & N. Vincent (eds), 397–426. Cambridge: Cambridge University Press. Rothweiler, M. 1993. Der Erwerb von Nebensätzen im Deutschen: Eine Pilotstudie. Tübingen: Niemeyer. Schmid, T. & Vogel, R. 2004. Dialectal variation in German 3 verb clusters: a surface-oriented Optimality theoretic account. Journal of Comparative Germanic Linguistics 7: 235–274. Schönenberger, M. 2001. Embedded V-to-C in Child Grammar: The Acquisition of Verb Placement in Swiss German. Dordrecht: Kluwer. Schönenberger, M. & Penner, Z. 1995. Cross-dialectal variation in Swiss German: doubling verbs, verb-projection raising, barrierhood and LF movement. In Studies in Comparative Germanic Syntax, H. Haider, S. Olsen & S. Vikner (eds), 285–305. Dordrecht: Kluwer. Schulz, P. 2003. Factivity: Its Nature and Acquisition. Tübingen: Niemeyer. Stromswold, K. 1990. Learnability and the Acquisition of Auxiliaries. PhD dissertation, MIT. Thornton, R. 1990. Adventures in Long-distance Moving: The Acquisition of Complex Wh-questions. PhD dissertation, University of Connecticut, Storrs. Thornton, R. 2004. Why continuity. In Proceedings of the 28th Boston University Conference on Language Development, A. Brugos, L. Micciulla & C.E. Smith (eds), 620–632. Somerville, Mass: Cascadilla Press. Thornton, R. & Wexler, K. 1999. Principle B, VP Ellipsis, and Interpretation in Child Grammar. Cambridge MA: MIT Press. Uhmann, S. 1998. Verbstellungsvarianten in ‘weil’-Sätzen: Lexikalische Differenzierung mit grammatischen Folgen. Zeitschrift für Sprachwissenschaft 17(1): 92–139. Vikner, S. 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxofrd University Press. Wegener, H. 1993. Weil – das hat schon seinen Grund: Zur Verbstellung in Kausalsätzen mit weil im gegenwärtigen Deutsch. Deutsche Sprache 4: 289–305. Westergaard, M. 2003. Word order in wh-questions in a North Norwegian dialect: some evidence from an acquisition study. Nordic Journal of Linguistics 26(1): 81–109. Westergaard, M. & Bentzen, K. 2007. The (non-)effect of input frequency on the acquisition of word order in Norwegian embedded clauses. In Frequency Effects in Language Acquisition: Defining the Limits of Frequency as an Explanatory Concept, I. GÜlzow & N. Gagarina (eds) 271–306. Berlin/New York: Mouton de Gruyter. Wexler, K. 1999. Very early parameter setting and the unique checking constraint: A new explanation of the optional infinitive stage. In Language Acquisition: Knowledge Representation and Processing, A. Sorace, C. Haycock & R. Shillock (eds). Special Issue of Lingua: 23–79. Wexler, K. To appear. Cues don’t explain learning: Maximal trouble in the determiner system. In Proceedings of CUNY Sentence Processing Conference, T. Gibson & N. Pearlmutter (eds.) Zwart, C.J.-W. 1997. Morphosyntax of Verb Movement: A Minimalist Approach to the Syntax of Dutch. Dordrecht: Kluwer.
Manuela Schönenberger
Appendix The data in the following tables are subdivided according to what I regard as developmental steps in Moira’s acquisition: (i) age 3;10–4;04: pronominal and nonpronominal subjects appear to be treated in the same way, (ii) age 4;05–4;11: pronominal subjects are treated differently from non-pronominal subjects, (iii) age 5;00–5;05: competition between a verb movement grammar and a non-verb movement grammar, and (iv) age 5;06–8;01: the non-verb-movement grammar wins over the verbmovement grammar. Table 7.1. Moira’s verb-placement patterns in clauses introduced by a complementizer containing an overt subject (SU) Age 3;10–4;04 4;05–4;11 5;00–5;05 5;06–8;01
comp V SU …
comp SU V …
comp … V
ambiguous
93.6% (161) 49.5% (196) 8.1% (28) 0.6% (6)
4.1% (7) 28.8% (114) 9.8% (34) 5.8% (62)
1.7% (3) 3.5% (14) 61.6% (213) 74.3% (787)
Total
0.6% (1) 172 18.2% (72) 396 20.5% (71) 346 19.3% (204) 1059
Table 7.2. Moira’s verb-placement patterns in relative clauses containing an overt subject (SU) Age
wo V SU …
wo SU V…
wo … V
3;10–4;04 4;05–4;11 5;00–5;05 5;06–8;01
100% (85) 76.9% (90) 32.0% (32) 5.6% (22)
0% (0) 2.6% (3) 1.0% (1) 2.8% (11)
0% (0) 0% (0) 48.0% (48) 59.4% (233)
ambiguous 0% (0) 20.5% (24) 19.0% (19) 32.1% (126)
Total 85 117 100 392
Table 7.3. Moira’s verb-placement patterns in wh-complements containing an overt subject (SU) Age
wh V SU …
wh SU V …
wh … V
3;10–4;04 4;05–4;11 5;00–5;05 5;06–8;01
100% (64) 76.9% (120) 33.3% (24) 6.4% (29)
0% (0) 3.8% (6) 0% (0) 1.5% (7)
0% (0) 0.6% (1) 33.3% (24) 47.6% (217)
ambiguous 0% (0) 18.6% (29) 33.3% (24) 44.5% (203)
Total 64 156 72 456
Table 8. Moira’s verb placement in clauses introduced by wil ‘because’ Age 3;10–4;04 4;05–4;11 5;00–5;05 5;06–8;01
V2 54.7% (70) 55.9% (118) 44.2% (106) 28.1% (124)
wil … Vfin
ambiguous
Total
0% (0) 1.9% (4) 27.1% (65) 51.7% (228)
45.3% (58) 42.2% (89) 28.7% (69) 20.2% (89)
128 211 240 441
The acquisition of universal quantifiers in Spanishi Linda Escobar & Vicenç Torrens This chapter deals with experimental research on Spanish children’s interpretation of universal quantifiers. A potentially difficult asymmetry is highlighted: quantifiers in preverbal or postverbal position in contexts of Clitic Left Dislocation. We show that five-year-olds easily obtain the wide scope interpretation of postverbal quantifier subjects, unless they are presented with biased matching pictures (with an impaired object or an impaired agent). We give support to the continuity assumption that children and adults share a common core of linguistic knowledge. The distribution of the chapter is as follows. In section 1, we present an interesting asymmetry observed in the literature of universal quantifiers in adult Spanish. In section 2, some insightful background on acquisition of quantifiers is provided. Section 3 deals with the methodology of the experiment. The results are left for section 4. Section 5 is intended as the final discussion.
1. On the distribution of subject quantifiers 1.1 Uribe-Etxebarria (1995) In search for some insight into the issue of the analysis of preverbal subjects in Spanish, where both SVO and VSO word orders alternate, Uribe-Etxebarria studies the scopal properties of quantifiers in both preverbal and postverbal subject positions.1 Her conclusion is that preverbal subjects are not base generated where they appear.
i. The experimental results reported in this paper form part of a larger project on the acquisition of binding in Spanish done together with Sergio Baauw and Bill Philips. We would like to thank the children that participated in the study, their parents and teachers at the Colegio Aldeafuente in Madrid. 1. In the Generative literature of the 80’s and early 90’s, most linguists working on the so-called pro-drop parameter adopted Koopman and Sportiche (1991) thesis that Nominative Case assignment takes place under government of V (cf. Solà 1992 and all references there-in).
Linda Escobar & Vicenç Torrens
In particular, the claim is supported on the intricate phenomenon of long wh-movement with respect to the scopal properties of two universal quantifiers: cada (every) and todo-el-mundo (everybody) embedded in preverbal or postverbal position. The observation is that while there are two possible readings for either quantifier in embedded postverbal subject position in (1a) and (1b), there is only one exclusive reading for them in embedded preverbal subject position in (1a’) and (1b’). (1) a. a’. b. b’.
¿A quién piensas que vio cada uno en el rally? Whom think-you that saw everyone at the rally ‘Who do you think that everyone saw at the rally?’ ¿A quién piensas que cada uno vio en el rally? Whom think-you that everyone saw at the rally ‘Who do you think that everyone saw at the rally?’ ¿Qué crees que ha comprado todo el mundo? What think-you has bought everybody ‘What do you think that everybody has bought?’ ¿Qué crees que todo el mundo ha comprado? What do you think that everybody has bought? ‘What do you think that everybody has bought?’
One of the two readings corresponding to the embedded postverbal subjects above implies the wide scope construal of the quantifier, by which the matrix wh-phrase “a quien” (whom) in (1a) or “que” (what) in (1b) are under the scope of “cada uno” (everyone) and “todo-el-mundo” (everybody) respectively. In both cases, the matrix wh-phrases do not have a unique reference. A possible answer for the question (1a) would be “It is John that Mary saw at the rally, and it is Peter that Carmen saw at the rally, etc.” A possible answer for the question (1b) might be “It is a pair of glasses that Mary bought, It’s a book that Ann bought, etc.” The second reading of (1a) and (1b) involves the narrow scope construal of the quantifiers, by which the matrix wh-phrases are no longer under the scope of the quantifiers. Hence, a second possible answer for (1a) would be “It is John that Mary, Peter and the rest (...) saw at the rally”, and a second possible answer for (1b) would be “It is a pair of glasses that Mary, and Ann and the rest (...) bought”. For the sentences (1a’) and (1b’), on the other hand, it is the wide scope construal of the quantifier that seems to be exclusively available. In order to explain such an asymmetry, Uribe-Etxebarria (1995) argues that overt movement of the quantifier to the preverbal subject position is constrained by the S-structure generalization in (2).
(2) A quantifier that has been moved overtly at S-structure cannot enlarge its scope by undergoing covert movement at LF.
For a language like English, on the other hand, the fact that an ambiguous interpretation is obtained with the embedded preverbal quantifier subjects in (3) seems to suggest that the S-constraint in (2) is inapplicable for this language. (3) a. Who do you think everyone saw at the rally ? b. What do you think everybody bought ?
The acquisition of universal quantifiers in Spanish
However, this linguist proposes another way to address such a parametric difference. In particular she assumes that it is the nature of [SpecIP] which varies from language to language. That is, [SpecIP] is an A-bar position in Spanish, but not in English, as formulated in (4):
(4) [SpecIP] is an A-position in English but an A-bar-position in Spanish.
1.2 Some problems for Uribe-Etxebarria’s solution In what follows, some theoretical and empirical problems for Uribe-Etxebarria’s analysis will be discussed. According to minimalist assumptions (cf. Chomsky 1995, chapter four), one should abandon the idea that structures are already constructed beforehand. In a nutshell, the only way to form structures is by means of merging linguistic expressions from numeration. Hence, Uribe-Etxebarria’s approach based on a functional distinction among languages should be reconsidered. For another thing, there are only two levels of syntactic representation: PF and LF. Hence, the above S-constraint does not hold in this new model of syntax. A second empirical problem for Uribe-Etxebarria is that it predicts that the quantifier subject freely A-bar binds a pronoun. Contreras (1991) in his study of Spanish resumptive pronouns shows that when the operator locally binds a pronoun the result is an ungrammatical sentence as in (5). The only way to save it is by replacing the pronoun with a syntactic variable: (5) *¿Qué libroi loi compró _ María? vs. ¿Qué libro compró _ María? ‘Which book did Mary buy it?’
(cf. Contreras 1991: 11a)
A similar analysis has been provided elsewhere for the cases of postverbal Operators. On the assumption that they involve Quantifier Raising at LF, they will end binding the pronoun in the subject, and this is why both the English sentence in (6a) and its Spanish counterpart in (6b) are ruled out:2
2. In the traditional generative literature, QR is considered to be a marked option, even though it must occur in order to explain certain similarities between constraints on overt movement and restrictions on the scope of elements which have not undergone visible movement. On the other hand, it has also been suggested for many other languages that it is sometimes difficult to perceive the readings Quantifier Raising generates (cf. Gil 1982). As Reinhart (1995) puts it, the conceptual problem with QR is the concept of markedness. If one tries to address the question whether the interpretation of quantification requires obligatorily an operation of QR, one has to avoid the idea that QR involves markedness. On the contrary, it seems to be required by logical necessity, otherwise certain interpretation will never be obtained. The intuition that QR is a marked operation rests on the fact that it is harder to obtain scope outside the c-command domain. However, it might be the case that all kinds of performance factors determine why one interpretation is preferred over the other. Notice that it has been noted over the years that some examples are better than others and thus the wide scope interpretation of a quantifier outside its scope has remained as an empirical issue, rather than an theoretical issue.
Linda Escobar & Vicenç Torrens
(6) a. *His mother praised every speaker vs. Every speaker praised his mother b. *Su madre alabó a cada conferenciante vs. Cada conferenciante alabó a su madre
1.3 Quantifiers with Clitic Left Dislocation Without having been explicit on any claim with respect to the status of the preverbal subject position in a language like Spanish, Zubizarreta (1998) also observes an interesting asymmetry between preverbal and postverbal subjects in contexts of Clitic Left Dislocation. In a nutshell, Zubizarreta provides the minimal pair in (7) and argues that a left-dislocated pronoun anaphor can be bound by a quantifier subject unless this appears in a postverbal subject position: (7) a. A su hijo, cada padre lo acompañará el primer día de escuela ACC his son, every father ACC-CL will_accompany the first school day ‘Every father will accompany his son the first school day’ b. *A su hijo, lo acompañará cada padre el primer día de escuela ACC his son, ACC-CL will_accompany every father the first school day (cf. Zubizarreta 1998)
However, grammaticality judgements vary and many speakers accept the bound reading of the left-dislocated pronoun by the postverbal quantifier in (7b). This could be due to the fact that we are in front of a crossover configuration. In Zubizarreta’s analysis, we could not find any constraint for Quantifier Raising. In other words, the analysis does not explain what refrains the postverbal quantifier subject from raising to a c-commanding position at LF, and from binding the pronoun anaphor from this position. Recall the independent argument we offered in the previous section in favour of Quantifier Raising in the Spanish adult grammar. For the sake of the discussion, Cecchetto (1995) constructed a similar asymmetry with his Italian examples in (8). Cecchetto argues that covert movement of the postverbal quantifier as an instance of Quantifier Raising meets a structure proper of crossover and so the example (8b) is quite marginal: (8) a. b.
Il suo paper, ogni ragazzo l’ha discusso t ieri his paper, every boy it has discussed yesterday ‘every boy has discussed his paper yesterday’ *Il suo paper, l’ha discusso t ieri ogni ragazzo his paper, it has discussed yesterday every boy ‘every boy has discussed his paper yesterday’
The concept of markedness is also avoided by Chomsky (1995) in his minimalist program of chapter 4, where all syntactic machinery is reduced to the checking of some [+Interpretive] feature. In the next section, an approximation to QR in this minimalist framework is finally attempted.
The acquisition of universal quantifiers in Spanish
We then could assume the same for the Spanish CLLD-ed counterpart above, and argue that the ungrammaticality of Zubizarreta’s example in (7b) is due to the fact that it is an instance of crossover, which does not always produce the same grammaticality judgments in all native speakers, as is well known. Turning to the grammaticality of the Spanish example in (7a), we may assume with Ordóñez (1993) that preverbal subjects in Spanish must be analysed as left-dislocated and therefore they are not the result of movement from their canonical postverbal position into this preverbal position. Note that this alternative analysis also captures most Uribe-Etxebarria’s facts discussed before. In the first place, if Uribe-Etxebarria’s analysis is right and the preverbal subject position in Spanish (and presumably Italian) is an A-bar position, why does not the overt movement of the quantifier operator across a clitic pronoun or a reconstructed left-dislocated anaphoric expression considering Zubizarreta’s examples and Cecchetto’s analysis of preverbal subjects produce any grammatical effect, unlike the covert movement of the quantifier operator in the examples of postverbal subjects? Bearing these adult facts in mind, our main goal in our experiment on acquisition of universal quantifiers is to deal with the research questions in (9).
(9) a. whether children have any preference for a preverbal or postverbal quantifier subject in ordinary instances of Clitic Left Dislocation. b. whether children make errors in instances of Clitic Left Dislocation with postverbal quantifier subjects that correspond with biased pictures (cf. Freeman 1985, Philip 1995). c. whether children and adults differ in their interpretation of quantifiers in the two situations above.
Before dealing with the experiment, some background on acquisition of universal quantifiers is provided.
2. Some basics of acquisition of universal quantifiers 2.1 Five year-olds Although Chien & Wexler (1991) concluded that five-year-olds are able to understand ‘quantifier binding’ properly, Philip (1995) claimed that five-year-olds still make errors in interpreting the determiner quantifier ‘every’. The claim is that at this stage, children wrongly interpret quantifier determiners as the quantification adverb ‘always’ (cf. De Swart 1991). For C&W, the fact that five-year-olds do not make any errors and correctly interpret c-commanding quantifiers (e.g., every) as antecedents of pronouns is relevant to prove that they do not violate Principle B. In a nutshell, C&W claim that apparent Principle B violations sometimes reported in the literature of language acquisition are anything but a violation of a pragmatic principle that ensures coreference. On the
Linda Escobar & Vicenç Torrens
assumption that binding and coreference do not share the same grammatical module, the claim is that the module responsible for coreference relations has not been acquired by these subjects (principle P in C&W’s terms).3 On the other hand, coreference does not play any role in quantifier binding. So children are correctly predicted to make no errors here.4 According to Philip (1995) five-year-olds do not indeed make any errors when they are asked to respond to a matching picture, i.e., when ‘paired readings’ are all provided. However, if they are questioned about mismatch pictures, i.e., when asymmetric ‘paired readings’ are all provided, they do make some errors. The explanation for this is that unlike adults that prefer an interpretation of quantifiers based on ‘object quantification’, i.e., quantification over individual objects in the traditional sense, five-yearolds prefer an event quantification interpretation of the quantifiers, by which roughly speaking the quantification is extended to all the object participants.
2.2 Two types of errors Philip’s (1995) findings corroborate the two types of errors already reported in previous language acquisition literature of the universal quantifier cada ‘every’: Overexhaustive search (OS) errors and Underexhaustive search (US) errors (cf. Freeman 1985). In general, these errors are associated to the metalinguistic question of how children distribute the referents of one NP to the referents of a second NP. In most cases, these two errors are commonly elicited in picture verification tasks in which children are presented with a picture showing two groups of objects, for example, boys and horses, where each boy is uniquely paired with a different horse (cf. Philip 1995; Philip & Coopmans 1995; Drozd 1996). The picture either includes an extra horse, or an extra boy. First of all, OS errors are errors in which children incorrectly answer ‘no’ to the question Is every boy riding a horse?, when shown the picture in (10a). When asked to explain their answers, children usually point to the extra horse. US errors, on the other hand, are errors in which children incorrectly answer ‘yes’ to the same question Is every boy riding a horse? when shown picture (10b), as if the extra boy was irrelevant to the interpretation of the sentence. (10) a. boy1 riding horse1 b. boy2 riding horse2 boy3 riding horse3 horse4
boy1 riding horse1 boy2 riding horse2 boy3 riding horse3 boy4
3. However, see Grodzinsky & Reinhart (1993) for an alternative hypothesis addressing those errors found by Chien and Wexler (1991) as instances of lack of a processing load required to obtain the coreference interpretation. 4. See Grodzinsky & Reinhart (1993) for a similar argument on this.
The acquisition of universal quantifiers in Spanish
2.3 Awareness of WCO effects Crain & Thornton (1996) have shown, on the basis of one of their experiments, that Weak crossover structures as those exemplified in (11a) are never accepted by children, and so they only accept the bound interpretation of the pronoun with respect to the example in (11b). (11) a. *His rider brushed every horse b. Every rider brushed his horse
However, according to their results, adults sometimes accept the bound interpretation of the pronoun by the non c-commanding quantifier in those sentences exemplified in (11a).
3. The experiment 3.1 Rationale If Quantifier Raising is part of the early grammar of Spanish (the null hypothesis, given the assumption that Universal Grammar constitutes the initial stage), the choice for a postverbal or preverbal quantifier subject is in principle free (although it has an effect in the output). More crucially, subjects at all ages are expected to obtain the wide scope interpretation of postverbal quantifier subjects. Moreover, the fact that children make errors given a particular distribution of participants in the picture leads us to test both children and adults with respect to similar scenarios, to corroborate whether there are any differences between adults and children’s interpretation of quantifiers, as predicted by Philip (1995) and Philip and Coopmans (1995).
3.2 Subjects Thirty-seven children who ranged in age from 4;4 to 7;2, with a mean age of 5;4 took part in the experiment. Table in (12) gives the statistics of the 37 children included in (12)
Table: Age of subjects Age group
n
Age range
Mean age
4 yrs 5 yrs 6 yrs 7 yrs Total
3 15 16 2 1 37
4,42 to 4,10 5,2 to 5,5 5,6 to 5,9 6,17 to 6,25 7,25 4,42 to 7,25
4,4 5,1 5,7 6,1 7,2 5,4
Linda Escobar & Vicenç Torrens
the study. In addition, 29 adult subjects -most of them students in linguistics- participated as a control group in a collective session.
3.3 Method The experimental paradigm was essentially that of Chien & Wexler (1991), also followed by Philip & Coopmans (1995). The experimental task of the child was to answer a yes-no question, the CLLD sentence formulated as a question, with respect to a picture which illustrated a number of objects displayed in a distributive fashion, but not always symmetric.5 There were three different trials for a CLLD-Input with a preverbal quantifier subject, as illustrated by the control conditions in (13), and four different trials for a CLLDInput with a postverbal quantifier subject, as illustrated by the test conditions in (14). There was a total of 7 experimental items, presented in a single pseudo-random (maximally varied) order, each item consisting of a 21 x 29 cm color picture about which the relevant yes/no CLLD question was asked. The materials were counterbalanced as to whether a universal quantifier was used in the preverbal or in the postverbal position in the target input, and as to whether the expected adult response was “yes” or “no”. The used predicates were all transitive: recoger (collect), dar (give) and llevar (take):→ (13) Control Conditions target input a. b. c.
¿A su niña cada mamá la recoge M1→N1 en coche? M2→N2 (ACC her girl every mum ACC-CL M3→N3 collects by car) ¿Su cuaderno cada niña lo da a N1→C1→S1 la seño? N2→C2→S2 (ACC Her notebook every girl N3→C3→S3 ACC-cl gives to the teacher) ¿A su niña cada mamá la lleva en M1→ N1, N2 coche? M2 (ACC her girl every mum ACC-cl M3 → N3 takes by car)
(14) Test Conditions target input a.
picture-type adult response code yes
preY1
yes
preY2
no
preN
picture-type adult response code
¿A su niña la lleva cada mamá en M1→ N1 yes/no coche?’ M2→ N2 (ACC her girl ACC-CL takes every N3 mum by car)
postOS
5. For independent evidence on acquisition of Clitic Left Dislocation in Spanish, see Torrens and Wexler (2000) and Torrens, Escobar and Wexler (2006).
The acquisition of universal quantifiers in Spanish
b. ¿A su amigo lo lleva cada niña en N1 → A1 patinete? N2 → A2 (ACC her friend ACC-CL takes N3
yes/no postUS
every girl on skateboard) c. ¿A su niña la lleva cada mamá en M1 → N1, N2 no furgoneta? M2 (ACC her girl ACC-CL takes every M3 mum by van) d. ¿A su amigo lo lleva cada niña en N1 → A1 yes patinete? N2 → A2 (ACC her friend ACC-CL takes every girl on skateboard)
postN
postY
Each child was interviewed individually by two experimenters (Experimenter A and Experimenter B) in a separate room relatively free of distraction at the school. No tape recorder was used. Instead, Experimenter A took detailed notes of any relevant spontaneous productions made by the child, in addition to his yes/no answer. As for any special coding, the CLLD-input that corresponded on the answer-sheet to the picture with a ‘non-paired’ object or an extra agent variable condition was specially marked, so that a negative answer to this input was followed by the question ‘why did you say ‘no’?’ If the child then pointed to the extra object, the symbol PXO (points to an extra object) was circled in the answer-sheet. The child sat at a table. Experimenter A also sat at the table next to the child. Experimenter B sat at the table opposite the child and Experimenter A. The set of pictures was hidden face down. Experimenter A presented the experiment as a guessing game to the child, and discussed with the child what kind of objects were going to be shown in the picture: ‘mamás, niñas, niños, coches, furgonetas, patinetes’ (mums, girls, boys, cars, vans, skateboards...). Next, experimenter A showed each picture to the child pretending that Experimenter B was unaware of what was happening in the picture. Experimenter A, while pointing at each agent and patient in the picture, said the context setting statement to the child like ‘aquí hay tres mamas, tres niñas y tres coches’ (here you have three mums, three girls and three cars), in aloud voice, so that experimenter B was also aware of this. Next, experimenter A gave a kind of hint to Experimenter B by saying a transitive verb like: llevar (take home or drive home), recoger (pick up), and dar (give). Then, Experimenter B repeated the context setting statement of the picture (in fact, he read it on the back of the picture), and immediately he suggested an answer in form of a rhetoric question to the child (the relevant CLLD-input question). The child then agreed or disagreed by saying sí (yes) or no (no), respectively. Experimenter A recorded the child’s answer on the answer sheet, sometimes together with other interesting spontaneous productions.
3.4 The design of the experiment In this study of quantifiers, 66 subjects participated in the experiment: 37 children were interviewed individually, whereas 29 adults were interviewed in a collective session.
Linda Escobar & Vicenç Torrens
The two main variables to be measured were: first, a between-subjects variable (“Age”) with two levels: (“Child”, n = 37 and “Adult”, n = 29), and second a withinsubjects variable which we will refer to as the “Quantifier condition” variable with four, cf. (15) or seven levels for the Quantifier condition, cf. (16), depending on how we look at it. The term “Within-subjects” (or “repeated measures”) means that each subject received every level of the quantifier-condition variable. The order was randomized but all subjects received the same order. (15)
Four levels for the Quantifier condition: a. post-verbal with extra object, b. post-verbal with extra subject, c. post-verbal no frills, and d. pre-verbal no frills
(16)
Seven levels for the Quantifier condition: a. post-verbal with extra object (PostOS), b. post-verbal with extra subject (PostUS), c. post-verbal with “yes” response (postY), d. post-verbal with “no” response (postN), e. pre-verbal with “yes” response (preY1), f. pre-verbal with “yes” response (preY2), and g. pre-verbal with “no” response (preN)
Each data point was coded as 1 or 0 depending on whether the response was “correct” or “incorrect”, regardless of whether the correct answer consisted of a “yes” or a “no”.
4. Results The distribution of the data according to the seven-level version was as in table 1. As for the data concerning the four-level version the distribution was as in table 2. In either case, trials of the same type were averaged before statistics were calculated. The above data was analyzed according to an Omnibus ANOVA’s with all factors. For the seven-levels version, the age factor was coded under ‘between subjects’, or ‘children vs. adults’. According to our measures, children and adults performed Table 1. Statistical analysis of postverbal and preverbal quantifiers
postOS
postUS
postN
postY
preY1
preY2
preN
child (n=37) adult (n=29)
m = .78 sd = .417 m = .41 sd = .501
m = .59 sd = .498 m = .66 sd = .484
m = .92 sd = .277 m = .92 sd = .258
m = 1.0 sd = 0.00 m = 1.0 sd = 0.00
m = .97 sd = .164 m = .90 sd = .310
m = .92 sd = .277 m = .96 sd = .186
m = 1.00 sd = 0.000 m = .97 sd = .186
The acquisition of universal quantifiers in Spanish
Table 2. Statistical analysis of postverbal and preverbal quantifiers collapsed
postOS
postUS
post
pre
child (n=37) adult (n=29)
m = .78 sd = .417 m = .41 sd = .501
m = .59 sd = .498 m = .66 sd = .484
m = .96 sd = .138 m = .96 sd = .129
m = .96 sd = .104 m = .94 sd = .157
(significantly) differently from each other, i.e., children m = .88 (.321), vs. adult m = .83 (.374). Actually, children were more correct than adults overall (P < .05). As for the ‘Quantifier factor (within subjects)’, there was at least one significant difference among the quantifier condition types (P < 001). At least, there was also one significant interaction between quantifier condition and age of subject (P < .001). Since the omnibus ANOVA was not very informative for our specific hypotheses, we conducted an analysis based on focus contrast and we found eight different contrasts with the method of “L-scores” (cf. Rosenthal & Rosnow 1985) whereby each subject’s data point for each of the post- and pre-verbal subject conditions are assigned a contrast weight reflecting our hypothesis. For each subject an L-score was computed by formula L = S(xili), where xi the data point from a given cell and li is the contrast weight assigned to that cell’s column, as in Table 3. First, we wanted to establish Contrast 1, by which data of Post-verbal vs. Pre-verbal quantifier subjects were measured for all individuals. The null hypothesis in this case was that there is no difference for children between post-verbal and pre-verbal quantifier subjects. To test this, we used the method of “L-scores” (cf. Rosenthal & Rosnow 1985), whereby each subject’s data point for each of the post- and pre-verbal subject conditions are assigned a contrast weight reflecting our hypothesis as in Table 4. For each subject an L-score was computed by formula L = S(xili), where xi is the data point from a given cell and li is the contrast weight assigned to that cell’s column. For this first contrast, the post-verbal vs. pre-verbal difference was not significant. We obtained a mean L = .049 (i.e., small but slightly in the direction of favouring pre-verbal Table 3. L-scores for post- and pre-verbal subject conditions
postOS
L-scores n
postUS
postN
postY
preY1
preY2
preN
n
n
n
n
n
n
Table 4. Contrast weight for postverbal and preverbal quantifiers
postOS postUS
children 0
0
postN
postY
preY1
preY2
preN
–3
–3
2
2
2
Linda Escobar & Vicenç Torrens
subjects), and t(65) = –.3618 (df = n–1) P = .71866 (2-tailed). The conclusion was that subjects did not favour pre- or post-verbal quantifier subjects for their wide scope interpretation. Second, we established a Contrast 2 concerning the Extra object (PostOS) vs. other postverbal-quantifier conditions (PostY and PostN) for children. Here too, L-scores were used with weights –2, 0, 1, 1, 0, 0, 0 (see order above), where the Mean L = .351, and the t(28) = 2.494, whereas the P = .0174, value that we multiplied by 8 times, obtaining a figure of P =.1392, which meant that the difference was no longer significant. The conclusion we drew was that we can reject the null hypothesis that there is no difference between the extra object condition and the conditions with no extra (unpaired) items. Children did (significantly) worse in the condition with the extra object (means were .78 vs. .96 and this difference is significant). Next, we established a Contrast 3. This time, we were interested in the comparison between Extra object (PostOS) vs. other postverbal-quantifier conditions (PostY and PostN) for adults. We obtained that the same contrast weighted as above but for adult subjects. The relevant statistics consisted in a Mean L = 1.10, a t = 5.649, and a P < .0001 which we multiplied by 8 and obtained a value of P = .00003784, a still highly significant value. The conclusion we drew was that we could safely reject the null hypothesis that there was no difference between the extra object condition and the other postverbal condition for adults. They did significantly worse with an extra object. We continued exploring a Contrast 4, which involved the extra subject (PostUS) vs. other post-verbal quantifier conditions for children. Again, the L-scores used corresponded with weights 0, –2, 1, 1, 0, 0, 0 (see order above). We obtained a Mean L = 1.20, a t = 4.047, a P = .00026, that we multiplied by 8, and obtained a value of P = .00208, which was still highly significant. The conclusion was that children did worse in the extra subject condition than in the other post-verbal quantifier conditions. As for Contrast 5, we compared the Extra subject (PostUS) condition vs. other post-verbal conditions for adults. Again, the L-scores used consisted of the weights of 0, –2, 1, 1, 0, 0, 0 (see order above). We obtained a Mean L = .549, a t = 3.132, and a P = .004 which we multiplied by 8, and obtained a value of P = .032, which was still significant. The conclusion was that adults did worse with an extra subject in the picture, the prompt question having a post-verbal quantifier subject. Next, we attempted at measuring a Contrast 6. Now, children were compared vs. adults in the extra object condition only (PostOS). This meant a contrast within the between-subjects variable, so we treated it like any two-sample-test. The means can be garnered from the table of data above, but as we next show, the children had a higher
6. In interpreting each P-value, we keep in mind that a total of eight planned comparisons were performed. To correct for the possibility of finding differences due to chance, we mentally multiply the P-value found in each of our comparisons by 8, according to the so-called Bonferroni correction.
The acquisition of universal quantifiers in Spanish
mean than the adults, indicating that they did not make as many errors on this condition as the adults did: t = 3.3725, whereas P = .0013, which, as common practice, we multiplied by 8, obtaining a figure of P = .0104, still significant. The conclusion was that children did significantly better than adults in this condition. Yet, this could be an artifact. In any case, the fact that adults did not do it any better strongly argues against the hypothesis that the symmetrical interpretation response is a linguistic phenomenon of cognitive development (cf. Philip 1995, Philip & Coopmans 1995). In any case, it does not show any incomplete acquisition of an adult-like system of pragmatic principles. We also considered a Contrast 7, and children were compared vs. adults in the extra agent condition (PostUS) only. According to the data here, children and adults both did worse on this condition than in the other post-verbal quantifier conditions. Yet, looking back at the table above, adults did slightly better (66%) than children (59%). Here we test whether this difference is significant. We obtained a t = .81, and a P = .42, which as in the previous cases we multiplied by 8 times and obtained a value of P = 1, which meant that the difference between adults and children was not significant in this case. In this experiment, the correct response was varied between “yes” and “no” to control for a possible effect of subjects’ preferring one type of response over the other (e.g., children are said to have a “yes-bias”). In the following contrasts we measured whether this was true. Note that in the post-verbal quantifier conditions there was only one of each response type so it is impossible to conclude whether any differences (if found) reflected the answer type difference or some other difference between the two items. In the pre-verbal conditions there were two “yes” items and one “no” item so it may also be difficult to interpret any differences if found. We thus finally consider a Contrast 8 to measure all Yes vs. No responses for all conditions, and for all subjects. Again the L-scores were used with weights 0, 0, 3, –2, –2, –2, 3. We obtained a mean L = .027, a t = –.2245 and a P = .8231. The conclusion was that there were no differences between yes and no response trials for children.
5. Discussion As an answer to the research questions in (9), now in (17), and according to the results of the experiment, we can conclude that children at the mean age of 5;4 did not show any especial preference for a preverbal or postverbal quantifier subject in instances of CLLD. Moreover, children could interpret the wide scope reading of postverbal quantifier subjects. Actually, they only made mistakes when presented mismatched pictures, which is independently justifiable. Crucially, adults also did have some problems when presented mismatched pictures. This result is not fully predicted from a theory that distinguishes between children and adults with respect to their interpretation of quantifiers (cf. Philip 1995; Philip & Coopmans 1995).
Linda Escobar & Vicenç Torrens
(17) a. whether children have any preference for a preverbal or postverbal quantifier subject in ordinary instances of Clitic Left Dislocation. b. whether children commit errors in instances of Clitic Left Dislocation with postverbal quantifier subjects that correspond with biased pictures (cf. Freeman 1985; Philip 1995). c. whether children and adults differ in their interpretation of quantifiers in the two situations above.
First of all, like some children, some adults made OS errors when shown picture (18a). They incorrectly answered ‘no’ to the CLLD-Input (18b) below. In some additional sessions, they were asked to explain their answers, and it turned out that most of them pointed out at the lone girl, i.e., the ‘extra’ or ‘non-paired’ object (in our terms). (18) a. mum1 driving girl1 (home) mum2 driving girl2 (home) girl3 b. ¿A su niña la lleva cada mamá en coche? ACC her girl ACC-cl takes every mom by car? ‘Her girl every mom takes home?’
Secondly, some adults also made US errors when shown picture as in (18a). They also incorrectly answered ‘yes’ to the CLLD-Input as in (18b). In the same individual sessions, these adults when asked to explain their answer, did not show to put any attention to the ‘extra’ girl, i.e., the ‘non-paired’ agent (in our terms). For them, it was irrelevant to the interpretation of the sentence. (19) a. b.
girl1 carrying her friend1 by skateboard girl2 carrying her friend2 by skateboard girl3 ¿A su amigo lo lleva cada niña en patinete? (ACC her friend ACC-CL takes every girl by skateboard) ‘Her friend every girl takes by skateboard?’
As Freeman (1985) noticed for Child English, we did not find any special correlation between these two types of errors in Child Spanish or in Adult Spanish. Furthermore, no variation was found when subjects were shown non-ambiguous pictures, i.e., when neither ‘extra object(s)’ nor ‘extra agent(s)’ were included. For a CLLD-Input as provided in (20b), corresponding to the picture in (20a), a 100 % of the children and of the adult responses were ‘yes’. (20) a. b.
girl1 carrying her friend1 on skateboard girl2 carrying her friend2 on skateboard ¿A su amigo lo lleva cada niña en patinete? ACC her friend ACC-CL takes every girl on skateboard ‘every girl takes her friend on skateboard?’
The acquisition of universal quantifiers in Spanish
Previously we have seen that in adult grammar, the binding of a clitic pronoun by a quantifier operator results in an ungrammatical sentence as illustrated by Contrera’s (1991) example in (5), an instance of crossover. Likewise, the ungrammaticality of our example in (6b) was explained on the assumption that Quantifier Raising applies at LF, that is, in order for the quantifier to have scope over the clitic it has to move to the initial operator position. Similar crossover configurations were found with Zubizarreta’s (1998) example in (7b) and our example in (20b). The fact that both children and adults allow a wide scope reading of postverbal quantifiers in these cases suggests that like adults, children allow crossover, which is an unexpected result given Crain and Thornton’s (1996) claims for Child English. Notice that for the wide scope reading of the quantifier cada niña (every girl) over the clitic pronoun “lo” in the latter example, we require the quantifier to move across the clitic pronoun at LF. For the purpose of our investigation, we haven’t found any differences between adults and five-year-olds with respect to the reading of subject quantifiers. The position of quantifier subjects in Child Spanish doesn’t seem to differ much from the position of quantifier subjects in Adult Spanish in the CLLD-ed contexts tested here. Further research will only become more significant if the additional complexities of the Spanish clause structure we have discussed here are taken into account.
References Cecchetto, C. 1995. Reconstruction in clitic-left-dislocation. Ms DIPSCO. Milano: Instituto San Raffaele. Chien, Y.C. & Wexler, K. 1991. Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition 1: 225–295. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Contreras, H. 1991. On resumptive pronouns. In Current studies in Spanish linguistics, H. Campos & F. Martínez-Gil (eds). Washington DC: Georgetown University Press. Crain, S. & Thornton, R. 1996. Why child grammar is flawful. In Proceedings of Groningen Assembly on Language Acquisition, C. Koster & F. Wijnen (eds) 155–165. Groningen: University of Groningen. De Swart, H.E. 1991. Adverbs of Quantification: A Generalized Quantifier Approach. PhD dissertation, University of Groningen. Drozd, K.F. 1996. Quantifier interpretation errors as errors of distributive scope. In Proceedings of the 20th Annual Boston university Conference on Language Development 1, A. Stringfellow et al. (eds), 177–189. Somerville MA: Cascadilla. Freeman, N.H. 1985. Reasonable errors in basic reasoning. Educational Psychology 5: 239–249. Gil, D. 1982. Quantifier scope, linguistic variation and natural language semantics. Linguistics and Philosophy 5: 421–472. Grodzinsky, Y. & Reinhart, T. 1993. The innateness of binding and coreference. Linguistic Inquiry 24: 69–101 Koopman, H. & Sportiche, D. 1991. The position of subjects. Lingua 85: 211–258.
Linda Escobar & Vicenç Torrens Ordóñez, F. 1993. Postverbal asymmetries in Spanish. Ms, Graduate Center, City University of New York. Philip, W. 1995. Event quantification in the acquisition of universal quantification. Amherst MA: GLSA. Philip, W. & Coopmans, P. 1995. Symmetrical interpretation and scope ambiguity in the acquisition of universal quantification in Dutch and English. Ms, University of Utrecht. Reinhart, T. 1995. Interface Strategies [OTS Working Papers]. Utrecht: OTS. Rosenthal, R. & Rosnow, R. 1985. Contrast Analysis. Cambridge: Cambridge University Press. Solà, J. 1992. Subjects and Agreement. PhD dissertation, University of Barcelona. Torrens, V. & Wexler, K. 2000. The acquisition of clitic doubling in Spanish. In The Acquisition of Scrambling and Cliticization. S. Powers & C. Hamann (eds). Dordrecht: Kluwer. Torrens, V., Escobar, L. & Wexler, K. 2006. The analysis of experiencers and the external argument requirement hypothesis. In The Acquisition of Syntax in Romance Languages, V. Torrens & L. Escobar (eds). Amsterdam: John Benjamins. Uribe-Etxebarria, M. 1995. On the nature of SpecIP and its relevance for scope asymmetries in Spanish and English. In Contemporary Research in Romance Linguistics, J. Amastae et al. (eds) Amsterdam: John Benjamins. Zubizarreta, M.L. 1998. Topic, Focus, and Prosody. Cambridge MA: The MIT Press.
Appendix: Items for experiments Quantifiers in ordinary sentences a) Control Conditions ¿Lleva cada jirafa un globo? Does every giraffe carry a balloon? ¿Lleva cada cerdito una manzana? Does every pig carry an apple? b) Experimental Conditions ¿Lleva un niño todos los paraguas? Does a boy carry every umbrella? ¿Lleva un elefante todos los globos? Does an elephant carry all balloons?
picture-type
adult response
matched
yes
matched
yes
matched
no
matched
yes
c) Experimental Conditions on non-matched pictures ¿Monta cada niño un caballo? non-matched Does every boy ride a horse? “extra object” Overexhaustive errors ¿Ha cogido cada niño un cerdito? non-matched Has every pig boy picked up a pig? “extra agent” Underexhaustive errors
yes
yes
The acquisition of universal quantifiers in Spanish
¿Montan todos los niños un elefante? Do all the boys ride an elephant? Underexhaustive errors ¿Llevan todos los niños un paraguas? Do all the boys carry an umbrella? Overexhaustive errors ¿LLeva cada dinosaurio un globo? Does every dinosaur carry a balloon? Overexhaustive error ¿Bebe cada niño una limonada? Does every boy drink a limonade? Underexhaustive errors
non-matched “extra agent”
yes
non-matched “extra object”
no
non-matched “extra object”
no
non-matched “extra agent”
no
Experimental Yes/No-questions in CLLD sentences a) with matched pictures preverbal subject ¿Su cuaderno cada niña lo da a la seño? Her notebook, every girl gives to her teacher? ¿A su niña cada mamá la recoge en coche? P her duaghter, does every mom take by car? ¿A su niña cada mamá la lleva en coche? P her daughter, does every mom take by car? postverbal subject ¿A su amigo lo lleva cada niña en patinete? P her friend does every girl take on skateboard? ¿A su niña la lleva cada mamá en furgoneta? P her daughter, does every mom take by van? b) with non-matched pictures ¿A su niña la lleva cada mamá en coche? P her daughter does every mom take by car? ¿A su amigo lo lleva cada niña en patinete? P her friend, does every girl take on skateboard?
picture-type
adult response
matched
yes
matched
yes
matched
no
matched
yes
matched
no
non-matched “extra object” non-matched “extra subject”
yes no
Subject-object asymmetry in children’s comprehension of sentences containing logical words Sharon Armon-Lotem
This chapter investigates the influence of syntax, semantics and pragmatics on children’s understanding of sentences containing some and or, showing a subjectobject asymmetry. Chierchia, Crain, Guasti and Thornton (1998) showed that once the scalar implicature is erased (in the prediction mode), children accept that some could mean every (75%), and or could mean and (nearly 100%). 40 Hebrew-speaking children aged 3–6 were tested using eyzeshehu yeled for some boy and o for or: 22 in the description mode and 18 in the prediction mode. Children accepted the use of inclusive or in the prediction mode (96%) but far less in the description mode (37%). For some, a subject-object asymmetry was found in both modes. In the prediction mode, the implicature was always erased in direct object position, but in subject position the implicature was only erased 40% of the times. Similarly, in the description mode, the scalar implicature was derived in two thirds of the cases in subject position, but in only a third of the cases in object position. This suggests that syntactic position contributes to children’s comprehension of sentences containing logical words and that both the difference between some and or and the subject-object asymmetry stem from specificity, a pragmatic condition which affects reference.
1. Introduction Traditionally, it has been observed that there are mismatches between the way some operators and connectives are defined in logic (e.g.,∨, ∃x) and the meanings of their closest counterparts in ordinary English (e.g., or, some). Grice (1975) claimed that the differences between the meaning of the operators in logic and natural language arise as a result of pragmatic mechanisms interacting with the basic semantic interpretations. That is, the derived meanings are, in fact, not the basic meanings of the natural language operators, but conversational implicatures. Thus, semantically, A or B is true when A, B, or [A and B] are true, but pragmatically, or is usually used to imply that either A is true or B is true but not both. Similarly, semantically, some A is true when one A, two As . . . every A are true, but pragmatically, some A is usually used with the
Sharon Armon-Lotem
implication that some number of As smaller than all is true. That is, by Grice’s (1975) principle of cooperation, some implies not every in an extensional context (e.g., in the description mode). Nonetheless, the implicature can be suspended in special contexts including, for example, an intensional context (as in the prediction mode: I predict that some boy will leave), and the same applies to or. The question is whether children have access to the semantic (logical) interpretation of some and or, on the one hand, and whether they derive the implicature when required, on the other hand. That is, do children derive the pragmatic implicature in the appropriate context (e.g., in description mode) and do they recognize the implicature as pragmatic and cancel it in appropriate contexts (e.g., in prediction mode), thus using or and some with a ‘pure semantic’ interpretation? Beilin & Lust (1975) showed that children interpret or as having an exclusive reading while Braine & Rumain (1981, 1983) argue that children favour an exclusive reading for some. These findings indicate that children derive the scalar implicatures, but do not tell us whether they have access to the ‘pure semantic’ reading. Chierchia, Crain, Guasti & Thornton (1998) tested the understanding of sentences with logical words in Italian and English using a truth value judgement task (Crain & Thornton 1998), in the description mode and in the prediction mode. They showed that once the scalar implicature is erased (in the prediction mode), children, like adults, are likely to accept some in a context in which every could also be used (75%), and or in contexts in which and could be used (nearly 100%). This chapter tests for the universality of these findings by looking at a non-IndoEuropean language, and investigating the difference between some and or. It reports the results of a truth value judgement task administered to 40 Hebrew-speaking children aged 3–6, and 20 adults, using eyzeshehu yeled for some boy and o for or. The results of the Hebrew replication are found to be similar to the Italian and English results for or, supporting the universality of the English and Italian findings. However, the results for some, though similar to the English and Italian overall percentages, showed an interesting subject-object asymmetry not noted in Chierchia et al (1998), suggesting a possible structural interference. To anticipate the conclusions we reach, we suggest that the difference between some and or and the subject-object asymmetry stem from specificity, a pragmatic condition related to topicality and affecting reference (Groenendijk & Stokhof 1981; Cormack & Kempson 1991). The chapter shows that child language learners are highly sensitive to a variety of pragmatic conditions influencing interpretation, in addition to the semantic ones. The chapter starts with a presentation of the theoretical assumptions, focusing on the role of semantics and pragmatics in our understanding and the use of scalar implicatures, and the relation of specificity to scalar implicatures. Specificity is viewed as a pragmatic principle which applies to the interpretation of quantifiers (but not to coordinators) on the one hand, and makes a distinction between subjects and objects on the other hand. This is followed by a discussion of possible differences in the way adults and children interpret scalar implicatures. After presenting the predictions in Section 3,
Children’s comprehension of logical words
we set out the methodology and findings in sections 4 and 5, respectively. This is followed by a discussion, showing the contribution of the different pragmatic principles to children’s derivation and erasure of scalar implicatures in different contexts.
2. Theoretical assumptions Adult judgements of the truth value of sentences with logical words such as some and or result from an interaction between semantic knowledge and pragmatic knowledge. The semantic knowledge defines the full set of readings which could yield judgements of truth. The pragmatic knowledge, embodied by Grice’s (1975) principle of cooperation, yields a limitation on the truth value within the context excluding some of these readings. That is, semantically, A or B is true when A, B, or [A and B] are true. This is the inclusive reading. Pragmatically, A or B implies not [A and B]. That is, either A is true or B is true but not both. This is the exclusive reading, in which the conjoined interpretation of or is excluded. Similarly, semantically, some A is true when one A, two As . . . every A are true. This interpretation is intuitively parallel to the inclusive reading of or. Pragmatically, some A implies not (every A) – some number of As smaller than all is true. This interpretation is intuitively parallel to the exclusive reading of or, in which the reading of some as all is excluded. The set of situations in which the exclusive reading is true is a subset of those in which the inclusive reading is true. This subset relation can also be seen as a scale between stronger and weaker readings. The use of logical words such as some and or, which have scales of readings, in sentences creates implications which are called scalar implicatures. This section elaborates on the semantics and pragmatics of the logical words triggering scalar implicatures, focusing on some and or, discussing the influence of other pragmatic principles, i.e., specificity, on the truth value assigned to scalar implicatures. In the final subsection, we shall discuss what is known about children’s interpretation of scalar implicatures.
2.1 The semantics and pragmatics of scales: the case of some and or The semantics of logical words, such as, or, and some, stems from the way they are defined in logic, but their conversational interpretation is more restricted. Thus, the truth conditions for (1a) are assumed to be the same as those for (1b), but (1a) conversationally implies (1c): (1) a. John ate ice cream or pizza b. ate ice cream (j) ∨ ate pizza (j) c. ¬ (ate ice cream (j) ∧ ate pizza (j))
That is, since A or B is true if A is true, or B is true, or A and B are true, the sentence in (1) for example, is true in all the situations in which John ate ice cream, or those in
Sharon Armon-Lotem
which John ate pizza, or those in which John ate both. The interpretation in (1b) is the inclusive reading of or. Nonetheless, in many contexts we do not get this reading for or. The interpretation we tend to get for or is the exclusive reading as in (1c), that is, A or (exclusive) B is true if A is true, or B is true, but not when both A and B are true. Thus, for example, the sentence in (1a), is true in all the situations in which John ate ice cream, or those in which John ate pizza, but not in those in which he ate both. Similarly for some, the truth conditions for (2a) are assumed to be the same as those for (2b), but (2a) conversationally implies (2c): (2) a. Some dog swam in the pool b. ∃x[dog(x) ∧ swam in the pool(x)] c. ¬∀x [dog(x) → swam in the pool(x)]
That is, since some A is B is true if one A is B is true, if two As are B is true, if three As are B is true . . . and even if every A is B is true, the sentence in (2a) is true in the situations in which one dog swam in the pool, those in which two dogs swam in the pool, those in which three dogs swam in the pool, and even those in which every dog swam in the pool. This interpretation of some, which is intuitively parallel to the inclusive reading of or, is the weaker reading of some. This discrepancy between the semantic interpretation of or and some and their conversational interpretation, is typical of logical words which are parts of scales (Horn 1972, 1989). A scale is a set of elements arranged in an order α1 α2 α3 . . . αn such that αn entails αn–1 (in an upward directed scale) or αn+1 (in a downward directed scale). S1 entails S2 if any situation which makes S1 true also makes S2 true. Thus, for example, or is part of the informational scale
, while some is part of the positive quantifier scale <some, many, most, every> (Chierchia 2004). These scales are upward directed, that is, the higher elements entail lower ones as in (3): (3) every ⇒ some and ⇒ or
An asymmetrical entailment holds within these scales. That is, A and B entails A or B, but not the other way round, making A and B more informative (stronger) than A or B. Similarly, every A is B entails some A is B, but not the other way round, making every A is B more informative (stronger) than some A is B. This is demonstrated in (4) for every/some, and in (5) for and/or: (4) a. Every dog ate pizza ⇒ Some dog ate pizza b. Some dog ate pizza ⇒ ¬Every dog ate pizza (5) a. John has a boy and a girl ⇒ John has a boy or a girl b. John has a boy or a girl⇒ ¬John has a boy and a girl
Every entails some, but some does not entail every. The set of situations which satisfy the truth conditions of every is a subset of the set of situations which satisfy the truth
Children’s comprehension of logical words
conditions of some. They are in a subset/superset relation. Similarly, and entails or, but or does not entail and. The set of situations which satisfy the truth conditions for and is a subset of the set of situations which satisfy the truth conditions for or. Yet again, it is a subset/superset relation. If X asymmetrically entails Y, X, being more restricted, is stronger than Y. Grice (1975) suggests that the discrepancy between the semantic interpretation of or and some and their conversational interpretation, arises from the interaction between pragmatic mechanisms and the basic semantics of scales. The conversational interpretation, the implicatures, arises by the communicative needs of the speakers, that is, by the need to be as cooperative as is required at the stage in which the conversational contribution of the speaker is made. By the principle of cooperation, the speaker’s contribution must be as informative as is required, but not more than is required (Maxim of Quantity) and as relevant as possible (Maxim of Relation). Where scales are concerned, these maxims mean that the speaker’s contribution should be as strong as possible (i.e., most informative), related to the situation. Thus, uttering Y implicates that any element stronger than Y is probably false. For example, uttering (6a) implicates that the stronger sentence (6b) is probably false: (6) a. John has a boy or a girl b. John has a boy and a girl
Formally, uttering an element Y in the scale implicates the negation of any X which is higher in the scale (as in (2)). Uttering some implicates not every and uttering or implicates not and (i.e., not both). The listener assumes that the speaker adheres to the Maxim of Quantity and the Maxim of Relation, using the sentence that is most informative by making the strongest statement that is consistent with what he knows to be true. If he knew that X (which is higher in the scale) is true, he would say that, and not the weaker Y. So if he says Y, this is because X is probably not true. This general principle can immediately account for a wide range of implications with other lexical elements, which otherwise seem unconnected and arbitrary, e.g., three, not all, good, possible, all of which are parts of a scale (Horn 1972, 1989). Another advantage of this Neo-Grician account of scalar implicatures is that it can help explain why these implicatures tend to arise in assertions, but not in sentences expressing predictions or uncertainties. In contexts of uncertainty, where we cannot make assumptions about the speaker’s knowledge, the speaker cannot be expected to be as informative as in assertions, due to the lack of information. Compare, for example, the descriptions in (7a) and (7b) with the predictions in (8a) and (8b), respectively: (7) a. Some children ate pizza b. John has a son or a daughter (8) a. I bet some children ate pizza b. I bet John has a son or a daughter
Sharon Armon-Lotem
While the assertions in (7) adhere to the Maxim of Quality, yielding an exclusive reading, the predictions in (8) yield a possible inclusive reading, since the speaker has restricted knowledge and is uncertain about whether the event will take place. If one does not know the outcome, it is natural to use the weaker expression. These examples show that the implicature can be suspended when there is uncertainty about the outcome, that is, when we cannot assume that the speaker knows the outcome. Extensional contexts which trigger the implicatures are implicature raising contexts, whereas intensional contexts in which the scalar implicature does not arise are implicature erasing contexts. Conditions of uncertainty can arise in many ways, among which are making a bet or a prediction, i.e., situations in which the speaker is not expected to have the evidence on the basis of which he utters the statement. Thus, the hearer’s expectations about informativeness are different in conditions of uncertainty, which affects the strength of the prediction.
2.2 Specificity and reference The distinction between description and prediction can also be seen as a difference in information, i.e., a difference in the specificity of the reference. The speaker has different information about the event or the reference in the different modes. This difference can be captured by specificity. Hellan (1981) and Ioup (1977) define an indefinite NP as specific when the speaker has information about its referent. That is, an indefinite NP has a specific reference if the speaker perceives it as unique within a given universe of discourse (Polya & Tarnay 2003). As noted by Karttunen (1969), the same indefinite NP within the same sentence has two distinct interpretations, one specific and the other non-specific, as in (9):
(9) You should show this paper to a semanticist.
On the one hand, a semanticist in (9) can refer to a specific person the speaker has information about, or simply, under a non-specific reading, describe the kind of person it ought to be. This ambiguity crucially depends on the speaker’s intentions. Kripke (1977) and Ludlow & Neale (1991) view this as a pragmatic ambiguity. Groenendijk & Stokhof (1980) interpret specificity as a pragmatic condition affecting the identification of the referent. Cormack & Kempson (1991), using a Gricean model, argue that specificity arises from the way in which the speaker and hearer interact. That is, though some, for example, can be interpreted as the existential quantifier of predicate calculus, in many contexts, we, as hearers, assume that the speaker using the existential quantifier has evidence for what he says. That is, the speaker has some individual in his mind as a reference, even though the speaker did not make reference to this particular individual. Cormack and Kempson further point out, following Dowty, Wall & Peters (1981) that specificity is context sensitive. In contrast, Kratzer (1998) proposes a semantic theory of the specific/nonspecific distinction. She proposes that specific indefinites are interpreted by a choice function while non-specific ones
Children’s comprehension of logical words
are interpreted as truly quantificational. But again, pragmatic factors influence which interpretation is chosen. The details of these different approaches need not concern us here. For the present chapter, we are concerned (1) with the interaction between specificity and the derived implicatures of logical words in the different modes, and (2) with the interaction between specificity and the assignment of references to the subject and object of the sentence. For the former, we propose that specificity influences the hearer’s assumptions about the speaker’s knowledge when the speaker talks about entities, but not when the speaker talks about events. We assume that this is because specific knowledge about participants in events will directly affect assignment of denotations to nominals, but will not affect the evaluation of alternative situations. Thus, since some is part of the nominal phrase and quantifies over individuals, while or quantifies over events, some is expected to be more sensitive to properties of individuals, i.e., specificity of reference, while or which conjoins two nominal phrases, but quantifies over events, is expected to be more sensitive only to properties of events, i.e., conditions of uncertainty as to whether the event occurred. Therefore, the implicature is predicted to be erased more easily with or than with some in the prediction mode. As for the latter, in the unmarked case, the subject will be the topic of the sentence. (Foley & Van Valin 1984, Givón 1990, Lambrecht 1994). The connection between topichood and specificity has been noticed widely in the literature and follows from Reinhart’s (1982) observation that “the topic of a sentence . . . is the entity the sentence is about” (Ebert & Endriss 2002). The topic, the ‘Old Information’, represents something which has already appeared in the previous discourse and is likely to be identifiable. It is therefore more specific. On the other hand, the comment, the ‘New Information’ is less or not at all specific. Though specificity is a tendency, for the ease of presentation, we use in the present study [+specific] and [-specific] to reflect a higher and a lower degree of specificity. This connection between specificity and topichood is further discussed by Cresti (1995) and Portner & Yabushita (2001). Portner and Yabushita follow Enç (1991) who argues that specific indefinites quantify over a pre-established or contextually inferable set. They show that specificity is closely related to the speaker’s presuppositions, some of which may concern his degree of certainty about the likelihood of the particular event occurring or having occurred and the reason for it. Thus, even though specificity has to do with the interpretation of noun phrases and conditions of uncertainty are about events, they are not entirely independent. If specificity is indeed a pragmatic phenomenon related, in part, to speaker’s presuppositions about the certainty and uncertainty of events, we expect that it may interfere in the drawing of pragmatic implications which crucially rely on conditions of uncertainty, as in the case of scalar implicatures. The interaction between the conditions of uncertainty which apply to the whole sentence and the level of specificity of the noun phrase influences the level of uncertainty about the reference assigned to the particular NP. The two factors, [± topic]
Sharon Armon-Lotem
and [± intensional context], lead to four possible degrees of information about the reference of a noun phrase, which vary in their level of specificity/certainty: topic in an extensional context and non-topic (comment) in an extensional context as in (10a) and (10b), and topic in an intensional context, and non-topic in an intensional context as in (11a) and (11b): (10) a. Some dog ate the hotdogs b. The dogs ate some hotdog (11) a. I guess that some dog ate the hotdogs b. I guess that the dogs ate some hotdog
Since scalar implicatures are derived based on the knowledge about the real world reference of the NP, which the hearer attributes to the speaker, it is plausible that the implications differ depending on the degree of specificity of the NP. The combination of a noun phrase in topic position which is [+specific] and an extensional context which is also more specific, since it refers to events which took place in the real world, as in (10a), is likely to yield an implicature raising context. On the other hand, the combination of a noun phrase in a non-topic position, i.e., the object, which is [-specific], and an intensional context, which is also less specific since it refers to possible worlds, as in (11b), is likely to yield an implicature erasing context with a strong preference for the inclusive reading. This predicts that the level of specificity of the subject (being the topic) might be reduced by an intensional context in (11a), while the level of certainty in an extensional context might be lower in the object position which is part of the comment in (10b). That is, when the context and the noun phrase position match in their features, we expect the level of certainty or uncertainty to be enhanced by the features of the noun phrase. This will be reflected in the interpretation assigned to the scalar implicatures. On the other hand, when the context and the noun phrase position are marked with conflicting features, we expect the features of the noun phrase to reduce the influence of the context. This might offer a window into comparing the relative influence of the whole event and that of the individual.
2.3 Scalar implicatures in acquisition The question is whether children’s logical reasoning conforms to classical logic and whether they are sensitive to the principle of cooperation. That is, do children know, for example, that A or B entails A and B? Do children know that A or B implies ¬(A and B)? And do they recognize the implicature as pragmatic and are able to cancel it in appropriate contexts, using or (and some) with a ‘pure semantic’ interpretation? Beilin & Lust (1975), for example, found that, initially, children interpret or as having the exclusive reading with no access to the logical truth conditions associated with the inclusive reading. That is, when the children were asked to pick a car or an apple, they picked only one and not both. Similarly, Braine & Rumain (1981/3) actually argue that children favour the exclusive reading for some. These studies used implicature raising
Children’s comprehension of logical words
contexts. At the same time, it has been suggested by Paris (1973) that when children do get the inclusive reading, ignoring the conversational implicature, it is because they fail to distinguish between and and or. Smith (1980) also found that children ignore the conversational implicature, giving an affirmative response to questions such as Do some elephants have trunks? This was taken to show that they interpret some to mean all. It is not clear however whether it is felicitous to use such questions out of context. These studies, however, did not manipulate the pragmatic contexts, comparing explicitly implicature raising contexts with implicature erasing contexts. Since, as pointed out by Chierchia et al. (1998), the “pure” semantic interpretation is accessible only in implicature erasing contexts, the above experiments leave open the possibility that children know both the semantic entailments and the pragmatic implicatures, but were not given the suitable context for showing this knowledge. Chierchia et al (1998) were the first to address this possibility, presenting logical words for which scalar implicatures can be derived both in implicature raising contexts, e.g., the description mode, and in implicature erasing contexts, e.g., the prediction mode. In order to find out how children interpret scalar implicatures, they used a truth value judgement task (Crain & Thornton (1998)). To avoid carry-over effects between the two modes they never tested the same child in both modes. In the description mode, the puppet and the child watched a story that was actedout with toys and props. The puppet tried to describe the events of the story and the child was asked to judge whether the puppet’s description of the story was true or false. When the child judged the puppet’s description of the story to be false, she was asked to explain to the puppet “What really happened” (to check understanding of the events). An example is given in (12): (12) A troll contemplates what to eat for lunch. He considers 3 options: hamburger, slice of pizza, ice-cream. Eventually, he chooses pizza and ice-cream, but not hamburger. Kermit the frog says: “I know what happened. The troll ate a slice of pizza or an ice-cream”
In this mode, the description came after the story was acted-out, when there was no uncertainty about what the troll had eaten. Thus, the cooperative puppet (who embodies the speaker) is expected by the hearer to use the more informative logical word and rather than the less informative word or. In the prediction mode, the child heard part of the story, and then the magician puppet predicted how it would end. An example is given in (13): (13) A troll contemplates what to eat for lunch. He considers 3 options: hamburger, slice of pizza, ice-cream. The magician puppet says: “I know what will happen. The troll will eat a slice of pizza or an ice-cream”. Then the troll chooses pizza and ice-cream, but not hamburger.
The prediction was then repeated at the conclusion of the story, if necessary. Since the prediction is made at the point when both speaker (embodied by the puppet) and
Sharon Armon-Lotem
hearer are uncertain about the outcome, the less informative word or is a sensible choice. Chierchia et al (1998) tested children both in English and Italian and found that children’s semantic knowledge is consistent with classical logic. That is, in implicature erasing contexts, such as making predictions, children, like adults, accepted that some could mean every (75%), and or could mean and (nearly 100%). In addition they showed that in implicature raising contexts, such as a description, some children were “cooperative”, showing adult-like behaviour, while others were not. Even by the age of five, only half of the children derived the scalar implicatures in the description mode. They concluded that while children’s semantic knowledge is consistent with classical logic, they are sensitive to certain pragmatic principles which can raise and erase the conversational implicature. These findings generated a lengthy line of research testing both for the same scalar implicatures as well as others, in different languages in the different modes (Chierchia, Crain, Guasti, Gualmini & Meroni 2001; Chierchia, Guasti, Gualmini, Meroni, Crain, & Foppolo 2004; Foppolo, Guasti, & Chierchia 2004). The different studies all show that preschool children find it difficult to derive the pragmatic inference. Moreover, these studies show a consistent difference between some and or. According to these studies, about 50% of kindergarten children derive the scalar implicature for or (Chierchia et al 2001), but fewer children derive the implicature for some (Noveck 2001; Papafragou & Musolino 2003). Gualmini, Meroni, & Crain (2001), for example, show that children erase the implicature within the scope of modals, which create conditions of uncertainty, indicating that they know that and and or are associated with different truth-conditions. In addition to testing some, Papafragou & Musolino (2003) tested other scalar implicatures, such as numbers and aspectual verbs in Greek, and found that while adults abide by the principle of cooperation and reject the weaker statements in implicature raising contexts which justify the stronger ones, children violate the principle of cooperation, accepting aspectual verbs, such as start and begin, two thirds of the time in implicature raising contexts where a more cooperative way to describe the situation would have been using the stronger finish and end. On the other hand, with degree modifiers, such as half, the level of acceptance of the inclusive reading was reduced to a third of the responses, which is less than the adults (who scored at ceiling), but resembles Chierchia et al’s (1998) findings for some. They also reported a very low acceptance rate for some in contexts where all was more appropriate in the description mode (12.5%). When the implicature was cancelled, the inclusive reading was always accepted. These studies, while showing that children are able to assign the pure semantic interpretation in implicature erasing contexts, being sensitive to the conversational implicatures, raise the question why children differ from adults in implicature raising contexts. In an attempt to address this question, further studies, like Gualmini, Crain, Meroni, Chierchia & Guasti (2001); Chierchia et al (2001); Gualmini & Crain (2002);
Children’s comprehension of logical words
Musolino & Lidz (2006); and Guasti & Foppolo (2006) introduce modifications of the methodological aspects of the tasks used as well as attempts to modify the account for the differences found between child and adult responses in some of these studies for some of the scalar implicatures. These modified accounts introduce a variety of syntactic factors (Guasti & Foppolo 2006), pragmatic factors (Gualmini & Crain 2002) and processing costs comparing different alternatives (Gualmini et al 2001; Chierchia et al 2001) which could influence the interpretation of scalar implicatures. Guasti & Foppolo (2006), for example, argue that the difference between children and adults is an artifact of the design used in previous studies. They point out that in previous studies, children were presented with scenes which were so similar to each other that they could predict what the puppet would say, and would opt for a strategy for judging the truth value of the puppet’s statements. Thus, they modify that task so that every condition appears only once for each child, making it impossible to predict the puppet’s assertion, or develop a strategy for responding. Using this methodology they found that five year old children derived the scalar implicature for some in 70% of the appropriate contexts. They checked the implicatures for some both in subject and in object positions and found no difference between the two. Guasti and Foppolo’s findings suggest that children abide by the principle of cooperation and derive the implicature in implicature raising contexts. Nonetheless, even Guasti and Foppolo find a difference between children and adults who derive the implicature in 98% of the times.
3. Predictions Against this background, the following predictions can be made: 1. Following Chierchia et al (1998), Hebrew-speaking children are expected to derive the exclusive reading in the description mode and judge a weaker statement as false where a stronger one is more informative, but accept an inclusive reading in the prediction mode, judging the weaker statement as true in conditions of uncertainty. 2. Since or quantifies over events, it is expected to be more sensitive to properties of events than some which quantifies over individuals. Thus, we expect the difference in mode to influence the derivation of the implicature for or more than for some. That is, we expect children to judge the inclusive reading of or in the prediction mode as true, even when the use of and is possible, more readily than the weaker reading of some in the prediction mode when the use of every is also possible. 3. Since some quantifies over individuals, differences in specificity between the subject and the object are predicted to yield a subject-object asymmetry in the interpretation of some in the prediction mode, i.e., the implicature will be cancelled only in object position. That is, some is sensitive to the properties of the NP, such as
Sharon Armon-Lotem
specificity, while or is not. Since we know that specificity effects are stronger in subject position than in object position, we expect it to be more difficult to cancel the implicature in subject position than in object position. Thus, we expect the specificity of the subject/topic position to block the influence of the non-specificity of the prediction (intensional) mode. This does not apply to or since or is not sensitive to the pragmatic properties of the nominal phrase
4. Methodology 4.1 Subjects The subjects were 40 Hebrew-speaking children ranging in age from 3–6. This range of ages covers a period in which language is still undergoing change, but no age related differences were found. Therefore, they are presented as one group. All children come from a middle socio-economic status (SES) and attended one of two preschools in a suburb of Tel-Aviv. Children who showed language difficulties in a free play task were excluded. 22 children were presented with sentences in the description mode and 18 children were presented with sentences in the prediction mode, where the implicature can be suspended. We chose to test two different groups in order to avoid a carryover from one testing condition to the other. The whole task was usually presented in one session per child. A control group of 20 adult native speakers of Hebrew, all university students, was presented with an oral story version of the task (see method section), in order to test for patterns which might be unique to the administration of the task in Hebrew. They were all presented with the two modes in two different sessions a week apart.
4.2 Method A Truth-Value Judgement task (Crain & Thornton 1998) was used, adapting the task used in Chierchia et al (1998) into Hebrew. Eyzeshehu (yeled) was used for some (boy) and o was used for or. In Hebrew, some translates into eyzeshehu followed by a singular noun, or xelek me ‘some/part of ’ followed by a plural noun. The decision to use eyzeshehu was motivated by the need to use a linguistic form which morphologically supports the full scale, from 1 to n. There is no record in the literature of the use of eyzeshehu by children, but as the findings show, children had no difficulty understanding this form. Moreover, since eyzeshehu is closer to an indefinite article, it is predicted to be more sensitive to specificity. In the description mode, a puppet, manipulated by one experimenter, and a child watched a story that was acted-out by another experimenter with small props. After the act-out, the puppet tried to describe what had happened in the story. The child had to judge whether the puppet’s description of the story was true or false. When the
Children’s comprehension of logical words
child judged the puppet’s description of the story to be false, she was asked to explain to the puppet “What really happened” (in order to check that she had understood the events). In the prediction mode, the child heard part of the story, and then a magician puppet made a prediction about the rest of the story. The prediction was repeated at the conclusion of the story, if the child was not sure what it was. Here again, the child judged whether the puppet’s prediction was true or false, and explained her answer whenever she judged it as false. In order to test the first prediction, 22 children were presented with sentences in the description mode and 18 children were presented with sentences in the prediction mode, where the implicature can be suspended. We chose to test two different groups in order to avoid a carryover from one testing condition to the other. In order to test the second prediction, each child was presented with 5 sentences containing or, 5 containing some, and 5 fillers, a total of 15 stories per child. The target sentences always presented the weak, less informative interpretation, which children are expected to judge as false in the description mode, but true in the prediction mode. The fillers were used in order to counterbalance the children’s replies. If the child judged the puppet’s description or prediction as true, the filler triggered false as an answer, and vice versa. A list of all test items is given in the appendix. In order to test the third prediction, the logical word was in subject position in half of the sentences and in object position in the other half. Half of the children were presented with 3 sentences containing or in subject position and 2 sentences containing or in object position, and the other half was presented with 2 sentences containing or in subject position and 3 sentences containing or in object position (but no more than 5 sentences per child). The same applied to some. This yielded 55 test items in each of the four categories (or-subject, or-object, some-subject, and someobject) in the description mode, and 45 test items in each category in the prediction mode. The adults were presented with an oral story version of the tasks, using no props. They were asked to imagine that they were watching TV programs. In the description mode, they were asked to judge whether the reporter had described the scenes accurately. In the prediction mode, they were told that they were watching the program with a friend who is an avid gambler and likes to bet on everything. They had to decide by the end of each scene whether he had won the bet or not. In the adult version there were 3 stories each including 4 scenes which capture or and some in subject and object position. This yielded 60 test items in each category in the two modes. The findings are presented by the percentage of acceptance of the inclusive interpretation, first for all test items in the different modes, then comparing some and or in both modes and finally comparing the interpretation of some and or in subject and object position. A one-tailed paired t-test was performed for the within group analysis and a one-tailed independent t-test was performed for the across group analysis.
Sharon Armon-Lotem
5. Results 5.1 Scalar implicature in description and prediction contexts – Adults As predicted, the adults always allowed the use of the weaker less informative logical word in the prediction mode for both some and or. On the other hand, their responses varied along three trends in the description mode: eleven rejected the less informative word altogether, six accepted it for some but not for or, and three accepted it everywhere, though expressing discomfort with the choice of the weaker term rather than the stronger one on the scale. As noted by Guasti & Foppolo (2006), all those who rejected the weaker statement explained their rejection using the stronger term. 100%
Rate of acceptance of the weaker statement (adults)
80% 60%
Prediction Description
40% 20% 0%
Some
Or
Figure 1. Rate of acceptance of the weaker statement (adults).
Of special interest for this chapter are the responses of the six adults who made a distinction between some and or. These speakers found it necessary to motivate their responses, saying: “they all jumped, but he can say some horse and talk about one of them”, indicating that they interpret some NP as a specific NP with a referent, rather than an existential quantifier with the inclusive reading. Similar to Guasti and Foppolo’s findings, no subject-object asymmetry was found for the adults in Hebrew.
5.2 Scalar Implicature in description and prediction contexts – Children As predicted, the findings for the Hebrew-speaking children replicated the English and Italian findings in Chierchia et al (1998), with children accepting an inclusive reading for some and or in the prediction mode more than in the description mode. Figure 2 presents the rate of acceptance of the weaker statement for some and or (by percentage) in the description and prediction mode.
Children’s comprehension of logical words 100% 80% 60% 40% 20% 0% Description
Some 51%
Or 37%
Prediction
70%
96%
Figure 2. Rate of acceptance of the weaker statement (children).
As seen in Figure 2, children accepted the inclusive reading of or in the prediction mode (96%) but far less in the description mode (37%). Similarly, they accepted the inclusive reading of some in the prediction mode (70%) but far less in the description mode (51%). This difference was significant in the two modes both for some (t = 4.68, p < 0.05) and for or (t = 11.40 , p < 0.05). These findings are similar to the adult data and show a clear preference for rejecting the inclusive reading in the description mode. There was a significant difference between some and or both in the prediction mode (t = –2.03, p < 0.05) and the description mode (t = 5.41, p < 0.05). This shows, as predicted, a distinct difference between or and some, in both modes. A closer look at the responses, using a binomial test, shows that the children do not show a chance response pattern for or, but seem to show such a pattern for some. For both words, however, they often motivate the negative responses, using the stronger term in the scale, and accompany the positive responses by indicating that they were aware of the ambiguity. This chance pattern for some disappears when the findings for subject and object positions are separated.
5.3 Subject-object asymmetry The difference between some and or becomes clearer when comparing the acceptance of the weaker use of these words in subject position and object position. While or shows consistent behaviour in subject and object position (36% and 38% respectively in the description mode and 93% and 98% respectively in the prediction mode), some shows a subject-object asymmetry. Figure 3 presents the rate of acceptance of the inclusive reading of some (by percentage) in subject and object position in the description and prediction modes.
Sharon Armon-Lotem 100% 80% 60% 40% 20% 0%
Subject
Object
Description
37%
65%
Prediction
40%
100%
Figure 3. Rate of acceptance of the weaker statement: Some.
As seen in Figure 3, there is no difference between the description and the prediction mode in subject position, with children accepting the weaker use of some in contexts in which every is more informative in slightly more than a third of their responses. On the other hand, in object position, there is a distinct difference between the description and the prediction mode. That is, some NP was always accepted when used in object position in the prediction mode, while it was accepted in only two thirds of the responses when it was used in object position in the description mode.
6. Discussion and conclusion Similarly to Chierchia et al (1998), our findings show that children derive the exclusive reading of or in the description mode two thirds of the times and judge a weaker statement as false where a stronger one is more informative. They also accept an inclusive reading in the prediction mode, judging the weaker statement as true in conditions of uncertainty. These findings suggest that children know the semantics of or, on the one hand, and abide by the principles of cooperation, deriving the scalar implicature on the other hand. Nonetheless, the findings for some are somewhat different. In the description mode, children seem to show chance behaviour, rejecting the weaker statement only half of the time. That is, as Gualmini et al. (2001) point out, while children accept sentences which violate the implicature, they also show a preference for the alternative which doesn’t. Though this rate is higher than was found by Papafragou & Musolino (2003)
Children’s comprehension of logical words
and Noveck (2001), it cannot be evaluated as indicative of pragmatic knowledge, unless we accept Reinhart’s (2006) explanation of such findings as showing competition between two representations, where the chance response is indicative of limited working memory resources which are needed in order to complete the computation. In the prediction mode, however, children seem to erase the scalar implicature 70% of the time, accepting the weaker prediction. This suggests that they are able to access the semantics of some. Furthermore, the difference found between some and or shows, as predicted, that children erase the implicature of or accepting the inclusive reading in the prediction mode as true, even when the use of and is possible, more readily than they erase the implicature of some accepting the weaker reading in the prediction mode when the use of every is also possible. This, as suggested earlier, is because or quantifies over events, and is therefore expected to be more sensitive to properties of events than some which quantifies over individuals. Finally, the predicted subject-object asymmetry found for some, and the lack of such asymmetry for or, shows how specificity influences the derivation of scalar implicatures and their erasure. We found that in the prediction mode, it was harder to erase the implicature in subject position which is more specific (40%), than in object position which is less specific (100%). In fact, as predicted, the implicature was cancelled by all children in all the trials only in object position. That is, the specificity of the subject/topic position blocked the influence of the non-specificity (uncertainty) of the prediction mode. We also found that in the description mode, children derived the scalar implicatures more readily in subject position (63%) than in object position (35%). This again is explained by the higher specificity of the subject/topic position. Similar subject-object asymmetry in children’s interpretation of quantifiers is reported by Kang (2001) for quantifier spreading, showing that it is affected by other syntactic and pragmatic considerations too. No subject-object asymmetry was found for or, since or is not a quantifier and is therefore not sensitive to the pragmatic properties of the nominal phrase. We suggest that the difference between some and or and the subject-object asymmetry stems from the same source. We hypothesize that both are due to specificity, which, as a pragmatic condition affecting reference, is a plausible mechanism for affecting pragmatic implications about reference (Groenendijk & Stokhof 1981; Cormack & Kempson 1991). The subject which is the topic of the sentence refers to old information and is therefore [+specific], whereas the object which is part of the comment refers to new information, and is therefore [–specific]. This distinction interacts with the lack of specificity in conditions of uncertainty. The description mode, being an extensional context describing the real world, encourages [+specific] interpretation for noun phrases, while the prediction mode, being an intensional context describing possible worlds, encourages [–specific] interpretation for noun phrases. Thus, specificity can interact with both individuals and events. Conversational implicatures are likely to arise in more specific contexts and be cancelled in less specific contexts.
Sharon Armon-Lotem
Or introduces operations on events. Therefore, it is sensitive to properties of events but not to properties of individuals. This makes it sensitive to the difference in mode, but not to the difference in the specificity of the two positions. That is, some, being a part of the nominal phrase, quantifies over individuals, while or although it conjoins two nominal phrases, has the semantic effect of introducing an operation on events. Moreover, some is a lexical determiner, and we explicitly choose to use it, rather than saying every, presumably with the pragmatic implication with which it is associated. Existential quantification over events is a default null operator that we have no choice but to use. Therefore, the pragmatic implications coming with the lexical existential determiner can be expected to be stronger, as can be seen by the results. The interaction between the level of specificity of the different modes and the level of specificity of the noun phrase which is quantified by some is summarized in Table 1, showing the percentage of acceptance of an inclusive reading in each combination: Table 1. Degree of acceptance (in percentage) of inclusive reading in each combination Topic [+specific] Comment [–specific]
Description mode [+specific] 37% 65%
Prediction mode [–specific] 40% 100%
As seen in Table 1, the combination of old information in topic position which is [+specific] and the description mode which is also [+specific], since it refers to events which took place in the real world, is a combination in which the conversational implicature is bound to yield a preference for the exclusive reading. On the other hand, the combination of new information in the comment, which is [−specific], and the prediction mode which is also [−specific], referring to possible worlds, is a combination in which the conversational implicature is bound to be cancelled due to the high level of uncertainty yielding a strong preference for the inclusive reading. As predicted, the [+specific] combination yields a 63% rejection rate of the inclusive reading of some, whereas the [−specific] combination yields a 100% acceptance rate of the inclusive reading of some. While the above combinations are unified in terms of specificity and uncertainty, the other two combinations have conflicting values for the mode and for the noun phrase. One of these combinations is manifested as a close to chance response for some within the topic in the prediction mode. There, the specificity of the subject/topic position blocks the influence of the non-specificity of the prediction (intensional) mode. Therefore, the findings in subject position are identical in both description and prediction modes. The other combination, of a comment in the description mode, indicates the superiority of one domain over the other. Namely, the reduced specificity of the object, being the comment, generates a level of uncertainty, which allows implicature erasure. The combination of the object position, which is less specific, being part of the comment, and the non-specificity of the prediction (intensional) mode exploits the complexity of some enabling the less specific reading of some as all to emerge.
Children’s comprehension of logical words
Some inconsistencies between our findings and those of Guasti & Foppolo (2006) requires further discussion. First, Guasti & Foppolo (2006) found that children in previous studies were consistent in their responses, either deriving the implicature everywhere or never deriving it, while we found that children were not consistent in their responses. Guasti and Foppolo argued that the consistent responses were indicative of a response strategy adopted by the subjects due to the kind of task they were presented with. No such strategy was found in the present study, but the children consistently motivated their non-adultlike responses in a way which showed their awareness of the alternative response. Thus, it seems that our study does reflect children’s knowledge of the pragmatics of scalar implicatures rather than a response strategy. Second, Guasti and Foppolo did not find a subject-object asymmetry in the use of some, while the present chapter did. This can be explained by the differences in meaning between the Hebrew translation which was chosen for the quantifier and the Italian one. While the Hebrew phrase eyzeshehu N is a translation equivalent to the English some N, the Italian alcuni dei Ns is a translation equivalent to some of the Ns. That is, though both are translation equivalents to some, the Hebrew one is more sensitive to specificity since it is closer in meaning to an indefinite article, while the Italian one has a partitive meaning which is less sensitive to specificity. Despite these inconsistencies, the present chapter seems to support Guasti and Foppolo’s major observation that children adhere to the pragmatic principles which influence the derivation and erasure of scalar implicatures. Another contribution of this chapter is to the ongoing discussion of the principle of Modularity. By the principle of Modularity (cf. Crain, Gualmini & Meroni 2000; Gualmini, Crain & Meroni 2000), the language faculty, as well as its different levels, are modular. Thus, decisions reached outside the language faculty cannot influence operations within the language faculty, [and] decisions reached at higher-levels within the language processing system cannot influence operations at lower levels (Gualmini et al 2000, 374–375).
The above mentioned studies support the Modularity principle by showing, for example, that pragmatic principles, which are higher level operations, can cut out some of the options given by the semantic module, a lower level module, but not vice versa. As can be seen from the findings reported and discussed in this chapter, different principles within the pragmatic module can influence each other while there is no evidence of cross-modular influence. The different pragmatic principles integrate to prune the outcome of the semantic modules, but never influence the operation at that level. To conclude, the present chapter shows that children’s knowledge of the semantics of logical words is universal as is evident by their ability to erase the scalar implicature in the prediction mode. It also shows that they have the ability to derive the scalar implicatures in the description mode. Both abilities are sensitive to conditions of uncertainty which apply to the nominal phrase (difference in specificity) and to the whole event (the difference in mode). Differences between some and or, as well as the
Sharon Armon-Lotem
subject-object asymmetry found in the interpretation of some, are also indicative of the pragmatic awareness among these children. Further research is necessary in order to see whether the difference between some and or would be found in the use of other nominal modifiers such as degree expression.
Acknowledgement This chapter emerged from the work I had done with the help of Stephen Crain during my stay at the University of Maryland, College Park (1996–1999). Earlier versions of this article were presented at the 30th Boston University Conference on Language Development (2005), the annual meeting of the Israel Association of Theoretical Linguistics, The Technion, Haifa (2005), and the Child Language Seminar, Bristol, UK (2004). We thank the audiences for useful comments. We also thank Susan Rothstein from Bar-Ilan University for most valuable help and three anonymous reviewers for many valuable comments and suggestions.
References Beilin, H. & Lust, B. 1975. A study of the development of logical and linguistics connectives: Linguistics data. In Studies in the Cognitive Basis of Language Development, H. Beilin (ed.), 217–284. New York NY: Academic Press. Braine M.D.S. & Rumain, B. 1981. Children’s comprehension of ‘or’: Evidence for a sequence of competencies. Journal of Experimental Child Psychology 31: 46–70. Braine, M.D.S. & Rumain, B. 1983. Logical reasoning. In Handbook of Child Psychology, Vol. 3: Cognitive development. J. Flavell & E. Markman (eds.), 261–340. New York NY: Academic Press. Chierchia, G. 2004. Scalar implicatures, polarity phenomena, and the syntax/pragmatics interface. In Structures and Beyond: The Cartography of Syntactic Structures, Vol. 3, A. Belletti (ed.), 39–103. Oxford: OUP. Chierchia, G., Guasti, M.T., Gualmini, A., Meroni, L., Crain, S. & Foppolo, F. 2004. Semantic and pragmatic competence in children’s and adults’ comprehension of or. In Experimental Pragmatics, I. Noveck & D. Sperber (eds.), 283–300. New York NY: Palgrave. Chierchia, G., Crain, S., Guasti, M.T., Gualmini, A. & Meroni, L. 2001. The acquisition of disjunction: Evidence for a grammatical view of scalar implicatures. In Proceedings of the 25th Annual Boston University Conference on Language Development, A.H.-J. Do, L. Dominguez & A. Johansen (eds), 57–168. Somerville MA: Cascadilla. Chierchia, G., Crain, S., Guasti, M.T. & Thornton, R 1998. Some and or: A study on the emergence of logical form. In Proceedings of the 22nd BUCLD, A. Greenhill, M. Hughes, H. Littlefield & H. Walsh (eds.), 97–108. Somerville MA: Cascadilla. Cormack, A. & Kempson, R. 1991. On specificity. In Meaning and Truth, J.L Garfield & M. Kitely (eds), 546–581. New York NY: Paragon. Crain, S., Gualmini, A. & Meroni, L. 2000. The acquisition of logical words. LOGOS and Language 1:49–59 Crain, S. & Thornton, R. 1998. Investigations in Universal Grammar: A guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge MA: The MIT Press.
Children’s comprehension of logical words Cresti, D. 1995. Indefinite Topics. PhD dissertation, MIT. Dowty, D.R., Wall, R.E. & Peters, S. 1981. Introduction to Montague Semantics. Dordrecht: Reidel. Ebert, C. & Endriss, C. 2002. Topic interpretation and wide scope indefinites. NELS 34. Enç, M. 1991. The semantics of specificity. Linguistic Inquiry 22: 1–25. Foley, W. & Van Valin, R.D. 1984. Functional Syntax and Universal Grammar. New York NY: CUP. Foppolo, F., Guasti, M.T. & Chierchia, G. 2004. Pragmatic inferences in children’s comprehension of scalar items. Second Lisbon Meeting on Language Acquisition, Lisboa. Givón, T. 1990. Syntax: A Functional Typological Introduction. Amsterdam: John Benjamins. Grice, H.P. 1975. Logic and conversation. In Syntax and Semantics 3: Speech Acts, P. Cole & J. Morgan (eds.), 41–58. New York NY: Academic Press. Groenendijk, J. & Stokhof, M. 1981. A pragmatic analysis of specificity. In Ambiguities in Intensional Contexts, F. Heny (ed.), 153–190. Dordrecht: Reidel. Gualmini, A., Meroni, L. & Crain, S. 2001. The inclusion of disjunction in child grammar: Evidence from modal verbs. In Proceedings of NELS 30, 247–257. Amherst MA: University of Massachusetts, GLSA. Gualmini, A., Crain, S. & Meroni, L. 2000. Acquisition of disjunction in conditional sentences. In BUCLD 24 Proceedings, S.C. Howell et al. (eds), 367–378. Sommerville MA: Cascadilla. Gualmini A., Crain, S., Meroni, L., Chierchia, G. & Guasti, M.T. 2001. At the semantics/pragmatics interface in child language. In Proceedings of SALT XI, 231–247. Ithaca NY: Cornell University. Gualmini, A. & Crain, S. 2002. Why no child or adult must learn De Morgan’s laws. In Proceedings of the 26th Annual Boston University Conference on Language Development, 243–254. Summerville MA: Cascadilla. Guasti, M.T. & Foppolo, F. 2006. Scalar Implicatures in children: Failures or skilful strategies? In Proceedings of NELS 35. L. Bateman and C. Ussery (eds.), Charleston SC: Booksurge. Hellan, L. 1981. On semantic scope. In Ambiguities in intensional context, F. Heny (ed.), 47–81 Dordrecht: Reidel. Horn, L.R. 1972. On the Semantic Properties of Logical Operators in English. PhD dissertation, UCLA. Horn, L.R. 1989. A Natural History of Negation. Chicago IL: Chicago University Press. Ioup, G. 1977. Specificity and the interpretation of quantifiers. Linguistics & Philosophy 1: 233–245 Kang, H.K. 2001. Quantifier spreading: Linguistic and pragmatic considerations. Lingua 111: 591–627. Karttunen, L. 1969. Problems of Reference in Syntax. Bloomington IN: Indiana University doctoral dissertation. Kratzer, A. 1998. Scope or pseudoscope? Are there wide-scope indefinites? In Events and Grammar, S. Rothstein (ed.), 163–196. Dordrecht: Kluwer. Kripke, S. 1977. Speaker’s reference and semantic reference. In Midwest Studies in Philosophy, Vol. II: Studies in the Philosophy of Language, P.A. French, T.E. Uehling, Jr. & H.K. Wettstein (eds.), 255–276. Morris MN: University of Minnesota. Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge: CUP. Ludlow, P. & Neale, S. 1991. Indefinite descriptions: In defense of Russell. Linguistics and Philosophy 14: 171–202. Musolino, J. & Lidz, J. 2006. Why children aren’t universally successful with quantification. Linguistics, 44: 817–852. Noveck, I. 2001. When children are more logical than adults: Experimental investigations of scalar implicatures. Cognition 78: 165–188.
Sharon Armon-Lotem Papafragou A. & Musolino, J. 2003. Scalar implicatures: Experiments at the semantics-pragmatics interface. Cognition 86: 253–282. Paris, S.G. 1973. Comprehension of language connectives and propositional logic relationships. Journal of Experimental Child Psychology 16: 278–291. Polya, T. & Tarnay, L. 2003. The capacity to keep track of specificity: The genuine vs. the linguistic. Paper presented at the workshop Direct Reference and Specificity at ESSLLI, Viena. Portner, P. & Yabushita, K. 2001. Specific indefinites and the information structure theory of topics. Journal of Semantics 18: 271–297. Reinhart, T. 2006. Interface Strategies-Optimal and Costly Computations. Cambridge MA: The MIT Press. Reinhart, T. 1982. Pragmatics and linguistics. An analysis of sentence topics. Philosophica 27: 53–94. Smith, C. 1980. Quantifiers and question-answering in young children. Journal of Experimental Child Psychology 30: 191–205.
Appendix Test items in the description mode - some Subject 1. 2. 3.
eyzoshehi yalda kafca la-mayim some (fm) girl jumped into-the-water eyzeshehu sus kafac meal ha-xaviyot some (ms) horse jumped over the-barrels eyzeshehu yeled axal sukariya some (ms) boy ate candy
Object 1. 2. 3.
ha-pil axal eyzoshehi prusa the-elephant ate some (fm) slice-of-bread (fm) hayalda kar’a eyzeshehu sefer the-girl read some (ms) book (ms) ha-sus nitka be eyzoshehi gader the-horse bumped into some (fm) fence (fm)
Test items in the description mode - or Subject 1. 2. 3.
ha-sus ha-lavan o ha-sus ha-xum kafcu meal ha-safsal the-horse the-white or the-horse the-brown jumped (pl) over the-bench ha-kelev o ha-xatul axlu ugiya the-dog or the-cat ate (pl) cookie ha-yeled o ha-yalda axlu vaflim the-boy or the-girl ate (pl) waffles
Children’s comprehension of logical words
Object 1. 2. 3.
ha-sus nitkal ba-safsal o ba-ec the-horse bumped into-the-bench or into-the-tree ha-kelev axal gvina o ugiya the-dog ate cheese or cookie ha-yeled sixek ba-kadur o ba-dubi the-boy played in-the-ball or in-the-teddy bear
Test items in the prediction mode - some Subject 1. 2. 3.
ani I ani I ani I
menaxesh she eyzoshehi yalda tikfoc lamayim guess that some (fm) girl will-jump to-the-water menaxesh she eyzeshehu sus yikfoc meal ha-xaviyot guess that some (ms) horse (ms) will-jump over the-barrels menaxesh she eyzeshehu yeled yoxal sukariya guess that some (ms) boy will-eat candy
ani I ani I ani I
menaxesh she ha-pil yoxal eyzoshehi prusa guess that the-elephant will-eat some (fm) slice-of-bread (fm) menaxesh she ha-yalda tikra eyzeshehu sefer guess that the-girl will-read some (ms) book (ms) menaxesh she ha-sus yitaka be-eyzoshehi gader guess that the-horse will-bump into-some (fm) fence (fm)
Object 1. 2. 3.
Test items in the prediction mode - or Subject 1. 2. 3.
ani menaxesh she ha-sus ha-lavan o ha-sus ha-xum I guess that the-horse the-white or the-horse the-brown yikfecu meal ha-safsal will-jump (pl) over the-bench ani menaxesh she ha-kelev o ha-xatul yoxlu ugiya I guess that the-dog or the-cat will-eat (pl) cookie ani menaxesh she ha-yeled o ha-yalda yoxlu vaflim I guess that the-boy or the-girl will-eat (pl) waffles
Object 1. 2. 3.
ani I ani I ani I
menaxesh she ha-sus yitkel ba-safsal o ba-ec guess that the-horse will-bump into-the-bench or into-the-tree menaxesh she ha-kelev yoxal gvina o ugiya guess that the-dog will-eat cheese or cookie menaxesh she ha-yeled yesxek ba-kadur o ba-dubi guess that the-boy will-play in-the-ball or in-the-teddy bear
On the “vulnerability” of the left periphery in French/German balanced bilingual language acquisition Matthias Bonnesen
In this chapter I investigate the question of whether there is evidence for crosslinguistic influence in German/French bilingual first language acquisition, using data from two children of the DuFDE-corpus (see Köppe 1994), Annika and Pierre. The theoretical framework on which my analysis is based is the Principles and Parameters approach as in Chomsky (1986).1 Recently Platzack (2001) has proposed that, in contrast to the IP, the CP, i.e., the left periphery of the clause, is vulnerable. He states that the structure of the left periphery of one language might influence the structure of the other language. Since German and French differ clearly as far as the left periphery is concerned, a German/French bilingual L1 corpus offers an excellent opportunity to verify the “vulnerability of the left periphery” hypothesis. German is analyzed as a verb-second (V2) language (den Besten 1977, Platzack 1983) which means that the finite verb always occupies the second position in main clauses via movement to C0, whereas in French, a fronted object or an adverb is adjoined to the IP and therefore results in a V3-structure. I will show that the children nearly always use target-like V2-constructions in German and that the general I-to-C-movement rule of German is not transferred to French. Nevertheless, in Annika’s data, several V2-constructions are attested in French with the adverbs ‘là’ and ‘ici’ and the object pronoun ‘ça’ in the left periphery. But this kind of lexically–driven V2-structure in French does not support the “vulnerability of the left periphery” hypothesis. In contrast, I will argue that only some lexical items can trigger V2-constructions.
1. Introduction Several studies in bilingual first language acquisition have revealed that bilingual children can separate the two languages from an early stage onwards (see for example 1. In Bonnesen (2004) I argue that Rizzi’s (1997) analysis of the left periphery is rather problematic. For other authors who argue against Rizzi’s “Split-CP” approach see for example de Cat (2002, 2004), Labelle (2000).
Matthias Bonnesen
Genesee, Nicoladis & Paradis 1995 for English/French; De Houwer 1994 for Dutch/ English; Lanza 1997 for English/Norwegian; Meisel 1990; Meisel & Müller 1992; Parodi 1990 for French/German). In the recent discussion, it was proposed that cross-linguistic influence is only possible in certain language domains. The approach of vulnerable domains is discussed by several authors in Müller (2003) for different syntactic projections. One such domain is proposed by Hulk & Müller (2000) and Müller & Hulk (2001). The authors argue that the interface between syntax and pragmatics is a vulnerable domain. This assumption is based on the observation that the grammatical omission of an argument, e.g., the subject in a null subject language, is motivated by pragmatic considerations. Furthermore, the authors propose that cross-linguistic influence could be expected when there is a surface overlap between the two languages for a given structure, whereas the syntactic representation is different. In order to verify this assumption the authors investigate the use of object omissions in monolingual acquisition of French and Italian compared to French/German and Italian/German bilingual first language acquisition. In German the object can occupy the first position in V2-constructions. In this case it can be omitted as a so-called topic drop (Huang 1984). This option does not exist in the Romance languages. Indeed, Hulk & Müller observe that the bilingual children exhibit more ungrammatical object omissions in French and Italian than the monolingual children. This result can be accounted for by cross-linguistic influence in the syntaxpragmatics interface. In contrast to Hulk & Müller (2000) Valian (1991) investigates the distribution of null subjects in English/Italian bilingual children and presents different results. At the age of two, the children already separate the two languages with respect to the nullsubject parameter. In the null subject language Italian, the children show a subject omission rate of 70%, whereas in the non-null subject language English, the rate is only 31%. Serratrice (2002) confirms Valian’s study. In her investigation of a bilingual English/Italian child, the subject omission rate in Italian is about 75%. This percentage corresponds to the omission rate in monolingual acquisition of a null subject language as stated by Grinstead (1998, 2000) for Spanish and Catalan (80%). A similar percentage is calculated by Bates (1976) for Italian adults at 70%. Paradis & Navarro (2003) include the language use of the parents in their study of two Spanish monolingual children and one English/Spanish bilingual child who lives in England. The bilingual child produces less subject omissions in Spanish than the monolingual children. The most interesting result of their study is that this high rate of realized subjects is also attested by the parents of the child. Hence the authors conclude that the subject omission rate in Spanish of the bilingual child reflects the input of the language provided by the parents. In accordance with Silva-Corvalán (1994) they explain that the change of the subject omission rate in Spanish takes place in the adult language due to the contact with English, i.e., the adults develop a contact variety of Spanish and the child consequently acquires this variety.
On the vulnerability of the left periphery
In contrast to the studies mentioned above, Platzack (2001) focuses on a different aspect of the interface between syntax and pragmatics, the left periphery of the clause. Since the pragmatic elements topic and focus are realized in the left periphery due to pragmatic considerations, this structural domain can also be investigated to verify the vulnerability of this interface. Keeping in mind the results obtained by Hulk & Müller (2000) and Paradis & Navarro (2003) according to which the interaction between syntax and pragmatics is vulnerable, the left periphery as a further interface domain between syntax and pragmatics could be vulnerable as well. In his study Platzack investigates the languages German and Swedish by very young L1 learners, aphasics, children with Specific Language Impairment (SLI2) and in L2 acquisition. His results indicate that there is indeed a certain vulnerability in the left periphery. However, since he does not use undisturbed bilingual first language acquisition data, it remains an open question whether the left periphery is vulnerable in this kind of acquisition and if so up to which age. This chapter focuses exactly on this question. I analyze the data of two German/French bilingual children of the DuFDE-project3 (see below) with respect to the use of the left periphery. The simultaneous first language acquisition of French and German offers an excellent opportunity to test Platzack’s hypothesis. The left periphery refers to the position(s) before the subject and the finite verb, i.e., positions higher than SpecIP. German is a verb-second (V2) language, i.e., the finite verb always occupies the second syntactic position in main clauses. French is an SVO language, i.e., when an element is placed at the left periphery, the verb remains in I° and an XSV sequence occurs, in contrast to German where a fronted element triggers subject-verb inversion. Hence, the syntactic representations of the left periphery in French and German are very much different. In German, the maximal projection at the left periphery occupies the SpecCP position and the verb is located in C0, whereas in French elements in the left periphery are adjoined to the IP (see below), i.e., the CP is not involved at all. In second language acquisition several studies have revealed that L2 learners of a V2 language like German exhibit problems acquiring the V2 structure, in contrast to L1 (see for example Clahsen & Muysken 1986, 1989; Hakansson, Pienemann & Sayehli 2002). As far as I know, no research has been carried out on the vulnerability of the left periphery in bilingual first language acquisition. Hence this chapter, inspired by Platzack’s work, is the first investigation in this field and contributes to the more general questions whether and if so in which fields cross-linguistic influence occurs in bilingual first language acquisition. 2. SLI stands for Specific Language Impairment (see Leonard 1998) 3. The DuFDE (Deutsch und Französisch doppelter Erstspracherwerb, “German and French bilingual first language acquisition”) project was made possible by research grants from the “Deutsche Forschungsgemeinschaft (DFG)” to Jürgen M. Meisel. The project is described in Köppe (1994).
Matthias Bonnesen
In the following section, I present the syntactic analysis on which my discussion is based.
2. The V2-structure in German Due to the complementary distribution of V2-constructions and complementizers in German, den Besten (1977) proposes that in main clauses the finite verb occupies the position in which the complementizer is located in subordinates. In a generative framework, this means that the finite verb moves to C0. Bayer (1984) suggests that the I-toC-movement is triggered by a finiteness feature in C0. In order to give evidence for this approach, he shows that in Bavarian, a dialect spoken in southern Germany, inflected complementizers exist, i.e., complementizers which agree with the subject in certain Person and Number features. Since the SpecCP position may not be empty in German declaratives, a maximal projection must occupy SpecCP. This projection can be almost any part of the sentence, e.g., the subject, an object, an adverb etc.. This approach accounts for the fact that the finite verb always occupies the second position in main clauses:4 (1) a. b. c.
Ich ging schnell ins Kino. I went quickly to the cinema ‘I went quickly to the cinema.’ Ins Kino ging ich schnell. To the cinema went I quickly ‘To the cinema I went quickly.’ Schnell ging ich ins Kino. Quickly went I to the cinema ‘Quickly I went to the cinema.’
If, for example, an object is fronted, it occupies the SpecCP position and the finite verb is raised to C0. The subject remains in SpecIP, which accounts for the postverbal subject position. To illustrate this analysis, I present a tree diagram. In German, the CP is always activated in main clauses. When for example an adverb or an object is fronted, subject-verb inversion occurs. In French, on the other hand, the activation of the left periphery in declaratives does not trigger inversion. In fact, the CP is not involved at all, as I will show below.
3. The left periphery in French In French the verb remains in I° in declaratives. If an element is placed in the left periphery, XSV word order results. Following Carroll (1981, 1982) and Chomsky (1977), 4. Actually, the first position can be occupied by an empty category, i.e., an empty topic, due to the topic drop-principle which operates in German (see for example Huang 1984, Dürscheid 1989).
On the vulnerability of the left periphery CP
(2) Spec
C´ C°
IP Spec
I´ VP
I°
Ichi gingn ti schnell ins Kino tn Ins kinoj gingn ich schnell tj tn Schnellk gingn ich tk ins kino tn
I assume that the constituent in the left periphery is base-generated in an adjoined position, i.e., it is adjoined to the IP.5 Carroll (1981) demonstrates that several elements can be adjoined to the IP: (3) Moi, Marie, je l´ ai vu manger un bouleau entier. I Mary I her have seen to eat a birch whole ‘Me, Mary I saw eating a whole birch.’ (Carroll 1981: 84)
The adjunction approach can account for the fact that several elements may appear in the left periphery. (4) displays the structure on which my analysis in the remainder of this chapter is based. (4)
IP IP
XP Spec
I´ I°
Une pomme Jean l’a An apple
VP mangéei
John it has eaten
‘An apple John has eaten.’
5. For a similar approach within the Minimalist Program (Chomsky 1995) see for example de Cat (2004).
Matthias Bonnesen
Hence in French the CP is not activated when an element is topicalized and therefore I-to-C-movement is not possible and subject-verb inversion does not show up. In the following, I will discuss what kind of construction can be considered as evidence for or against cross-linguistic influence.
4. Possible influence of French on German Since in French, XSV order is used when the left periphery is activated, such constructions in German could be interpreted as a sign of French influence: (5) a. b.
*Jetzt Michael geht ins Kino. now Michael goes to the cinema ‘Now Michael is going to the cinema.’ *Ins Kino Michael geht jetzt. to the cinema Michael goes now ‘to the cinema Michael is going now.’
In (5) the initial constituent and the finite verb are not in a specifier-head relation as they should be according to the German structure. Therefore the constructions are ungrammatical. The word order corresponds to the French structure. If such constructions occur in the language of the bilingual children, they can be considered as evidence for language influence.
5. Possible influence of German on French If the German V2-structure was transferred to French, we should expect to find an activation of the CP, due for example to an initial object or adverb, resulting in Ito-C-movement, i.e., subject-verb inversion. Nevertheless, we should keep in mind that, according to Roberge (1986), Kaiser & Meisel (1991) and Kaiser (1992) among others, clitic subjects in French are verbal affixes. In Bonnesen & Meisel (2005) further evidence for the affix hypothesis is offered. We observe that in Spoken French, inversion in Wh-questions with a fronted Whphrase always occurs with non-clitic subjects, but virtually never with clitic subjects.6 Based on this observation, we argue that the clitic moves along with the verb as a ver6. Wh-constructions in which inversion with a DP subject is not allowed do not occur, instead, a clitic subject is used and the DP subject appears dislocated as in où il a mis le livre Jean? The complex inversion belongs to written French and is neither attested in Behnstedt’s corpora (1973) nor in Bonnesen and Meisel’s corpora (2005). (see Bonnesen and Meisel 2005 for details about inversion in spoken French). Hence, if a fronted Wh-phrase is used with a DP subject, inversion always occurs in the corpora.
On the vulnerability of the left periphery
bal affix to C0, whereas non-clitic subjects are located in SpecIP, thus accounting for word order regularities in Spoken French. Hence we should expect the same regularity in child language meaning that I-to-C-movement in V2-structures will only be detectable on surface order if a non-clitic subject is used:7 (6) a. b.
*Maintenant chante Jean. now sings John ‘Now John is singing.’ Maintenant il chante. now he sings ‘Now he is singing.’
In (6a) the subject is located in postverbal position, hence the verb must occupy a position higher than the subject position SpecIP, i.e., the verb has been moved to C0 and the adverb is placed in SpecCP. In contrast to that, in (6b) the verb can be located in I0 or have been moved to C0. As a verbal prefix, the subject clitic il is always located in preverbal position, independent of the position of the verb (see Bonnesen & Meisel 2005). Hence construction (6b) can be accounted for both in a French and a German analysis. The tree diagram below illustrates the German structure: (7)
CP C´
Spec
IP
C0 Spec
I´ I0
Maintenant [il chante]i
ti
Since it is not possible to determine in which position the verb is located in constructions like (6b), only utterances in which non-clitic subjects are used provide clear evidence for or against language transfer. If in such constructions the verb is moved to C0, the subject will occur in postverbal position:
7. Even if one does not analyze clitics as affixes, the fact has to be taken into account that, in spoken French, inversion always occurs in Wh-questions with non-clitic subjects and fronted Wh-phrases, but hardly with a clitic, i.e., the word order of Wh-questions with a fronted Wh-phrase is Wh-SCL-V (e.g., où il habite?, “where he lives”) and Wh-V-DP (e.g., où habite Jean? “where lives John”). Hence, independently of the underlying analysis, one should not find subject-verb inversion with clitics.
Matthias Bonnesen CP
(8) Spec
C′ C°
IP Spec
I′ I°
∗Maintenant chante Jean
now
ti
sings John
Having presented the structure on which my analysis is based, I now turn to the analysis of the child language.
6. The acquisition data The bilingual data was collected by the DuFDE research group which studied the simultaneous acquisition of German and French by preschool children who grew up in Germany longitudinally. The children’s speech was recorded on video every second week from the age of about one and a half to at least the age of five years. The recordings were transcribed and entered into a computer. French is the native language of the mothers, while the fathers’ native language is German. The parents use their respective native language when communicating with the children. Both children are rather balanced bilinguals (see Köppe 1994). They acquire finiteness, questions and subordinates in both languages at about the same age (see Bonnesen 2005; Meisel 1990; Möhring 2004). Furthermore, Pierre’s MLU8 is rather similar for both languages in all recordings and Annika’s MLU is about the same in both languages until the age of three years. Later her MLUs are always a bit higher in German than in French (see Köppe 1994 for the MLUs). In addition to the bilingual corpus I will refer to the data of the monolingual French child Philippe (Suppes, Smith and Leveillé 19739) in order to be able to compare the utterances of the bilingual children with a monolingual database. Philippe has been audio-recorded almost every week from the age of 2;01,19 to 3;03,12 (see MacWhinney 2000a, b for details).
8. MLU means mean length of utterances (see Köppe 1994). 9. The corpus is available under http://childes.psy.cmu.edu/ .
On the vulnerability of the left periphery
7. The use of the left periphery in German In this section I will examine whether the children use target-like constructions when the left periphery is activated in German. I will begin with Pierre.
7.1 The use of the left periphery in the corpus of Pierre’s German First of all, it is important to note that, compared to other bilingual and monolingual children, Pierre is relatively slow in his grammatical development (see for example Clahsen 1982, Meisel 1994). According to Meisel (1990), Pierre has acquired finiteness in German at the age of 2;11-3;0. He uses target V2-structures from the beginning of his production of finite utterances, with subjects and non-subjects in the first position as in: (9) a. b. c. d. e. f. g. h.
da hinten kann (2;10,12) over there can ‘over there can’ nu is’ er weg (3;00,07) now is it gone ‘now it is gone’ da fehlt aber was (3;00,21) there misses but something ‘but there is something missing’ da (=das) mach ich kaputt (3;02,23) this make I out of order ‘I will break this’ so ka’ man das [ne:] (=nicht) umdreh’n (3;03,15) like this can you it not turn around ‘you cannot turn it around like this’ gar kein schlüssel hat Sarah (3;06,00) at all no key has Sarah ‘Sarah has no key at all’ des kann man (xx) so machen (3;06,28) this can you like this make ‘you can do this like this’ ein mann brauch’ ich (3;09,05) a man need I ‘I need a man’
There are only three XSV-constructions in the whole corpus corresponding to French word order: (10) a.
gleich die fliegen (3;09,05) soon they fly ‘soon they are going to fly’
Matthias Bonnesen
b. c.
doch (ededede) hier es gibt kein rosa (3;10,04) yes here it exists no pink ‘yes there is no pink here’ ey hier ich kletter auch da hoch und (hol den) (4;08,20) hey here I climb also there up and fetch it ‘hey here I climb up there and fetch it’
In colloquial German the adverb hier (“here”) (14b, c) can be used in order to attract the attention of the hearer to the following utterance. In this context the adverb is dislocated and therefore the verb occurs in the third position. The German monolingual interviewer of Pierre also produces such constructions as in: (11) hier ich bin zu klein, brauch eine tasche (Interviewer, Pierre recording) here I am too small, need a bag (50 = 3;05,17) ‘here, I am too small, I need a bag’ In all other cases, Pierre uses the V2-structure with the adverb hier (“here”)as in: (12) a. b. c.
hier kommt das darüber (...) (3;09,05) here comes it over ‘here you put it over’ (x) hier is’ kein platz hier (...) (3;10,18) here is no room here ‘there is no room here’ hier is noch ein (4;04,25) here is more one ‘here is one more’
In (10a) an XSV-construction occurs similar to the ones with the adverb hier (“here”). The adverb gleich (“soon”) can also be used in V3-constructions in Spoken German. Clahsen (1982) also observes that monolingual German children occasionally produce V3-constructions, even some which are, in contrast to the examples in (10), clearly ungrammatical. The utterances in (10) therefore do not represent unambiguous evidence indicating language transfer. Furthermore from the age of 3;09 to 4;08 in which the only XSV utterances in the whole corpus are produced, 186 target AdvSV constructions occur. Hence the three XSV constructions which Pierre uses and which are also attested in monolingual German children and German adults represent only 1,6% of all constructions which contain a fronted adverb at this age. Other XSV constructions do not appear. Pierre for example produces 41 utterances with a preposed object until the age of 4;08. In all of them the target V2 construction occurs. Hence, Pierre’s German data does not confirm the hypothesis of a vulnerable left periphery.
On the vulnerability of the left periphery
In the following I will analyze Annika’s German data.
7.2 The use of the left periphery in the corpus of Annika’s German Annika produces V2-structures almost categorically from the beginning when using finite verbs as in: (13) a. b. c. d. e. f.
nä, so eins, das nehme [is] =(ich) (2;03,30) well such one, that take I ‘well one of those I take’ das, das hol [is] (=ich) (2;04,18) this this fetch I ‘this one I fetch’ da sitz is (=ich) (2;04,18) there sit I ‘there I sit’ hier ist da/das} (2;06,01) here is that ‘here is that’ dann mach ich ein haus (2;06,18) then build I a house ‘then I build a house’ drehen kann man das (2;08,19) to turn can you it ‘you can turn it’
In her corpus, one finds three XSV constructions up to the age of 3;00,15 which could be interpreted as evidence for French influence: (14) a. b. c.
doch, doch schneiden ich kann (2;07,01) yes yes to cut I can ‘yes I can cut’ das ich mach so, das brot (2;09,31) that I make this way the bread ‘that I make this way the bread’ ketten [is] habe keine (3;00,15) chains I have none ‘I don’t have chains’
In the same recording, Annika produces a similar utterance as in (14a), but with the correct word order. This constitutes evidence that the error in (14a) should be interpreted as a result of performance factors. (15) schneiden will ich (2;07,01) to cut want I ‘I want to cut’
Matthias Bonnesen
Throughout the same period, in which the three ungrammatical XSV constructions in (14) occur, Annika produces 273 target-like V2 sequences. In 60 of these, a non-subject is placed in first position. That means that in only 3 out of 63 utterances in which the first position is not occupied by the subject the ungrammatical French-like XSV sequence occurs which leads to a total of 4,8%. In other words, the target V2 construction occurs in 95.2% of all utterances in which an element other than the subject is fronted. Hence the error rate is so low that French influence is hardly a plausible explanation, especially because Clahsen (1982) occasionally observes similar constructions as in (14) in monolingual acquisition of German. After the age of 3;0 years, ungrammatical V3-constructions occur even less frequently than before. In most utterances Annika produces the target-like XVS sequence when a non-subject is used in the left periphery. Some of these very few ungrammatical V3/V4 sequences indicate that French influence is rather improbable as an explanation for these constructions: (16) a. b. c.
jetzt noch den löffel muss ich jetzt (3;01,26) now still the spoon must I now ‘now I still must the spoon’ so viele dadrauf sind so viele ne (3;04,05) so many on it are so many aren’t there ‘there are so many on it, aren’t there’ und dann oft spiele ich bei ihn (3,04,05) and then often play I at him ‘and then I often play at his home’
In the constructions listed in (16), the subject is located in postverbal position which does not correspond to the French SV word order in main clauses. She produces dislocated constituents followed by the target-like V2 structure in which the verb precedes the subject. Actually, between the age of 3;0 and 3;06 there is only one utterance like the examples given in (14) which could be taken as evidence for French influence out of over 300 correct V2 constructions. Considering all these aspects, we can conclude that no unambiguous evidence can be found speaking in favour of language transfer from French to German, as far as the aspect of the left periphery on which I focus here is concerned. I now turn to the French corpus.
8. The use of the left periphery in French In this section I will analyze the French data in order to find out if the children use ungrammatical V2 constructions in French which correspond to German word order. As in the section before, I will begin with Pierre.
On the vulnerability of the left periphery
8.1 The use of the left periphery in the corpus of Pierre’s French As expected within an affix analysis of the clitics, Pierre does not use a single example where subject clitics appear in ‘inverted’ position when the left periphery is activated. Instead, he uses SCL-V sequences as in: (17) a. b. c. d.
ap- après on vois vois dadapeau après (3;00,07) afterwards you see see flag afterwards ‘afterwards you see flag afterwards’ et avec ça on descend (3;08,21) and with that you go down ‘and with that you go down’ deux cr/o/p/e/s(=trompettes) il a apporté (3;11,28) two trumpets he has carried ‘two trumpets he has carried’ ça /i/ faut mettre au poubelle (4;01,10) this you must put in the dustbin ‘this you must put in the dustbin’
Consequently, based on an analysis according to which clitics are verbal affixes, one has to conclude that it is impossible to decide whether the clitic-verb complex is located in I° or in C0 (see above).
Nevertheless, Pierre also produces XSV sequences. Only one exception can be found, i.e., an example where he produces a V2-construction: (18) là est la kélo (=vélo) (2;11,17) there is the bike ‘there is the bike’
I will focus on V2-constructions with the adverb là (“there”) in the next section. As far as the analysis of Pierre’s data is concerned, utterance (18) is negligible since he produces the XSV sequence in all of the 30 other constructions in which an element which is not the subject is used in the left periphery and a DP subject occurs as in: (19) a. nainant (=maintenant) la pot (=porte) est fermée (3;02,11) now the door is closed ‘now the door is closed’ b. m: mainant la cochon veut pas sorti (=sortir) (3;02,11) now the pig wants not to go out ‘now the pig does not want to go out’ c. après le docteur veut donner le médicament le docteur (3;11,02) afterwards the doctor wants to give the medicine the doctor ‘afterwards the doctor wants to give the medicine the doctor’
Hence Pierre’s data provides clear evidence against the vulnerability of the left periphery hypothesis.
Matthias Bonnesen
Now I turn to the corpus of Annika.
8.2 The use of the left periphery in the corpus of Annika’s French As expected, neither Annika nor Pierre produce a single construction with inverted subject clitics. In most cases Annika also produces the target XSV sequence when nonclitic subjects are used, as in:10 (20) a. b. c. d. e. f.
(le) collier Annika a (2;08,18) the necklace Annika has ‘the necklace Annika has’ ça moi descend (2;10,13) this I go down ‘this I go down’ et puis moi a rien du tout (3;02,24) and then I have nothing at all ‘and then I have nothing at all’ parce que un fois Daniel l’ a aidé à ouvri (3;06,16) because once Daniel her has helped to open ‘because once Daniel has helped her to open’ et puis Daniel Daniel a tiré tiré avec moi (3;07,13) and then Daniel Daniel has drawn drawn with me ‘and then Daniel Daniel has drawn with me’ maintenant le gâteau est fini (4;00,29) now the cake is finished ‘now the cake is finished’
In (20) Annika uses target-like XSV structures. These constructions can thus be considered as evidence against German influence in the left periphery. In addition to target XSV sequences, she also produces V2-constructions. In fact, she uses them consistently when the adverbs là (“there”) and ici (“here”) or, until the age of 3;11, when the object pronoun ça is located in the left periphery and a non-clitic subject occurs. See examples in: (21) a.
ça a toi fait? (2;10,27) that have you done? ‘have you done this?/you have done this?’
10. Note that Annika uses the pronouns moi (“I”) and toi (“you”) as DP subjects without coreferent clitics during a certain stage in her grammatical development. This does not conform to the target norm. In Bonnesen (2004), I show that moi and toi are used in the same way as other DP subjects at this stage. In Wh-questions, moi and toi are always inverted, as are DP subjects in general. Therefore, utterances with these pronouns as well as other DP subjects indicate whether Annika uses the I-to-C-movement.
On the vulnerability of the left periphery
b. c. d. e. f. g. h.
là est la maman (3;02,24) there is the mummy ‘there is the mummy’ là vient le père noël (3;05,04) there comes the Santa Claus ‘there comes Santa Claus’ là sont les lettres (3;06,16) there are the letters ‘there are the letters’ là est toi (3;07,02) there are you ‘there are you’ ça a fait moi (3;07,02) this have made I ‘this I have made’ ici sont ses pieds (3;10,26) here are his feet ‘here are his feet’ ici est le nouveau (4;05,26) here is the new one ‘here is the new one’
Apart from the lexemes là, ici and ça, Annika produces a 100% correct rate (16/16) of X-S-V sequences as in (20). According to Lahousse (2002), V2-constructions with the adverbs là and ici are grammatical in French. However, it has to be noted that this observation refers to written language. In the corpus of the monolingual child Philippe (Suppes, Smith & Leveillé 1973), not a single inversion can be found in a main clause. The monolingual French interviewers of Annika, Pierre and Philippe do not use such V2-constructions either, and French native-speakers also regard these constructions as being unusual. In contrast to là and ici, the object pronoun ça is ungrammatical in the left periphery in V2-constructions in written language as well as in spoken language. It is striking that only the object pronoun ça is attested in the corpus of Annika in V2constructions, but no other objects. Annika not only uses ungrammatical V2 constructions. Möhring (2004) observes that she produces some ungrammatical OV sequences as well. Rather interesting with respect to Annika’s V2 sequences in French is the lexical item which occurs as an object in the incorrect OV constructions. According to Möhring (2004), during the period in which Annika has only acquired one functional projection in which finiteness is located, one finds only two ungrammatical non-finite constructions with OV sequences. In addition to these constructions, only four ungrammatical SVfinOVinfin sequences are attested in the whole corpus. All of them contain the object pronoun ça:
Matthias Bonnesen
(22) a. b. c. d. e. f.
non ça seul cherch[e] là (2;04,18) no this only to search there ‘no search only this’ ça enlev[e] ça enlev[e] (2;06,01) this to pick up this to pick up ‘to pick up this’ parce que toi toi a ça cassé (2;10,27) because you you have this broken ‘because you have broken this’ il faut ça faire (3;00,29) you must this do ‘you must do this’ je peux ça faire (4;05,26) I can this do ‘I can do this’ mais /s/ /s/ je veux ça faire comme ça (4;09,28) but I want this do like this ‘but I want to do it like this’ (examples (22a-e) in Möhring 2004: 196, 201)
It is striking that in all the utterances in (22) the object is the pronoun ça. Since the lexical item ça occurs in a position apparently reflecting German word order in the left periphery as well as in the object position, one can suppose that this pronoun is responsible for the ungrammatical word order. Both OV and ungrammatical V2 constructions are obviously lexically restricted. Hence in contrast to Platzack’s hypothesis (2001), in the data I investigated, the left periphery is not a vulnerable domain; after all the general I-to-C-movement rule of German is not transferred to French. In general Annika produces the X-S-V sequence when an element in placed in the left periphery as in the examples in (20). Only three lexemes occur in German-like V2 constructions. One of these items also occurs in a German-like OV construction which is ungrammatical in French. Apart from these items, Annika produces 100% correct X-S-V sequences instead of X-V-S. Overall, even though we find certain V2 constructions in the French corpus of Annika, the data of both children does not support the hypothesis of the vulnerability of the left periphery as far as the structures investigated in this chapter are concerned.
9. Discussion of the results To summarize the results obtained in this study, a table is given below. The German data used in this chapter provides clear evidence against the hypothesis that the left periphery is vulnerable for language influence. The French data of Pierre does not indicate language influence either.
On the vulnerability of the left periphery
Table 1. Results in percent Construction
Correct V2
Incorrect XVS
Pierre 98.4 1.6 German Annika 95.2 4.8 German Pierre French Annika French
Correct XSV
Incorrect V2
96,8
3,2
(100%, only lexically (0%, only lexically restricted V2) restricted V2)
Only in the French corpus of Annika do we find certain V2-constructions which provide evidence for German influence. But it is obvious that these constructions are lexically restricted. Apart from the lexemes là, ici and ça Annika produces 100% target XSV constructions when elements are located in the left periphery, meaning that the general I-to-C-movement of German is not transferred to French. As far as the V2-constructions are concerned, we should keep in mind that in standard French, V2 is possible with certain adverb as, according to Confais (1980), peut-être and à peine, or, according to Lahousse (2002), là and ici. 11 Hence most of the V2-constructions used by Annika, i.e., all V2-utterances with là and ici, are grammatical in a certain register of French. However, since this register rather reflects written language, we would not expect children acquiring their mother tongue to use such constructions. The observation that the other bilingual child, Pierre, and the monolingual child, Philippe, do not use V2-constructions in French can be seen as evidence for this. Since, so far, I have only investigated the data of one monolingual and two bilingual children and since V2 in French occurs only in the language of one of the bilingual children, it is unclear whether German influence can account for these constructions. After all one can assume that French children receive a certain amount of V2 constructions with specific adverbs in their input since they are probably exposed to stories read to them by their caretakers. The question is whether there are at least a few monolingual French children who integrate such constructions into their grammar on the basis of their probably very little V2 input. A possible assumption is that German/French bilingual children need less V2constructions in their input to integrate them into their grammar because they have V2 available in their German, i.e., German may function as a kind of boot-strapping for the acquisition of V2 in French. However, according to this hypothesis, a certain amount of French V2 input, maybe very little, is needed in order to integrate V2 into French. This is all the more plausible as Annika acquires finiteness in French and in German at about the age of 2;03–2;04 (see Bonnesen 2005, Möhring 2004) and from then on uses V2 regu11. For other cases of V2 in French in declaratives see for example Bonami (1999).
Matthias Bonnesen
larly in German. In contrast, V2 in French occurs more than seven months later, i.e., she only produces XSV sequences in the first seven months after the acquisition of finiteness when the left periphery is activated. If Annika integrated V2 into her French only due to her German, why does it take seven months for her first V2 construction to occur in French? If we assume that at least a little V2 input in French is needed, we can account for why V2 in French appears later than in German. Since such a construction is rather rare in French, it takes a lot longer than in German to receive the necessary input. If the assumption of the boot-strapping is true, one would rather expect to find V2 constructions in French/German bilingual than in monolingual French children. A different explanation for Annika’s V2 constructions in French is that they are indeed triggered by her German, i.e., that they constitute a case of language influence. But since most of her V2 constructions are grammatical in written language, it is speculative to consider them as evidence for language influence. Only Annika’s V2-constructions with ça are clearly ungrammatical and hence may more likely indicate German influence than the V2 with là and ici. However, there are only three of these constructions in the whole corpus, all of them produced before the age of 4;0 years. Furthermore, the lexical item ça is not only used incorrectly in a V2construction, but also in a German OV sequence which is ungrammatical in French. This provides further evidence for a lexically restricted phenomenon which is, in the case of ça, not limited to the left periphery. The fact that only certain lexemes are attested in a German word order, one of these even in different constructions, goes against Platzack’s hypothesis of a vulnerable left periphery. If Platzack was right, one would expect to find all kinds of elements in the left periphery in a V2 construction in French. Since this does not show up, Annika’s French corpus, as well as the other corpora investigated in this chapter, rather provides evidence against a general vulnerability in the left periphery. If we take Annika’s French data as evidence for language influence, then it can only be seen as a lexically restricted phenomenon. This observation points in a new direction which has been neglected within the discussion of language influence. The results of this chapter indicate that we should not only focus on language interfaces as vulnerable domains as proposed in Hulk & Müller (2000); Platzack (2001) and Müller (2003), but also consider the role of the lexicon in bilingual language acquisition. Especially Annika’s use of the pronoun ça is rather striking. This lexeme seems to function as a bridge to the structure of the other language since it is attested in ungrammatical German OV constructions as well as in ungrammatical German V2 sequences in French. But since there is only one lexeme in the data of one child, it is also possible that the OV and V2 constructions are not connected, i.e., that the lexical item is not the trigger of the ungrammatical construction in both cases. Further research focusing on the lexicon in bilingual language acquisition is needed to see if there are other lexical items which seem to function as bridges to the structure of the other language. If one assumes that the ungrammatical V2 and OV sequences are indeed lexically triggered, another important question must be addressed: Why do only certain
On the vulnerability of the left periphery
lexemes occur in the word order of the other language, or in other words, what are the features of these items which make them occur in ungrammatical positions? The adverbs là and ici may occur in V2 because of Annika’s French input. But the pronoun ça is ungrammatical in spoken and written language. Focusing on the lexemes which occur in V2, it is striking that these items share certain features. All three elements are deictic, they are phonetically very short and they are generally acquired very early in child language, which is also true for the German equivalents (see Bonnesen 2005 for details). Perhaps these features are the key to the V2-constructions or, in a more general sense, if we assume that Annika’s V2 constructions occur due to her German, are the key to language transfer. Further research is necessary to see if one of these features or the combination of all of them could be the basis for lexically restricted influence. If the hypothesis that Annika receives a certain amount of V2 in her French input with là and ici is correct, the shared features of both these lexemes and of ça may account for why she also uses three V2 constructions with ça, though she has probably never heard such a construction. In this case one could also find such constructions in monolingual children. Further research in the monolingual acquisition of French is needed to see if one can find certain V2 constructions in declaratives in monolingual children, a possibility which should not be excluded. With the data presented in this chapter one can only speculate. Furthermore, we must address the question of whether V2 could be transferred to a language which has no evidence for V2 at all. After all, as I have argued, in written French V2 with certain adverbs is grammatical in declaratives and in written and spoken language, V2 is used in questions.
10. Summary In this chapter I have shown that the left periphery is not vulnerable. In German there is no evidence for language influence in the left periphery. In French only one of the children, Annika, uses V2 constructions in declaratives. Since only the lexical items là, ici and ça can trigger V2, it is clear that the general I-to-C-movement rule is not transferred to French. The object pronoun ça is also used in a German like OV sequence in the VP. Hence, this item seems to create a bridge into the structure of German. As far as the V2 constructions are concerned, I have argued that Annika may have received at least a few V2 constructions with the adverbs là and ici in her input. Since in German she has already acquired V2, German may function as a kind of boot-strapping for lexically restricted V2 in French, i.e., a bilingual child may need less V2-constructions in the input than a monolingual to integrate V2 into French. This possibility implies that at least a little French V2 input is needed in order to use such constructions in French. As far as ça is concerned, I have proposed that the shared features of the adverbs and the object pronoun, all three elements are deictic,
Matthias Bonnesen
phonetically short, and very frequent, could lead to her using V2-constructions with this item. Nevertheless, further research will be required to shed more light on the domain of possibly lexically restricted language influence and on the question if one can find V2 in declaratives in monolingual children.
Acknowledgement This study has been carried out as part of the research project “Simultaneous and Successive Acquisition of Bilingualism”. It is one of several projects funded by the DFG (Deutsche Forschungsgemeinschaft) within the Collaborative Research Center on multilingualism established at the University of Hamburg. I wish to thank Jürgen M. Meisel, Regina Köppe, Robert Mensching, Anja Moehring, Noemi Quintana, Susanne Rieckborn, Nicholas Burke and the anonym reviewers for their useful comments.
References Bates, E. 1976. Language and Context: The Acquisition of Pragmatics. New York NY: Academic Press. Bayer, J. 1984. Comp in Bavarian syntax. The Linguistic Review 3: 209–274. Behnstedt, P. 1973. Viens-tu? Est-ce-que tu viens? Tu viens? Formen und Strukturen des direkten Fragesatzes im Französischen [Tübinger Beiträge zur Linguistik 41]. Tübingen: Niemeyer. den Besten, H. 1977. On the interaction of root transformations and lexical deletive rules. GAGL: Groninger Arbeiten zur Germanistischen Linguistik 20: 1–78. Bonami, O.D. 1999. Les constructions de verbe: Les cas des groupes prépositionnels argumentaux. PhD dissertation, University of Paris 7. Bonnesen, M. 2005. Der Erwerb der linken Satzperipherie bei zwei Französisch-Deutsch bilingual auf wachsenden Kindern. PhD dissertation, University of Hamburg. Bonnesen, M. & Meisel, J.M. 2005. Die ‘Subjekt-Verb Inversion’ in Interrogativkonstruktionen des gesprochenen Französischen: Zum Problem der syntaktischen Variation. In Deutsche Romanistik – generativ [Generative Studies of Romance Languages in Germany], G. Kaiser (ed.), 31–48. Tübingen: Narr. Carroll, S. 1981. Notions fonctionnelles en grammaire transformationnelle: Dislocations et structures topicalisées en Français contemporain. PhD dissertation, University of Montréal. Carroll, S. 1982. Redoublement et dislocation en franÇais populaire. In La Syntaxe Composée du Français Standard et Populaire: Approcher formelle et fonctionnele. Tome 1, C. Lefebvre (ed.), 291–357. Québec: Office de la langue française. De Cat, C. 2002. French Dislocation. PhD dissertation, University of York. De Cat, C. 2004. French dislocation without movement. A minimalist account. Ms, University of York. Chomsky, N. 1977. On Wh-movement. In Formal Syntax, P.W. Culicover, T. Wasow & A. Akmajian (eds.), 71–132. New York NY: Academic Press.
On the vulnerability of the left periphery Chomsky, N. 1986. Knowledge of Language: Its Nature, Origin and Use. New York NY: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Clahsen, H. 1982. Spracherwerb in der Kindheit. Eine Untersuchung zur Entwicklung der Syntax bei Kleinkindern. Tübingen: Narr. Clahsen, H. & Muysken, P. 1986. The availability of Universal Grammar to adult and child learners – a study of the acquisition of German word order. Second Language Research 2(2): 93–119. Clahsen, H. & Muysken, P. 1989. The UG paradox in L2 acquisition. Second Language Research 5(1): 1–29. Confais, J.-P. 1980. Grammaire Explicative. Ismaning: Max Hueber. Dürscheid, C. 1989. Zur Vorfeldbesetzung in deutschen Verb-Zweit-Strukturen. Trier: Wissenschaftlicher Verlag Trier. Grinstead, J. 1998. Subjects, Sentential Negation and Imperatives in Child Spanish and Catalan. PhD dissertation, UCLA. Grinstead, J. 2000. Case, inflection and subject licensing in child Catalan and Spanish. Journal of Child Language 27: 119–156. Hakansson, G. Pienemann, M. & Sayehli, S. 2002. Transfer and typological proximity in the context of second language processing. Second Language Research 18(3): 250–273. De Houwer, A. 1994. The separate development hypothesis: Method and implications. In The Cross-Linguistic Study of Bilingual Development, G. Extra & L. Verhoeven (eds.), 39–50. Amsterdam: KNAW. Huang, C. 1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15: 531–574. Hulk, A. & Müller, N. 2000. Bilingual first language acquisition at the interface between syntax and pragmatics. Bilingualism: Language and Cognition 3: 227–244. Kaiser, G. 1992. Die klitischen Personalpronomina im Französischen und Portugiesischen. Eine synchronische und diachronische Analyse. Frankfurt: Vervuert. Kaiser, G. & Meisel, J.M. 1991. Subjekte und Null-Subjekte im Französischen. In ‘Det, Comp und Infl’: Zur Syntax funktionaler Kategorien und grammatischer Funktionen [Linguistic Studies 263], G. Fanselow & S. Olsen (eds.), 110–136. Tübingen: Niemeyer. Köppe, R. 1994. The DuFDE Project. In Bilingual First Language Acquisition. French and German Development, J.M. Meisel (ed.), 15–27. Amsterdam: John Benjamins. Labelle, M. 2000. Explorations on the acquisition of the left periphery. Ms, University of Montréal. Lahousse, K. 2002. Licensing NP subject inversion in French. Paper presented at Going Romance 2002 in Groningen. Lanza, E. 1997. Language Mixing in Infant Bilingualism: A Sociolinguistic Perspective. Oxford: Clarendon Press. Leonard, L.B. 1998. Children with Specific Language Impairment. Cambridge MA: The MIT Press MacWhinney, B. 2000a. The CHILDES Project: Tools for Analyzing Talk. 3rd edn. Vol. I: Transcription Format and Programs. Mahwah NJ: Lawrence Erlbaum Associates. MacWhinney, B. 2000b. The CHILDES Project: Tools for Analyzing Talk. 3rd edn. Vol. II: The Database. Mahwah NJ: Lawrence Erlbaum Associates. Meisel, J.M. 1990. INFL-ection: Subjects and subject-verb agreement. In Two First Languages: Early Grammatical Development in Bilingual Children, J.M. Meisel (ed.), 237–298. Dordrecht: Foris. Meisel, J.M. 1994. Getting FAT: Finiteness, agreement and tense in early grammars. In Bilingual First Language Acquisition: French and German Grammatical Development, J.M. Meisel (ed.), 89–129. Amsterdam: John Benjamins.
Matthias Bonnesen Meisel, J.M. & Müller, N. 1992. Finiteness and verb placement in early child grammars: Evidence from simultaneous acquisition of French and German in bilinguals. In The Acquisition of Verb Movement, J.M. Meisel (ed.), 109–139. Dordrecht: Kluwer. Möhring, A. 2004. Erst- und Zweitspracherwerb im Vergleich (Französisch und Deutsch). Der Erwerb der Wortstellung bei bilingualen Kindern und erwachsenen Lernern. PhD dissertation. University of Hamburg. Müller, N. 1993. Komplexe Sätze. Der Erwerb von COMP und von Wortstellungsmustern bei bilingualen Kindern (Französisch/Deutsch). Tübingen: Narr. Müller, N. 2003 (ed.). (In)Vulnerable Domains in Multilingualism [Hamburg Studies on Multilingualism 1]. Amsterdam: John Benjamins. Müller, N. & Hulk, A. 2001. Cross-linguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition 1(4): 1–21. Paradis, J. & Navarro, S. 2003. Subject realization and cross-linguistic interference in the bilingual acquisition of Spanish and English: What is the role of input? Journal of Child Language 30: 1–23. Parodi, T. 1990. The acquisition of word order regularities and case morphology. In Two First Languages: Early Grammatical Development in Bilingual Children, J.M. Meisel (ed.), 237– 298. Dordrecht: Foris. Platzack, C. 1983. Germanic word order and the COMP/INFL parameter. Working Papers in Scandinavian Syntax 2. Platzack, C. 2001. The vulnerable C-domain. Brain and Language 77: 364–377. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer. Roberge, Y. 1986. On doubling and null argument languages. Proceedings of the North Eastern Linguistic Society 16: 388–402. Serratrice, L. 2002. Overt subjects in English: evidence for the marking of person in an EnglishItalian bilingual child. Journal of Child Language 29(2): 327–356. Silva-Corvalán, C. 1994. Language Contact and Change. Spanish in Los Angeles. Oxford: Clarendon Press. Suppes, P., Smith, R. & Leveillé, M. 1973. The French syntax of a child’s noun phrases. Archives de Psychologie 42: 207–269. Valian, V. 1991. Syntactic subjects in the early speech of American and Italian children. Cognition 40: 21–81.
The subjects of unaccusative verbs in bilingual Basque/Spanish children Pilar Larrañaga The upsurge of interest in word order rules amongst generative scholars over the last ten years has led to approaches which combine syntax, pragmatics and phonology in a highly persuasive explanatory way (Zubizarreta (1998); Cinque (1993) to name but a few). This chapter examined the position of the subjects of unaccusative verbs and their pragmatic entailments in the speech of two Basque/Spanish children brought up bilingually. The longitudinal data were analyzed with respect to the position of the subject and its information status. The results showed that there were two clear patterns of acquisition depending on the language concerned and have been discussed in the light of old and new hypotheses. Our discussion will be centred around the issues of the left periphery (Rizzi 1997) and cross-linguistic influence (Müller & Hulk 2001) since the domain of preverbal subjects meets the criteria proposed by the latter.
1. Introduction The information structure of unaccusative constructions in Basque and Spanish is of particular interest for scholars working in the field of early bilingualism as the languages are in many respects diametrically opposed from a typological point of view. With respect to the position of the head, Basque is a left-branching and Spanish is a right branching language. In this chapter I will adhere to Burzio (1986) and assume that the subjects (S) of unaccusative verbs (V) are base generated in the complement position. This implies that subjects of unaccusative verbs will appear post-verbally in Spanish and pre-verbally in Basque in the so-called neutral readings. However, both languages share some crucial properties; both have free word order. This means that both languages will allow for SV and VS. I will postpone the discussion on information structure and word order until chapter 3. This chapter will investigate the issue of cross-linguistic influence with unaccusative verbs in early bilingual (Basque/Spanish) language acquisition. Recent research has been able to show that bilinguals keep their grammars separate, yet signs of crosslinguistic influence can be found at the interface of syntax and pragmatics (Müller & Hulk 2001; Hulk & Müller 2000; Serratrice et al. 2004). The question is whether the children use SV and VS constructions in the appropriate contexts in the respective languages or whether one language influences the other resulting in deviant word
Pilar Larrañaga
order-pragmatic entailments. The chapter is organised as follows. A literature review is the topic of chapter one. General properties of the unaccusative verbs are discussed in chapter two. Chapter three deals with the definitions of topic and focus. The method is explained in chapter 4. Chapters 5 and 6 deal with the results and discussion. A conclusion closes the chapter.
2. Previous research in acquisition of pragmatics Although a considerable amount of research has been devoted to the early acquisition of formal aspects of Spanish in the two last decades, it is only in more recent years that pragmatic concerns have attracted the attention of researchers working in a generative framework (Grinstead 2000; Serratrice et al. 2004). Most of this research has been centred around the issue of subject realization as opposed to subject drop (Grinstead 2000). Various studies summarized in Grinstead have reported that overt subjects are used by children at the age they implement the functional category I in their grammars irrespective of whether the language in question is [+pro-drop] or [–pro-drop]. Grinstead, working on Catalan/Spanish bilinguals, did not provide an in-depth analysis of the pragmatic facts ruling the use of subjects in these languages. As for unaccusative verbs, Bel (2003) who studied Catalan Spanish monolingual children, made the observation the position of the subject of the unnaccusative verbs is governed by pragmatic rules that are linked to the information status and not to the pro-drop parameter as it has been widely proposed in the past. A more elaborate account of pragmatic issues surrounding the acquisition of the pro-drop parameter has been put forward by Serratrice et al. (2004) who studied bilingual English/Italian as well as monolingual English and Italian children. They assessed the contexts in which subjects were dropped according to seven features; Person, absence, activation, contrast, differentiation in discourse, query and the transitivity of the predicate. They allocated points for the subject being dropped according to the principle of pragmatic principle of informativeness, i.e., a subject that refers to a person that is present receives 0 points as opposed to 1 point if the referent is absent. The idea behind this was that, if the referent is absent, it is more informative and the subject is, therefore, less likely to be dropped. I will not discuss the other features in any depth here, because they are not relevant to the present study, but all the other features were allocated points in the same way. The noteworthy issue is that very little has been written in the area of pragmatic issues in spoken language. More recently, Thornton (2002) studied early utterances of one monolingual child acquiring English. She drew special attention to so called genitive subjects in early English such as in (1):
(1) No, my do that (A.L. 1;10)
Subjects of unaccusative verbs
She claims that most of the subjects of this sort have a contrastive reading since they are used in contexts in which a presupposition is negated. The remaining subjects have been analyzed as emphatic according to a recent approach by Echepare (1998) who has proposed on the basis of evidence from Basque that contrastive focus and emphasis have to be dealt with differently. Emphasis is more topic like as opposed to contrast, which involves syntactic movement in Basque. Thornton does not deal with the issue of verb class in her paper. It is interesting, however, that most of the examples quoted are transitive verbs. Volterra & Taeschner’s (1978) who proposed a 3 stage development in which children would first have one lexicon, secondly two lexicons and one grammar, to achieve grammar differentiation at a subsequent stage has been widely rejected at this stage. However, recent work in the field has proposed that the two languages a bilingual individual is acquiring can influence each other if certain requirements are met. Whilst some researchers claim that cross-linguistic influence is likely to be due to language external reasons (Genesee & Paradis 1996) such as language dominance, other authors support the idea that there are language internal reasons for it (Müller & Hulk 2001). Paradis & Genesee propose that if one language is more dominant than the other, this may result in delay, acceleration or transfer in the other one. Müller & Hulk point out that, if dominance had an impact on linguistic output, it would happen only in one direction: the stronger language would affect the weaker one. Assuming that crosslinguistic influence can take place for language-internal reasons seems more promising. Platzack (2001) has identified language domains located at the syntax-pragmatics interface , i.e., in the CP (Complementizer Phrase) that cause serious problems even to monolinguals. Working on the premises that Platzack has set, Hulk & Müller (2000) and Müller & Hulk (2001) have proposed that cross-linguistic influence may occur if the following requirements are met: 1. Cross-linguistic influence occurs at the interface between two modules of grammar, and more particularly at the interface between pragmatics and syntax in the so-called C-domain, since this is an area which has been claimed to create problems in L1 acquisition also. 2. Syntactic cross-linguistic influence occurs only if language A has a syntactic construction which may seem to allow more than one syntactic analysis and, at the same time, language B contains evidence for one of these two possible analyses. In other words, there has to be a certain overlap of the two systems at the surface level. The domain of language that will be studied in this chapter meets Hulk & Müller’s requirements (s. chapter 3). The SV-VS order in Spanish and Basque pertains to the C-domain (see below). Note that the neutral word order is VS in Spanish and SV in Basque, both involving the IP domain. Moreover, SV with a topic or contrast reading of S involves a domain which is higher than IP in both languages, i.e., CP. Moreover, SV
Pilar Larrañaga
can have a wide focus reading in Basque. VS is the word order in a wide focus context in Spanish, but the linear order VS is also allowed in if the most natural position for the focused entity in Basque is the pre-verbal one, traditional grammars (Euskaltzaindia 1984) acknowledge that focused entities may appear in post-verbal position. This is possible in story telling contexts. Therefore, Basque and Spanish overlap at the surface level both languages allowing SV as well as VS. Nevertheless, those word orders have different analyses, in the sense that Basque allows for VS only in story telling contexts and Spanish uses VS in wide and narrow focus contexts. Summarizing, a domain in grammar at the syntax-pragmatics interface of Basque and Spanish has been identified which is susceptible for cross-linguistic influence in early bilingual language acquisition, crucially involving the left periphery of the sentence. Therefore, this chapter aims at providing an exhaustive analysis of word orders and their pragmatic entailments in both languages. The issue of cross-linguistic influence is the second research question. If there is cross-linguistic influence, and this is internally caused as proposed by Müller & Hulk, here are the two possible scenarios; 1. If Basque influences Spanish, SV should appear in all contexts (wide and narrow focus) in Spanish, if on the contrary Spanish influences Basque, 2. VS structures with narrow and wide focus in Basque in non story-telling situations should occur as well as StopV.
3. The unaccusative verbs Before we concentrate on unaccusative verbs it is necessary to draw some attention to the issue of intransitive verbs, currently divided into unaccusative and unergative. It is relatively easy to draw the line between unaccusative and unergative verbs in Basque since Basque has different auxiliary types. All unaccusative verbs are inflected with the mono-argumental auxiliary of the verb izan ‘to be’, the unergatives with the auxiliary edun ‘to have’. There is a vast literature that has reported on the difficulty of establishing a clear borderline between unaccusative and unergative verbs (Levin & Rappaport 1995) especially in those languages lacking different auxiliaries, i.e., Spanish, one of the main reasons for this is the fact that this class of verbs is semantically very heterogeneous. Nevertheless, some tests that will be presented below have turned out to be rather strong devices in determining to which category the verb in question belongs. Nominalisation and participial constructions are amongst the strongest. Unaccusative verbs (2) disallow nominalizations whereas the latter are possible with unergatives (3). Moreover, unaccusative verbs allow participial constructions with post-verbal subjects (4), whereas this possibility is banned for unergative verbs (5). (2) *El caedor *Eroriraria *The faller
Subjects of unaccusative verbs
(3) El bailaor Dantzaria The dancer (4) Terminada la reunión, se tomaron todos un refresco (Once) the meeting was over, everyone drank a refreshment (5) *hablado Juan, nos fuimos al teatro *spoken Juan, we went to the cinema
As to the heterogeneity of this verb class, researchers (Levin & Rappaport 1995; Mendikoetxea 1999) have identified four main subclasses within the class of unaccusative verbs; verbs of change of state (internal and external) (6, 7), verbs of appearance (8) and verbs of motion (9). (6)
Se ha estropeado el coche Autoa-Ø ondatu egin da Car-abs break make is The car broke down
(7)
Se ha sonrojado Lorea Lorea-Ø gorritu da Lorea-abs blush is ‘Lorea blushed’
(8)
Ha aparecido el dinero Dirua-Ø azaldu da Money-abs appear is ‘The money appeared’
(9)
Se ha ido Lorea Lorea-Ø1 joan da Ana-abs go is ‘Ana lefT’
However, not all unaccusative verbs behave alike with respect to the position of their subjects. Pinto (1997) has shown for Italian that only those unaccusative verbs that are subcategorised for a [+locative] argument do not allow for VS with a neutral reading because this argument occupies the complement position. The same seems to apply to Spanish. This is relevant, insofar as children may use both types of unaccusative verbs, thus, possibly complicating the issue of information status. Although the type of NP does not seem to influence the position of S with unaccusative verbs, it is worth mentioning some trends that Dixon (1994) has observed in various ergative languages that led him to posit a nominal hierarchy. He notes that lexical NPs are more likely to appear as subjects of unaccusative verbs or objects of 1. The subject of unaccusative verbs is zero marked (absolutive case) as are direct objects.
Pilar Larrañaga
transitive verbs as opposed to pronouns and demonstratives that tend to occur as subjects of transitive verbs. Hence, the probability of finding subjects in the form of personal or demonstrative pronouns with unaccusative verbs in child language is rather low. Moreover, the fact that both languages are pro-drop (Basque is a subject and object drop language) makes it even less likely.
4. Topic and focus: a syntactic approach In this section, information focus will be briefly defined and will be linked to the issue of neutral word order for unaccusative verbs. The section concludes with a brief definition of contrastive focus and topic, addressing the issue of the representation in the left periphery. Because the domain of pragmatics that studies word order is overloaded with terminology that is not always used in the same terms of reference, for the sake of clarity I will briefly explain how focus has to be understood in this chapter. To simplify matters, I will adhere to the terminology used by Zubizarreta (1998, 1999) and her definition of focus while mentioning at times other terminology that seems to be relevant. Focus is the nonpresupposed part of the sentence (Zubizarreta 1998:1). In terms of information structure many researchers (Vallduví 1993; Zubizarreta 1998, 1999) have proposed that new information is located in the last sentence position in Spanish, a distribution that is found in other Romance languages as well. Focus is left-adjacent to the verb in left branching languages such as Basque. New information is prosodically more prominent than old information, whereby a phonological stress rule is redundant because focus correlates with the basic word structure (Cinque 1993). Since information and contrastive focus are different in nature, Zubizarreta follows the question/answer strategy in order to define neutral or information focus. To the extent that the answer to a wh-question has the same presupposition as the question, the focus in a statement can be identified as the part of the statement that substitutes for the wh-phrase in the context question. Zubizarreta (1998:2)
Depending on the extension of focus, wide and narrow focus can be identified. Narrow focus (10–11) refers to a part of the sentence which is new, crucially the part that replaces the question being asked, often the object or the whole VP, the subject is, therefore, the topic. In terms of information structure the sentence is categorical (Lambrecht 1994). The whole sentence contains new information in the case of wide focus (12–13), and is thetic in Lambrecht’s terms. In the latter case the underlying question is ‘what happened’, in the former ‘who . . . . .’ with unaccusative verbs. (10) ¿Quién ha llegado?/ Ha llegado [LoreaF] ‘Who has arrived?’/‘Lorea has arrived’
Subjects of unaccusative verbs
(11) Nor etorri da? [LoreaF] iritsi da Who come has? Lorea come has ‘Who has arrived?’/‘Lorea has arrived’ (12) ¿Qué ha pasado?/ [Ha llegado LoreaF] ‘What happened?’/‘Lorea has arrived’ (13) Zer gertatu da?/ [Lorea iritisi daF] What happen has/ Lorea arrived has ‘What happened?’/‘Lorea has arrived’
Examples (10) and (12) show that subject-verb constructions with unaccusative verbs have a VS order with a wide focus as well as with narrow focus reading in Spanish. The constituent word order with unaccusative verbs is always SV (11,13) in Basque irrespective of the information status of S because Basque is left-branching. The sentences that are labelled by some scholars (Lambrecht 1994) as being categorical have, on the contrary, an SV word order, where S is topic and V is focus in both languages. Recall that Basque is a typical SOV language whereas Spanish is an SVO language (RAE 1989). Both orders are believed to have a ‘neutral’ reading in wide focus contexts which has been briefly discussed above. With respect to the subjects of unaccusative verbs, there is wide consensus amongst researchers that these are generated in the complement position, and that verbs that do not assign Case to their objects cannot assign a theta role to their subjects (Burzio 1986). Burzio’s generalization explains why subjects appear in post-verbal position with this type of verb in Spanish. As a matter of fact, those subjects are deep objects (Belletti 2001:77). Because of all the above, VS in Spanish and SV in Basque with wide focus reading are located in the IP domain. The narrow focus of the subject (SfocV) in Basque pertains to the C-domain. The issue of agreement and Case licensing of these post-verbal subjects has been a matter of much debate in recent years. Both nominative Case and agreement must be assigned by a head located in the I projection. Since I cannot govern the constituents within VP, a chain in which the post-verbal subject is linked to a null expletive in SpecIP has been widely assumed, although this results in violation of principle B. The distribution of subjects with unaccusative verbs in Basque has been straightforward until this stage. Recall that all 3 readings discussed until now allowed only SV. However, Basque allows for post-verbal focused entities in story telling contexts (Euskaltzaindia 1984). Although Basque scholars working in the generative framework are aware of this issue, to my knowledge, no one has attempted to put forward a syntactic analysis for it. The issue that arises is whether these entities are licensed syntactically or pragmatically. In order to deal with pre-verbal subjects, the concepts of topic and contrastive focus will be introduced in what follows. I will adhere to the definition of ‘aboutness’ as regards sentence topic (Zubizarreta 1998:7ff.). The topic is what the sentence is about, and the rest is the focus, whereby the former usually precedes the latter. It goes without saying that sentences in wide focus are topic-less. All the above is valid for
Pilar Larrañaga
both languages. A contrastive focused element may precede the verb in both languages irrespective of its syntactic status. Contrastive focus is a different type of focus. Zubizarreta (1998:7) points out that ‘contrastive focus negates certain aspects of the assertion introduced by its context statement’. Both topic and contrastive focus are anchored in the preceding discourse. Summarizing, both topics and contrastive focused elements may precede the verb in both languages. Whereas a wide range of studies assumes that topic and focus are pragmatic phenomena that ought to be dealt within LF (Zubizarreta 1998 amongst many others), recently, Rizzi (1997) has proposed that topic and focus have a representation at the syntactic level, in an articulated left periphery involving the CP. Leaving aside the arguments that lead him to make this proposal, he suggests that the left periphery of a sentence is complex much in the same way as the IP of Pollock (1989). Simplifiying his proposal somewhat, the CP is split into two compulsory components; ForcePhrase on the top and FinP-IP located below, and TopicPhrase and a Focusphrase which are sandwiched between the former and the latter. Topic is recursive, whereas focus does not allow recursion. Another innovation of Rizzi is that he introduces a topic and a focus criterion very much in the vein of the wh-criterion which have to be satisfied. The whole CP is discourse related but, whereas ForceP is compulsory and hosts conjunctions, TopP and FocP are optional phrases that are projected solely when an entity is marked with the feature [+topic] or [+focus] respectively. This point requires some more attention because Rizzi is not very explicit about the notion of topic. In this respect, Casielles-Suárez (1997:122f.) argues that “subject raising in Spanish is related to the notion of topic”. This implies that subjects that are marked with the feature [+topic] must leave the focus domain, i.e., the IP. The issue is where these subjects land. Adopting Rizzi’s proposal, Ordóñez & Treviño (1999) suggest that a sentence with an overt topic subject cannot fill the Spec-Agr position, having to project a Top-projection above IP, whereas a sentence with a covert subject would only project an IP. But, what about the subjects with a contrastive reading? As it has been shown above, contrastive focus is discourse anchored and belongs to the left periphery according to Rizzi. However, can contrastive subjects be located in SpecIP or must they move to FocusP in order to satisfy the focus-criterion? Contrast can be realized by means of phonological prominence in Spanish, an option that is disallowed in Basque. Theoretically, this type of subject in Spanish could stay in SpecIP and get its contrastive reading by means of phonological prominence. Nevertheless, setting a contrastive focused entity either subject or object in front of the verb in Spanish means to reverse the neutral order of topic-comment into contrast-presupposition irrespective of its syntactic function. The question that arises now is, if a marked pragmatic order is involved with both subjects and objects, should they have the same syntactic representation? If we assume that the IP domain is linked to information structure, which is related to the basic word order of a language, it seems plausible to assume that the CP is a better candidate to host all discourse related entities such as contrastive focused and topic DPs irrespective of the entity involved.
Subjects of unaccusative verbs
As for Basque, Ortiz de Urbina (1999a, 1999b) proposes that an articulated left periphery in Rizzi’s terms adequately explains the facts linked to contrast emphasis and topic in Basque (s. also Elordieta 1999). Therefore, both topics and contrastive subjects will be analysed following Rizzi’s proposal in both languages. For the subject of unaccusative verbs, this entails that subjects marked with the features [+contrastive focus] and [+topic] must be located above IP in both languages in the respective projections of Spec-FocP and Spec-TopP. Recall that the articulated left periphery is discourse related involving movement, whereas the projections below are linked to information structure, and hence to the basic word order.
5. Method Two male (Basque/Spanish) subjects brought up bilingually from birth in the Basque Country were videotaped (P 1;11,00-4;00,02, M 1;07,14-3;11;17)2 and transcribed by experienced transcribers. A second transcriber checked all the data. The Data belong to the Busde/HEGEHJ3 project carried out at the universities of the Basque Country (Spain) and Hamburg (Germany) under the direction of J.M. Meisel and I. Idiazabal. Methodological details can be found in Mahlau (1994) who provides an extensive overview of the data used in the present paper. A subsequent accurate perceptive analysis has been carried out by the author of this chapter on the relevant data in order to identify contrast and topic by looking at sentence stress and pitch. Moreover, the entire context of the relevant utterances has been analyzed in order to distinguish between topics, contrastive entities and topicalized entities. However, it was not always easy to make a decision on the topic-focus structure. Hence, some of the available data have not been taken into account for the information structure analysis since it was impossible to determine which pragmatic entailment they had (s. Thornton 2002 on this issue as well). The data have been divided into two stages: Stage I until 3, stage II from 3 onwards because Larrañaga (2000) made the observation that topicalizations appeared at around that age in both children. The widely accepted MLUm as been left out since utterance length in morphemes does not determine the information status of constituents, despite the trend of longer constituents to prefer a position after the verb. Also, and as Scarborough (1990:2) puts it 2. P stands for Peru and M stands for Mikel and will be referred to as P and M respectively in the remainder of this chapter. 3. This study was conducted as part of the research project “Simultaneous and successive acquisition of multilingualism’, under the supervision of Prof. Jürgen M. Meisel at the University of Hamburg (Germany). This project was one of the projects carried out at the Research Centre on Multilingualism established at the University of Hamburg and funded by the Deutsche Forschungsgemeinschaft (German Science Foundation) in a grant awarded to the first investigator Prof. Dr. Dr.hc. Jürgen M. Meisel.
Pilar Larrañaga
content validity has been seriously challenged [] on the grounds that qualitative aspects of syntactic development are not directly evaluated.
Moreover, one very important argument against the use of MLU is its cross-linguistic comparability as Thordardottir & Weismer (1998:3) state MLU levels cannot be compared directly across languages.
In the same vein, Hickey (1991:569) points out that: such comparisons may be useless at best and deceptive at worst. It seems advisable to regard MLU as a purely intra-linguistic device, allowing comparisons of the same child’s language over time, and between children acquiring the same language.
In order to decide whether a certain construction had a wide, narrow or contrastive reading the whole context has been carefully scrutinized in both languages. All utterances occurring in ‘out of the blue’ contexts were classified as having wide focus reading. If the verb was introduced, the subject was classified as having narrow focus reading. All those subjects in which a given presupposition was negated, have been classified as contrastive.
6. Results This section is organized as follows: To start with, I will analyze the information structure of the relevant structures. Tables 1 and 24 give a general overview of all the subjects used by both children in pre- and post-verbal position. Figures 1 and 2 show the position of the subjects of the unaccusative verbs in Basque and Spanish. Both pre-verbal and post-verbal subjects are found across the age-range in both languages, although to a different extent. The data clearly show that there are two well defined stages in Spanish, but subjects generally appear in pre-verbal position in Basque throughout the entire period in both children. Subjects in postverbal position in Spanish are in the majority in Stage I, whereas they become a minority in stage II. There are only a few pre-verbal subjects in Stage I in both languages in both children. The number of preverbal subjects increases in stage II. Moreover, a few postverbal subjects are used by both children in Basque. Figures 3–4 show the position of Spanish subjects broken down in pronouns (personal and demonstrative) and DPs. The rationale of this distribution is that, according to Dixon’s (1994) nominal hierarchy, nouns are more likely to be subjects of intransitive 4. The tables 1a-b and 2a-b are an adaptation of Larrañaga’s (2000) result and are located in the appendix . These tables (1–2) show that the verbs used are all verbs that are subcategorized for a [+locative] argument, thus all allowing for VS with a wide focus reading. The only exception to this is the verb haundia egin ‘‘to become older’’. The verbs used do not contain internally caused unaccusative verbs.
tokens
Subjects of unaccusative verbs 100 90 80 70 60 50 40 30 20 10 0
P M
SV
VS
SV
VS Stage II
Stage I
tokens
Figure 1. SV and VS structures by P and M in Basque.
100 90 80 70 60 50 40 30 20 10 0
P M
SV
VS
SV
Stage I
VS Stage II
Figure 2. SV and VS structures by P and M in Spanish.
verbs (unaccusatives in Burzio’s terminology) than of transitive verbs. In the same vein, Du Bois (1987) has shown that subjects of unaccusative verbs are more likely to be overt than subjects of transitive verbs. The present data reveal a clear pattern. In stage I, subjects in post-verbal position outweigh those in pre-verbal position in Spanish. These postverbal subjects are DPs in the majority of cases. There are only a handful of subjects in post-verbal position that belong to the pronominal class. Subjects in pre-verbal position are in the minority in stage I and belong to the pronoun class in most cases. In fact, the few subjects in pre-verbal position are almost all (M 7 out of 8 and P 10 out of 11) pronouns. Both children use a large number of pre-verbal subjects in stage II, most of which are pronouns as well. Pronouns and DPs in VS are evenly distributed in Stage II in both children.
Pilar Larrañaga 70 60
tokens
50 40
P
30
M
20 10 0
Pronouns
DPs Stage I
Pronouns
DPs
Stage II
Figure 3. Pronouns and DPs in subject position in Spanish SV constructions.
70 60
tokens
50 40
P M
30 20 10 0
Pronouns
DPs Stage I
Pronouns
DPs
Stage II
Figure 4. Pronouns and DPs in subject position in Spanish VS constructions.
Pronouns in post-verbal position are very rare in either stage in both children. This distribution seems to suggest that there is a strong tendency for pronouns to appear in pre-verbal position which is almost in complementary distribution with the DPs, which preferably appear in post-verbal position in stage I. Moreover, this trend becomes relevant for information structure as well, as it will be shown below. The Basque data (Figures 5–6) reveals that no two stages can be found, neither for SV nor for VS. Most subjects in both stages appear in pre-verbal position. Pronouns and DPs are equally distributed in the pre-verbal position. Subjects in post-verbal
Subjects of unaccusative verbs 70 60 tokens
50 40
P M
30 20 10 0
Pronouns
DPs Stage I
Pronouns
DPs
Stage II
Figure 5. Pronouns and DPs in Basque SV constructions by P and M. 70 60 tokens
50 40
P M
30 20 10 0
Pronouns
DPs Stage I
Pronouns
DPs
Stage II
Figure 6. Pronouns and DPs in Basque VS constructions by P and M.
position are very scarce, whereby P uses significantly more VS than M. DPs are slightly more frequent in the post-verbal position in stage II.
6.1 Information status of subjects 6.1.1 Pre-verbal subjects: Basque Figure 7 summarizes the information status of the preverbal subjects in Basque. In stage I, only wide (14a,c) and narrow focus (14b) readings occur. (14) a.
[Kotxie apu(r)tu daF] Car broken has ‘The car has broken down’ Context: I: zer diozu?/‘what are you saying?’
(M 2;02,11)
tokens
Pilar Larrañaga 40 35 30 25 20 15 10 5 0
emphasis wide narrow topic
P
M Stage I
P
M Stage II
Figure 7. Information status of the subjects in S Vunacc constructions in Basque by P and M.
b. c.
[AnaF] eto(rr)i da Ana come has ‘Ana arrived’ Context: They are talking about who has come with whom Eta [hau, gurpila mugitzen daF] And this wheel move aux ‘And this, the wheel moves’ P shows to the interlocutor how the wheel works
(M 2;00,00)
(P 2;09,05)
The corpus in stage II is much richer in terms of pragmatic entailments, allowing for topic (15 a-b) and two emphatic readings as well. (15) a. b.
[JurgiT] bai kabitzen da Jurgi yes fit aux ‘Jurgi does fit in’ [BambiT] egin da haundixe Bambi make aux big ‘Bambi has grown’ Interlocutor: Bambi be bai? Bambi as well?
(P 3;05,09)
(M 3;04,12)
There are, however, a few preverbal (M 4 in stage I and 9 in stage II, P 14 in stage II ) subjects with either constituents following the subject, verb or constituents preceding the subject that require a closer look. The instances with S XP V or XP S V are target like, because the subject is topic and the XP is focus in the first case and the opposite is true for the latter case (M 6, P 5) (16–17). Analyzing M’s data first, there are two utterances (18a–b) with wide focus interpretation in stage I and 3 in stage II. They all occurred in out of the blue contexts. The remaining (2)(19) involve negation, in which a post-verbal constituent is target-like as the negation is the focus of the sentence (Laka 1990) (21). As to P, all instances appear in stage II. Three of the subjects with postverbal constituents have been uttered in story telling contexts (18c) which is target like.
Subjects of unaccusative verbs
Four of the pre-verbal subjects have a topic reading, 2 of them have been classified as having wide focus reading. (16) a. b.
Ni, atzerunke joango naz I,5 backwards go-fut aux ‘I, go backwards’ Kotxeak, Migelengana joaten dira ‘The cars, go to Migel’
(M 3;04,12)
(17) a. b.
Eta ni, bestera noie ‘And I, went to the other one’ Nirea hemendik joaten da ‘Mine goes from here’
(M 3;05,10)
(18) a. b. c.
Hau doie beste garajera ‘This goes to the other garage’ Ni sartuko naz hemendik ‘I will go in this way’ Holako batean ni montatu naiz tren askotan ‘Once upon a time, I got in many trains’
(M 2;07,01)
(19) Ni ez noie zure etxera ‘I will not go home’
(P 3;04,04)
(P 3;07,27)
(P 3;07,27) (P 3;06,17) (M 2;09,04)
Some of the utterances that have been classified as having a wide focus reading do have a post-verbal constituent, which is not target like in Basque unless it is used in a story telling context (20–21). (20) Somorroa dator hemendik Bug comes here-from ‘The bug is coming from here’
(P 3;09,02)
(21) Kotxea, gelditzen da martxan Car, remain aux movement ‘The car kept going’
(P 4;00,15)
Summarizing, wide and narrow focus overweigh in the pre-verbal position, topic and emphatic readings remain in the minority. Moreover, contrastive focused subjects are virtually non existent.
6.1.2 Pre-verbal subjects: Spanish Turning now to the information status of the subjects in Spanish, Figure 8 summarizes the results. There are no pronominal subjects that have been classified as having
5. ,the comma stands for a pause in the speech flow.
tokens
Pilar Larrañaga 40 35 30 25 20 15 10 5 0
emphasis wide narrow topic P
M Stage I
P
M Stage II
Figure 8. Information status of the subjects in S Vunacc constructions in Spanish by P and M.
contrastive reading by both children. The children do use emphatic subjects in form of lexical DPs and demonstratives with these verbs, the verb faltar, ‘to be missing’ is involved in most cases though (22–25). M has 2 examples from 2;07 onwards and P9 from 2;09 onwards. The results show that emphasis is a very limited option with unaccusative verbs in Spanish, although there are slightly more examples than in Basque. (22) Uno falta ‘One is missing’
(P 2;09,01)
(23) Esto falta ‘This one is missing’
(P 3;04,26)
(24) Esto más- me falta ‘This one is missing (me)’
(M 2;07,28)
(25) Las zapatillas faltan ‘The slippers are missing’
(M 3;11,17)
The metagrammatic function of correction is a very seldom phenomenon in normal discourse despite (26), in which case the example could be analyzed as contrastive. (26) (s)teF queda ‘This one is left’ Interlocutor: no te queda ninguno ‘You haven’t got any left’
(M 3;04,26)
If we turn now to topics, with the only exception of one example at 2;00, subjects with topic reading appear at the age of 3;00 in Mikel. In Peru’s case they occur after the age of 3;00 (21–26), although P speaks very little from 2;04 to 3;00. Note that a few of the Basque examples in stage II are S PP V, reinforcing the idea that there is a position in front of the focus position, as it is the case in Basque. A handful of examples have been analyzed as having wide and narrow focus reading, which is unexpected. Those examples require a closer examination.
Subjects of unaccusative verbs
(27) a. b. c. d. e. f. g. h.
Y eso también se va6: la pia (playa) ‘And this one also goes’ Cuando se brompe (=rompe) un coche, se va aquí ‘When it breaks, a car goes here’ Ése ha caído ‘This one has fallen down’ Y éste vive aquí ‘And this one lives here’ Y éste va subir al aeropuerto ‘And this one goes to the airport’ Para echar gasolina este coche va ‘In order to fill up with petrol, this car goes’ El señor sube por ahí ‘The man goes up this way’ Éste sale por aquí ‘This one goes out this way’
(28) a. b. c. d.
Y tú te vas en avión ‘And you go by plane’ Jo, esto se ha caído ‘Oops, this one has fallen down’ Esto esto se va ‘This one goes’ Éste se ha rompido ‘This one has has broken down’ Este va par (=por) ahí igual ‘This one goes this way, maybe’
(P 3;02,14) (P 3;02,14) (P 3;03,07) (P;3,04,26) (P 3;09,25) (P 3;09,25) (P 3;09,25) (P 3;11,17) (M 2;06,11) (M 2;11,18) (M 3;02,07) (M 3;03,15) (M 3;09,08)
All these examples (27–28) have been uttered in out of the blue contexts. There are two common denominators in these examples; either they have a post-verbal PP or the subject is a demonstrative pronoun. The standard assumption in this chapter is that these kind of utterances have a wide focus reading. However, these examples show how difficult it is to judge the information status of any given utterance because contextual information might not be sufficient to decide what the speaker assumes that the hearer knows. Therefore, the examples with a realized postposed PP might well be intended as utterances with a narrow focus reading on the postposed prepositional phrase Pinto (1997). On the contrary, in 27b, 28 b–d a preposed demonstrative pronoun is involved with no post-verbal constituents. Although a demonstrative refers to an entity that is present in the discourse world, and is therefore discourse old, it might well represent new information because it is not ‘active’ at the moment of the utterance (Vallduví 1993). Echepare (1997:114) contends that a shared presupposition between speaker
6. : stands for final lengthening.
Pilar Larrañaga
and hearer is not necessary in the case of emphasis. Emphatic elements are topic-like, but occur under unexpected circumstances. The typical context is 29 (29) (you know what?) JOHN invited Bill
These few instances of pre-verbal subjects in stage I can easily be accommodated with Echepare’s approach.
6.2. Post-verbal subjects
tokens
Figure 9 shows the distribution of post-verbal subjects in Spanish. As expected, these have either wide or narrow focus reading. There are no subjects with a topic or emphatic reading.
40 35 30 25 20 15 10 5 0
emphasis wide narrow topic P
M
P
Stage I
M Stage II
tokens
Figure 9. Information status of the subjects in Vunacc S constructions in Spanish by P and M.
40 35 30 25 20 15 10 5 0
emphasis wide narrow topic P
M Stage I
P
M Stage II
Figure 10. Information status of the subjects in Vunacc S constructions in Basque by P and M.
Subjects of unaccusative verbs
VS is a very limited option in Basque (Figure 10). Firstly, M has only a few examples of VS in Basque in both stages. P uses a few VS in stage I and its number increases in stage II. Some of the subjects have a topic reading, which is target-like. Some others have a wide or narrow focus reading. These will be dealt with in more detail. P’s VS with narrow focus reading in Stage I do not involve story telling contexts (30 a–b). In contrast, those VS’s in stage II all involve a story telling context, which is target like (30 c–e). The same applies to M. He uses VS in Basque in story telling contexts. (30) a. b. c. d. e.
Joan da Peru Gone aux Peru ‘Peru has gone’ Puti (=busti) da brotxa become wet aux brush ‘The brush has become wet’ Eta erori(=etorri) zen katua eta txakurra And come aux cat and dog ‘And the cat and the dog came’ Etorriko da, bere amatxo Come-fut aux, his mother ‘His/her mother will come’ Eta etortzen da, Jiman And come aux, Himan ‘And Himan arrives’
(P 2;05,26)
(P 2;07,09)
(P 3;06,17)
(M 2;09,04)
(P 4;02,29)
7. Discussion This chapter had two main goals; firstly, to give an exhaustive description of word order structures with unaccusative verbs in the early language of Basque/Spanish bilinguals. Secondly, and since increasing evidence for cross-linguistic influence in the syntax-pragmatics interface is being found in the literature, the issue of cross-linguistic influence will be addressed. Unaccusative verbs have been defined as informationally rich, since they introduce new events in the world of discourse. Their subjects often introduce new referents in the world of discourse as well, especially if they occur in form of a lexical NP. These two facts might explain why there are virtually no contrastive focused subjects with unaccusative verbs in either language, an option that is possible but is not frequent in the current data. The overall results can be summarized as follows; VS is dominant in a first stage in Spanish, whereas its frequency decreases over time. In contrast, SV is dominant in Basque throughout the entire age span. Interestingly, Grinstead (2000) reports that both pre-verbal and post-verbal subjects occur at the same age in bilingual
Pilar Larrañaga
Catalan-Spanish children. This has also been reported for transitive verbs in Spanish by Larrañaga (2000) on bilingual Basque/Spanish children. However, the data on unaccusative verbs shows that pre-verbal subjects occur several months later than the post-verbal counterparts in Spanish in M whereas they occur at the same age in P, but become regular in the pre-verbal position after several months. Hence, there is a first long stage in Spanish where pre-verbal subjects are scarce. To sum up, all pre-verbal subjects in Basque and Spanish are IP internal at stage I. Topic and emphatic readings in pre-verbal position appear in stage II in both languages, suggesting that the necessary projections to check those features were not available in stage I. Another piece of evidence for IP peripherial projections comes from the various utterances with two pre-verbal constituents in stage II. Because of all the above, the data discussed in this chapter fully support Rizzi’s left periphery approach As to the issue of cross-linguistic influence, recall that two scenarios were possible:
if Basque influences Spanish, S in SV with a narrow or wide focus reading in Spanish should occur, and if Spanish influences Basque, S with a narrow or wide focus reading in VS in Basque as well as S with topic reading in SV should be attested..
Both children use a VS linear order in stage I, starting to use SV in stage II in Spanish. S with a narrow focus reading in Spanish is not found in either of the children at stage I. The few early instances of SV have been analyzed as having topic reading. Moreover, no target deviant instances of SV with a wide focus reading were found. This clearly shows that Basque does not influence Spanish. The Spanish word order pattern does not apply for preverbal subjects in Basque, where SV is dominant from the start and has a wide variety of pragmatic readings. Moreover, the information status analysis has shown that most subjects in stage I in Basque have a wide focus reading, suggesting that they are not IP peripheral. Subjects in SV constructions with target like and target deviant topic reading appear at stage II, suggesting that the projection TopP is missing in stage I. In contrast, both children use some VS in Basque, whereby P makes more use of them than M. However, a thorough analysis has shown that all the instances of VS in Basque were target like in stage II. The issue that arises now is whether those few examples in stage I with a topic reading are enough evidence to claim that Spanish influences Basque. I will claim that the scant data do not support Müller & Hulk’s hypothesis, although the language internal requirements are met. However, it is noteworthy that the few target deviant utterances occur in an early stage of development in both children, suggesting that it takes longer to master the pragmatics of available syntactic structures, a finding that is not new. There were no SV structures in Basque with a topic reading of S. In
Subjects of unaccusative verbs
addition to these structures, I analyzed some utterances containing either preposed or postposed [+ locativ] adverbials in Basque. It is in these cases that a trend towards a narrow focus interpretation of the postposed element or a wide focus interpretation of the whole sentence is found. It seems as if the presence of another constituent makes children rearrange the linear order of the constituents, in a way that very much resembles Spanish. Yet the data are scarce and do not provide enough evidence for cross-linguistic influence. This finding is surprising since word order in so called free word order languages seems to be the domain per se prone to cross-linguistic influence and Basque and Spanish have been in contact for centuries. The issue is why this very sensitive domain of grammar is so resistant to cross-linguistic influence. My claim is that children acquiring Basque and Spanish simultaneously do not have any difficulty in dealing with word order with the verbs studied in this chapter because the grammatical subject was involved. Research by Larrañaga (2000) has shown that, both children use the SVO in wide focus contexts in Basque and where the object is focused, where an S with narrow focus reading is expected. In sum, if the object is involved, cross-linguistic influence is more likely to arise than if the deep object (subject of unaccusative verb) is at stake. Moreover, some of the utterances that have been analyzed in this chapter contained other post-verbal constituents in Basque. Although there were not many of these, they tended to have a ‘Spanish-like’ reading.
8. Conclusions This chapter set out from Müller & Hulk’ (2001) proposal that, although bilingual children separate their two grammars, there could, under certain conditions, be cross-linguistic influence between the two emerging grammars. In order to test their hypothesis, SVunacc-VunaccS structures in early Basque and Spanish and their pragmatic entailments have been analyzed. The data clearly show that children can separate both languages in a grammatical domain that is highly susceptible to crosslinguistic influence. None of the cross-linguistic influence scenarios outlined in the predictions was borne out. The few examples that could count as evidence for Müller & Hulk’s proposal could be explained assuming that the children were using means that their languages make available, ie. emphasis and use of non-standard word orders in restricted pragmatic environments. However, the fact that the grammatical subject was the constituent in question seems to be the reason why no influence was found. Further research needs to be conducted in order to confirm or disconfirm the results in the present chapter, especially a standardized methodology has to be developed to produce results which can be compared.
Pilar Larrañaga
References Bel, A. 2003. The syntax of subjects in the acquisition of Spanish and Catalan. Probus 15: 1–26. Belletti, A. 1988. The case of unaccusatives. Linguistic Inquiry 19(1): 1–34. Belletti, A. 2001. Inversion as focalization. In Subject Inversion in Romance and the Theory of Universal Grammar, A. Hulk & J.-Y. Pollock (eds), 60–90. Oxford: Oxford University Press. Du Bois, J.W. 1987 The discourse basis of ergativity. Language 63(4): 805–854. Bosque, I. & Demonte,V. 1999 (eds.), Gramática descriptiva de la lengua española. Madrid: Espasa Calpe. Burzio, L. 1986. Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel. Casielles-Suárez, E. 1997. Topic, Focus and Bare Nominals in Spanish. PhD dissertation, University of Massachusetts, Amherst. Cinque, G. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24(2). Dixon, R.M.W. 1994. Ergativity. New York NY: Cambridge University Press. Echepare, R. 1997. Two types of focus in Basque. WCCFL 15: 113–128. Echepare, R. 1998. The Grammatical Representation of Speech Events. PhD dissertation, University of Maryland. Elordieta, A. 1999. Verb Movement and Constituent Permutation in Basque. PhD dissertation, University of Leiden. Euskaltzaindia. 1985. Euskal Gramatika: Lehen Urratsak I-II. Iruñea. Genesee F. & Paradis, J. 1996. Syntactic acquisition in bilingual children: Autonomous or interdependent. SSLA 18: 1–25. Grinstead, J. 2000. Case, inflection and subject licensing in child Catalan and Spanish. Journal of Child Language 27: 119–155. Hernández Pina, F. 1984. Teorías psicosociolingüísticas y su aplicación al español hablado. Madrid: Siglo XXI. Hulk, A. & Müller, N. 2000. Bilingual first language acquisition at the interface between syntax and pragmatics. Bilingualism: Language and Cognition 3(3): 227–244. Laka, I. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. PhD dissertation, MIT. Lambrecht, K. 1994. Information Structure and Sentence Form Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Larrañaga, M.P. 2000. Akkusative Sprachen, ergative Sprachen: der Erwerb des Kasus bei bilingualen Kindern. Frankfurt:Vervuert. Levin, B. & Rappaport Hovav, M. 1995. Unaccusativity. At the Syntax-Lexical Semantics Interface. Cambridge MA: The MIT Press. Mahlau, A. 1994. El proyecto BUSDE: Corpus y metodología. In La adquisición del vasco y del castellano en niños bilingües, J.M. Meisel (ed.), 21–34. Frankfurt:Vervuert. Meisel, J.M. 1994b (ed.), La adquisición del vasco y del castellano en niños bilingües. Frankfurt: Vervuert. Mendikoetxea, A. 1999. Construcciones inacusativas y pasivas. In Gramática descriptiva de la lengua española, Vol. 2, I. Bosque & Demonte (eds.), 1575–1629. Madrid: Espasa Calpe. Müller, N. & Hulk, A. 2001. Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition 4: 1–21. Ordóñez, F. & Treviño, E. 1999. Left dislocated subjects and the pro-drop parameter. A case Study of Spanish. Lingua 107: 39–68.
Subjects of unaccusative verbs Ortiz de Urbina, J. 1989. Some Parameters in the Grammar of Basque. Dordrecht: Foris. Ortiz de Urbina, J. 1999a. Focus in Basque. In The Grammar of Focus, G. Rebuschi & L. Tuller (eds.) 311–334. Amsterdam: John Benjamins. Ortiz de Urbina, J. 1999b. Force phrases, focus phrases and left heads in Basque. In Grammatical Analyses in Basque and Romance Linguistics J. Franco, A. Landa & J. Martín (eds.) 179–194. Amsterdam: John Benjamins. Pinto, M. 1997. Licensing and Interpretation of Inverted Subjects in Italian. PhD dissertation. University of Utrecht. Platzack, C. 2001. The vulnerable C-domain. Brain and Language 77: 364–377. Pollock, J.-Y. 1989. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20(3): 365–424. Rebuschi, G. & Tuller, L. (eds.). 1999. The Grammar of Focus. Amsterdam: John Benjamins. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer. Serratrice, L., Sorace, A. & Paoli, S. 2004. Crosslinguistic influence at the syntax-pragmatics interface: Subjects and objects in English-Italian bilingual and monolingual acquisition. Bilingualism: Language and Cognition 7: 183–205. Thornton, R. 2002. Lets Change the subject: Focus movement in early grammar. Language Acquisition 10(3): 229–271. Vallduví, E. 1993. The Informational Component. Philadelphia PA: University of Pennsylvania. Volterra, V. & Taeschner, T. 1978. The acquisition and development of language by bilingual children. Journal of Child Language 5: 311–326. Zubizarreta, M.L. 1998. Prosody, Focus and Word Order. Cambridge MA: The MIT Press. Zubizarreta, M.L. 1999. Las funciones informativas: Tema y foco. In Gramática descriptiva de la lengua española, I. Bosque & V. Demonte (eds), 4215–4244. Madrid: Espasa Calpe.
Appendix A Table 1a. Unaccusative verbs in Spanish used by M and P in stage I and stage II Stage I SV
VS
M
Faltar, ir, caerse, Caer, caerse, soltarse, parar pasar, venire, quitarse, irse, acabarse, pincharse, romperse
P
Bajar, salir, caerse, irse, ir, entrar, romperse, faltar
Salir, ir, caerse, entrar, romperse, moverse, caber, pasar
Stage II SV
VS
Caer, subir, irse, pasar, faltar, estropearse, caerse, salir, romperse, enganchar, conducir, chocar, caber Ir, andar, venir, subir, pasar, caber, irse, venir, esconderse, entrar, salir, venir, subir, mover, esconder, quedar, romperse, faltar, aparcar, chocarse, perderse
Escaparse, venire, caerse, pasar, faltar
Salir, venir, moverse, irse, caerse, abrirse, escaparse, guardarse, faltar, aparcar, caerse, gastarse, caber
Pilar Larrañaga Table 1b. Unaccusative verbs in Spanish used by M and P (translation) Stage I
Stage II
SV
VS
SV
VS
M
be missing, go, fall, stop
fall, go up, leave, occur, be missing, break down, go out, break, hook up, drive, bump into, fit in
escape, come, fall, occur, be missing
P
go down, go out, fall, leave, go, come in, break down, be missing
fall, fall, open, occur, come, move, go, be over, have a flat tyre, break go out, go, fall, get in, break down, move, fit in, occurr
go, work (function), come, go up, occur, fit in, leave, come, hide, come in, go out, come, go up, move, be leftover, break down, be missing, park, collide, get lost
go out, come, move, leave, fall, open, escape, hide, to be missing, park, fall, to be over, fit in
Table 2a. Unaccusative verbs in Basque used by M and P in stage I and II Stage I M
P
Stage II
SV
VS
SV
VS
Zikindu, apurtu, etorri, ibili, sustatu, pasatu, joan, kabitu, haserratu, falta, Sartu, mugitu, montatu
Sartu, pasa, joan?, apurtu,
etorri, haundia egin, jausi
Etorri, gelditu, pasatu, falta, jausi, igo
Apurtu, zikindu, joan, busti, bizi, jaiki, atera
Apurtu, etorri, mugitu, etorri, hil, bete, galdu, atera, erori, montatu, ailatu, kabitu, joan, egon, igo, , gertatu, jausi, montatu, falta, sartu, hasi, egon agertu, izkutatu, ibili, soltatu, jausi, ito, gelditu, askatu, galdu,
Subjects of unaccusative verbs Table 2b. Unaccusative verbs in Basque used by M and P in stage 1 and stage 11 Stage I M
P
Stage II
SV
VS
SV
VS
Become dirty, break, come, work, frighten, occur, go, fit in, become angry, be missing Come in, move, get into
Come in, occur, go, break
Come, grow, fall
Come, stop, occur, be missing, fall, go up
Break, become dirty, go, become wet, live, get up, get out
Break, come, move, fill up,get lost, go out, fall, fit in, be, go up, get in, be missing, come in, appear, hide, go, get loose, fall, drawn, stop, get free, get lost
Come, die, go up, arrive, occur, fall, start, be
Dominance, mixing and cross-linguistic influence On their relation in bilingual development Tanja Kupisch In this chapter I explore the relation between language dominance, mixed language utterances and cross-linguistic influence based on data from two German/French bilingual children. The children differ in terms of the extent to which one language can be considered dominant, but both produce fewer mixed utterances in their stronger language. Cross-linguistic influence is examined with respect to determiner acquisition. Monolingual German and French children acquire determiners at different rates, the process being faster in French. The bilinguals use more determiners in German than monolinguals of comparable ages, which is interpreted in favour of positive influence from French. I suggest that these results are inconsistent with mixing patterns, arguing that influence and mixing represent different types of contact phenomena.
1. Introduction During the 1970s and 1980s many researchers have reported initial mixed languages states (e.g., Leopold 1970, Volterra & Taeschner 1978). This has given rise to the Unitary Language Systems Hypothesis (ULSH) (or: Single System Hypothesis), i.e., the assumption that children start out with one unitary language system for both languages. Two different types of contact phenomena typically served as evidence in favour of the ULSH, while their absence was taken to imply that children are able to keep the two languages separate. One type is code-mixing (or simply: mixing), i.e., the use of utterances containing elements from more than one language, e.g., das voiture rouge (the car red), a DP containing a German determiner and a French adjective and noun. Another type of contact phenomenon is the occurrence of cross-linguistic influence in monolingual utterances, e.g., a French utterance such as la rouge voiture (the red car), produced by a German/French child in analogy to German das rote Auto (the red car) instead of la voiture rouge (the car red) with a postnominal adjective.1 Both contact 1. It seems uncontroversial that such an utterance could be attributed to influence from German. One has to acknowledge, however, that a few French adjectives may occur prenominally
Tanja Kupisch
phenomena may be defined with respect to various linguistic levels, i.e., phonetic, morphological, syntactic or semantic aspects of language use. Case studies on many different language combinations and a variety of different grammatical domains have given rise to the (currently dominant) view that the ULSH is no longer tenable (Meisel 1986, 1989, 1990, 1994; Meisel & Müller 1992; Müller 1993; Genesee 1989; Genesee, Nicoladis & Paradis 1995; De Houwer 1990; Quay 1993; Pearsson, Fernandez & Oller 1993; Larrañaga 2000, inter alia). Proponents of this view do not deny the existence of mixed language utterances, but they attribute them to reasons other than the lack of language separation, such as lexical borrowing (e.g., Genesee 1989; Meisel 1989), rule-governed code-switching (e.g., Meisel 1989), the presence of other bilingual speakers in the communicative setting (e.g., Grosjean 2001), mixing in the (parental) input (e.g., Bergman 1976), or a (temporary) dominance in the one of the two languages (e.g., Grosjean 1982; Genesee, Nicoladis & Paradis 1995; Gawlitzek-Maiwaldt & Tracy 1996; Hulk 1997; Schlyter 1993, 1994; Bernardini & Schlyter 2004). Hence, unitary systems are no longer considered to be explanatory of young bilingual children’s code-mixings. A similar view has emerged with respect to cross-linguistic influence in monolingual utterances. Its occurrence is no more taken to imply that bilingual children have a fused language system (Hulk & Van der Linden 1996; Hulk 1997; Hulk & Müller 2000; Müller & Hulk 2001; Müller, Cantone, Kupisch & Schmitz 2002, inter alia). On the contrary, the debate about cross-linguistic influence in bilingual first language acquisition implies the assumption of a dual system, since transfer of structure could not occur if there was no host or recipient system (Paradis & Genesee 1996). If unitary systems are not explanatory of contact phenomena, other explanations have to be found. According to Gawlitzek-Maiwald & Tracy (1996), influence is likely to occur if a grammatical phenomenon is harder to acquire in one language than in the other, i.e., if there is a developmental asynchrony between the two languages. They quote mixed language utterances such as Kannst du move a bit? (with a German IP and an English VP) from a bilingual German-English child, arguing that the girl borrows the German IP because it is easier to acquire than the English IP. Many other scholars have related contact phenomena with language dominance. Schlyter & Bernardini (2004) assume that children borrow functional structure from their stronger language
as well, e.g., la grand chanteuse ‘the famous singer’. Therefore, such constructions may also result from the overgeneralization of a marginally possible option in French, which is strengthened by the German input. In that case, it appears more appropriate to speak of an indirect, quantitative influence (Hulk & Van der Linden 1996), while cases in which the transferred structure is inexistent in the recipient language may be referred to as qualitative influence. In this article, the term cross-linguistic influence (Müller & Hulk 2001) shall be used to cover both quantitative and qualitative influence.
Dominance, mixing and cross-linguistic influence
when speaking in their weaker language. The view that the stronger language influences the weaker language is also emphasized by Grosjean (1982: 188ff.). According to him, “[t]he main effect of dominance is not only that the stronger language is more developed than the weaker one (more sounds are isolated, more words are learned, more grammatical rules are inferred), but also that the stronger language interferes with or influences the weaker one. In this sense, dominance retards differentiation by imposing aspects of the dominant language on the weaker one.” (Grosjean 1982:190, my emphasis). In this study, I compare two German/French bilingual children who have been studied longitudinally in terms of language dominance, code-mixing, and crosslinguistic influence. The study shows, first, that dominance does not retard differentiation. On the contrary, the fact that mixing occurs to different extents in the two language contexts indicates that the children can well distinguish between the two languages. Second, although the amount of mixing appears to be related to language dominance, the data indicate that mixing is not unidirectional and does not exclusively affect the language developing more slowly. Third, I show that cross-linguistic influence does not go hand in hand with the mixing rate and with dominance, arguing that cross-linguistic influence and mixing constitute two different types of contact phenomena. Cross-linguistic influence will be examined with respect to determiner acquisition. Determiners are a good testing ground for the occurrence of cross-linguistic influence because in monolingual studies productive use of determiners has been observed from a very early age in children acquiring French (productive use before 2 years), but comparatively late by children acquiring German (productive use after the age of 2). I shall examine monolingual data to further substantiate this observation. As for the bilinguals, if they behave like two monolingual children in each of their languages, the same contrast in the rate of determiner acquisition should emerge. If the latter is not borne out, cross-linguistic (quantitative) influence may be stipulated. The chapter starts with a comparison of determiner use in German and French with a focus on prenominal articles in Section 2. Section 3 introduces the database. Section 4 examines the bilingual data with respect to language dominance and language mixing, while Section 5 is concerned with cross-linguistic influence. The chapter concludes with a discussion and conclusion in Sections 6 and 7.
2. Determiner use in French and German Both German and French have pronominal definite and indefinite determiners but there are inter-language differences concerning frequency of occurrence, morphological load, and the prosodic structure of determiners. It is quite plausible to assume that these factors have an effect on the rate of determiner acquisition.
Tanja Kupisch
Let’s first look at the syntactic distribution of determiners. French and German both have definite and indefinite articles, which are obligatory with singular count nouns. (1) a. Fr. Je lis un / le livre. vs. *Je lis livre. b. Ge. Ich lese ein / das Buch. vs. *Ich lese Buch. I read a / the book I read book
It should be added that both languages mark certain semantic and pragmatic distinctions by means of appropriate article choice. More particularly, both the distinction between specific vs. non-specific2 referents and the distinction between referents that are commonly known to speaker and hearer vs. referents that are only known to the speaker is marked by the use of definite DPs as opposed to indefinite or zero marked DPs. Articles are also used with plural and mass nouns when they have specific reference. (2) a. Fr. J’ achète les tomates et le vin. b. Ge. Ich kaufe die Tomaten und den Wein. I buy the tomatoes and the wine
With reference to non-specific entities, the languages show considerable differences. Here, determiners are commonly used in French, while nouns remain largely undetermined in German.3 Examples are provided in (3) through (6). (3) a. Fr. J’aime le vin. (generic object) b. Ge. Ich mag Wein. ‘I like wine.’ (4) a. Fr. À la maison nous avons l’ADSL et le cable. (non-specific object) b. Ge. Zu Hause haben wir DSL und Kabelfernsehen. ‘At home we have DSL and cable TV.’ (5) a. Fr. Je vais acheter du pain et du lait. (non-specific object) b. Ge. Ich gehe Brot und Milch kaufen. ‘I’m going to buy (some) bread.’
2. I consider a noun phrase to be specific if it refers to a particular entity, and non-specific if it does not refer to a specific entity. 3. The French article system may be considered further grammaticalized than the German article system. The diachronic process of article grammaticalization is commonly viewed in terms of a gradual spread of articles towards an increasingly higher number of contexts. The dimension of specificity was shown to be an important parameter in this process. Articles first emerge in contexts of specific reference, and come to be used in contexts of non-specific reference only in later stages of grammaticalization.
Dominance, mixing and cross-linguistic influence
(6) a. Fr. Il a les yeux bruns. (inalienably possessed object) b. Ge. Er hat braue Augen. ‘He has brown eyes.’
These differences have a direct bearing on the token frequency of nouns with articles (and other determiners) as compared to bare nominals. According to Kupisch (2004), it amounts to 94% in French as compared to 82% in German. It is plausible to assume that the higher token frequency of articles in French increases French children’s awareness of these elements at an earlier age than in German. The German determiner system is also more complex in terms of morphological load, because it encodes a larger number of semantic and syntactic distinctions. While French determiners are marked for Number (singular and plural) and for two different genders in the singular (masculine and feminine), German encodes three genders (neuter in addition to masculine and feminine), Number, and four different Cases. For example, la taupe in the examples in (7) translates into four different articles in German, depending on the syntactic function of the DP. (7) a. Fr. La taupe est dans le jardin. b. Ge. Der Maulwurf ist im Garten. (nominative) ‘The mole is in the garden.’ c. Fr. J’ai vu la taupe dans le jardin. d. Ge. Ich habe den Maulwurf im Garten gesehen. (accusative) ‘I saw the mole in the garden.’ e Fr. J’ai donné à manger à la taupe. f. Ge. Ich habe dem Maulwurf zu essen gegeben. (dative) ‘I gave the mole to eat.’ g. Fr. On voyait seulement la tête de la taupe. h. Ge. Man sah nur den Kopf des Maulwurfes. (genitive) ‘One could only see the head of the mole.’
In addition, German determiners have reduced variants which are commonly used in spoken speech. Hence, there is a high number of formally different articles (and other prenominal determiners) in German. I will not be concerned with morphologically correct production of determiners in in this study, but it is conceivable that morphological complexity has an indirect impact on the rate of omission, making it more difficult for the child to establish the correct linguistic representation. This could lead children to the strategy of omitting determiners rather than using incorrect forms. The third difference concerns the prosodification of determiners. French determiners tend to be prosodified with the following noun, and intervening adjectives are rare, as most adjectives occur postnominally in French (exceptions are high frequency adjectives like grand ‘big’, petit ‘small’ and bon ‘good’). In German, determiners do not always form a prosodic unit with the following noun, as shown in (8b) (see Lléo 2001).
Tanja Kupisch
(8) a. Fr. J’ai vue [PW une coccinelle ]. b. Ge. Ich [PW habn ] [PW Käfer ] gesehen. ‘I saw a beetle.’4
These three differences could be the source of a developmental asynchrony in the rate of determiner acquisition. More specifically, it appears plausible that the high token frequency of determiners and the isomorphism between syntax and prosody draws the French-learning child’s attention to the presence of prenominal determiners at an earlier age than the German-learning child. In addition, German-learning children may need more time to establish the morphosyntactically correct representation of determiners.
3. Data The bilingual data presented in this article were collected as part of the project Bilingualism in early childhood: Comparing Italian/German and French/German (see also Müller & Pillunat, this volume).5 The children Alexander and Céline grew up in bi-national families. Alexander’s mother is German, his father is French, while Céline has a German mother and a French father. Both children have elder brothers. The children’s language development was followed regularly from the age of 2;0 (Céline) and 2;2 (Alexander) until the age of 5;0. The period analyzed for determiner omission covers the age before 3;0. In Céline’s French it was extended to 3;6 because her French developed more slowly. The children were video-recorded twice monthly for roughly half an hour in each language. The recordings took place at the children’s homes, while the interlocutors played with the children. The play context created a monolingual situation: While the German-speaking (monolingual) person was interacting with the child, the French-speaking (monolingual) person was recording in the background. The investigators took turns after half an hour. The children rarely addressed the person behind the camera. For comparative purposes, additional monolingual data will be analyzed with respect to determiner acquisition. For French, I used the GRÉGOIRE-corpus, collected by
4. The prosodic word (PW) is one element in a series of hierarchically ordered phonological constituents known as the prosodic hierarchy. PWs behave as units of pronunciation according to rules which may differ from one language to another. They have one primary stress. 5. The research project has been funded by the German Science Foundation (DFG) as part of the Research Centre on Multilingualism in Hamburg. Data collection and transcription started before funding was granted. I would like to express our gratitude to the researchers involved in data collection and to the University of Hamburg for a research grant during the pilot phase of the project. Members of the research team were: Katja Cantone ([email protected]), Tanja Kupisch ([email protected]), Katrin Schmitz ([email protected]); cf. Müller et al. (2002, 2006a). Address for correspondence: Prof. Dr. Natascha Müller, Bergische Universität Wuppertal, Fachbereich A Romanistik, Gaußstraße 20, D-42119 Wuppertal, [email protected].
Dominance, mixing and cross-linguistic influence
Christian Champaud, and parts of the corpora MAX and ANNE, collected under the direction of Bernadette Plunkett as part of the York-corpus. (The data of Max is representative of Canadian French, but there are no major differences in article use between these two varieties of French). For German, I analyzed parts of the corpora EMELY and FALKO (Szagun 2001). All five corpora are available at Childes. The Chantal-corpus was collected and transcribed by the author together with N. Müller. In addition, two crosssectional speech corpora were analyzed. Both were collected under the direction of N. Müller (Müller, Hulk and Jakubowicz 1999, Müller et al. 2006b). These two corpora will not be used in MLU-based comparisons, because some of the recordings were comparatively short and no MLU has been established. An overview of all corpora is provided in Table 1. Table 1. Overview of the data French corpora Age-span
No. of recordings
Alexander 2;2,6–2; 11,20 18 Céline 2;0,9–3; 6,12 21 Anne 1;10,12 2; 5,18 15 Max 1;9,19–2; 4,18 15 Grégoire 1;9,18–2; 5,27 10 cross- 2;2–2;9 13 sectional
German corpora Age-span Alexander 2;2,6–2; 11,20 Céline 2;0,9–2; 11,29 Chantal Emely 1;6,29–3;0,7 Falko 1;8–2;5,16 cross- 2;4–2;9 sectional
No. of recordings 18 21 29 14 8 8
4. Language dominance and mixed language utterances 4.1 Language dominance The children’s development in the two languages has been measured in terms of mean length of utterances in words (MLU), upper bound (UB), number of utterances (UTT), and increase of noun (NOU) and verb types (VER) (Kupisch 2006; Müller et al. 2006a). All measures have been established for each recording session in each language. For ease of illustration, mean values for a three-month-period will be represented here (Figure 1). Through this method, the inter-language contrasts are clearly visible and the developmental perspective is not completely suppressed despite taking means. For verb and noun types, which were analyzed incrementally, the value of the last recording session comprised in the respective phase will be indicated (Figure 2). The figures indicate higher values in Céline’s German throughout the observation period, although the contrast between the languages diminishes with increasing age. The latter is especially clear with respect to the amount of utterances produced in each language. On the whole, Céline may be considered dominant in German. Unlike Céline, Alexander shows higher values in French, but the contrast between the languages
Tanja Kupisch 5
5 Alex
MLU
4 3
3
2
2
1
1
0
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11
UB
Alex
12
8
8
4
4 2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11
Céline
0
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11 3;0-3;2 3;3-3;5
400
400 300
Alex
Céline
300
200
200
100
100
0
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;113;0-3;2 3;3-3;5
16
12
0
UTT
0 20
20 16
Céline
4
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11 French
0
2;0-2;2 2;3-2;5 2;6-2;82;9-2;113;0-3;2 3;3-3;5 German
Figure 1. MLU, UB and UTT in Alexander and Céline.
is less pronounced as in the case of Céline. His French may be considered slightly stronger, but dominance seems to too strong a term to use here.
4.2 Mixed language utterances This section looks at the amount of mixed language utterances in the two children. In the analysis, three categories of utterances were distinguished. Following Genesee (1989),
Dominance, mixing and cross-linguistic influence 600
NOU
500
600 Alex
400
400
300
300
200
200
100
100
0
VER
0
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11 Alex
Céline
400
300
300
200
200
100
100
0
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11 3;0-3;2 3;3-3;5
500
500 400
Céline
500
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11
0
French
2;0-2;2 2;3-2;5 2;6-2;8 2;9-2;11 3;0-3;2 3;3-3;5 German
Figure 2. Increase of noun types and verbs types in Alexander and Céline.
intra-utterance mixings (utterances containing words from both languages) were distinguished from inter-utterance mixings (monolingual utterances in the language not spoken by the interlocutor). The third category constitutes monolingual utterances matching the target language (i.e., the language of the interlocutor). Utterances addressed to the person behind the camera were not included. The results are illustrated in Figures 3 through 6. The values for Céline and Alexander are from Cordes (2002) and Müller & Kupisch (2003). Both children produce monolingual utterances in the target language, i.e., the language of the interlocutor, in more than 95% of all cases in one of their languages (Figures 4 and 5), whereas in the other language they produce more intra- and inter-utterance mixings. Both types of mixings together constitute 14% in Alexander’s German (Figure 3) and 48% in Céline French (Figure 6). Note also that there is a clear decrease of mixing as Céline grows older. In fact, the amount of mixing amounts to 73% during the period before age 3. In summary, both Alexander and Céline have a language in which their language choice is more consistent with the target language: French in the case of Alexander and
Tanja Kupisch
100% 80% French utterances
60%
Mixed utterances German utterances
40% 20%
2;11,6
2;10,2
2;8,28
2;7,27
2;6,25
2;5,25
2;4,6
2;2,20
0%
Figure 3. Alexander, language choice in German (Müller & Kupisch 2003). 100% 80% French utterances mixed utterances German utterances
60% 40% 20%
3;5,29
3;4,23
3;3,26
3;1,10
3;0,13
2;11,15
2;10,18
2;9,20
2;8,16
2;7,19
2;6,7
2;5,8
2;4,5
2;1,14
2;0,24
0%
Figure 4. Céline, language choice in German (Cordes 2002).
German in the case of Céline. These mixing patterns go hand in hand with distance between children’s two languages in terms of proficiency, Alexander being slightly more advanced in French than in German, Céline being dominant in German. The results suggest that dominance is an explanatory factor in determining language choice, as found by Genesee et al. (1995), Deuchar & Quay (2000), Nicoladis (1998), Nicoladis & Secco (2000). This does not imply, however, that intra-language mixing cannot occur in the dominant language, as indicated by the data of Alexander. Illustrative examples are provided in (9) with German words in italics. Instances mostly consisted in the insertion of single German words into otherwise French utterances, i.e., cases of lexical borrowing.
Dominance, mixing and cross-linguistic influence 100% 80% German utterances mixed utterances French utterances
60% 40% 20%
2;11,6
2;10,2
2;8,28
2;7,27
2;6,25
2;5,25
2;4,6
2;2,27
2;2,6
0%
Figure 5. Alexander, language choice in French (Müller & Kupisch 2003).
100% 80% German utterances mixed utterances French utterances
60% 40% 20%
3;6,12
3;5,15
3;4,9
3;3,12
3;0,27
2;11,29
2;11,3
2;10,5
2;8,29
2;8,2
2;6,21
2;5,24
2;4,19
2;3,15
2;1,6
2;0,9
0%
Figure 6. Céline, language choice in French (Cordes 2002).
(9) a. b. c.
moi me c’est it’s tu you
je mange tout les Salzstangen (Alexander 2,7,6) I eat all the pretzel sticks kaputt ma maison (Alexander 2;8,12) broken my house prends mon Schlauch (Alexander 2;8,28) take my hose
Tanja Kupisch
The fact that the children choose the language of their interlocutor in the majority of cases indicates their speaker-sensitivity. The fact that they do so less when their interlocutor speaks their weaker language points to a relation between dominance and mixing. Hence, dominance may affect language mixing, although it is not the only source for it. The fact that the children show different mixing patterns in their two languages, suggests that they separate their languages.
5. Cross-linguistic influence in determiner acquisition This section examines the occurrence of cross-linguistic influence with respect to the use of prenominal determiners. For two reasons, determiner acquisition is particularly suited to investigate cross-linguistic influence. First, there is a partial overlap of the target-systems, which has been considered a precondition for cross-linguistic influence (Hulk & Müller 2000). This overlap may be defined with respect to syntax, such that both languages have prenominal definite and indefinite articles (and other determiners), which are obligatorily used in certain contexts, and whose use is determined by similar pragmatic and semantic rules. On the other hand, the overlap must be considered partial, because French is more restrictive with respect to the occurrence of bare nouns, i.e., nouns without determiners (see above). Furthermore, the results of previous studies indicate a developmental asynchrony, i.e., monolingual French children acquire determiners earlier and omit them less extensively than monolingual German children. While the former produce their first article-like elements between the ages of 1;6 and 1;10 (although these articles may appear in a phonologically reduced form, such as [e] or [a]), the latter tend to produce their first articles around the age of 2 or later (see Bassano & Eme (2001); Van der Velde, Jakubowicz & Rigaut (2002) for French; Lleó (2001); Eisenbeiss (2002); Kupisch (2006) for German).6 The lack of such a difference in the bilingual data can be taken as indicative of cross-linguistic influence, because if there were no influence, the bilingual German/French child should mirror the developmental patterns of a monolingual French child in one language, and of a monolingual German child in the other language.
6. The observation carries over to other Romance and Germanic languages. Cross-linguistic studies comparing determiner omission in a Germanic and a Romance language have consistently shown that children acquiring a Romance language start to use determiners at earlier ages than children acquiring a Germanic language (see Chierchia et al. (1999) for English, French, Italian; Lleó & Demuth (1999) for German and Spanish; Guasti & Gavarró (2003) for Italian and Catalan; Van der Velde (2003) for Dutch and French; Kupisch (2006) for German, French and Italian).
Dominance, mixing and cross-linguistic influence
5.1 Determiners in the monolingual acquisition of French and German Before analyzing the bilingual data, I shall compare three longitudinal corpora and one cross-sectional corpus in each language with respect to determiner use in obligatory contexts. The percentage of determiner omissions corresponds to the relative number of contexts in which a determiner was absent although it should have been used from the total of contexts in which a determiner should have been used. The analysis was restricted to illicit determiner omissions, since German has more contexts in which determiners may be omitted. In other words, the inclusion of licit omissions in the analysis would have been disadvantageous for the German-learning children. Prenominal determiners other than articles (demonstratives, possessive pronouns, etc.) were included in the counts. The results are illustrated in Figure 7 (the cross-sectional data is represented as FRE (French) and GER (German)). The results mirror those of previous studies on determiner acquisition, showing that monolingual German children start to use determiners later than monolingual French children and omit them to greater extents.
rate of D-omission (%)
100 MAX
80
ANN GRÉ
60
FRE FAL
40
EME CHA
20 0
GER 1;7
1;10
2;1
2;4
2;7
2;10
Figure 7. Determiner omission in monolingual German children (FAL, EME, CHA, GER) and monolingual French (MAX, ANN, GRÉ, FRE) children.
5.2 Bilingual data Let’s now take a look at the bilingual data, previously analyzed in Kupisch (2006). There are two possible approaches to determine whether the languages of a bilingual child are subject to cross-linguistic influence with respect to a particular grammatical phenomenon. The first constitutes an intra-individual comparison of the development of the grammatical domain. Such a comparison may indicate whether the rate of development is similar or different in the two languages. As mentioned above, there is a developmental asynchrony in determiner acquisition
Tanja Kupisch
between monolingual French and German children, and bilinguals should mirror this asynchrony, if the languages do not influence each other. If no such asynchrony occurs, this may be suggestive of cross-linguistic influence. In the latter case, the question arises whether the observed development resembles the French pattern more closely, or the German pattern. It may also be different from either. To find this out, an additional comparison with monolingual data will be made. Furthermore, with children who develop noticeably slower in one of the languages, an age-based comparison seems insufficient, because the appearance of a particular grammatical element may be retarded with respect to age, but normal or early with respect to the general linguistic development, measured in terms of MLU. Therefore, Alexander’s and Celine’s development of determiners will be compared in three steps: (i) intra-individually in terms of age; (ii) with respect to monolingual children; (iii) in terms of MLU.
5.2.1 Comparing French and German Under the assumption of a developmental asynchrony in determiner acquisition between monolingual French and German, one would expect omission patterns for the omission rates in the bilinguals’ two languages to be distinct and lower in French, provided that they behave like monolingual children in each of their two languages. This prediction is not clearly borne out in the case of Alexander, as shown in Figure 8. He practically ceases to omit determiners in obligatory contexts by 2;4 in French and only slightly later in German. Except in the first two recordings, the rate of determiner omission is very similar in both languages. Thus, the monolingual situ-
rate of D-omission (%)
100
German French
80 60 40 20 0
2;1
2;2
2;3
2;4
2;5
2;6
Figure 8. Rate of determiner omission, Alexander.
2;7
2;8
2;9
2;10 2;11
Dominance, mixing and cross-linguistic influence
ation is not mirrored.7 In the case of Céline, the omission rates are initially visibly lower in German than in French, which is exactly the opposite of what is found in the monolingual situation. Given that Céline’s French develops more slowly overall, this does not come as a surprise, though. (Note that due to the low overall number of nouns in the early French recordings data of two sessions have been collapsed to attain more representative percentages (e.g., 2;1,6 and 2;1,14 have been unified to 2;1)). Further below, Céline’s development of French determiners will be discussed in terms of MLU rather than age. How can we determine now whether there is cross-linguistic influence? One way is to compare the rate of omission in French and German in each child and see whether there is a statistically significant contrast. This was done for individual periods based on the absolute number of determiner use and omission. As for Alexander, these values in German and French do not differ significantly between 2;2 and 2;3 (χ2 = 3, df = 1, p = > 0.05), between 2;4 and 2;6 (χ2 = 3.05, df = 1, p = > 0.05) and between 2;6 and 2;11 (χ2 = 0.18, df = 1, p = > 0.05). For Céline, the contrast is significant between 2;0 and 2;3 (χ2 = 5.34, df = 1, p < 0.05), highly significant between 2;4 and 2;6 (χ2 = 18.87, df = 1, p < 0.001) but not significant between 2;7 and 3;0 (χ2 = 1.16, df = 1, p >0.05).
rate of D-omission (%)
100
French German
80 60 40 20 0
2;0
2;2
2;4
2;6
2;8
2;10
3;0
3;3
3;5
Figure 9. Rate of determiner omission, Céline.
7. Unfortunately, in the case of Alexander the onset of determiner use has been missed, and we cannot exclude the possibility of a language contrast in determiner use during earlier stages of development. However, even if there were such a contrast, it would not weaken the assumption of influence at a later stage.
Tanja Kupisch
The absence of significant differences in Alexander may be taken to suggest that the two languages influence each other. Significant contrasts in Céline, on the other hand, do not automatically imply the reverse, i.e., the absence of cross-linguistic influence. If Céline behaved like a monolingual child in each language, we would expect the omission rates in French to be lower than in German. This is not borne out, possibly because of the relatively slow development of her French. Here, only a comparison with monolingual children matched in terms of age and MLU can provide further clues. A comparison with monolinguals is also desirable for Alexander, as, so far, we have only seen that the languages seem to influence each other, while the direction of influence is still unclear.
5.2.2 Comparing bilingual and monolingual data In what follows, the omission rates of Alexander and Céline will be compared to those of the monolingual children investigated in 5.1. Figure 10 shows that monolingual French children tend to reach omission rates below 50% between 1;9 and 2;0 and that determiner omission decreases below 10% between 2;2 and 2;7 (with a few exceptions in the cross-sectional corpus). Of course, there is individual variation, e.g., Gregoire appears to be a comparatively fast learner. During these periods, Céline has noticeably higher rates of omission, while Alexander largely patterns with the monolingual children. Figure 11 illustrates the results for German. Chantal and Emely reach 50% of omission only at age 2;6 and cease to omit article considerably later than monolingual French children. Falko appears to be a fast learner, who reaches 50% by age 2;0. If we take Chantal and Emely to be more representative of monolingual German learners than Falko, which would be consistent with previous observations in the literature, we may conclude that both Céline and Alexander have lower rates of omission than the monolingual children at most times of their development.
rate of D-omission (%)
100 80
CÉL ALE
60
MAX ANN
40
GRÉ FRE
20 0
1;6
1;10
2;1
2;3
2;6
2;8
2;11
3;1
3;5
Figure 10. Rate of determiner omission in French (bilingual and monolingual).
Dominance, mixing and cross-linguistic influence
rate of D-omission (%)
100 80
CÉL ALE
60
EME FAL
40
CHA GER
20 0
1;6
1;9 1;11 2;0
2;2
2;3
2;5
2;6 2;8
2;9 2;11
Figure 11. Rate of determiner omission in German (bilingual and monolingual).
Summarizing the results for French, we may observe that Céline omits more determiners than monolingual children of comparable age, while Alexander’s development is consistent with that of monolingual children. In German, both bilingual children omit determiners noticeably less than monolingual children of comparable age. Can this be taken to mean that French positively influences the acquisition process in German? An answer along these lines is immediately plausible for Alexander. But how about Céline: Can her French, which develops so slowly, have a positive influence on her German? Let’s take a closer look at Céline. In fact, if we consider the production of prenominal French determiners from the perspective of MLU, the picture looks different than in the age-based comparison.
rate of D-omission (%)
100 80 ALE 60
CÉL
40
ANN
MAX GRÉ 20 0
1;0
2;0
3;0
4;0
5;0
6;0
Figure 12. Rate of determiner omission in French (bilingual and monolingual), MLU-based.
Tanja Kupisch
Alexander reaches an MLU beyond 2 in the first recording at 2;2. With an MLU between 2 and 2.5 his mean rate of omission is 25%, mostly similar to monolingual children (at times even lower). With a comparable MLU (between 2 and 2.5), Céline’s rate of omission is already below the 10% level, which suggests that many of her twoword utterances are nouns with determiners. She shows noticeably lower omission rates than the monolinguals. Thus, Céline’s development of French determiners is by no means retarded in terms of MLU. Rather, she acquires determiners relatively early with respect to her general linguistic development in French (if MLU is taken to be representative of general linguistic development). It must be noted that an MLU-based comparison is to be considered with some caution, because the MLU increases with the use of articles becoming productive. However, this also holds for the monolingual samples. In summary, we have seen that Alexander shows similar acquisition patterns for determiners in his two languages. Since the omission rates in German are unusually low for a German-learning child, positive influence from French to German seems likely. Céline omits fewer articles in German than in French, but this is hardly surprising, since when she starts to produce German determiners, she is still in the one-word stage in French. Yet, her omission rates in French are not low from the perspective of MLU. In German, both children omit fewer determiners than monolingual Germanlearning children at the same stage of development, i.e., they show quantitative differences in determiner omission. This being so, an interesting question to be raised is what this means in qualitative terms. There are three possibilities:
(i) bilinguals are less likely than monolinguals to omit determiners when these are required; (ii) bilinguals are more likely than monolinguals to use determiners in optional contexts; (iii) bilinguals overuse determiners in contexts where they would be used in French but not in German.
As for (ii) there are no contexts in German where determiner use is optional. Whenever the absence of a determiner does not result in an ungrammatical utterance, there is a change in meaning. Hence, (ii) can be ruled out.8 The interesting question is whether
8. One may argue that articles with proper names constitute an exception. In Southern German dialects, proper names tend to be used with articles, while in Northern German dialects proper names are usually bare. Northern German speakers do not consider articles with proper names to be ungrammatical, but they perceive them as belonging to a different dialect or even attribute a slightly different meaning to the noun phrase, e.g., that the person referred to has some property which makes him/her more easily identifiable among potential referents with the same name.
Dominance, mixing and cross-linguistic influence
bilinguals overuse determiners in the sense of (iii). A closer look at such contexts (see examples (3–6)), provides abundant evidence that the children resort to zero determination when a noun must be bare in German but should be preceded by a determiner in the corresponding French utterance. Examples are presented in (10). (10) a. b. c. d. e. f. g. h. i. j.
du mußt da Batterian [=Batterien] (Alexander 2;4,20) you must there batteries ‘you have to (put) batteries there’ Platz is da nich (Alexander 2;6,8) space is there not ‘there’s no space’ die sin noch Fotos (Alexander 2;7,27) ‘those are more pictures’ ja ich hab da Waggons (Alexander 2;8,12) yes I have there waggons ‘yes I have waggons there’ aber doch nich Lego das is (Alexander 2;8,28) but INT not lego that is (INT=intensifier) ‘but that’s no lego’ für Klebe is das (Céline 2;0,24) for glue is that ‘that’s for glue’ den nimm ich nach Hause (Céline 2;1,6) it take I to home ‘I take it home’ hat schöne Haare (Céline 2;1,14) has nice.PL hair.PL ‘has nice hair’ will Vanille- Eis (Céline 2;4,5) ‘want vanilla ice-cream’ kann auch Musik machen (Céline 2;6,8) ‘can also music make’
On the basis of such examples, it can be ruled out that children overuse German determiners where determiners would be required in French. I conclude from this that the observed influence is purely quantitative such that German bilinguals are less likely than monolinguals to omit determiners when these are required.
6. Cross-linguistic influence from weaker to stronger language The results presented in the previous section suggest that there is no cross-linguistic influence from German to French, which means that the simultaneous acquisition of German does not cause a delay in the development of French determiners, not even in
Tanja Kupisch
the case of Céline, whose French develops at a lower rate. I have proposed elsewhere (Kupisch 2003), that the French DP seems to be invulnerable to delays and crosslinguistic influence, possibly due to the high frequency of filled D-positions (as opposed to bare nouns) in the target-language. Frequency related arguments have also been put forward in the domain of prosody (e.g., Lléo & Demuth 1999) and may be carried over to the domain of French determiners (see Goad & Buckley 2006 for a prosody-oriented account of French articles in acquisition). I am not intending to go into more detail with respect to the invulnerability of the French DP, because the focus of this chapter is the relation between mixing, cross-linguistic influence and language dominance. The issues to be discussed in the remainder of this section are (i) whether the low omission rate in the children’s German is suggestive of positive influence from French to German, and (ii) whether this argument can be carried forward to Céline, whose French constitutes the weaker language. As for Alexander, we know that his overall development in German is rather slow compared to that in French. Nevertheless, and despite the fact that French-learning monolinguals tend to produce determiners later than German-learning monolinguals, the data show that his rate of omission in French does not differ significantly from his rate of omission in German, which I take to be suggestive of cross-linguistic influence. Céline’s case is more difficult to interpret. Unlike Alexander, her German develops relatively fast. The question arises whether the rate of omission in German is low because she is a fast learner. An MLU-based comparison may shed light on this question. Figure 13 compares the bilingual children’s rate of omission to that of the other German-learning children in terms of MLU. In the MLU-based comparison of the children’s use of determiners in German, the children’s rates of omission are more homogeneous as in the age-based account. Supposedly this may be due to the fact that MLU-development and determiner development are not completely independent, as
rate of D-omission (%)
100 ALE CÉL CHA EME FAL
80 60 40 20 0
1,0
2,0
3,0
Figure 13. Rate of determiner omission in German, MLU-based.
4,0
5,0
Dominance, mixing and cross-linguistic influence
mentioned above. At the same time, the comparison shows that, as in the age-based account, Alexander omits fewer determiners than monolingual children with a comparable MLU. Céline reaches higher MLU values at earlier ages than the other two children, but yet her rate of determiner omission is lower than Chantal and Emely’s at comparable MLU-stages. With an MLU between 3 and 4, there are a few recordings in which Céline produces an unusually high number of bare nouns (around 15%), but, most relevant tokens are instances of naming, which constitute borderline cases where adults would be more inclined to use articles, but may be found to drop them occasionally. (11) a. adult: der is ein Bär / oder was is das? / ne? ‘he’s a bear / or what’s that? / mh?’ child: Bär ‘bear’ b. adult: und das weißt du ne? ‘and that (one) you know, don’t you?’ child: Krokodili (single referent) ‘crocodile’ c. adult: was is das denn für’n Tier eigentlich? ‘and what kind of animal ist hat?’ child: Eichhörchen ‘squirrel’ d. adult: weißt du was das für einer is? / der da? ‘you know what kind that on is? / that one?’ child: Pinguini (single referent) ‘penguin’ (Céline 2;10,18)
Overall, the findings presented here suggest that the acquisition of German determiners is anticipated through the influence of the children’s second first language — French. Of course, further monolingual data should be analyzed to strengthen this point and to exclude that the bilingual children’s relatively fast development in German is not a case of individual variation. Provided these conclusions were along the right lines, the question arises, with respect to Céline, why the weaker language can influence the stronger language. Perhaps, the question must be posed in a different way. In fact, it may not be the weaker language that influences the stronger language, but the child’s exposure to it. From what we know about Céline, she was exposed to both languages from birth. Her father addressed her exclusively in French and French was the family language. The fact that Céline initially spoke little French does not mean that she perceived the French input in a way different from other children. It is therefore quite plausible that the French input the child received simultaneously with German had an accelerating effect on the acquisition of German determiners (Paradis & Genesee (1996: 3) for a definition of acceleration). In other words, it is not the weaker language per se, but the input in the weaker language that strengthens the acquisition of the German DP. Considering these
Tanja Kupisch
arguments, I would like to conclude that there is positive cross-linguistic influence in Alexander’s and Céline’s acquisition of German determiners.
7. Conclusions This article has compared the linguistic development of the bilingual children Alexander and Céline with respect to several aspects. The rate of their development in German and French differs in terms of MLU, upper bound, absolute number of utterances, increase of verb and noun types, i.e., parameters reflecting fluency and proficiency in a language. The language developing faster is German in the case of Céline, and French in the case of Alexander, the difference between the two languages being more pronounced in the case of Céline. Mixing patterns are consistent with these findings. Céline produces more mixed language utterances in French, while Alexander does so in German, but the overall percentages are higher in Céline’s speech production. Thus, in these two cases, strong(er) language and mixed language utterances go hand in hand, which confirms earlier findings by Berman (1979), Petersen (1988), Genesee, Nicoladis & Paradis (1995), and others. This does not imply that mixed language utterances are excluded in the language developing faster. They are merely noticeably fewer in number. Thus, instead of positing that mixing is unidirectional, as claimed by Petersen (1988), I would like to conclude that children are likely to produce more mixings in the language that develops more slowly, which does not exclude that mixing occurs also in the stronger language. In the second part of this chapter I have argued that French is resistant to crosslinguistic influence with respect to determiner omissions. In German, by contrast, both children cease to omit determiners early compared to other German-learning children, which suggests that French positively influences German in that consistent input in French may anticipate determiner production in German. Similar observations were made by Granfeldt (2003: 108). He notes that the Swedish-French bilingual child Mimi’s bare noun stage in Swedish is very short, possibly due to influence from French. Since Swedish was diagnosed as the child’s weaker language, this finding matches the view that the dominant language influences the weaker one. The present study suggests, however, that cross-linguistic influence is independent of language dominance. While the amount of mixed language utterances correllates with language dominance, cross-linguistic influence correlates neither with dominance nor with mixing. French, the (slightly) stronger language of Alexander, positively influences the acquisition of determiners in German. But German, the stronger language of Céline, does not negatively influence her acquisition of determiners in French. What is more, Céline’s exposure to French seems to have a positive effect on the development of determiners in German, although French is her weaker language. All of this suggests that mixed language utterances and cross-linguistic influence are two different kinds of contact phenomena. While the children produce more
Dominance, mixing and cross-linguistic influence
mixed language utterances in their strong(er) language, cross-linguistic influence occurs independently of whether a language is strong or weak. This finding runs counter to previous claims in the literature according to which influence operates unidirectionally from stronger to weaker language, while it is consistent with the claim that cross-linguistic influence is independent of language dominance (e.g., Müller & Hulk 2001). It seems important, though, to clearly define what we mean by cross-linguistic influence and to clarify what exactly the difference is between the two types of contact phenomena: mixing on the one hand, and cross-linguistic influence on the other hand. At present, we know little about the sources of cross-linguistic influence. Therefore, my concluding remarks should be seen as tentative. The observations made in this article appear to tally well with the idea, put forward by Paradis & Genesee (1996), that mixed language utterances are a performance phenomenon, while cross-linguistic influence is a competence phenomenon. This view gains plausibility if we consider recent findings by Cantone (2004), who observes that children mix elements of which they have already acquired the translation equivalent. Consequently, if a bilingual child inserts words from the more advanced language into utterances in the less advanced language, it cannot be taken for granted that these words are absent from the lexicon of the weaker language. Rather, they may be merely more easily accessible in the more advanced language (which probably corresponds to the language in which the child speaks more often). One should keep in mind that mixing also occurs with adult bilinguals who are competent users in both languages, and where it is largely dependent on the communicative situation. Cross-linguistic influence in the acquisition of grammar is different from mixing. It is independent of the communicative setting, because a grammatical category is either acquired or not. If we think of the DP in terms of a parameter distinguishing between languages that have a grammaticalized position for definiteness and languages without such a position, as proposed by Lyons (1999), the acquisition task would be for the child to determine the setting for the language he/she is exposed to. Since both German and French have grammaticalized this category, the settings are the same. This does not mean that D is always projected (i.e., also in the absence of overt determiners). Rather it is obligatorily projected with particular semantically and pragmatically determined readings, the range of which differs across languages. The path towards setting the DP-parameter may be language-dependent and slowed down or facilitated by syntaxexternal factors, such as prosody, overall frequency, and morphological load. For the reasons mentioned in section 2, it is likely that French is favourable to an early setting while German is not. The data presented here suggest that the simultaneous acquisition of French promotes an early setting of the parameter in German in German/French bilinguals. This does not mean that the bilingual transfers French structures to German. It means that exposure to French helps the child attain a certain syntactic competence earlier in German than would be the norm in monolingual children. To conclude, unlike mixing, cross-linguistic influence seems to be a phenomenon of syntactic competence. Syntactic competence with respect to a particular grammatical
Tanja Kupisch
domain may be attained earlier in one language of a bilingual child than is the norm in monolingual development through the influence of another language. This may be the case independently of whether his language constitutes the stronger or the weaker language, provided that the child was consistently exposed to both languages. A child’s performance in a language may appear to be weaker in one language than in the other, because s/he produces shorter utterances and a smaller variety of words. In such cases we are inclined to speak of language dominance. This weaker performance may correlate with a higher number of mixed language utterances. Yet such phenomena must not be taken as indicative of the child’s competence in a language.
References Bassano, D. & Eme, P.E. 2001. Development of noun determiner use in French children: lexical and grammatical bases. In Research on Child Language Acquisition, The Study of Child Language, M. Almgren, A. Barrena. M.J. Ezizabarrena, I. Idiazabal & B. MacWhinney (eds), 1207–1220. Somerville MA: Cascadilla. Bergman, C.R. 1976. Interference vs. independent development in infant bilingualism. In Bilingualism in the bicentennial and beyond, G.D. Keller, R.V. Taeschner & S. Viera (eds), 86–96. New York NY: The Bilingual Review Press. Berman, R. 1979. The re-emergence of the bilingual: A case study of a Hebrew-English speaking child. Working Papers on Bilingualism 19: 157–179. Bernardini, P. & Schlyter, S. 2004. Growing syntactic structure and code-mixing in the weaker language: The Ivy-Hypothesis. Bilingualism: Language and Cognition 7(1): 49–69. Cantone, K. 2004. Code-Switching in Bilingual Children. PhD dissertation, University of Hamburg. Chierchia, G., Guasti, M.T. & Gualmini, A. 1999. Nouns and articles in child grammar and the syntax/semantics map. Presentation given at GALA, Potsdam, Germany. Cordes, J. 2002. Zum unausgewogenen doppelten Erstspracherwerb eines deutsch-französisch aufwachsenden Kindes: Eine empirische Untersuchung. MA thesis, University of Hamburg. De Houwer, A. 1990. The Acquisition of Two Languages from Birth: A Case Study. Cambridge: Cambridge University Press. Deuchar, M. & Quay, S. 2000. Bilingual Acquisition: Theoretical Implications of a Case Study. Oxford: Oxford University Press. Eisenbeiss, S. 2002. Merkmalsgesteuerter Grammatikerwerb: Eine Untersuchung zum Erwerb der Struktur und Flexion der Nominalphrase. PhD dissertation, Heinrich Heine Universität, Düsseldorf. Gawlitzek-Maiwald, I. & Tracy, R. 1996. Bilingual bootstrapping. Linguistics 34: 901–926. Genesee, F. 1989. Early bilingual development: One language or two? Journal of Child Language 16: 161–179. Genesee, F., Nicoladis, E. & Paradis, J. 1995. Language differentiation in early bilingual development. Journal of Child Language 22: 611–631. Goad, H. & Buckley, M. 2006. Prosodic structure in child French: Evidence for the foot. Catalan Journal of Linguistics 5 (Special issue on the acquisition of Romance languages as first languages): 109–142 Granfeldt, J. 2003. L’acquisition des catégories fonctionnelles. Étude comparative du développement du DP franσais chez des enfants et des adultes. PhD dissertation, University of Lund.
Dominance, mixing and cross-linguistic influence Grosjean, F. 1982. Life with Two Languages. Cambridge MA: Harvard University Press. Grosjean, F. 2001. The bilingual’s language modes. In One Mind, Two Languages, L. Nicol & T.D. Langendoen (eds), 1–22. Oxford: Blackwell. Guasti, M.T. & Gavarró, A. 2003. Catalan as a test for hypotheses concerning article omission. In Proceedings of the 24th Annual Boston University Conference on Child Language Development, B. Beachley, A. Brown & F. Conlin (eds), 288–298. Somerville MA: Cascadilla. Hulk, A. 1997. The acquisition of French object pronouns by a Dutch/ French bilingual child. In Language Acquisition: Knowledge, Representation, and Processing. Proceedings of the GALA 1997 Conference on Language Acquisition, A. Sorace, C. Heycock & R. Shillcock (eds), 521– 526. Edinburgh: University of Edinburgh. Hulk, A. & van der Linden, E. 1996. Language mixing in a French-Dutch bilingual child. In Eurosla 6. A Selection of Papers, E. Kellermann, B. Weltens & T. Bongaerts (eds), 89–101. Utrecht: Vereniging voor Toegepaste Taalwetenschap. Hulk, A. & N. Müller 2000. Cross-linguistic influence at the interface between syntax and pragmatics. Bilingualism: Language and Cognition 3(3): 227–244. Kupisch, T. 2003. The DP – a vulnerable domain? In (In)vulnerable Domains in Acquisition, N. Müller (ed.), 1–39. Amsterdam: John Benjamins. Kupisch, T. 2006. The acquisition of determiners in bilingual German/Italian and German/ French children. Lincam: Europa. Larrañaga, M.P. 2000. Ergative Sprachen, akkusative Sprachen: der Erwerb des Kasus bei bilingualen Kindern. Frankfurt: Vervuert. Leopold, W. 1970. A child’s learning two languages. In Second Language Acquisition: A Book of Readings, E. Hatch (ed.), 54–69. Rowley MA: Newbury House. Lleó, C. 2001. The interface of phonology and syntax: The emergence of the article in the early acquisition of Spanish and German. In Approaches to Bootstrapping: Phonological, Lexical, Syntactic and Neurophysiological Aspects of Early Language Acquisition, Vol. 2, J. Weissenborn & B. Höhle (eds), 23–44. Amsterdam: John Benjamins. Lleó, C. & K. Demuth 1999. Prosodic constraints on the emergence of grammatical morphemes: Cross-linguistic evidence from Germanic and Romance languages. In Proceedings of the 23rd Annual Boston University Conference on Child Language Development, A. Greenhill, H. Littlefield & C. Tano (eds), 407–418. Somerville MA: Cascadilla. Lyons, C. 1999. Definiteness. Cambridge: Cambridge University Press. Meisel, J.M. 1986. Word order and case marking in early child language. Evidence from simultaneous acquisition of two first languages. Linguistics 24: 123–183. Meisel, J.M. 1989. Early differentiation of languages in bilingual children. In Bilingualism across the Lifespan: Aspects of Acquisition, Maturity and Loss, K. Hyltenstam & L.K. Obler (eds), 13–40. Cambridge: Cambridge University Press. Meisel, J.M. (ed.). 1990. Two First Languages. Early Grammatical Development in Bilingual Children. Dordrecht: Foris. Meisel, J.M. (ed.) 1994. Bilingual First Language Acquisition. French and German Grammatical Development. Amsterdam: John Benjamins. Meisel, J. & Müller, N. 1992. Finiteness and verb placement in early child grammars. Evidence from simultaneous acquisition of two languages: French and German. In The Acquisition of Verb Placement: Functional Categories and V2-Phenomena in Language Acquisition, J. Meisel (ed.), 109–138. Dordrecht: Kluwer. Müller, N. 1993. Komplexe Sätze. Tübingen: Günther Narr. Müller, N. 1998. Transfer in bilingual first language acquisition. Bilingualism: Language and Cognition 1(3): 151–171.
Tanja Kupisch Müller, N. & Hulk, A. 2001. Cross-linguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition 4: 1–21. Müller, N. & Kupisch, T. 2003. Zum simultanen Erwerb des Deutschen und des Französischen bei (un)ausgeglichen bilingualen Kindern. Vox Romanica 62: 145–169. Müller, N., Hulk, A. & Jakubowicz, C. 1999. Object omissions in bilingual children: Evidence for cross-linguistic influence. In Proceedings of the 23rd Annual Boston University Conference on Child Language Development, A. Greenhill, H. Littlefield & C. Tano (eds), 482–494. Somerville MA: Cascadilla. Müller, N., Cantone, K, Kupisch, T. & Schmitz, K. 2002. Zum Spracheneinfluss im bilingualen Erstspracherwerb: Italienisch – Deutsch. Linguistische Berichte 190: 157–206. Müller, N., Kupisch, T., Schmitz, K. & Cantone, K. 2006a. Einführung in die Mehrsprachigkeitsforschung. Tübingen: Narr. Müller, N., Schmitz, K., Cantone, K. & Kupisch, T. 2006b. Null-arguments in monolingual children: A comparison of Italian and French. In The Romance Turn, V. Torrens & L. Escobar (eds), 69–93. Amsterdam: John Benjamins. Nicoladis, E. 1998. First clues to the existence of two input languages: Pragmatic and lexical differentiation in a bilingual child. Bilingualism: Language and Cognition 1: 105–116. Nicoladis, E. & Secco, G. 2000. The role of a child’s productive vocabulary in the language choice of a bilingual family. First Language 58: 3–28. Paradis, J. & Genesee, F. 1996. Syntactic acquisition in bilingual children: Autonomous or interdependent? Studies in Second Language Acquisition 18: 1–15. Pearson B.Z., Fernández, S.C. & Oller, D.K. 1995. Cross-language synonyms in the lexicons of bilingual infants: One language or two? Journal of Child Language 22: 345–368. Petersen, J. 1988. Word-internal code-switching constraints in a bilingual child’s grammar. Linguistics 26: 479–493. Quay, S. 1995. The bilingual lexicon: implications for the study of language choice. Journal of Child Language 22: 369–387. Schlyter, S. 1993. The weaker language in bilingual Swedish-French children. In Progression and Regression in Language: Sociocultural, Neuropsychological and Linguistic Perspectives, K. Hyltenstam & A. Viberg (eds), 289–309. Cambridge: Cambridge University Press. Schlyter, S. 1994. Early morphology in French as the weaker language in French-Swedish bilingual children. Scandinavian Working Papers in Bilingualism 9: 67–87. Schlyter, S. & Håkansson, G. 1994. Word order in Swedish as the first language, second language and weaker language. Scandinavian Working Papers on Bilingualism 9: 49–67. Szagun, G. 2001. Learning different regularities: The acquisition of noun plurals by Germanspeaking children. First Language 21: 109–141. Van der Velde, M. 2003. Déterminants et pronoms en néelandais et en français: syntaxique et acquisition. PhD dissertation, Université de Paris-5/ Université Paris-8. Van der Velde, M., Jakubowicz, C. & Rigaut, C. 2002. The acquisition of determiners and pronominal clitics by three French-speaking children. In The Process of Language Acquisition [Proceedings of the 1999 GALA Conference], I. Lasser (ed.), 115–132. Frankfurt: Lang. Volterra V. & Taeschner, T. 1978. The acquisition and development of language by bilingual children. Journal of Child Language 5: 311–326.
A cross-linguistic analysis of binding in Down syndrome Alexandra Perovic This chapter reports an investigation of the knowledge of binding in young adults with Down syndrome (DS), speakers of Serbo-Croatian, and compares it to that of English-speaking individuals with DS and typically developing controls. In line with data reported for English, Serbo-Croatian speakers with DS showed difficulties interpreting anaphoric elements. It is argued that these results point to a specific deficit in the grammar of DS: an inability to establish a syntactic relationship of binding between an anaphor and its antecedent. These results present evidence against the characterisation of language development in DS as delayed, since the pattern observed in this population cross-linguistically has not been observed during any stage of typical language development.
1. Introduction Down syndrome is a genetic disorder, accompanied by impaired cognitive abilities and often very limited language abilities. Despite widely reported individual differences, both in the realm of cognition and language, grammatical achievement in individuals with this disorder appears more severely limited than other aspects of language, such as pragmatic or vocabulary knowledge. Inconsistent use of grammatical morphemes such as articles, auxiliaries, copulas, pronouns, conjunctions, prepositions, verbal and nominal inflection has earned the speech of individuals with DS the characteristic ‘telegraphic’. Usually only simple syntactic structures are present, with complex structures such as passives, possessive forms, negation or interrogatives being out of reach for both children and adults (Fowler 1990; Rondal 1995). Speech intelligibility is reduced by inappropriate phonological processes such as final consonant deletion, consonant cluster reduction, various substitutions and omissions (Dodd 1976). Despite the observed grammatical deficits, comparisons of linguistic development in DS and in the typical population often come to the conclusion that children with DS acquire language just like typically developing (TD) children, the only difference being that the process of acquisition occurs at a much slower pace (Bridges & Smith 1984; Fowler 1988; Rutter & Buckley 1994, amongst others). However, if their acquisition process is the same as that of TD children, why is the end achievement so limited? The ‘slow but normal’ characterisation of language development in DS does not fit easily with reports of the
Alexandra Perovic
ultimate level of linguistic attainment in this population, which is comparable to that of a TD 2 or 3 year old (Fowler 1988; Rondal 1995). This inexplicable plateau, which a large number of individuals with DS may never exceed, inevitably points to the conclusion that language development in DS must be different. Our goal in this paper is to establish whether the knowledge of the syntactic module of binding in adults with DS differs to that of TD children, matched on a particular measure of cognitive and language development, and if so, how this difference surfaces in two distinct languages, English and Serbo-Croatian.1 Binding Principles guide the distribution and interpretation of nominal expressions, and as such constitute a major component of adult syntactic knowledge (Chomsky 1981, 1986). In standard Binding Theory (Chomsky 1981, 1986), Principle A governs the distribution and interpretation of reflexives, whereas Binding Principle B is concerned with pronominals. Typically developing children are known to acquire Principle A early and with few difficulties, whilst their acquisition of Principle B is significantly delayed. Using different experimental paradigms (most often the act out task, where participants are asked to act out the sentences presented by an experimenter, or truthvalue judgment task, where they are asked to provide yes/no answer to questions accompanying picture stimuli), English speaking children as old as 5 or 6 would accept (1) as grammatical around 50% of the time (Jakubowicz 1984; Chien & Wexler 1990):
(1) *Maryi is washing heri.
This phenomenon, often referred to as the ‘Delay of Principle B Effect’ (DPBE), has been reported for a variety of languages in addition to English: Dutch (Philip & Coopmans 1996), Russian (Avrutin & Wexler 1992), Icelandic (Sigurjónsdóttir 1992). The DPBE has been explained by invoking the distinction between binding and coreference (Grodzinsky & Reinhart 1993; Chien & Wexler 1990). The reference of a pronoun can be determined either syntactically (binding) or extra syntactically (coreference), and it has been argued that TD children have no difficulty with syntactic binding, but have problems establishing appropriate coreferential interpretation of pronouns, for which they have to rely on their processing (Grodzinsky and Reinhart 1993) or pragmatic (Chien & Wexler 1990) abilities which have not yet matured (see Perovic 2004 for a review of literature on DPBE). If language development in DS is merely delayed, then investigations of the availability of the Binding Principles in this population should demonstrate parallels with
1. ‘Serbo-Croatian’ is no longer the term officially used in the countries of former Yugoslavia – the language is now referred to as Bosnian, Croatian or Serbian. However, the author does not regard the differences between the three sufficient enough to subscribe to only one of the three variants. Furthermore, the aspect of syntactic knowledge discussed here does not differ in speakers of the three geographical variants of the language.
A cross-linguistic analysis of binding in Down syndrome
normal language development – Principle A should pose few problems, but Principle B should yield interpretive difficulties until later stages of development. This prediction is proven incorrect for English speakers with DS in Perovic (2002), who found that young adults with this disorder had specific difficulties assigning appropriate interpretation to reflexives; but none in assigning interpretation to pronouns. This unique pattern, which has not been evidenced in TD children cross-linguistically, is argued to reveal a deficit syntactic in nature, namely, a deficit in establishing the syntactic relation between the anaphor and its antecedent. The pattern of difficulties with anaphors but not pronouns in the population with DS has been replicated in teenage English-speaking children with DS (Ring & Clahsen 2005) but no data as yet exist in other languages. If there exists a grammatical deficit in a genetic syndrome such as DS, it should surface cross-linguistically, although there may be interesting differences in the actual patterns of deficiencies, due to different grammatical properties of languages under investigation. This is exactly what this study will show: Serbo-Croatian speaking adults with DS exhibit a pattern of deficiencies parallel to that seen in their English counterparts, but with some differences in how they interpret the relevant grammatical elements. Importance of cross-linguistic studies cannot be overestimated in our quest to understand the nature of language impairments in developmental disorders such as DS. Most of the literature on grammar in DS has centred on the English-speaking DS population, with the notable exception of several studies on Dutch, French and Italian. The present study is the first of its kind in focusing on speakers of a Slavic language with DS, Serbo-Croatian (SC), and comparing this knowledge to that of English speakers with this disorder. A caveat is in place here: the results of the two studies will be discussed in parallel, but not as the results of one same experiment conducted in two languages. The two studies, while parallel in investigating the same syntactic phenomenon in DS and using the same experimental task, were conducted as separate experiments, thus the methodology used was not identical: they were conducted during different time periods (within several years of each other), used different measures to assess participants’ general abilities and different matching methods between participants with DS and controls, on the basis of what resources were available (especially with regard to the study conducted in Serbia/then Yugoslavia), and finally, they employed a different number of participants with DS, one wider age range than the other, so the English speakers with DS and SC-speakers with DS, while of similar ages and cognitive abilities, were not matched in the two experiments. The paper is organized as follows: section 2 presents a brief overview of the pronominal system in SC to familiarize the reader with basic facts needed for understanding the task presented to children in the acquisition of binding in this language. Section 3 introduces the theoretical framework of Reinhart & Reuland (1993) which will be used to discuss the experimental data, and outlines specific predictions for the knowledge of binding in DS. Section 4 gives details of a previously unpublished experimental study on binding in six SC-speaking participants with DS and their matched TD controls, while section 5 gives a summary of the experiment with English speakers
Alexandra Perovic
with DS, reported at more length in Perovic (2002, 2004). In section 6 we discuss the findings of the two studies and their implications for the knowledge of grammar in DS. Here we also discuss the differences in how the observed deficit surfaces in English and SC speakers with DS. Section 7 concludes the paper.
2. Properties of the pronominal system in Serbo-Croatian (SC) Unmarkedly SVO, SC is known for extensive scrambling. Each argument is marked by morphological Case affixes, which denote the argument’s function in a clause, and each modifier agrees with its head with regard to Case, Gender and Number. Not surprisingly, it is a (subject) pro drop language. Both tensed verbs and participles agree with the subject in Person and Number; participles also agree with the subject in Gender (there is no participle/object agreement). The complete paradigm of pronominal elements, the focus of our investigation, is given in Table 1. Table 1. Full and clitic forms of pronouns and reflexives in SC (only Accusative forms given) Person/ Gender
full pronoun
pronominal clitic
full reflexive
reflexive clitic
1st 2nd 3rd masc 3rd fem 3rd neut
mene tebe njega nju njega
me te ga je ga
sebe “ “ “ “
se “ “ “ “
2.1 Personal pronouns and pronominal clitics Both full pronouns and pronominal clitics are fully specified for Case, Person, Gender and Number features. Full and clitic forms have been argued to have a different X’ status: full pronouns are generally considered to be XPs and clitics X0s.2 With SC having a rather flexible word order, the distribution of pronouns seems virtually unrestricted. It is interesting however that they rarely surface in the theta position itself, save for purposes of deictic use or contrastive stress (see 2).
2. The issue of the X’ status of pronouns and clitics, as well as their base generation, is a contentious one. Basing her claim on the distribution of pronouns when modified by a particular adjective, Progovac (1998) claims that SC pronouns are heads, generated in a lower (N) head projection and then moved to D.
A cross-linguistic analysis of binding in Down syndrome
In unmarked contexts, they are usually found scrambled to the left of the verb (examples 3 and 4).3 (2) Marija često zove njega (contrastive stress) ‘Mary often calls him.’ (3) Marija njegai često zove ti (unmarked) Mary him often calls ‘Mary often calls him.’ (4) Njegai Marija često zove ti (unmarked) him Mary often calls ‘Mary often calls him.’
In traditional discourse-functional terms, such organisation of the sentence involves the NP in the base-generated position carrying information focus, which refers to ‘new information’, whereas the fronted NP refers to ‘old information’.4 In contrast to the full pronouns, the distribution of pronominal (and other types of) clitics is highly restricted in SC: they occupy a rigidly fixed ‘second position’ in a clause. In fact the clitic can either follow a full constituent, (5), or appear after the first ‘word’, even if the latter involves ‘splitting’ a syntactic constituent, as seen in (6). The obligatory 2nd position requirement implies that SC clitics are never found in the argument position, or clause initially. (5) Ovaj čovek ga poznaje. this man him-cl knows ‘This man knows him.’ (6) Ovaj ga čovek poznaje. this him-cl man knows ‘This man knows him.’
The phenomenon of second position clitic placement, especially their apparent ability to split a constituent, has attracted a great deal of attention, mainly over which components of grammar have greater influence in determining clitic placement: morphology,
3. See Stojanović (1997) for an argument that this movement is clause-bounded, unless the pronoun is in the external topic position. 4. Stojanović (1997) treats this as object shift, in the fashion of Scandinavian. However, she adopts semantic, rather than morphological motivation (Holmberg 1986) for this movement, as proposed by Diesing & Jelinek (1995), who argue that all definite/specific nominal elements that do not introduce novel referents must scope out of the VP.
Alexandra Perovic
syntax or/and phonology.5 Whatever weight is given to each of the components, it is generally accepted that second position cliticization is motivated by a phonological requirement for the clitic to be attached to a host on its left, whilst the actual movement happens in syntax (Wilder & Ćavar 1994; Mišeska Tomić 1996; Progovac, 1996, amongst others). Following the standard approach(es), we thus assume that pronominal clitics in SC are heads, base-generated in the argument position and head-moved to some topmost position in the clause, e.g., C0, where they form a cluster with other types of clitics. The claim that SC pronominal clitics are arguments generated within VP, rather than just verbal inflection, is supported by the fact that SC does not allow clitic doubling (Halpern & Fontana 1994; Franks & Holloway King 2000).
2.2 The full reflexive pronoun and the reflexive clitic In striking contrast to the pronominal paradigm, there is only one full reflexive pronoun, ‘sebe’ (7). (7) Marijai smatra sebei pametnom. Mary self-Acc considers clever ‘Mary considers herself clever.’
Although it is marked for Case (it has no nominative form), the reflexive is not marked for Gender, Person or Number, and thus is able to take a range of antecedents. The SC anaphor is subject oriented: it can only relate to an antecedent that bears the grammatical relation of subject.6 (8) Marijai je pričala Suzanij o sebii/*j. Mary aux told Susan about self ‘Mary told Susan about herself.’
Similar to pronouns, ‘sebe’ can be found in a variety of overt positions, with the position to the left of the verb being least marked (cf. 2, 3, 4).
5. For an overview of clitic phenomena and different accounts of their distribution in Slavic, see Wilder & Ćavar (1994), Halpern & Zwicky (1996), Boškovic (1997), Franks & Holloway King (2000), amongst others. 6. In terms of Pica (1987), subject orientation of the SC anaphor can be explained by its movement at LF: morphologically simple anaphors move at LF, are subject-oriented and allow long distance antecedents, in contrast to complex anaphors, such as those in English. See Progovac (1992) for an account of anaphor distribution of SC without invoking movement: she captures the same properties of SC anaphors (morphological simplicity, subject orientation and some long distance characteristics) in terms of her ‘Relativized SUBJECT’ analysis, where the binding domain for a reflexive pronoun must contain a potential antecedent which has to be X-bar compatible with the reflexive in order to bind it.
A cross-linguistic analysis of binding in Down syndrome
Like the full reflexive, there is a single form of the reflexive clitic, used for all Persons, Number and Gender: (9) Marko se kupa. Marko se bathes ‘Marko is having a bath.’
The ‘se’ clitic occupies the second position all SC clitics are restricted to; all other position are ungrammatical: (10) *Marko kupa se. Marko bathes se ‘Marko is having a bath.’
Like in Romance and other Slavic languages, ‘se’ can also appear with a number of other constructions e.g., unaccusatives, impersonals, passives, middles, but here we shall discuss only its reflexive use. Following traditional approaches to reflexive cliticization, we assume that SC reflexive clitic is not derived in an argument position and subsequently moved, but base generated as a head in some position left of the verb, from which it moves to the second position. It has been argued that the reflexive clitic is associated with internal (object) theta role reduction/absorption (Wehrli 1986; Chierchia 1989; Reinhart 1996). In Reinhart (1996) and Reinhart and Siloni (2005) inherent reflexive marking is viewed as a lexical operation of reducing the internal theta role, where a two-way relation, λxλy (xRy)), is reduced to a property, λx(xRx)), rendering the predicate intransitive. The arity operations (valence-affecting operations) can apply either in the lexicon or in the syntax (see Reinhart & Siloni 2005 for the full account of where different languages fall on their Lexicon-Syntax parameter). Languages differ as to how these operations are marked: by the use of reflexive clitic, in Romance languages where the parameter is set to syntax, or by verb morphology in Hebrew, or even absence of marking altogether in English, where the arity operation takes place in the lexicon. In Dutch, another language where this operation takes place in the lexicon, a simplex anaphor ‘zich’ appears with inherently reflexive predicates, not as a morphological marker of reflexivity, but to absorb the residue of accusative Case after the relevant lexical operation takes place. The status of the SC clitic is not exactly clear. According to Marelj (2004) and Reinhart & Siloni (2005), SC ‘se’ behaves like the Romance clitic, signalling that the arity operation has taken place in the syntax. However, there are contexts where the reflexive clitic ‘se’ seems to behave more like Dutch ‘zich’ than the Romance reflexive clitic.7 For instance, ‘se’ seems restricted to inherently reflexive predicates as in (11), and cannot be used to reflexivize just any transitive predicate: (12a) and (12b) are
7. I am grateful to a reviewer who pointed out the similarities between SC ‘se’ and Dutch ‘zich’.
Alexandra Perovic
questionable or outright ungrammatical for a number of speakers of SC, who prefer the full reflexive ‘sebe’ in these contexts:8 (11) Marko se brije. Marko se shaves ‘Marko shaves.’ (12) a.
Marko *se/sebe voli. Marko se loves ‘Marko loves himself.’
(12) b.
Marko *se/sebe crta. Marko se/self draws ‘Marko draws himself.’
In contrast, Romance clitics show no such restrictions as seen in the French examples below (examples from Reinhart and Siloni 2005):9 (13) a.
Jean se dessine. Jean se draws ‘Jean draws himself.’
8. The use of the clitic is fine with some transitive, non-inherently reflexive perceptual verbs, e.g., ‘see’: (i) Marija se vidi u ogledalu. Mary se sees in mirror ‘Mary sees herself in the mirror.’ However, it is feasible that such verbs are a special case. Other uses of ‘see’ do not allow the clitic: in the dialect of the author and her informants, the following example is unacceptable and the full reflexive ‘sebe’ is preferred: (ii) Marija *se/sebe vidi u direktorskoj fotelji za nekoliko godina. Mary se/self sees in manager’s chair in several years ‘Mary sees herself as a manager in several years.’ 9. There are other instances where SC patterns with Dutch, and not with Romance. In the following instances of Dutch nominalization (Everaert 1986), ‘wash’ usually implies a reflexive reading, whereas ‘hate’ does not. (i) (ii)
Wassen is gezond. Washing (oneself) is healthy Haten is niet gezond. Hating (only someone else) is unhealthy
We find parallel examples in SC: (iii) (iv)
Kupanje je zdravo. Bathing (oneself) is healthy Mrznja je nezdrava Hatred (someone else) is unhealthy.
A cross-linguistic analysis of binding in Down syndrome
(13) b. Jean s’aime. Jean se loves ‘Jean loves himself.’
The precise status of ‘se’ as a marker of the lexical operation of reflexivization needs to be established, but such an endeavour is out of scope of this paper. It can very well be that there are dialectal differences with regard to its nature and uses, and, in the framework of Reinhart & Siloni (2005), where the arity operation takes place in SC. Cases of dialectal variation for Dutch ‘zich’ have also been reported: standard Dutch ‘zich’ has no clitic-like properties, but in Heerlen Dutch ‘zich’ is reported to behave more like the French reflexive clitic ‘se’ (Cornips & Hulk 1996). For the purpose of this paper, I will assume that, like Dutch ‘zich’, the SC reflexive clitic appears with inherently reflexive predicates only, and, not being an argument, it does not enter the relation of syntactic binding. Its exact nature requires further research. Following Everaert (1986) we assume that both the transitive and inherently reflexive form of a predicate are listed as separate entries in the lexicon: when they are inherently reflexive, they appear with the clitic (e.g.,‘prati se’ – wash) and when ordinary transitive verbs, without the clitic (‘prati’ – wash). In the experiments discussed in section 4, we employ verbs which can be used both as inherently reflexive and transitive. The ensuing section discusses the theoretical approach to binding which will be adopted to account for the experimental data obtained in sections 4 and 5, followed by predictions about the knowledge of binding in DS.
3. A Reflexivity approach to Binding (Reinhart & Reuland, 1993) To account for the distribution of the anaphoric elements in SC briefly reviewed above, we adopt an approach developed by Reinhart & Reuland (1993) (thereafter In Spanish, as an anonymous reviewer pointed out, the reflexive clitic is obligatory (see vii below) for a reflexive reading to be obtained, otherwise the non-reflexive reading is only available (vi): (vi) Lavar es saludable. *Washing oneself is healthy (vii) Lavare es saludable Washing oneself is healthy Note however that some language-specific grammatical properties also seem to be at play: if the infinitive form, rather than the nominal, is used in the example (iii) above, the clitic must be used: (viii) *Kupati je zdravo/Kupati se je zdravo Bath-inf is healthy/ Bath-inf se is healthy To bath oneself is healthy
Alexandra Perovic
R&R), which differs to strictly syntactic approaches to interpretive dependencies, such as standard Binding Theory of Chomsky (1981, 1986). In standard Binding Theory, interpretive dependencies crucially rely on structural conditions (Binding Principles A and B), however, there are instances of extra syntactic dependencies that cannot be regulated by strictly locality condition, such as coreference. The Binding Theory adopted here incorporates the central notion of Reinhart (1986) that the reference of a pronoun can be determined in two ways: syntactically and extra syntactically. Syntactic binding, which we are concerned with here, pertains to bound variable anaphora only, where the anaphoric relation involves a dependency mediated by grammar internal processes. We assume this relation to be regulated by conditions proposed by R&R on the reflexivity of predicates and the formation of syntactic chains. R&R’s version of Binding Theory argues for two types of conditions on A-binding: conditions on binding as conditions on reflexivity of predicates, and the condition on chain formation. Together they account for the distribution of pronominals and anaphors cross-linguistically. Conditions on binding are given in (14), and the generalised condition on A-chains in (15). (14) Condition A: A reflexive marked syntactic predicate must be reflexive. Condition B: A reflexive semantic predicate must be reflexive marked. (where the predicate is reflexive if two of its arguments are co-indexed). (15) A maximal A-chain (a1,. . ., an) contains exactly one link – a1 – that is both [+R] and Case-marked (fully specified for pronominal features, including structural Case).
In this framework, anaphors and pronominals are classified according to two characteristics: referential independence [±R], and their ability to reflexive-mark a predicate, [+/–SELF]. Referential properties reflect whether an element is fully specified for grammatical features of Gender, Number, Person and structural Case; these properties determine whether the element can enter the chain formation. Elements lacking in referential features, [–R],10 can occupy the foot of an (extended) A-chain, whilst those with full specification for referential features, [+R], are excluded from this position. Elements that have the function of reflexive-marking, [+SELF], are able to reflexivize transitive predicates, those that are not reflexive-markers cannot appear as arguments of transitive, lexically non-reflexive predicates. In (16), the two arguments of the predicate are co valued – the complex anaphor reflexivizes the predicate, satisfying Condition A; the pronoun is ruled out by Condition B: (16) Johni loves himselfi/*himi.
10. The [–R] property of anaphors is related to their incomplete specification for structural Case, i.e., the lack of Nominative.
A cross-linguistic analysis of binding in Down syndrome
As mentioned earlier, inherently reflexive predicates are marked so in the lexicon (cf. the discussion on inherently reflexive predicates in section 2.2 in relation to reflexive clitic in SC). Languages have different ways of expressing inherent reflexivity: in English, they are intransitive (see 17).11 In Dutch, inherently reflexive predicates appear with a simplex anaphor zich whose function is to absorb the residue of accusative Case, after the lexical operation has taken place (see 18). (17) Johni behavesi /himself/*him. (18) Jani gedraagt zichi/*zichzelfi/*hemi. (Dutch) John behaves zich/himself/him
Since the predicates in (17) and (18) are inherently reflexive, there is no need for them to be reflexive-marked in syntax by the complex anaphor (‘himself ’ in English and ‘zichzelf ’ in Dutch); the conditions on reflexivity are satisfied. The revised Chain Condition excludes pronouns from the same contexts. The important consequence of Binding Conditions, as stated in R&R, is that they account for the distribution of anaphors and (together with the revised chain theory) pronouns, without stating any restrictions on their structural domains. Classification of SC personal pronouns and pronominal clitics along the R&R parameters is straightforward. None of these elements have a reflexive-marking function, being [–SELF], and so cannot appear as arguments of reflexive predicates: (19) *Marijai jei/njui voli. Mary her-cl/her loves ‘Maryi loves heri.’
Since pronouns and pronominal clitics have full Gender, Person, Number and Case paradigms, they are excluded from appearing in the foot of the chain by the Chain Condition (even if Binding Conditions of R&R do not prevent them from occurring as arguments of inherently reflexive predicates). (20) *Markoi njegai brije. Marko him shaves ‘Markoi shaves himi.’
The SC full reflexive pronoun ‘sebe’ is [–R]. It lacks most referential features, not being marked for Number, Person or Gender. Despite a full paradigm of inherent Case,
11. Some English verbs seem ambiguous between being inherently reflexive and purely transitive predicates (e.g. behave, shave, wash). When inherently reflexive, they are intransitive and appear with one argument only (John shaves). When transitive, they can take an object distinct in reference from the subject (John shaved Peter), or become reflexivized just like any other transitive predicate, with their internal argument appearing in the form of a complex anaphor (Johni shaves himselfi). Each occurrence of the predicate is listed separately in the lexicon.
Alexandra Perovic
it is not fully specified for structural Case: it lacks Nominative. Thus being [–R], it is able to occupy the foot of the A-Chain. The reflexive-marking function makes ‘sebe’ a SELF anaphor, which means that it can appear as a co argument of a transitive reflexive predicate. Both are illustrated in (21). (21) Markoi sebei voli. Marko self loves ‘Markoi loves himselfi.’
As discussed earlier, SC reflexive clitic ‘se’ is not an argument and does not enter the A-chain relation. It occurs only with inherently reflexive predicates. It is not an anaphor and therefore it is unable to reflexive-mark a transitive predicate in syntax the way ‘sebe’ can: (22) *Marko se voli. Marko se loves ‘Marko loves himself.’
In summary, the key characteristics of SC anaphors and pronouns, in terms of properties that govern their distribution, are as follows: a. Like the English reflexive, the only ‘true’ anaphoric element that has the function of reflexive marking in syntax and lacks referential features is ‘sebe’([–R] and +SELF); b. pronouns and pronominal clitics do not reflexive-mark predicates in syntax, and are fully specified for referential features ([+R] and –SELF) c. the reflexive clitic ‘se’ is not an anaphor but a lexical marker of inherent reflexivity and does not enter anaphoric relations.
3.1 Knowledge of binding in DS: Predictions If there exists a deficit in the ability to establish a syntactic dependency between an anaphor and its antecedent in DS, as argued by Perovic (2002) on the basis of difficulties in the comprehension of anaphors reported for English-speaking adults with DS, a parallel deficit should surface in individuals with DS who are speakers of SC. It is hypothesised that the problem with interpreting anaphors does not extend to the knowledge of variable binding or coreference in DS, nor to the conditions on binding as given in the framework of Reflexivity of R&R. However, there may be differences in the actual patterns of performance on different pronominal elements, due to differences in the pronominal systems in English and SC. With its use of the reflexive clitic with inherently reflexive predicates, SC offers a unique opportunity to investigate the knowledge of reflexivity without the use of anaphor.12 12 See Perovic (2004) for an additional experiment with the English participants with DS who were tested on their knowledge of inherently reflexive verbs used without the reflexive forms, i.e., wash, shave. Their good performance revealed knowledge of reflexivity conditions independent of anaphoric binding.
A cross-linguistic analysis of binding in Down syndrome
With the properties of the SC pronominal system in mind, we can make concrete predictions about the patterns our SC-speaking participants will exhibit. Perovic (2002) argued that the English speakers with DS revealed knowledge of Binding Condition B but revealed a lack of the relation of anaphoric binding. This deficit in anaphoric binding made it unfeasible to probe for any awareness of Binding Condition A: this condition states that reflexive predicates have to be interpreted reflexively. Note however that the SC version of the experiment can test lexically reflexive predicates, via constructions involving the reflexive clitic, thus providing an opportunity to test for Condition A without resorting to anaphoric binding. Thus, on the basis of the interaction between the presence of binding principles and the absence of anaphoric binding, the following predictions concerning the patterns which SC-speaking participants will show, can be made: i.
Knowledge of Condition A: If participants have the knowledge of conditions on reflexivity (know the semantic properties of reflexive predicates), they are expected to correctly interpret constructions containing inherently reflexive predicates accompanied by the reflexive clitic ‘se’, as reflexive. ii. Knowledge of Condition B: If participants have the knowledge of conditions on reflexivity (know the semantic properties of reflexive predicates), they are expected to rule out constructions depicting reflexive action but containing predicates that are not reflexive-marked, where the argument is a pronoun. iii. An inability to use anaphoric binding: If participants have the knowledge of conditions on reflexivity, but are unable to establish the syntactic relation of binding between the anaphor and its antecedent, they are expected to show difficulties interpreting constructions containing the strong reflexive form, the true anaphor ‘sebe’. However, due to distinct grammatical properties of SC and English anaphors, attempts to interpret the anaphor may result in different strategies. Consequently, performance patterns different to those employed and shown by English participants may be expected. In short, if the principles of binding are available, but the relation of anaphoric binding is not, participants will struggle to interpret the anaphor ‘sebe’, but will have no difficulties interpreting constructions containing pronouns or the reflexive clitic ‘se’.
4. Experiment: Binding in Serbo-Croatian-speaking individuals with DS To establish whether the deficit in the knowledge of binding in DS is present in this population cross-linguistically, a study was conducted with six young adults with DS, individually matched on non-verbal abilities to TD controls. The experimental method used was a SC version of the experiment reported in Perovic (2002, 2004) with English-speaking individuals with DS (we will be referring to this study throughout the paper; a summary of the relevant data is given in section 5).
Alexandra Perovic
4.1 Participants Six female adults with DS, aged 19–29, and six TD children,13 aged 5–6;11, participated in the study. They were recruited with the help of the Down Syndrome Association of Yugoslavia, Belgrade, and the secondary school for students with learning disabilities SŠ ‘Milan Petrović’ in Novi Sad, Serbia. Control participants came from a kindergarten in Novi Sad. Four participants with DS were similar in chronological age and level of cognitive functioning to the four English participants with DS studied in Perovic (2002). Two more somewhat older participants (23 and 29), functioning at similar cognitive and language levels as the original four participants, were included in the experimental group, in order to exhaust the population available in the two establishments where the testing was being conducted. The aetiological subtype of DS in the participants was not confirmed, but is suspected to be standard trisomy 21. All participants were considered by their tutors to be highly functioning in comparison to average adults with DS. Their cognitive abilities ranged from average to high (with regard to the range typical for individuals with DS), as shown in Table 2: their verbal IQ, as measured by the Yugoslav standardisation of Wechsler Adult Intelligence Scales, VITI, ranged between 57 and 59, whilst their performance IQ ranged from 61 to 64. Their MLU, measured in words,14 ranged between 3.41 and 5.84.15 Their scores on Raven’s Coloured Progressive Matrices (RCPM), another test assessing non-verbal reasoning, are reported here as age equivalents here (they were all below the 1st percentile). The noticeable disparity between verbal and non-verbal IQ of WAIS (confirmed by a paired samples t-test: t (5) = 4.919, p = 0.004) is reminiscent of that reported in the literature on DS, however no details of their mastery of grammar and vocabulary are given since no standardised tests of grammar and vocabulary comparable to TROG or BPVS are available in SC. All participants lived with their families, of average to high social status background. Five out of six participants came from a monolingual background; DS3 has undergone schooling in SC whilst also exposed to Slovak at home. All participants attended primary and secondary schooling for students with learning
13. In order to first establish when binding is acquired by SC-speaking TD children, the experiment was conducted with 37 children aged 3;3 to 6;11, divided into 4 age groups: 3–4, 4–5, 5–6 and 6–7 yr old. See Perovic (2004) for complete data and a discussion of observed patterns. 14. In highly inflected languages, MLU is best calculated in words, as counting all the morphemes would overestimate the complexity of an utterance (see Rivero & Goledzinowska, 2002, for Polish). The speech sample was elicited with the aid of the picture book ‘Frog, where are you’, used in narrative elicitation in both English and languages other than English (Slobin & Berman 1994). 15. On the same measure (narrative elicitation), the word MLU for a non-impaired SC-speaking adult was 6–7. For 3–7 year old children it ranged between 3.5 and 7.
A cross-linguistic analysis of binding in Down syndrome
Table 2. Scores on standardised tests and MLU for SC speaking participants with DS Age MLU in words Perfor IQ* Verbal IQ**
RCPM*** rs AE
DSI DS2 DS3 DS4 DS5 DS6
22 15 15 18 22 13
22;3 20;4 23;1 19;11 19;6 29;4
4.48 5.5 3.41 4.89 4.78 5.84
64 61 62 62 64 68
59 57 58 58 59 57
8 5;6 5;6 7 8 >5;6
Note: AE = Age Equivalent; rs = raw score *Performance scale of VITI (Vekslerov Individualni Test Inteligencije), Yugoslav standardisation of Wechsler Adult Intelligence Scales (WAIS). **Verbal scale of VITI, as above. ***RCPM Raven’s Coloured Progressive Matrices
disabilities in Serbia. DS1, DS2 and DS5 attended a creative arts workshop, set up especially for young adults with DS. DS3, DS4 and DS6 attended a centre for people with learning disabilities, where they were encouraged to engage in activities such as drawing, sewing, embroidering, or ceramics. Six control TD children (see Table 3), aged between 5 and 6;11 (chosen from the larger group of 37 children in Perovic, 2004), were individually matched to participants with DS on the raw scores on RCPM, within –2 to +1 point.16 This measure has been used in the literature to match individuals with DS to both TD children and those with SLI (Laws & Bishop 2003).
16. Since the focus of our investigation is language development, it would have been more desirable to match the participants with DS to controls on a linguistic measure, rather than a measure of non-verbal cognitive abilities. An ideal measure, often used in the literature would be a measure of receptive vocabulary, also used in Perovic (2002) for the English speaking adults with DS. Using such a measure would also have allowed easier comparisons between English speaking adults with DS in Perovic (2002) and SC-speaking adults with DS in this study. However, the lack of relevant standardized measures in SC made this unfeasible. As suggested by an anonymous reviewer, a way of enabling a more direct comparison between the two groups with DS would be to match the English group to controls on the same measure the SC-speaking group with DS were matched to controls – on Raven’s CPM. It was unfortunately not possible to recruit new control participants to match them to English-speaking adults with DS for the purpose of this paper. While a direct comparison would have been desirable, especially as the two groups of adults with DS did show parallels in cognitive abilities (e.g., a similar mean and ranges of raw scores on Raven’s), the two studies should be treated as separate due to additional differences in methodology, as outlined in the introductory section of the paper.
Alexandra Perovic
Table 3. Details of ages and standardized test scores for SC-speaking control children Age CS1 CS2 CS3 CS4 CS5 CS6
6;3 5;8 6;7 6;7 5;2 5;3
RCPM rs
RCPM Percentiles
23 15 15 18 20 14
95 50 25 50 90 50
Note: rs = raw score
4.2 Materials and procedure The task used in the study was a SC version of the truth-value judgement task used in Perovic (2002) with English speaking adults with DS, itself an adaptation of the task used with English TD children in Chien & Wexler (1990), but with original materials created and used in Perovic (2002). Its basic design, involving the elicitation of yesno answers to experimental questions accompanying picture stimuli, has been used extensively in studies on the acquisition of pronominal reference across languages and populations during the past decade: Hamann, Kowalski & Philip (1997) for French, Baauw (2000) for Dutch and Spanish, Norwegian (Hestvik & Philip 2000), van der Lely & Stollwerck (1997) for English speaking children with SLI, Clahsen & Almazan (1998) for English speaking children with Williams syndrome. Its strengths lie in eliciting both yes and no questions, to control for the positive bias that is often present in young TD children, and possibly with impaired populations, where participants answer ‘yes’ to please the experimenter, or whenever in doubt about the correct answer. Like in the English study with adults with DS (Perovic 2002), the same verbs (touch – ‘dirati’, wash – ‘prati’ and dry – ‘brisati’) were used.17 The same sentence types/experimental and control conditions were used, with the addition of test conditions involving clitic forms for each of the original conditions in the English version, and omitting one control condition that was deemed unnecessary. Sentences were constructed to reflect the unmarked order in SC, with full pronouns
17. In relation to our discussion of the SC clitic and the issue of inherently reflexive vs. noninherently reflexive verbs having separate entries in the lexicon, it is important to note that these verbs are considered inherently reflexive when used with the reflexive clitic, and ordinary transitive when other pronominal elements (and are able to be reflexivized with the help of the anaphor ‘sebe’ as any other transitive verb). However, when used with the reflexive clitic, the verb ‘dirati’ (touch) allows a non-reflexive reading where it in effect means ‘to touch someone else’. See footnote 20 for the implications of the existence of this reading in the participants’ performance on the reflexive clitic experimental conditions.
A cross-linguistic analysis of binding in Down syndrome
and full reflexives in the unmarked preverbal position. The clitic forms occupied the obligatory second position. There were 21 types of questions/sentence types, with 8 items in each question, except for one (CAX) which had 12 items, totalling 172 items. The examples for each experimental and control sentence type/question are given in Table 4. Table 4. Example questions used in SC experiment (English gloss) Experimental conditions
MATCH No. items MIS-MATCH No. items Example
Name-Pronoun NPM 8 NPX 8 Name-Pron. Clitic NPMcl 8 NPXcl 8 Name-Reflexive NRM 8 NRX 8 Name-Refl. Clitic NRMcl 8 NRXcl 8 Quant.-Pronoun QPM 8 QPX 8 Quant.-Pron. Clitic QPMcl 8 QPXcl 8 Quant.-Reflexive QRM 8 QRX 8 Quant.-Refl. Clitic QRMcl 8 QRXcl 8
Is Snow White her washing? Is her-cl Snow White washing? Is Snow White self washing? Is se Snow White washing? Is every bear him washing? Is him-cl every bear washing? Is every bear self washing? Is se every bear washing?
Control conditions Name-Name CNNM 8 CNNX 8 Quant.-Name CQNM 8 CQNX 8 Attention — — CAX 16
Is Snow White washing Cinderella? Is every bear touching Peter Pan? Is Father Christmas sleeping?
Sixteen sentence types/experimental conditions aimed to probe the knowledge of all types of pronouns in SC used in the object position: full reflexive, reflexive clitic, full personal pronouns, pronominal clitics. Half of them were ‘match’, where the question matched the picture, eliciting a correct ‘yes’ answer, and half of them were mismatch, where the question did not matched the picture, and the target answer was ‘no’. For instance, to assess the knowledge of the full SC reflexive ‘sebe’ when preceded by a referential (non-quantified) antecedent, two experimental conditions were used: NRM (name reflexive match) and NRX (name reflexive mismatch). Both involved the same question but the picture shown would be different: after introducing the characters shown in the picture, e.g., ‘This is Cinderella. This is Goldilocks’,
Alexandra Perovic
the participant was asked ‘Is Cinderella washing herself?’ In the match condition, NRM, the picture showed Cinderella washing herself, with Goldilocks standing by looking at Cinderella. The target answer to this question was ‘yes’. In the mismatch condition, NRX, the picture showed Cinderella washing Goldilocks, thus the target answer to the same question when presented with this picture was ‘no’. The full reflexive was also tested with quantified antecedents. For the QRM condition (quantifier reflexive match), a picture showing three witches, each drying herself, and Little Mermaid standing by, was accompanied by a question: ‘Is every witch drying herself?’. The correct answer to this question was ‘yes’. A different picture showing three witches drying Little Mermaid was used for QRX condition (quantifier reflexive mismatch), to the same question, ‘Is every witch drying herself?’, the target answer was ‘no’. Exactly the same pictures and questions were used for the reflexive clitic ‘se’, of course, with the clitic instead of the full reflexive: NRMcl (name reflexive match clitic) and NRXcl (name reflexive mismatch clitic). QRMcl (quantifier reflexive match clitic) and QRXcl condition (quantifier reflexive match clitic). Parallel questions, match and mismatch conditions, in referential and quantified contexts, were used to assess the knowledge of pronouns and pronominal clitics: NPM (name pronoun match) and NPX (name pronoun mismatch) with referential antecedents, and QPM (quantifier pronoun match) and QPX (quantifier pronoun mismatch) for quantified antecedents; for pronominal clitics: NPMcl (name pronoun match clitic) and NPXcl (name pronoun mismatch clitic) with referential antecedents, and QPMcl (quantifier pronoun match clitic) and QPXcl (quantifier pronoun mismatch clitic) with quantified antecedents. The same experimental pictures were used for the pronoun conditions, match (NPM –name pronoun match) and mismatch (NPX – name pronoun mismatch), where, after introducing the character in exactly the same way, a question ‘Is Cinderella washing her?’ could be used with two pictures: one showing Cinderella washing Goldilocks, for which the correct answer is ‘yes’. If the picture showing Cinderella washing herself is present with this question, the target answer is ‘no’. Of the five control conditions, four used exactly the same syntactic construction as the pronoun/reflexive conditions, where parallel questions were used but without any pronominal elements: CNNM (control name name match), CNNX (control name name mismatch), CQNM (control quantifier name match) and CQNX (control quantifier name mismatch). The last sentence type, CAX, control attention mismatch, was included to control for the well-known attention deficits in individuals with DS, it had 12 items, and was used only in mismatch form. It involved questions of the type ‘Is Santa Claus sleeping?’, presented with the picture when Santa Claus is standing next to a bed. The sentences were randomised, but presented to all the participants in the same order. Each of the three verbs was presented in each of the experimental as well as the control condition, thus familiarization effects could not be observed. Each participant was presented with all the sentences; due to their large number, they were administered in three blocks, in the course of three experimental sessions (64, 64 and 44 sentences).
A cross-linguistic analysis of binding in Down syndrome
The validity of the task was confirmed by ceiling performance of six non-impaired adults on all experimental and control conditions; they completed all three blocks in one session.
4.3 Results A mixed repeated measures analysis of variance (ANOVA) was performed with sentence type/test condition as within subjects factor, with 21 levels (16 experimental conditions and 5 control conditions) and group as a between subjects factor (two levels: DS and controls). It revealed main effects of test condition/sentence type (F(20,200) = 12.4, p < 0.001), group (F(1,10) = 8.687, p = 0.015) and a significant group X sentence type interaction (F(20,200) = 2.194, p = 0.003). To establish where the differences between the groups stem from, two subsequent separate mixed repeated measures ANOVA were run: on all match conditions, and on all mismatch conditions. An ANOVA with match conditions as within subjects factor, with 10 levels (NRM, NRMcl, NPM, NPMcl, QRM, QRMcl, QPM, QPMcl, CNNM, CQNM) and group as a between subjects factor (DS group, controls) revealed no significant main effects on test condition F(9,90) = 1.19, NS) or group (F(1,10) = 1.17, NS) and no interaction between the two (F(9,90) = .079, NS). In contrast, an ANOVA with mismatch conditions as within subjects factor (11 levels: NRX, NRXcl, NPX, NPXcl, QRX, QRXcl, QPX, QPXcl, CNNX, CQNX, CAX) and group as a between subjects factor (DS group, controls) revealed main effects of sentence type: F(10, 100) = 17.65, p < 0.001, group F(1,10) = 31.24, p < 0.001, and significant interaction between the two: F(10, 100) = 2.36, p = 0.015. Both groups performed well on control conditions (Figure 1), match and mismatch, which confirmed their understanding of the task and sentence types without the use of pronominal elements. 100
% correct
80 60
DS TD
40 20 0
CNNM
CQNM
CNNX
CQNX
CAX
Figure 1. Percentage correct on control match and mismatch conditions: SC speaking groups.
Alexandra Perovic 100
% correct
80 60
DS TD
40 20 0
NPM
NRM
QPM
QRM
Figure 2. Percentage correct on experimental full pronoun and reflexive match conditions: SC speaking groups.
High performance for both groups is again seen on experimental match conditions involving full pronouns and the reflexive (Figure 2). A different picture emerges on the mismatch conditions (Figure 3). Participants with DS show a distinct performance between conditions involving pronouns and those involving the reflexive pronoun. On mismatch conditions involving pronouns, NPX and QPX, they achieved high 92% correct and 90% correct, respectively. However, their performance on NRX and QRX, mismatch conditions involving the full reflexive, is markedly different: they attain only 60.4% correct for NRX, and 66.6% correct for QRX. In stark contrast, controls reached over 97% correct on all four mismatch condition. The mismatch experimental conditions reveal interesting individual variation in the group with DS (for individual data on all match experimental conditions, where not much variation is seen – the group mean is 90% or higher – see Perovic 2004). Individual scores for participants with DS given in Table 5 show wide differences mostly on conditions NRX and QRX. Although the group score for both conditions is 60% and 66.6% correct, respectively, participants DS2, DS3 and DS4 hover below or just above chance on these conditions, whereas DS518 and DS6 seem to show a stronger performance. Whilst DS1 does not perform badly, scoring 6 out of 8, or 75% correct, this is the first drop recorded in her performance, which had thus far been exclusively 100% on all control and experimental conditions. The participants’ performance is much more uniform on conditions involving pronouns, where the lowest score was 75% correct. 18. With her performance ranging between 87.5 and 100% correct, DS5 obviously does not share the difficulties in the interpretation of reflexive elements with the other participants. Note however, that she showed an extremely poor comprehension of pronominal clitics (Table 6), revealing perhaps some other deficit, unrelated to the deficit in the interpretation of reflexives.
A cross-linguistic analysis of binding in Down syndrome 100
% correct
80 60
DS TD
40 20 0 NPX
NRX
QPX
QRX
Figure 3. Percentage correct on experimental mismatch full pronoun and reflexive conditions: SC speaking groups.
Table 5. Percentage of individual correct responses, mismatch conditions: SC speaking group with DS Pronouns
Item code
DS1
DS2
DS3
DS4
DS5
DS6
Name Pron Quant. Pron.
NPX QPX
100 100
87.5 75
87.5 100
100 100
87.5 87.5
87.5 75
NRX QRX
75 75
37.5 62.5
37.5 62.5
37.5 37.5
87.5 100
87.5 62.5
Reflexive Name Refl. Quant. Refl.
Participants’ scores on clitic conditions, match and mismatch are given in figures 4 and 5.19 On the match clitic conditions, NRMcl, QRMcl, NPMcl and QPMcl, both groups scored around 90% correct or above (Figure 4). On all the mismatch conditions involving pronominal and reflexive clitics (Figure 5), TD controls show a superior performance
19. Due to a possible confounding reading that allows both reflexive and non-reflexive reading with the verb ‘touch’, any errors on the verb ‘touch’ in the reflexive clitic conditions NRMcl, QRMcl, NRXcl and QRXcl were treated as correct in the performance of both controls and participants with DS. If the results for ‘touch’ were excluded, both groups would still average 80% correct or higher. This does not affect other experimental conditions. See Perovic (2004) for this effect in young TD SC-speaking children and some of the adults with DS and its explanation, based on Rivero (2000) and Stojanović (2002). The same effect was observed in Stojanović (2002).
Alexandra Perovic 100
% correct
80 60
DS TD
40 20 0
NPMcl
NRMcl
QPMcl
QRMcl
Figure 4. Percentage correct on match clitic conditions: SC speaking groups.
100
% correct
80 60 DS TD
40 20 0
NPXcl
NRXcl
QPXcl
QRXcl
Figure 5. Percentage correct on mismatch clitic conditions: SC speaking groups.
of over 97% correct. The scores revealed by the group with DS were noticeably lower. Participants with DS showed a very good performance on the reflexive clitic condition, NRXcl, over 90% correct, but just under 80% on QRXcl. Their performance on pronominal clitic condition, NPXcl and QPXcl, was 83% and 70% correct, respectively. As seen in Table 6 showing individual percentages on clitic mismatch conditions, the lower group mean on these conditions was due to the poor performance of two participants, DS5 and DS6.
A cross-linguistic analysis of binding in Down syndrome
Table 6. Percentage of individual correct responses, mismatch clitic conditions: SC speaking group with DS Pronouns
Item code
DS1
DS2
DS3
DS4
DS5
DS6
Name Pron. Cl. Quant. Pron Cl.
NPXcl QPXcl
100 100
87.5 75
100 100
100 100
50 12.5
62.5 37.5
NRXcl QRXcl
100 100
62.5 87.5
100 75
87.5 75
100 87.5
100 50
Reflexive Name Refl. Cl. Quant. Refl Cl.
Independent samples t-test confirmed that the group with DS performed significantly worse than the control group on both mismatch conditions involving the full reflexive: NRX (t(10) = –3.8, p = 0.003) and QRX, (t(10) = –3.63, p = 0.005), and one condition involving the full pronoun, NPX (t(10) = –3.16, p = 0.01). They also performed significantly worse on the condition where the reflexive clitic is bound by a quantified antecedent, QRXcl (t(10) = –2.57, p = 0.028). No other statistically significant differences on test conditions were revealed.
5. Knowledge of binding in English speakers with DS This section provides a summary of the experiment reported in Perovic (2002) for four English-speaking young adults with DS, aged between 17 and 21. The participants were all highly functioning and able to attend daily classes at a further education college, with one participant undergoing training for a vocational diploma. The aetiological subtype of DS was not confirmed, but was suspected to be standard trisomy 21. All participants came from the Twickenham area in greater London, UK, from families of average to high economic status. Table 7 gives details of their performance on a variety of assessments. Table 7. Scores on standardized tests for English participants with DS Age DS1 DS2 DS3 DS4
17;9 17;2 19;3 20;7
BPVS rs 83 66 71 51
BPVS TROG AE rs 8;1 6;6 7;0 5;0
8 10 7 5
TROG AE 4;9 5;3 4;5 4;0
Perform RCPM IQ* rs AE 65 55 57 62
21 17 15 16
7;8 6;6 5;6 6;0
Note: AE = Age Equivalent; rs = raw score *Performance Scales of Wechsler Adult Intelligence Scales (WAIS).
In line with Laws & Bishop (2003), participants’ scores on receptive vocabulary (measured by British Picture Vocabulary Scales, BPVS-2) and grammar (Test of Reception
Alexandra Perovic
of Grammar, TROG) are given as age equivalents, since their standard scores were at floor. Participants’ non-verbal reasoning abilities were assessed by the performance scales of Wechsler Adult Intelligence Scales (WAIS) and Raven’s Coloured Progressive Matrices (RCPM), the only two measures that were available to use for the participants with DS in Yugoslavia. English participants’ standard scores on WAIS ranged between 55–65, with the mean of 59.75 (SD = 4.57) slightly lower than for the Yugoslav participants with the DS (M = 63.5, SD = 2.5). Their range of raw scores (15–21, M = 17, SD = 2.82) and age equivalents on RCPM were similar to the scores of Yugoslav participants (range 13–22, M = 17.5, SD = 3.83). Again, raw scores and age equivalents are given because participant’s performance on RCPM was below the 1st percentile. Control participants, recruited from North London, UK, aged between 5;11 and 7;10, were individually matched to participants with DS on the raw scores of the receptive vocabulary measure, BPVS, within –3 to +3 points. Their details are given in Table 8. Table 8. Details of ages and standardized test scores for English control children Age CS1 CS2 CS3 CS4
7 6;1 7;10 5;11
BPVS rs
BPVS SS
83 64 68 54
111 102 90 96
Note: rs = raw score
The experimental task used was the English version of the truth-value judgement task discussed in section 4.2 above, involving the same three verbs: ‘wash’, ‘dry’ and ‘touch’, and identical picture stimuli. It employed the same 8 experimental conditions as the SC experiment (no clitic conditions): NRM, NPM, QRM, QPM, NRX, NPX, QRX, QPX, and six control conditions: the same five conditions as in the SC study, CNNM, CNNX, CQNM, CQNX, CAX, plus an extra control condition: CNAX (control name action mismatch).20 There were 8 items in each condition, except for CAX which had 16 items (the SC version of CAX had 12 items). To establish the validity of the task,
20. CNAX (control name action mismatch) was included in the English version of the experiment as an additional condition to test participants’ attention, but was dropped from the SC version since having 5 other control conditions was deemed sufficient. CNAX used a parallel set up and syntactic construction as the other conditions, but the action in the picture did not match the action in the question. For example, the question ‘Is Cinderella washing Goldilocks?’, was accompanied by a picture showing Cinderella drying Goldilocks. There was no match version of this condition. Both the group with DS and the controls did well on this condition, over 90% correct.
A cross-linguistic analysis of binding in Down syndrome
four adult non-impaired controls were recruited, matched to participants with DS in age and sex; their performance on all experimental and control conditions was at ceiling. Results for the group with DS and the control group on experimental match conditions are given in Figure 6, and experimental mismatch conditions in Figure 7. The results obtained reveal a number of clear patterns. Typically developing children showed an excellent performance on all the experimental (and control) conditions except for the conditions NPM and NPX, where they reached between 75 and 80% correct, respectively. The difficulty with this condition can be interpreted as the remnants of ‘Delay of Principle B Effect’, mentioned in section 1 earlier. 100
% correct
80 60
DS TD
40 20 0
NPM
NRM
QPM
QRM
Figure 6. Percentage correct on experimental match conditions: English groups. 100
% correct
80 60
DS TD
40 20 0
NPX
NRX
QPX
QRX
Figure 7. Percentage correct on experimental mismatch conditions: English groups.
Alexandra Perovic
In the data presented by DS participants, a strikingly different pattern emerged. On the experimental match and mismatch conditions involving pronouns, in a local relation with either a referential or quantified antecedent (NPM, NPX, QPM, QPX), they performed near or at ceiling level. They correctly rejected locally bound pronouns in the mismatch condition and accepted a referent distinct from the local subject for the pronoun in the match condition. This performance is in fact slightly better than that of controls’. In contrast, on conditions involving reflexive pronouns, all four participants performed below, or around chance – both in referential and quantified contexts. Their mean performance on NRM reaches 59%, but dips to 25% on the quantified version of this condition, QRM. Similarly, on the mismatch conditions, they reach 56% correct on NRX and 47% on QRX. A mixed repeated measures ANOVA with sentence type as within subjects factor (14 levels: 8 experimental conditions and 6 control conditions) and group as between subjects factor (DS, controls) revealed a significant main effect of test condition/sentence type (F(13,78) = 29.13, p < 0.001), an effect of group approaching significance (F(1,6) = 5.88, p = 0.051) and a significant group X sentence type interaction (F(13,78) = 5.89, p = < 0.001). A subsequent ANOVA was run with all six match conditions (NRM, QRM, NPM, QPM, CNNM, CQNM) as within subjects factors and group (DS, controls) as between subjects factor. A significant main effect of test condition/sentence type (F(5,30) = 10.19, p < 0.001), group (F(1,6) = 10.6, p = 0.017), and a significant interaction between group and test conditions (F(5,30) = 17.11 p < 0.001) were revealed. Another mixed repeated measures ANOVA, run with mismatch conditions as within subjects factor and group as between subjects factor, showed a significant main effect of test condition (F(7,42) = 35.85, p < 0.001), but not of group (F(1,6) = 2.33, p < 0.177), although the interaction between the two was significant (F(7,42) = 3.29, p = 0.007). Independent samples t-test confirmed that the participants with DS performed significantly worse than controls on the match conditions involving reflexives, NRM (t(6 = –3.43, p = 0.014)) and QRM (t(6) = –8.48, p < 0.001). The tests did not confirm the observed differences between the groups on other experimental conditions to be statistically significant. However, the difference between the DS group and controls on NRX (47% correct for the group with DS vs. 80% correct for the controls) approached significance: (t(6) = –2.06, p = 0.085). The difference between the groups on NPX also approached significance (t(6) = 2.19, p = 0.071), where the participants with DS reached 100% in comparison to 84% correct for controls.
6. Discussion The results reported above revealed that Serbo-Croatian-speaking participants with DS achieved high scores on a number of test conditions. No differences on the control conditions were found between the participants with DS and matched TD controls. Similarly, on all match conditions, the scores of participants with DS did not statistically
A cross-linguistic analysis of binding in Down syndrome
differ from those of controls. They showed a good understanding of pronouns and a (relatively) good understanding of pronominal clitics and the reflexive clitic on both match and mismatch conditions. However, their performance on mismatch conditions involving the anaphor ‘sebe’ revealed a considerable difficulty with the interpretation of this element. How do these data fare with regard to the predictions on the knowledge of binding in DS made in section 3.1 above? Recall that the predictions stated that the knowledge of the purely syntactic component of anaphoric binding is impaired in DS, but that the conditions on reflexivity are available. Knowledge of Condition A is seen in participants’ good performance (80% correct and higher) on conditions NRMcl, QRMcl, NRXcl and QRXcl, indicating that they are able to correctly interpret constructions containing inherently reflexive predicates accompanied by the reflexive clitic ‘se’ as reflexive. With regard to their knowledge of Condition B and the Chain Condition, like the English speaker with DS, SC-speaking participants correctly ruled out co valuation between a pronoun and a local antecedent on the name-pronoun mismatch (NPX) and quantifier-pronoun mismatch (QPX) over 90% of the time. The third prediction stated that a deficit in anaphoric binding, if present, should show up as difficulties in interpreting the anaphor ‘sebe’. The match and mismatch conditions are equally important here as both require participants to construct an anaphoric binding relation in order to interpret the structure. Condition A is also involved, but participants need to be able to establish the relation of anaphoric binding first in order to be able to reflexive-mark the predicate. The Chain Condition rules all the constructions in, as the anaphor is referentially defective [–R]. The participants’ poor performance on the mismatch conditions involving the reflexive ‘sebe’ supports this prediction: SC-speaking participants showed a mean score of 60% for the name-reflexive mismatch (NRX) and 66% correct for the quantifier-reflexive mismatch (QRX).21 However, their performance is faultless on both match conditions involving the reflexive. How can this be explained? Recall that English participants showed difficulties interpreting anaphors in both match and mismatch conditions. At first sight, the SC-speaking participants’ high scores on the match conditions involving the anaphor ‘sebe’ may seem as if the deficit in anaphoric binding is not present in SC-speaking individuals with DS to the same extent as it is in the English population. We argue that the parallel deficit in the knowledge of anaphoric binding is present in both English and SC-speaking population with DS, however, SC-speaking participants’ high scores on the match conditions will be shown to be a result of a particular strategy the participants use in order to give some
21. Taking into an account the individual variation showed in some but not all the conditions, the deficit in anaphoric binding proposed in the grammar of DS is expected to show its effects on a continuum: it can show up to a greater extent in some individuals, and to a lesser extent in others. As with the English participants, there was a much greater variation within the anaphor conditions than on any other experimental condition.
Alexandra Perovic
interpretation to the (otherwise uninterpretable) anaphor. We discuss these strategies in section later, in 6.1. The pattern observed amongst our participants with DS has recently been replicated in two groups of English speaking teenagers with DS (Ring & Clahsen, 2005). Using the experimental paradigm of van der Lely & Stollwerck (1997) (also based on Chien & Wexler, 1990, but used with children with SLI), this study reports that participants with DS had more difficulties interpreting experimental questions containing reflexives, bound either by a referential or a quantified antecedent, than those containing pronouns in the same contexts. The SC-speaking group with DS showed some individual variation with regard to their performance on the mismatch experimental conditions. Two out of the six participants showed a very poor performance on pronominal clitics (at chance, or below chance for the NPXcl and QPXcl conditions). This pattern was not seen in TD controls. It is possible that these results show another area vulnerable to impairment in DS. Comprehension and production of pronominal object clitics is also deficient in other language-impaired populations, for instance SLI, in French (Jakubowicz et al 1998) and Greek (Varlokosta 2001). If clitics are viewed as functional heads that consist of pure grammatical features, difficulties with this element should be treated as symptomatic of a specific grammatical/morph osyntactic deficit, both in DS and SLI. This area certainly warrants further investigation. The results obtained in our studies have important implications for the accounts of language development in the population with DS. If the popular characterisation of linguistic development in DS in terms of the ‘delay, with no deviance’ hypothesis is correct, the acquisition patterns of particular linguistic modules in individuals with DS should be comparable to those found in TD children at some stage of language development. The results of the two studies reported above run against this prediction: both English and Serbo-Croatian speaking participants with DS showed an inordinate difficulty in the interpretation of reflexives, but not pronouns. This pattern was not observed in either of the control groups. It has not been documented at any stage of typical language development reported in the literature for English (Jakubowicz 1984; Wexler & Chien 1985; Chien & Wexler 1990; Grimshaw & Rosen 1990). For SC, none of the 37 TD children in Perovic (2004), from which the controls here were recruited, showed the pattern of poor performance on the reflexive pronoun, not even the youngest ones who were between three and four years old. These findings thus provide strong evidence against the characterisation of language development as severely delayed but essentially non-deviant.
6.1 Cross-linguistic differences in anaphor interpretation Being unable to link the anaphor to its syntactic antecedent, it is possible that participants would still attempt to give some interpretation to the anaphor. However, different strategies could have been adopted by speakers of English and SC, which may result in different patterns on the match and mismatch conditions involving anaphors. Unable
A cross-linguistic analysis of binding in Down syndrome
to interpret anaphoric elements, English-speaking participants could have treated the anaphor as a pronoun. As well as incorrectly accepting the non-local interpretation for the anaphor on mismatch conditions NRX and QRX, they also incorrectly rejected the local interpretation of the anaphor on match conditions NRM and QRM. This second pattern in fact reveals English participants’ knowledge of Condition B: assuming that they treat the anaphor as a pronoun, this pattern shows that the participants know that (a) pronouns are not reflexive-markers; and that (b) local co valuation of a pronoun and a referential antecedent must be ruled out. This strategy does not seem to have been employed by the SC-speaking participants: employing this strategy would result in a high percentage of negative answers to NRM and QRM, as these questions would have to be interpreted as violations of both Condition B and the Chain Condition. We saw that this was not the case. Why this difference? Crucially, English anaphors are complex SELF anaphors, consisting of both the ‘self ’ element and the pronoun marked for all phi features. As a result, English participants could have been ignoring the ‘self ’ element and merely paying attention to the pronoun (ruling it out in accordance with the Condition B and the Chain Condition). In contrast to English SELF anaphors, the SC anaphor is morphologically simplex. It does not contain a pronoun and is totally devoid of referential features. In addition, although marked for Case, it is not marked for Number, Person or Gender. These properties of ‘sebe’ make it impossible for the SC-speaking participants with DS to treat it as a pronoun. The lack of phi features of the SC anaphors is an important property of another type of pronominal elements that seem to have more in common with the SC anaphor than pronouns: logophoric elements. Underspecified for phi features like an anaphor, but able to refer to a non-local DP despite a legitimate antecedent DP in close proximity, like a pronoun, logophors are not subject to Condition B. They can appear in contexts involving a local legitimate antecedent DP but refer to a different, non-local antecedent. One lexical item that fits this description is the Japanese reflexive element ‘zibun’. It resembles SC ‘sebe’ in the sense that it is morphologically simple and subject oriented, but its distribution is not that of a classical anaphor. Crucially, it is able to appear in a local relation with a DP that is not its antecedent, in violation of Condition B: (23) Maryi-ga Annaj-ga zibun-oi,j semeta to omotta. Mary-NOM Anna-NOM self-ACC blamed COMP thought ‘Maryi thought Annaj blamed heri,j.’ (Hitoshi Shiraki, p.c.)
Thus one way of interpreting the pattern shown by the SC-speaking participants with DS could be that they treat the anaphor ‘sebe’ as if it were Japanese ‘zibun’: not subject to Condition B, ‘zibun’ need not be ruled out from contexts involving a local legitimate antecedent DP. Our participants rarely ruled out ‘sebe’ from contexts involving a local antecedent, thus showing a high percentage of (correct) ‘yes’ answers on the NRM and QRM conditions. Such an explanation of course relies on the assumption that anaphoric and logophoric binding are distinct grammatical dependencies. The issue to be addressed still is how this logophoric element is interpreted in the grammar of DS. It is likely that the participants would resort to some kind of strategy
Alexandra Perovic
in order to determine the reference for ‘sebe’. Their performance on the match and mismatch conditions discussed above provides a clue: it can be interpreted as a strategy where an antecedent is chosen on the basis of its accessibility in the discourse22 (in terms of hierarchy of the discourse prominence of the antecedent of Ariel, 1990). The participants select the most accessible antecedent – the one that has been provided in the experimental question. Recall that there were two possible referents provided in the experimental set up: the experimental picture depicted two characters, one performing the action and one standing by, both of which were introduced immediately before the test question is presented: ‘This is Snow White. This is Cinderella. Is Snow white drying herself?’ However, the nearest antecedent is given in the actual experimental question, either match or mismatch: (24) NRM: Da li Snežana sebe briše? comp prt.cl Snow White self drying ‘Is Snow White drying herself?’ (picture: SW drying herself) (25) NRX: Da li Petar Pan sebe briše? comp prt.cl Peter Pan self drying Is Peter Pan drying himself? (picture: PP drying X)
This strategy would explain both the high rate of incorrect ‘yes’ answers on the mismatch conditions, NRX and QRX, and the high rate of seemingly correct answers on the match conditions involving the anaphor, NRM and QRM: their acceptance of a local antecedent for the anaphor need not reveal the knowledge of anaphoric binding, but the same strategy of choosing the most accessible/discourse prominent antecedent. Our SC-speaking participants therefore reveal the same deficit in anaphoric binding as their English counterparts, but distinct grammatical properties of anaphors in SC and English force them to adopt different strategies in attempting to interpret the anaphoric element, resulting in slightly different performance patterns on our experimental task.
7. Conclusion This paper reports data obtained in the experiments on binding in the populations with DS that reveal a specific deficit in establishing the syntactic dependency of anaphoric
22. A strategy on the basis of linear proximity also seems possible, however, the current experimental setting only provided one possible antecedent. It is of course possible that the most accessible antecedent is the ‘nearest’ antecedent. To determine whether the participants are relying on some linearity strategy, examples similar to (i) could be tested:
(i)
The boy near the bear is washing himself.
A cross-linguistic analysis of binding in Down syndrome
binding. This was seen in speakers of two different languages, English and SC. The fact that the same deficit was observed in speakers of two distantly related languages adds considerable weight to our hypothesis about the knowledge of binding in DS. The data for the English participants with DS have recently been replicated by Ring & Clahsen (2005) who found a similar pattern in two groups of teenage children with DS. The main question of what exactly is the nature of the deficiency in the grammar of DS remains to be answered. More research is needed in order to determine whether there are other areas of grammar in DS that are deficient and how these deficiencies could be related to the impairment in establishing the binding relation observed in this study. Some hints have been provided by our participants’ performance on other complex syntactic structures: they all showed poor performance on passives, relative clauses and clefts on the TROG, for instance. The final point concerns the implications of the exposed deficit for the organisation of grammar in DS. The English and participants with DS employed different strategies in their attempt to interpret the anaphor, due to distinct grammatical properties of these lexical elements in the two languages. English-speaking participants sometimes interpreted the anaphor as a pronoun, while SC speakers with DS treated ‘sebe’ as if it were a logophoric element. The strategy used by the SC speakers is in fact in line with how other languages treat elements that are referentially defective, e.g., Japanese ‘zibun’. Note that the participants did not, at any point, violate a syntactic principle. The ‘deviance’ in their grammar only concerned a lack of a syntactic principle, but not its violation, suggesting that their grammar still seem to be constrained by the same principles as those seen in TD.
Acknowledgement This work is based on my Ph.D. thesis, completed in 2004 at the Department of Phonetics and Linguistics, University College London. I am indebted to my thesis supervisors, Professors Neil Smith and Ad Neeleman, and to all the children and adults who took part in the two studies in England and Yugoslavia. Thanks also to Peter Ackema, Sergey Avrutin, Karen Froud, Martina Graanin-Yuksek, Celia Jakubowicz, Vikki Janke, Marijana Marelj, Oystein Nilsen, Eric Reuland, Danijela Stojanović, Kriszta Szendroi, Nada Vasic and Ken Wexler for sharing their ideas and to two anonymous reviewers for very helpful comments on the previous version of this paper.
References Ariel, M. 1990. Accessing Noun-phrase Antecedents. London: Routledge. Avrutin, S. & Wexler, K. 1992. Development of Principle B in Russian: Coindexation at LF and coreference. Language Acquisition 2: 259–306.
Alexandra Perovic Baauw, S. 2000. Grammatical Features and the Acquisition of Reference: A comparative Study of Dutch and Spanish. PhD dissertation, Utrecht Institute of Linguistics. Bridges, A. & Smith, J. 1984. Syntactic comprehension in Down’s syndrome children. British Journal of Psychology 75: 187–196. Chien Y-C. & Wexler, K. 1990. Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition 1: 225–295. Chierchia, G. 1989. A semantics for unaccusatives and its syntactic consequences. Ms, Cornell University. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986. Knowledge of Language: Its Nature, Origin and Use. New York NY: Praeger. Clahsen, H. & Almazan, M. 1998. Syntax and morphology in Williams syndrome. Cognition 68: 167–198. Cornips, L. & Hulk, A. 1999. Affected Objects in Heerlen Dutch and Romance. Languages in Contrast 1(2): 191–210. Diesing, M. & Jelinek, E. 1995. Distributing arguments. Natural Language Semantics 3: 123–176. Dodd, B. 1976. A comparison of the phonological system of mental age matched normal, severely subnormal and Down’s syndrome children. British Journal of Disorders of Communication 11: 27–42. Everaert, M. 1986. The Syntax of Reflexivization. Dordrecht: Foris. Fowler, A. 1988. Determinants of rate of language growth in children with Down syndrome. In The Psychobiology of Down Syndrome, L. Nadel (ed.). Cambridge MA: The MIT Press. Fowler, A. 1990. Language abilities in children with Down syndrome: Evidence for a specific delay. In Children with Down Syndrome: A Developmental Perspective, D. Cicchetti & M. Beeghley (eds). Cambridge: CUP. Franks S. & Holloway King, T. 2000. A Handbook of Slavic Clitics. Oxford: OUP. Grimshaw J. & Rosen, S. 1990. Knowledge and obedience: The developmental status of the Binding Theory. Linguistic Inquiry 21: 187–222. Grodzinsky, Y. & Reinhart, T. 1993. The innateness of binding and coreference. Linguistic Inquiry 24: 69–101. Halpern, A. & Fontana, J. 1994. X0 and Xmax clitics. In Proceedings of the Twelfth WCCFL, E. Duncan & D. Farkas (eds), 251–266. Stanford CA: CSLI. Hestvik, A. & Philip, W. 2000. Binding and coreference in Norwegian child language. Language Acquisition 8(3): 171–235. Holmberg, A. 1986. Word Order and Syntactic Features in Scandinavian Languages and English. PhD dissertation, University of Stockholm. Jakubowicz, C. 1984. On markedness and binding principles. In NELS 14, C. Jones & P. Sells (eds). Amherst MA: University of Massachusetts. Jakubowicz, C., Nash, L., Rigaut, C. & Gerard, C-L. (1998) Determiners and clitic pronouns in French-speaking children with SLI. Language Acquisition 7: 113–160. Laws, G. & Bishop, D. 2003. A comparison of language impairment in Down syndrome and Specific Language Impairment. Journal of Speech, Language, and Hearing Research 46: 1324–1339. Marelj, M. 2004. Middles and Argument Structure Across Languages. PhD dissertation, University of Utrecht. Mišeska-Tomić, O. 1996. The Balkan Slavic clausal clitics. Natural Language and Linguistic Theory 14: 811–872.
A cross-linguistic analysis of binding in Down syndrome Perovic, A. 2002. Knowledge of binding in Down syndrome. In Proceedings of the 26th Annual Boston University Conference on Language Development, B. Skarabela, S. Fish & A. H.-J. Do (eds). Somerville MA: Cascadilla. Perovic, A. 2004. Knowledge of Binding in Down Syndrome: Evidence from English and SerboCroatian. PhD dissertation, University College London. Philip, W. & Coopmans, P. 1996. The double Dutch Delay of Principle B Effect. In Proceedings of the 20th Annual Boston University Conference on Language Development, A. Stringfellow, D. Cahana-Amitay, E. Hughes & A. Zukowski (eds). Somerville MA: Cascadilla. Progovac, L. 1998. Determiner phrase in a language without determiners. Journal of Linguistics 34: 165–179. Reinhart, T. 1983. Anaphora and Semantic Interpretation. London: Croom Helm. Reinhart, T. 1996. Syntactic effect of lexical operations: Reflexives and unaccusatives. UiL-OTS Working Papers. Reinhart, T. & Reuland, E. 1993. Reflexivity. Linguistic Inquiry 24: 657–720 Reinhart, T. & Siloni, T. 2005. The syntax-lexicon parameter. Linguistic Inquiry 36: 389–436. Ring, M. & Clahsen, H. 2005. Distinct patterns of language impairment in Down syndrome, Williams syndrome and SLI: The case of syntactic chains. Journal of Neurolinguistics 19: 479–501. Rivero, M. 2000. On impersonal sie in Polish: A simplex expression anaphor. Journal of Slavic Linguistics 8(1): 199–237. Rivero, M. & Goledzinowska, M. 2002. The acquisition of constructions with reflexive clitic in Polish. Cahiers Linguistiques d’Ottawa 30: 60–90. Rondal, J. 1995. Exceptional Language Development in Down Syndrome. Cambridge: CUP. Rutter, T. & Buckley, S. 1994. The acquisition of grammatical morphemes in children with Down’s syndrome. Down Syndrome Research and Practice 2:76–82. Sigurjónsdóttir, S. 1992. Binding in Icelandic: Evidence from Language Acquisition. PhD dissertation, University of California. Stojanović, D. 1997. Object shift in Serbo-Croatian. In Clitics, pronouns and movement [Current Issues in Linguistic Theory 40], J. Black & V. Montapanye (eds). Amsterdam: John Benjamins. Stojanović, D. 2002. The Acquisition of Pronominal Categories. Ms. University of Ottawa. van der Lely, H.K.J. & Stollwerck, L. 1997. Binding Theory and specifically language impaired children. Cognition 62: 245–290. Wehrli, E. 1986. On some properties of French clitic se. In Syntax and Semantics 19: The Syntax of Pronominal Clitics, H. Borer (ed.). Academic Press. Wexler, K. & Chien, Y-C. 1985. The development of lexical anaphors and pronouns. Papers and Reports on Child Language Development 24: 138–149. Wilder, C. & Ćavar, D. 1994. Long head movement? Verb movement and cliticization in Croatian. Lingua 93: 1–58.
Balanced bilingual children with two weak languages A French/German case study Natascha Müller & Antje Pillunat Research on bilingual first language acquisition has reported on bilingual children who develop the two languages at an equal pace and like monolingual children. Recent research also includes children with a dominant/weak language; for these children Müller & Kupisch (2003) have shown that the weak language develops in a qualitatively similar way to the languages in balanced bilinguals or the strong/dominant language in unbalanced bilinguals, the difference being only quantitative in nature. The weak language exhibits a delay in the development of grammatical phenomena. The present study investigates a bilingual German/ French child with two weak languages. The general result is that in a child with two weak languages delay effects can be observed to the same extent as in normally developing bilingual children. However, contrary to what other researchers have found in normally developing bilingual children, acceleration effects are absent in a child with two weak languages. Both findings are obtained on the basis of an age-matched and an MLU-matched comparison of three bilingual children. The grammatical domains investigated are determiner omissions and the realization of subjects and objects.
1. Introduction Research on bilingual first language acquisition has often studied balanced bilingual children, i.e., children who develop their two languages at an equal pace. Some researchers, for example. Schlyter (1993, 1994, 1995), have investigated bilingual children with one weak language, i.e., children who develop one language at a slower rate than the other one. Recently, Müller & Kupisch (2003) and Müller (2004) have shown that the weak language in bilingual children develops qualitatively like the languages in balanced bilinguals or the strong/dominant language in unbalanced bilinguals. In essence, they show that a weak language exhibits the qualitative characteristics of bilingual first language acquisition to the same degree as a strong language. Balanced bilingualism is diagnosed on the basis of the comparison of the two languages in the bilingual. If both languages develop at an equal pace (no or minimal distance between the
Natascha Müller & Antje Pillunat
two languages), researchers speak of balanced bilingualism. If one language develops faster/slower than the other, researchers use the term unbalanced bilingualism or language dominance (articulated language distance). The concept of dominance, although still insufficiently defined, has become a major explanatory factor for cross-linguistic influence as observed in bilingual children. It seems to be generally accepted that unbalanced bilingualism is the reason for cross-linguistic influence, i.e., the strong language influences the weak one. Müller; Kupisch; Schmitz & Cantone (2006) present and discuss evidence against this view: (1) Balanced bilinguals also exhibit cross-linguistic influence, (2) the same grammatical phenomena in the same languages are affected by the influence, i.e., for the grammatical phenomenon A language X and for the grammatical phenomenon B language Y, (3) the weak language can influence the strong one (Kupisch 2004), (4) the influence can take place during the same stage of development into language X for the grammatical phenomenon A and into language Y for the grammatical phenomenon B. The authors conclude that language dominance should be studied separately from cross-linguistic influence and that both are unrelated. Determining the weak language opens another dimension though, namely the comparison with monolingual children or with a norm established by pooling the data from many bilinguals for one language. Most researchers suppose that the weak language in a bilingual child does not develop like the language of a monolingual child, i.e., the acquisition process is delayed not only with respect to the other (the dominant) language but also with respect to monolingual peers. The present study will enable us to separate the two perspectives, the comparison of the two languages in the individual (language distance) and the comparison of bilingual and monolingual children (language norm), since the focus will be a balanced bilingual child with two weak languages. We will investigate the development of three grammatical domains in the bilingual German/French child Caroline. These grammatical domains are: determiners, the realization of subjects and that of objects – all well-studied grammatical domains in early bilingualism and domains which differ as to whether they show signs of crosslinguistic influence and as to which language is affected by the influence. Caroline is an interesting case since she is a rather balanced bilingual child. However, both her languages are weak if compared to other bilingual and monolingual children. Caroline’s acquisition of the three grammatical domains will be compared to that of two other German/French bilingual children. One of the children, Céline, speaks French as a weak language, “weak” being defined both on the basis of a comparison with her German and with monolingual French children. Her German, if compared to monolinguals, is well within the norm, as observed by Müller; Kupisch; Schmitz & Cantone (2006). The other child, Alexander, is a fairly balanced bilingual child with a slight preference for French. Both languages, however, range within the norm, as shown by Cantone, Kupisch, Müller & Schmitz (2008). Since Caroline will be shown to develop her two languages with a delay, the comparison between her and Céline will be particularly revealing. The present study will discuss the possibility that a delay in both of
Balanced bilingual children
the two languages may give rise to properties in grammatical development as observed for monolingual children with a language disorder. Section 2 summarizes the main characteristics of the adult systems with respect to the three grammatical domains studied. Since the main idea is to study grammatical phenomena which differ as to their degree of cross-linguistic influence, the presentation has to restrict itself to those characteristics being most important. Section 3 summarizes the acquisition literature. Again, the presentation of the different grammatical domains in acquisition is far from being complete. In the fourth section, the empirical results are presented. First, the three bilingual children will be characterized as to their language distance and the languages will be measured across the different individuals. Second, the different grammatical domains will be presented with a focus on those languages which show cross-linguistic influence in other bilingual children. Section 5 discusses the observations in relation to bilingual first language acquisition and language disorder.
2. The adult systems The following sections will give a short overview of the three grammatical domains in the two adult systems.
2.1 The realization of the subject and object position in French and German Both languages, French and German, can choose between several options to realize subjects and objects, depending on pragmatic factors and on the grammatical status of the particular categories provided by the language. Where pragmatic factors control the realizations, we expect no difference between the languages. This is the case for the use of pronouns versus DPs. The subject position, mostly conveying old information, is rarely realized as a DP, whereas objects – generally representing new information – are frequently realized as full DPs. This contrast is not language-specific as shown in figure (1) taken from Schmitz & Müller (2004).1 In both languages, omissions are rare and identified via pragmatics.2 If they occur, the omitted element must be the topic. 1. Schmitz & Müller (2004) looked at a corpus of spontaneous interactions by French and German adults with children (i.e., input data). All data come from monolingual adults who interacted with children aged 3 to 4 years. We suppose that the adults do not adapt their speech to the children at that age. For French, 2168 finite sentences with a subject and 1072 finite sentences with an object were analyzed. In German, the number of analyzed sentences for subjects amounts to 2204, for objects to 1648. Unintelligible or interrupted utterances as well as imperative clauses were not counted. 2. Omissions are however licensed differently in the two languages. French licenses empty subjects and objects only in situ, whereas in German, the omitted element has to be moved to the front of the clause, having the effect that only one and the first (topicalized) element per clause can be omitted.
Natascha Müller & Antje Pillunat
(1) a. b.
Voulez-vous que je vous donne mon numéro de téléphone? Want-you that I you give my number of telephone ‘Do you want me to give you my telephone number?’ Non, je connais déjà __ . No, I know already ‘No, I already know it.’ Möchten Sie, dass ich Ihnen meine Telefonnummer gebe? Want-you that I you my telephone number give ‘Do you want me to give you my telephone number ?’ ‘Nein, __ kenne ich bereits.’ No, know I already ‘No, I already know it.’
Subject omissions in French are further restricted since they can only occur with impersonal verbs, i.e., in constructions in which the expletive il ‘it’ has been omitted: il faut chanter ‘one has-to sing’ - ø faut chanter. The main difference between French and German is due to the inventory of pronominal functional categories. French has clitic pronouns both in subject position (like je) and in object position (like me) whereas German does not have clitics for neither position,3 but a weak pronoun es.4 As figure (1) shows, subject clitics amount to 85% in French. German uses strong pronouns (der, die, das, sie, etc.) with a similar frequency. Object clitics amount to 35% in French. The amount of realizations as clitics and strong pronouns (like moi) for objects in French is similar to that of weak and strong pronouns for objects in German. Thus, the general result is that the two languages do not differ as to whether subjects and objects are realized as full DPs or
percentage
100
omission
80
DP
60
clitic/weak
40
strong pron
20 0
Subject Fr.
Subject Ger.
Object Fr.
Object Ger.
Figure 1. Realization (in %) of subject and object position in French and German. 3. It is generally assumed that phonologically reduced pronouns like ’n for ihn ‘it’ (accusative Case) in German do not figure as syntactic clitics. 4. According to Cardinaletti & Starke (1999), pronouns are devided into three different classes, strong, weak and clitic pronouns. German es figures as a weak pronoun according to their typology.
Balanced bilingual children
pronouns or if they are being omitted (with the exception of subjects which can only be omitted with impersonal verbs in French). The difference is related to the usage of the particular functional elements provided by each language.5
2.2 Bare nouns in French and German Kupisch (2006) investigated the acquisition of determiners in bilingual children and reviewed the literature on the adult systems. French is generally treated as a language with obligatory determiner realization, i.e., nouns have to be preceded by a determiner. German, on the other hand, is a language which has well-defined contexts in which the determiner must be absent, like with generic noun phrases of the type bäume können sehr alt werden ‘trees can become very old’. Proper names are undetermined in both languages. Kupisch also analyzed a corpus of monolingual adults and found out that bare nouns – nouns which are not preceded by a determiner – are rare in French (5%) and amount to about 17% in German.6 She also differentiated between articles (Art) – definite and indefinite – and other determining elements (Det) like demonstratives, possessives etc. Figure 2, taken from Kupisch (2006), shows the results for the adult systems. To summarize, articles are more frequent in French than in German.
percentage
100
bare nouns
80
proper names
60
Det+N
40
Art+N
20 0
French
German
Figure 2. (Un)determined nouns in French and German.
3. The acquisition literature Research on the development of pronouns in French and other Romance languages has shown that object clitics are acquired with much more effort than other 5. There are other differences which are however not important for the presentation of the acquisition data, e.g., the syntactic difference that only the first constituent can be omitted in German (topic drop), thus giving rise to a verb-initial structure. This excludes topic drop in German subordinate clauses. 6. The base is 13 recordings with 30 minutes each, the absolute number of nouns is at least 2500 in each language.
Natascha Müller & Antje Pillunat
syntactic categories. Object clitics are realized later in development than subject clitics and strong subject and object pronouns. Research on early child French has shown that object clitics are omitted until the age of approximately 2;4 (Cipriani, Chilosi, Bottari & Pfanner 1993; Clark 1986; Guasti 1993/1994; Hamann 2002, Hamann, Rizzi & Frauenfelder 1994, 1996; Jakubowicz, Müller, Kang, Riemer & Rigaut 1996; Jakubowicz, Müller, Riemer & Rigaut 1997, Jakubowicz & Rigaut 2000, Schmitz & Müller 2004). Both the delay of object clitics and the subject-object asymmetry between the different types of clitics are robust across individual learners and acquisition types (L1, 2L1, early child L2, and SLI7; see Hulk 1997; Kaiser 1994; Müller, Crysmann & Kaiser 1996; Müller 2002, 2004; Müller, Cantone, Kupisch & Schmitz 2002; Müller & Hulk 2000, 2001; Müller, Hulk & Jakubowicz 1999; Müller & Kupisch 2003, 2006; and Schmitz & Müller 2004 for 2L1, White 1996, Grondin & White 1996, Prévost 1997 for early child L2, Hamann et al. 2002, Jakubowicz, Nash, Rigaut & Gérard 1998; and Paradis, Crago & Genesee 2003 for SLI). Object clitics have been proposed as a clinical marker for SLI. Jakubowicz et al. (1997) and Hamann (2002) observe (although for different ages) that German children leave out subjects and objects to a very high degree (cf. also Müller & Hulk 2001). For children between the ages of 2 to 3, omissions amount to about 40%. In contrast to French, however, no asymmetry between subjects and objects has been found, i.e., German children omit arguments as much in subject as in object position. Whether German children with SLI acquire this domain like normally developing children is unclear. Since there do not exist any studies which particularly focus on argument omissions in SLI children, it is reasonable to believe that these children, like the normally developing ones, omit arguments and that no asymmetry is to be found between subjects and objects (Cornelia Hamann, p.c.). Determiners – especially in French – are acquired fast and effortlessly, compared with subject clitics, if we look at their realization in contexts where their presence is obligatory (e.g., Pizzutto & Caselli 1992; Bottari et al. 1993/1994; Antelmi 1997; Chierchia et al. 1999; Kupisch 2006 for Italian, Jakubowicz et al. 1998, Chierchia et al. 1999, Van der Velde et al. 2002, Kupisch 2006 for French). Jakubowicz and her research team have observed that French children with SLI do not have problems with determiners in obligatory contexts. Thus, the effortless acquisition of determiners in French is robust across individual learners and acquisition types. German children have been shown to omit determiners (in a target-deviant fashion) frequently and for a much longer time than French children. In a study which compares Romance languages with German, Chierchia, Guasti & Gualmini (1999) conclude that German children are the latest ones to converge with the adult system. Looking at German children with SLI, it may again be deduced from the fact that there are no particular studies on determiners that these 7. We will not discuss the different explanations which have been advanced in the literature for SLI. In the present study, we are interested in the existence of similar properties in the grammatical development of children with SLI and the bilingual child with two weak languages.
Balanced bilingual children
children omit these functional elements just like their normally developing peers and that they, as well, converge late to the adult system (Cornelia Hamann, p.c.).
4. Bilingual first language acquisition with two delayed languages 4.1 The children under investigation: language balance and language norm The children analyzed in the present study have all been bilingual from birth (German/ French). They were born in Hamburg and raised following the principle one person one language.8 Alexander and Caroline have a French mother and a German father, Céline has a German mother and a French father. Caroline will be the main interest of the following study. Her language development is compared with that of Céline and that of Alexander (cf. Cantone, Kupisch, Müller & Schmitz 2008). In order to measure the languages of the bilingual children, two perspectives should be taken: the comparison of the two languages in the bilingual individual and the comparison of one language across individuals (monolinguals included). Due to space limitations, we will not be able to measure Caroline’s two languages with respect to a norm, but since Cantone, Kupisch, Müller & Schmitz (2008) have measured Alexander’s and Céline’s languages, a comparison across individuals will be revealing. Figure 3 shows Alexander’s MLU values from the first recording at age 2;2,8 until the age of 4. The figure illustrates that his French MLU is slightly higher at all points in
7.0
German French
6.0
MLU
5.0 4.0 3.0 2.0 1.0 3;8,11
3;6,21
3;5,24
3;4,19
3;3,22
3;2,16
3;1,22
2;11,6
2;10,2
2;8,28
2;7,27
2;6,25
2;4,6
2;5,25
2;2,27
2;2,6
0.0
age Figure 3. MLU: Alexander.
8. The project has been financed by a grant given to Natascha Müller by the Deutsche Forschungsgemeinschaft.
Natascha Müller & Antje Pillunat German French
300 250 200 150 100
3;8,11
3;6,21
3;5,24
3;4,19
3;3,22
3;2,16
3;1,22
2;11,6
2;10,2
2;8,28
2;7,27
2;6,25
2;5,25
2;4,6
0
2;2,27
50 2;2,6
number of utterances
350
age Figure 4. Number of utterances: Alexander.
his development as compared to the values reached in his German. Therefore, we may conclude that Alexander is a fairly balanced child with a slight preference for French. Cantone, Kupisch, Müller & Schmitz 2008 and Müller & Kupisch (2003) use other criteria, such as the number of verb types, in addition to MLU in order to measure language dominance or language balancy. All these criteria are qualitative in nature, i.e., they are interpreted as criteria which reflect the pace of development of qualitative aspects in the respective languages. Following Müller & Kupisch (2003), we have also counted the number of utterances used in each language in each recording in order to determine whether the bilingual children have a higher readiness to use one of the two languages. The absolute number of utterances is interpreted as one of the criteria which measure readiness to speak a language. Figure 4 shows that Alexander uses both languages to a similar extent during the recordings. Cantone, Kupisch, Müller & Schmitz (2008) show that both of Alexander’s languages are well within the norm. Therefore, Alexander can be regarded as a balanced bilingual child who develops both languages within the norm and who likes to use both languages equally frequent. In contrast to Alexander, Céline has a dominant language, namely German. Her French MLU values differ considerably from those in German, as can be seen in figure 5. In some recordings, the difference between the two languages amounts to more than 1. The recordings started at age 2;0,9 and the MLU was counted until the age of 4. Again contrasting with Alexander, Céline does not use French a lot until the age of 3;3, as graphically shown in figure 6. She is much more ready to speak German than to speak French. Cantone, Kupisch, Müller & Schmitz (2008) and Müller, Kupisch, Schmitz & Cantone (2006) show that Céline’s German ranges well within the norm (for some
Balanced bilingual children 6.0
German French
5.0 MLU
4.0 3.0 2.0 1.0 3;7,17
3;6,12
3;5,15
3;4,9
3;3,12
3;0,27
2;11,29
2;11,3
2;10,5
2;8,2
2;8,29
2;6,21
2;5,24
2;4,19
2;1,6
2;3,15
2;0,9
0.0
age
3;8,0
3;6,12
3;4,23
3;3,12
3;0,13
2;11,3
2;9,20
2;8,2
2;6,7
2;4,19
German French
2;1,14
500 450 400 350 300 250 200 150 100 50 0
2;0,9
number of utterances
Figure 5. MLU: Céline.
age Figure 6. Number of utterances: Céline.
criteria such as MLU her German is above the norm) whereas her French is much below it. Céline can be categorized as an unbalanced bilingual child who develops one language much slower than the other one and who also does not like to use the slower developing language in spontaneous conversation. In a study on the acquisition of determiners and (subject / object) clitic pronouns in French, Müller & Kupisch (2003) have shown that Céline’s development of these grammatical domains proceeds in a similar way as compared with other bilingual children. However, the development is delayed in French. The interesting finding was that there were no qualitative differences between Céline’s acquisition of determiners and clitics in French when she was compared with Alexander and other French/German bilinguals.
Natascha Müller & Antje Pillunat
The next child to be presented is Caroline. Caroline has been recorded from the age of 1;9 onwards. However, the first transcripts are available from age 2;8 onwards in French and from age 2;10 onwards (until 3;7) in German. The study ended at age 3;8. In her case, we would like to start with the figure which shows the absolute number of utterances per recording. Figure 7 clearly shows that Caroline is a talkative child and that she likes to speak both languages until the age of 3;3. However, during the end of the study, starting at age 3;3, the number of utterances in German nearly always stays below 100, which is very low as compared with Alexander and Céline. Her French does not decline as much as her German and one may argue that Caroline is more ready to speak French. If we look at Caroline’s MLU in figure 8, we can summarize that it oscillates between 2 and 3 in both languages during the time studied here. Such an MLU is German French
number of utterances
300 250 200 150 100 50 3;7,28
3;7,0
3;7,14
3;6,10
3;5,26
3;5,10
3;4,27
3;3,22
3;3,14
3;3,0
3;1,0
2;11,19 3;0,2
2;11,5
2;10,16
2;8,5
2;8,26
0
age Figure 7. Number of utterances: Caroline. 4
German French
3.5
MLU
3 2.5 2 1.5 1
age Figure 8. MLU: Caroline.
3;7,28
3;7,14
3;7,0
3;6,10
3;5,26
3;5,10
3;4,27
3;3,22
3;3,14
3;3,0
3;1,0
3;0,2
2;11,19
2;11,5
2;10,16
2;8,5
0
2;8,26
0.5
Balanced bilingual children
representative of Céline & Alexander until the age of 2;6 and 2;8 respectively. We might argue that Caroline is delayed in both languages, as Céline is in French. However, there is a qualitative difference between Caroline and the other two children: In Caroline, the MLU curve is not progressive. It is representative of a development which seems to stagnate at some point (MLU 3) for the period under investigation. The MLU curve reveals that Caroline is a balanced bilingual child. The following figures 9 for French and 10 for German compare the MLU of the three children. It becomes clear that Caroline’s French is as weak as Céline’s French for which we know that it does not conform to the norm. As for German, figure 10 shows that Caroline has the lowest MLU values if compared with Céline and Alexander. Remember that Alexander’s German has been classified as ranging within the norm. To summarize, Caroline is a balanced bilingual 6
Céline Alexander Caroline
5
MLU
4 3 2 1 3;10
3;8
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
2;0
0 age Figure 9. MLU all children: French. 6 Céline Alexander Caroline
5
MLU
4 3 2
age Figure 10. MLU all children: German.
3;10
3;8
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
0
2;0
1
Natascha Müller & Antje Pillunat
child with two weak languages. In what follows, we will look at the development of the grammatical domains in question.
4.2 Subject / object omissions in bilingual children: Having one or two weak languages does not matter Pillunat, Schmitz & Müller (2006) have studied subject omissions in bilingual French/ German children in relation to monolingual French and monolingual German children and have found that there was no cross-linguistic influence. The children under investigation developed each language like a respective monolingual child. In what follows, we will compare the development of subject omissions in French and German in the three bilingual children. All children omit subjects in both languages to a high extent not predicted by the two target systems. An example for a target-deviant subject omission in French and in German is given under (2). (2) a. b.
Éteinds la lumière, maman ! Switch-off the light, mom ! ‘I will switch off the light, mom !’ ‘J´éteinds la lumière, maman !’ Wo hab Katzi? Where have kitty? ‘Where do I have my kitty?’ ‘Wo habe ich mein Katzi?’
(Caroline; 3;4,27)
(Caroline, 3;7,0)
With respect to French subjects, an age-match comparison reveals that Caroline patterns like Céline, as can be seen in figure 11, which shows the percentage of target-deviant subject omissions for all three children. Both children, Caroline and Céline, stop to omit subjects at around age 3;8. 80
Caroline Alexander Céline
percentage
60 40 20
age Figure 11. Target-deviant subject omissions in French.
3;10
3;8
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
2;0
0
Balanced bilingual children
An MLU-matched comparison for French underlines the similarity in development of the three bilingual children. We may deduce from the comparison of figures 11 and 12 that although Céline and Caroline stop to use target-deviant subject omissions at a far more advanced age, they, just like Alexander, converge to the adult system at an MLU of 3/3.5. In German, Caroline is delayed with respect to subject omissions if compared with the other two children (figure 13) on the basis of age. This result was expected since we assumed that German is acquired by Caroline as a weak language. Interestingly, the MLU-matched comparison in figure 14 gives the same result for German as we have obtained for French, namely that Caroline behaves exactly like the other two bilingual children. 50
Caroline Alexander Céline
percentage
40 30 20 10 5–5,49
4,5–4,99
4–4,49
3,5–3,99
3–3,49
2,5–2,99
2–2,49
1,5–1,99
1–1,49
0
MLU Figure 12. Target-deviant subject omissions in French.
50
Caroline Alexander Céline
percentage
40 30 20 10
age Figure 13. Target-deviant subject omissions in German.
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
2;0
0
Natascha Müller & Antje Pillunat 40
Caroline Alexander Céline
percentage
30 20 10 0
1,5–1,99
2–2,49
2,5–2,99 3–3,49 MLU
3,5–3,99
4–4,49
Figure 14. Target-deviant subject omissions in German.
We may therefore conclude that a rather robust grammatical domain, such as the realization of the subject position, develops in a bilingual child with two weak languages in just the way it does in a balanced and an unbalanced bilingual child with one weak language. The property of a weak language, as already noticed by Müller & Kupisch (2003), is that it develops in a much more time-consuming fashion. If compared on an MLU-base, the weak language develops like any other language in early child bilingualism. Let us turn to objects. Müller & Hulk (2001) studied object omissions in bilingual and monolingual children and found that this grammatical domain is likely to be affected by cross-linguistic influence in bilinguals (for indirect objects cf. Schmitz 2006). Monolingual French children stop to omit objects around the age of 2;4–2;6, whereas monolingual German children still have about 40% of target-deviant object omissions until the age of 3. In the bilingual children studied by Müller & Hulk (2001), the Germanic language (German, Dutch) influences the Romance language (French, Italian) in the sense that object omissions occurred with a much higher frequency in the Romance language of the bilinguals than in the monolinguals (the number of targetdeviant object omissions in bilingual French is as high as observed for monolingual German children) and omissions were used by the bilinguals for a much longer time than by the monolinguals. In other words, the effect of cross-linguistic influence for this grammatical domain is a delayed development of the Romance language. Müller & Hulk (2001) defend the view that object omissions are prone to influence because object clitics represent a complex category in the acquisition process. Examples for target-deviant omissions in French and in German are given under (3). (3) a.
On va remettre. We will put-back.
[shoes]
Balanced bilingual children
b.
‘We will put them back.’ ‘On va les remettre.’ Ja, ich in [t]ule lernt. Yes, I in school learned. ‘Yes, I have learned that in school.’ ‘Ja, das habe ich in der Schule gelernt.’
(Caroline, 2;10,16)
(Caroline, 3;3)
All three bilingual children omit objects in a target-deviant way for a longer time and to a higher extent in French than observed with subjects and than exhibited by monolingual French children (Müller & Hulk 2001). Interestingly, Caroline again patterns like Céline, as can be seen in figure (15), which shows the percentage of target-deviant object omissions for all three children. Both children, Caroline and Céline, do not stop omitting objects until the last available recording. 80
Caroline Alexander Céline
70 percentage
60 50 40 30 20 10 3;10
3;8
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
2;0
0 age Figure 15. Target-deviant object omissions in French.
The MLU-based comparison of the three children in figure 16 shows a similar development for objects as for subjects, namely that all three children pattern the same. The number of target-deviant object omissions in bilingual French has dropped below 20% at an MLU of 3.0 / 3.5. In other words, after this MLU stage, there is still a lot to be acquired. In German, Caroline is delayed with respect to object omissions if compared with the other two children (figure 17) on an age-base. This result was expected since we assumed that German is acquired by Caroline as a weak language. Due to space limitations and since German is not affected by cross-linguistic influence for the grammatical domain of object omissions, we will not present an MLU comparison for German objects. We can summarize the results for objects as follows: A grammatical domain which is affected by cross-linguistic influence – delay for object omissions in French – develops in a bilingual child with two weak languages just as
Natascha Müller & Antje Pillunat 100
Caroline Alexander Céline
percentage
80 60 40 20 5–5,49
4,5–4,99
4–4,49
3,5–3,99
3–3,49
2,5–2,99
2–2,49
1,5–1,99
1–1,49
0
MLU Figure 16. Target-deviant object omissions in French. 100
Caroline Alexander Céline
percentage
80 60 40 20
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
2;0
0 age Figure 17. Target-deviant object omissions in German.
in a balanced and an unbalanced bilingual child with one weak language. If compared on an MLU-base, the weak language develops like a language which is in balance with the respective other language and which conforms to a pre-established norm in early child bilingualism.
4.3 Determiner omissions in bilingual children: Having two weak languages matters Kupisch (2006) studied Alexander and Céline for determiner omissions in French and German (cf. also Kupisch this volume). She also investigated the possibility of
Balanced bilingual children
cross-linguistic influence in the acquisition of determiner realization. For Alexander and, more interestingly for Céline, she found that their development of articles (use of articles in obligatory contexts) is accelerated in German, if one compares their development with that of monolingual peers. Thus, in German the children reach a point at which they cease to omit determiners in obligatory contexts earlier than monolingual German children. Kupisch hypothesizes that the French system positively influences the German one, the effect of cross-linguistic influence being acceleration. Remember that Céline’s French has been classified as a weak language. If Kupisch was right, a weak language (French) in a bilingual child could positively influence a strong language (German). Some examples for target-deviant determiner omissions in French and in German are given under (4). (4) a. b.
Crocodile est pas là. Crocodile is not there. ‘The crocodile is not there.’ ‘Le crocodile n’ est pas là.’ Telefon klingelt. Telephone rings. ‘The telephone rings.’ ‘Das Telefon klingelt.’
(Caroline, 3;0,2)
(Caroline, 3;5,10)
The following figures, figures 18 and 19, compare the development of target-deviant determiner omissions on an age-base in the three children (Alexander, Céline and Caroline) in French and in German. Alexander and Céline omit determiners in a target-deviant way in both languages in the beginning, that is around age 2 (omissions are target-deviant omissions in the figures). At 2;11, the children converge to the target-system: Alexander in both languages and Céline in German only. In French, 100
Caroline Alexander Céline
percentage
80 60 40 20
age Figure 18. Target-deviant determiner omissions in French.
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
2;0
0
Natascha Müller & Antje Pillunat 80
Caroline Alexander Céline
percentage
60 40 20
3;6
3;4
3;2
3;0
2;10
2;8
2;6
2;4
2;2
2;0
0 age Figure 19. Target-deviant determiner omissions in German.
target-deviant article omissions cease to occur at 3;6. Kupisch concludes that Céline’s development in French is delayed with respect to age. She does not show a different developmental path, however. Figures 18 and 19 show for the child Caroline that she does not reach the point of convergence to the adult system during the period of investigation in neither of her languages. We may therefore conclude that Caroline is delayed in both languages with respect to article realization. Let us now turn to Caroline’s German, the target of cross-linguistic influence in the case of article realizations. Figure 20 shows that her German is delayed for obligatory determiner realizations also if her development is compared with the accelerated acquisition process of the other two children on an MLU-basis. In other words, Caroline’s weak language – French – cannot positively influence her second weak language – German. From what Kupisch has concluded for Céline, we know that the absence of 80
Caroline Alexander Céline
percentage
60 40 20 0
1,5–1,99
2–2,49
2,5–2,99 3–3,49 MLU
3,5–3,99
Figure 20. Target-deviant determiner omissions in German.
4–4,49
Balanced bilingual children
an acceleration effect in Caroline’s German cannot be due to the fact that her French is weak. The fact that Caroline has two weak languages seems to be the important factor.
4.4 A link to SLI and second language acquisition?: Having two weak languages matters Let us very briefly turn to a grammatical domain for which Caroline’s development seems to differ qualitatively from that of the other bilingual children. Due to space limitations, we will have to concentrate on the realization of the subject position in French. In a recent article on a comparison between normally developing French monolinguals, French children with SLI and child second language learners of French, Paradis (2004) and Paradis & Crago (2004) found that the learner groups have different preferences as to how to supply the object if the clitic is not used (we have mentioned that usage of object clitics represents a problem for all learner groups). Interestingly, monolingual French children opt for lexical DPs (cf. also Müller, Schmitz, Cantone & Kupisch 2006), SLI children strongly prefer to omit the object, and second language learners use strong pronouns and ça ‘this’ instead of the clitic. If we consider the realization of the subject position in Alexander’s and Céline’s French (cf. Schmitz & Müller 2008; Schmitz 2006), subjects are most frequently realized as clitics. Since Céline’s French is delayed, the figure for French, figure 22, includes the time span until age 4. Let us turn to Caroline’s subjects. Figure 23 shows that she opts for subject omissions quite frequently, as we have shown in section 4.2. Furthermore, the figure reveals that a lot of subjects are realized as strong pronouns: toi manges la soupe ‘you eat the soup’. If we compare these results with the ones obtained by Paradis (2004) for objects, we may conclude that Caroline combines the two strategies which Paradis discusses
100
omission clitic pronoun strong pronoun DP
percentage
80 60 40 20
age Figure 21. Subjects: Alexander French.
3;1,22
2;11,6
2;10,2
2;8,28
2;7,27
2;6,25
2;5,25
2;4,6
2;2,27
2;2,6
0
Natascha Müller & Antje Pillunat 100
omission clitic
percentage
80
strong pronoun DP
60 40
3;11,29
3;11,1
3;10,4
3;8,28
3;8,0
3;6,26
3;5,29
3;4,23
3;3,26
0
2;0-3;0
20
age Figure 22. Subjects: Céline French.
100
omission clitic
percentage
80
strong pronoun
60
DP
40
3;7,28
3;7,0
3;5,26
3;4,27
3;3,14
3;10
2;11,19
2;10,16
0
2;8,5
20
age Figure 23. Subjects: Caroline French.
as specific for particular acquisition types: Like Paradis’ children with SLI, Caroline omits a lot of subjects. In addition, she uses strong pronouns in subject position, which is what Paradis found with respect to objects in child second language learners. Can Caroline’s pooling of strategies be said to be expected? As in a monolingual child with SLI, her French is much delayed in development and like a second language learner, Caroline acquires more than one language. Therefore the languages might influence one another. Space limitations do not allow us to either present her strategies to
Balanced bilingual children
supply objects in French or to show how she develops the subject and object position in German.
4.5 Summary of results Summarizing sections 4.2 and 4.3, we conclude that an otherwise robust or delayed development in balanced bilinguals with two languages within the norm and in unbalanced bilinguals with one weak language is mirrored in a bilingual child with two weak languages. The weak language can be defined as a language which shows delay effects for otherwise robust grammatical phenomena and which gives rise to an even more articulated delayed development on an age-based dimension. If an MLU dimension is taken into consideration, robust and delayed grammatical phenomena develop in a nearly identical fashion, also in a bilingual child with two weak languages. Having one weak language does not seem to matter for grammatical domains which are accelerated in bilinguals due to cross-linguistic influence; the weak language can positively influence the strong one with respect to that particular grammatical property. However, if both languages of the bilingual are weak, acceleration effects are absent for accelerated grammatical phenomena. The absence of accelerating effects is visible both on an age-based and on an MLU-based dimension. The general result is that in order for acceleration to be visible, a child needs to have at least one language which ranges within the norm. Put differently, at least one language has to develop like in a monolingual child in order to speed up the acquisition process in the bilingual child. Section 4.4 revealed that in terms of strategies in order to cope with problematic grammatical domains, a bilingual child with two weak languages may pool the different strategies observed in children with SLI and child second language learners. We will discuss the consequences of this observation in the discussion section.
5. Discussion The goal of the present research was rather ambitious: We have entered a still unexplored territory of early child bilingualism, namely the development of grammar in a balanced bilingual child with two weak languages. The grammatical phenomena studied in the present chapter differ with respect to whether or not they are candidates for cross-linguistic influence in bilingual children. The comparison of robust grammatical phenomena with grammatical properties which are sensitive to crosslinguistic influence has shown that a bilingual child with two weak languages displays cross-linguistic influence as far as vulnerable grammatical phenomena are concerned and lacks this kind of influence as regards robust domains. Research on monolingual children with SLI has revealed that some aspects of language are acquired in the same way as in monolingual normal children. The difference between children with SLI and normally developing children has been claimed to be of a quantitative nature (delay).
Natascha Müller & Antje Pillunat
Only recently, bilingual children with SLI have been studied with respect to the same aspects of language as monolingual normally developing children and children with SLI; Paradis et al. (2003). It seems to be the case that bilingual children with SLI develop morpho-syntactic aspects of language in a similar way as their bilingual normally developing peers, showing signs of cross-linguistic influence in the two languages. The comparison of early child bilingualism with two weak languages and bilingual children with SLI is particularly interesting, since both groups of children are probably severely delayed with respect to (age of acquisition of) certain aspects of morpho-syntax. We may therefore conclude that cross-linguistic influence is demonstrable for balanced bilinguals with two languages which range within a pre-established norm, unbalanced bilinguals, bilinguals with two weak languages and bilingual children with SLI. Put differently, the weakness of a language is completely unrelated to the occurrence of cross-linguistic influence. The study has also shown that one effect of cross-linguistic influence, namely acceleration, is absent in the bilingual child with two weak languages. Interestingly, acceleration effects are not absent in unbalanced bilinguals for whom one of the languages ranges within the norm. Very tentatively, we may conclude that for acceleration effects to occur, at least one of the two languages (not necessarily the influencing language for the grammatical phenomenon in question) has to develop “normally” in relation to other bilinguals and to monolingual children. Why is this result interesting for a theory of early child bilingualism? Future research should focus on accelerating effects in bilingualism, since the proof of such effects may help to exclude one reason for the occurrence of cross-linguistic influence in bilingual children often mentioned in the literature. Researchers have defended the view that cross-linguistic influence is due to limitations of the performance system of the bilingual child: Speaking two languages is more costly in terms of performance conditions and thus slows down the acquisition process in relation to monolinguals. The present analysis is in line with investigations which presuppose that the nature of the grammatical phenomenon is the reason for cross-linguistic influence. If future research succeeds in further substantiating the claim that acceleration effects are one possible outcome of cross-linguistic influence, this will prove it unlikely to attribute influence in early child bilingualism to the performance of more than one language. The child under investigation exhibited an age-based delay in the acquisition of some aspects of morpho-syntax. Is there a link between Caroline’s case and children with SLI/child second language learners? Caroline has not been diagnosed as an SLI child, nor can she be considered a second language learner, since she heard both languages from birth on. Some aspects of her development, especially the observation that she does not seem to be able to pass an MLU of 3 for quite a long period, fit into the picture of current SLI research. Additionally, Caroline has immense phonological problems and therefore is sometimes hard to understand. Furthermore, Caroline has been shown to omit functional categories in her utterances, a characteristic of the speech of children with SLI. Another parallel is that Caroline exhibited immense
Balanced bilingual children
problems with objects in French (a high number of omissions for a long period), which can be interpreted as problems with object clitics. On the other hand, Caroline’s development also showed parallels with child second language acquisition, namely the overuse of strong pronouns in subject position. The case of Caroline may lead to the conclusion that the contrast normally developing children, children with SLI, and child second language learners has to be rethought, at least if used as an explanation for different developmental paths. The case in question may lead to the conclusion that it is the weak language which gives rise to a different developmental path and that the reasons for why the language develops as a weak one are simply irrelevant for a language acquisition theory. This conclusion has to await more empirical evidence; certainly, the difficulty to find more individuals like Caroline will complicate the research situation. The weak language in bilinguals has also been argued to develop qualitatively like a second language acquired in adulthood, not like a second first language. Schlyter (1993, 1994, 1995) defended this view. Of course, this result rests on the assumption that a second language develops qualitatively different from a first language, a view which we do not support. However, our study of Caroline’s development has revealed particular properties which other researchers have found to be characteristic of child second language acquisition, like the overuse of strong pronouns in subject position. Schlyter did not differentiate between children with one weak language and another language which ranges within the norm and children with two weak languages which both do not range within the norm. It might be the case that only children like Caroline exhibit properties in their language development which are characteristic of child second language learners. It is unclear why this should be so, but it again strengthens the importance of future investigations of the weak language. Our study has distinguished an age-based and an MLU-based analysis of grammatical development in bilingual children. It has been shown that Caroline’s delay is always age-related, not MLU-related for those grammatical phenomena which developed in a similar way / with a delay in bilinguals as in monolinguals. Chronological age is a concept which is used in a lot of studies in order to explain differences among learners. The present study may contribute to a critical view on age as an explanatory concept for differences in language development.
References Antelmi, D. 1997. La Prima Grammatica Italiana. Bologna: Il Mulino. Bottari, P., Cipriani, P. & Chilosi, A.M. 1993/94. Protosyntactic devices in the acquisition of Italian free morphology. Language Acquisition 3: 327–369. Cardinaletti A. & Starke M. 1999. The typology of structural deficiency: A case study of the three classes of pronouns. In Clitics in the Languages of Europe, H. van Riemsdijk (ed.), 145–233. Berlin: Mouton de Gruyter. Chierchia, G., Guasti, M.T. & Gualmini, A. 1999. Nouns and articles in child grammar and the syntax/semantics map. Presentation at GALA 1999, Potsdam.
Natascha Müller & Antje Pillunat Cipriani, P., Chilosi, A.M., Bottari, P. & Pfanner, L. 1993. L’ acquisizione della morfosintassi in italiano: fasi e processi. Padua: UniPress. Clark, E. 1986. Acquisition of Romance, with special reference to French. In The Cross-linguistic Study of Language Acquisition, Vol.I: The Data, D.I. Slobin (ed.), 687–782. Hillsdale NJ: Lawrence Erlbaum Associates. Grondin, N. & White, L. 1996. Functional categories in child L2 acquisition of French. Language Acquisition 5: 1–34. Guasti, M.T. 1993/1994. Verb syntax in Italian child grammar. Finite and nonfinite verbs. Language Acquisition 3(1): 1–40. Hamann, C. 2002. From Syntax to Discourse. Pronominal Clitics, Null Subjects and Infinitives in Child Language. Dordrecht: Kluwer. Hamann, C., Rizzi, L. & Frauenfelder, U. 1994. On the acquisition of the pronominal system in French. Geneva Generative Papers 2(2): 91–103. Hamann, C., Rizzi, L. & Frauenfelder, U. 1996. On the acquisition of subject and object clitics in French. In Generative Perspectives on Language Acquisition. H. Clahsen (ed.), 309–334. Amsterdam: John Benjamins. Hamann, C. et al. 2002. Aspects of grammatical development in young French children with SLI. Ms, University of Geneva, Centre Hospilatier Universitaire de Lausanne, University of Sienna. Hulk, A. 1997. The acquisition of French object pronouns by a Dutch/French bilingual child. In Language Acquisition: Knowledge, Representation and Processing. Proceedings of the GALA 1997 conference on language acquisition, A. Sorace, C. Heycock & R. Shillcock (eds), 521–526. Edinburgh: University of Edinburgh. Jakubowicz, C., Müller, N., Kang, O.-K., Riemer, B. & Rigaut, C. 1996. On the acquisition of the pronominal system in French and German. In Proceedings of the 20th annual Boston university conference on language development, A. Springfellow, D. Cahana-Amitay, E. Hughes & A. Zukowski (eds), 374–385. Somerville MA: Cascadilla. Jakubowicz, C., Müller, N., Riemer, B. & Rigaut, C. 1997. The case of subject and object omissions in French and German. In Proceedings of the 21rst Annual Boston University Conference on Language Development, E. Hughes, M. Hughes & A. Greenhill (eds.), 331–342. Somerville MA: Cascadilla. Jakubowicz, C. Nash L., Rigaut C. & Gérard Ch.-L. 1998. Determiners and clitic pronouns in French-speaking children with SLI. Language Acquisition 7 : 113–160. Jakubowicz, C. & Rigaut, C. 2000. L ’acquistion des clitiques nominatifs et des clitiques objects en français. Canadian Journal of Linguistics 45(1–2): 119–157. Kaiser, G. 1994. More about INFL-ection: The acquisition of personal pronouns in French. In Bilingual First Language Acquisition. French and German Grammatical Development, J.M. Meisel (ed.), 131–159. Amsterdam: John Benjamins. Kupisch, T. 2006. The Acquisition of Determiners in Bilingual German-Italian and German/ French Children. Müncher: Lincom Europa Cantone, Kupisch, Müller, Schmitz 2008. Redefining language dominance in bilingual children. Linguistische Berichte 213 in press. Müller, N. 2002. Cross-linguistic influence in early child bilingualism: Italian/German. In EUROSLA Yearbook 2, S. Foster-Cohen, T. Ruthenberg & M.-L. Poschen (eds) 135–154. Amsterdam: John Benjamins. Müller, N. 2004. Null-arguments in bilingual children: French topics. In The Acquisition of French in Different Contexts. Focus on Functional Categories, J. Paradis & P. Prévost (eds), 275–304. Amsterdam: John Benjamins.
Balanced bilingual children Müller, N., Cantone, K., Kupisch T. & Schmitz, K. 2002. Zum Spracheneinfluss im bilingualen Erstspracherwerb: Italienisch – Deutsch. Linguistische Berichte 190: 157–206. Müller N., Crysmann B. & Kaiser G. 1996. Interactions between the acquisition of French object drop and the development of the C-system. Language Acquisition 5(1): 35–63. Müller, N. & Hulk A. 2000. Cross-linguistic influence in bilingual children: Object omissions and root infinitives. In Proceedings of the 24th Annual Boston University Conference on Language Development. C. Howell, S. A. Fish & T. Keith-Lucas (eds), 546–557. Somerville MA: Cascadilla. Müller, N. & Hulk, A. 2001. Cross-linguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition 4(1): 1–21. Müller, N., Hulk A. & Jakubowicz, C. 1999. Object omissions in bilingual children: Evidence for cross-linguistic influence. In Proceedings of the 23rd Annual Boston University Conference on Language Development, A. Greenhill, H. Littlefield & C. Tano (eds), 482–494. Somerville MA: Cascadilla. Müller, N. & Kupisch, T. 2003. Zum simultanen Erwerb des Deutschen und des Französischen bei (un)ausgeglichen bilingualen Kindern. Vox Romanica 62: 145–169. Müller, N. & Kupisch, T. 2006. The French DP in bilingual first and second language development: Proficiency versus age as windows to language acquisition. In Processes and Outcomes: Explaining Achievement in Language Learning, S. Haberzettl (ed.), Berlin: Mouton de Gruyter to appear. Müller, N., Kupisch, T., Schmitz, K. & Cantone, K. 2006. Einführung in die Mehrsprachigkeitsforschung. Tübingen: Narr. Müller, N., Schmitz, K., Cantone, K. & Kupisch, T. 2006. Null-arguments in monolingual children: A comparison of Italian and French. In The Acquisition of Romance Languages, V. Torrens & L. Escorbar (eds), Amsterdam: John Benjamins. 69–93 Paradis, J. 2004. The relevance of specific language impairment in understanding the role of transfer in second language acquisition. Applied Psycholinguistics 25: 67–82. Paradis, J. & Crago, M. 2004. Comparing L2 and SLI grammars in French: Focus on DP. In The Acquisition of French in Different Contexts, P. Prévost & J. Paradis (eds.), 89–107. Amsterdam: John Benjamins. Paradis, J., Crago, M. & Genesee, F. 2003. Object clitics as a clinical marker of SLI in French: Evidence from French-English bilingual children. In Proceedings of the 27th Annual Boston University Conference on Language Development, B. Beachley et al. (eds.), 638–649. Somerville MA: Cascadilla. Pillunat, A., Schmitz, K. & Müller, N. 2006 Die Schnittstelle Syntax-Pragmatile: Subjektauslas sungen bei bilingual deutsch-französisch aufwachsenden kindern. LiLi 143, 7–24. Pizzuto, E., & Caselli, M.C. 1992. The acquisition of Italian morphology: Implications for models of language development. Journal of Child Language 19: 491–557. Prévost, P. 1997. Truncation in Second Language Acquisition. PhD dissertation, McGill University, Montréal, Québec. Schlyter S. 1993. The weaker language in bilingual Swedish-French children. In Progression and Regression in Language, K. Hyltenstam & A. Viberg (eds), 289–308. Cambridge: Cambridge University Press. Schlyter, S. 1994. Early morphology in Swedish as the weaker language in French-Swedish bilingual children. Scandinavian Working Papers on Bilingualism 9 : 67–86. Schlyter, S. 1995. Formes verbales du passé dans des interactions en langue forte et en langue faible. .AILE 6: 129–152.
Natascha Müller & Antje Pillunat Schmitz, K. 2004. Zweisprachigkeit im Fokus: Der Erwerb der Verben mit zwei Objekten durch bilingual deutsch-französisch und deutsch-italienisch aufwachsende Kinder. Tübingen: Narr. Schmitz, K. 2006. Indirect objects and dative case in monolingual German and bilingual German/Romance language acquisition. In Datives and Other Cases. Between Argument Structure and Event Structures, D. Hole, A. Meinunger & W. Abraham (eds), 239–268. Amsterdam: John Benjamins. Schmitz, K. & Müller, N. 2008. Strong and clitic pronouns in the Bilingualism Language and cognition monolingual and bilingual acquisition of French and Italian. 11 (1), in press. Van der Velde, M., Jakubowicz, C. & Rigaut, C. 2002. The acquisition of determiners and pronominal clitics by three French-speaking children. In The Process of Language Acquisition, I. Lasser (ed.), 115–132. Frankfurt: Lang. White, L. 1996. Clitics in L2 French. In Generative Perspectives on Language Acquisition, H. Clahsen (ed.), 335–368. Amsterdam: John Benjamins.
Afterword Neil Smith
Language acquisition has become the major focus of the linguistic research of the last few decades. Since Chomsky’s (1959) review of Skinner’s Verbal Behavior (Skinner, 1957) it has been apparent that accounting for how children acquire their first language and how they (and adults) learn further languages is both a daunting challenge and a source of deep insight. The current volume illustrates the richness of the field in all its remarkable variety. We have studies from typically developing and pathological populations; from the core grammar of syntax and morphology to the realm of pragmatics; from English and a host of other languages: Basque, French, German, Hebrew, Norwegian, Russian, Serbo-Croatian and Spanish; from naturalistic and experimental studies. Even this impressive list doesn’t exhaust the diversity. We have studies of children and of adults, of synchrony and diachrony, of comprehension and production, of monolingual and bilingual acquisition. Within this last category we have contrasts between simultaneous and successive bilingualism, and between dominant and subordinate languages. And all this is couched within the framework of explicit theory: typically the Principles and Parameters approach of Chomskyan Generative Grammar (e.g., Chomsky 1995), but also the Usage-based approach of Tomasello (2003) and others. This emphasis on theory is the crucial contribution of the generative enterprise of the last 50 years, but there is – perhaps surprisingly – a sense in which it doesn’t matter which theory. Obviously we each are convinced that our own theory is the best and will be prepared to argue long (and sometimes acrimoniously) about the advantages of mine and the disadvantages of yours. But the choice is a matter of ongoing empirical debate characteristic of ‘normal science’ in Kuhn’s (1962) sense. One can discuss the relative merits of the continuity and maturational hypotheses of first language acquisition on the basis of linguistic analyses framed in Construction Grammar, Head-driven Phrase Structure Grammar, Lexical Functional Grammar, Minimalism, Optimality Theory, and so on ad libitum. The details will vary and, most importantly, the different theories will cut up the cake differently, defining different ‘natural classes’ of data and making different predictions. As a consequence theory A will be supported (or undermined) by one class of results and theory B by another. Over time the results will gradually converge to show that one kind of theory is systematically supported vis-à-vis the others. The last half century has shown one such striking convergence: on an internalist perspective – the claim that the object of linguistic inquiry is the content of the mind rather than some autonomous external language.
Neil Smith
The concepts are not always as transparent as one might wish – we are after all doing science – but the core concept is what Chomsky calls ‘I-language’, the internalized largely unconscious knowledge of the individual, rather than the traditional socio-cultural domain external to the individual that the pre-generative community took for granted. Obviously much of the evidence for the nature of this I-language comes from externalized utterances or judgments of the kind beautifully illustrated in the preceding chapters. But the underlying aim is to account for knowledge rather than behaviour (for discussion, see Smith 2004). I will illustrate some of the breadth mentioned above and the wider internalist implications of the various chapters in what follows. Perovic’s [ch. 10] interesting demonstration that the language of Down Syndrome subjects is deviant rather than just delayed depends crucially on ingenious experimentation rather than simply naturalistic observation, and is part of a wider trend using pathological data to illuminate the nature of the human language faculty. Crucially, these data reveal something about the knowledge systems of the populations under discussion: they are evidence, not just facts. An example with a similar conclusion that simultaneously illustrates the advantage of one theory over another is provided by Ring & Clahsen’s (2005) comparison of Down Syndrome and William Syndrome children’s performance on passives and binding. They presuppose that both are instances of ‘A-chain formation’. This links the two construction types for (e.g.,) Reuland (2001), whereas nothing links them for (e.g.,) Lexical Functional Grammar (where passive is a lexical alternation and binding a principle of semantic interpretation). It follows that results confirming the link would support the relevant theory, whilst a negative result could be due to having made either the wrong theory choice or the wrong hypothesis. The results were clear: William Syndrome subjects, like controls, were roughly error-free on both constructions, Down Syndrome subjects had problems with both. Ring & Clahsen conclude by taking the discussion to a further, biological, level, observing that “different genetic etiologies [of William Syndrome and Down Syndrome] are associated with different specifically linguistic patterns of impairment” (Ring & Clahsen 2005: 479). Not all analyses are so clearly theory-dependent in this way, as is illustrated by a group of papers on bilingual acquisition. Bonnesen’s [ch. 7] paper dissecting the claimed ‘vulnerability’ of the left periphery owes its existence and cogency to Rizzi’s (1997) account of the internal structure of the CP domain. His analysis of bilingual data from children learning French and German casts the vulnerability hypothesis into serious doubt, but the general issue of the direction of influence in bilingual development could clearly be addressed from other theoretical perspectives. Larrañaga’s [ch. 8] paper on unaccusativity is similarly dependent on Rizzi and also requires that the theory provide the kind of categorial distinction exemplified by unaccusative and unergative verbs and allow the statement of correlations between these and particular word-orders. By contrast, Kupisch’s [ch. 9] and Müller & Pillunat’s [ch. 11] papers on dominance effects on bilingual acquisition are relatively theory-independent: indeed, they invite the kind of innovation found in Tamburelli’s (2005) theory of paradigm
Afterword
formation and ‘cascade updating’, whereby a newly acquired property such as I to C movement is generalized automatically to all and only members of the appropriate paradigm, be this within the same language or cross-linguistically. A persistent theme throughout the book, and one which raises problems of interpretation, is the debate between continuity and maturation: whether child grammars are ‘continuous’ with adult grammars or can have properties such as the absence of functional categories which sets them apart from any possible adult grammar. Gordishevsky & Schaeffer’s [ch. 2], Westergaard’s [ch. 3] and Escobar & Torrens’ [ch. 5] papers all address the issue, and all come down on the side of continuity, in opposition to the positions defended by Radford (1990), Tsimpli (1996), and a variety of others. The papers are convincing but highlight the difficulty of interpreting the results when universal constraints, assumptions about the status of underspecified categories, input differences from the two languages concerned, phonological salience and pragmatic appropriateness as well as the role of the core morpho-syntax may all play a role (see Paradis & Genesee 1997, for discussion). This multiplicity of variables is another reflection of theoretical sophistication and points to another strength of the contributions in this volume: the systematic appeal to both syntax and pragmatics. Clear examples are provided by Armon-Lotem’s [ch. 6] discussion of the interaction of grammar and pragmatics in children’s comprehension of logical words such as or and some, and Escobar & Torrens’ [ch. 5] discussion of the determinants of children’s acquisition of universal quantification in Spanish. As is now generally accepted (see Westergaard’s [ch. 3] and e.g., Kang 2002, for discussion) no purely syntactic or purely cognitive account is adequate to describe what is going on. The beauty of the current scene is that at last we have theories in both domains that are jointly adequate to describe and even explain the children’s behaviour. In this context two points are worth emphasizing: first, that notions of ‘economy’ may be matters of either competence or performance and pertain either to the syntactic or the pragmatic domain; second, that even some aspects of our pragmatic abilities may be innate – for instance, Sperber & Wilson’s (1995) principle of relevance. It is not only UG that is part of the genetically determined human faculty of language in the ‘broad sense’ (FLB) defined by Hauser et al. (2002). Finally, the most exciting phenomena of first language acquisition come from the creations of the children themselves – what Parisse [ch. 1] refers to as ‘originals’ where the child produces forms not found (or not found systematically) in the input. It is striking that his data are analysed in a usage-based framework where such examples are unexpected, whereas Schönenberger’s [ch. 4] shows a comparable phenomenon (where children generalize verb movement beyond the input) in a framework where it is precisely expected. That is, a central assumption of the ‘generative’ approach is precisely that UG will license children to produce examples which are impossible in the ambient language, the clearest examples being such sentences as “What do you think what pigs eat?” (Crain & Pietroski 2002) where the repeated what would be licit in (some varieties of) German.
Neil Smith
At the risk of being obsessional, I would end by re-emphasizing that the current linguistic scene is illuminated by the existence of and competition among explicit internalist theories, where these are most tellingly tested by the challenge of explaining the data of acquisition. The present book has illustrated this supremely well. Read it again.
References Chomsky, N. 1959. Review of Skinner, 1957. Language 35: 26–58. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Crain, S. & Pietroski, P. 2002. Why language acquisition is a snap. The Linguistic Review 19: 163–183. Hauser, M., Chomsky, N. & Tecumseh Fitch, W. 2002. The faculty of language: What is it, who has it, and how did it evolve? Science 298: 1569–1579. Kang, H.-K. 2002. Aspects of the Acquisition of Quantification: Experimental Studies of English and Korean Children. Seoul: Hankook. Kuhn, T.S. 1962. The Structure of Scientific Revolutions. Chicago IL: Chicago University Press. Paradis, M. & Genesee, F. 1997. On continuity and the emergence of functional categories in bilingual first language acquisition. Language Acquisition 6: 91–124. Radford, A. 1990. Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Reuland, E. 2001. Primitives of binding. Linguistic Inquiry 32: 439–492. Ring, M. & Clahsen, H. 2005. Distinct patterns of language impairment in Down’s syndrome and Williams syndrome: The case of syntactic chains. Journal of Neurolinguistics 18: 479–501. Rizzi, L. 1997. The fine structure of the left periphery. In Elements of Grammar, L. Haegeman (ed.), 281–337. Dordrecht: Kluwer. Skinner, B.F. 1957. Verbal Behavior. New York NY: Appleton Century Crofts. Smith, N.V. 2004. Chomsky: Ideas and Ideals. 2nd edn. Cambridge: Cambridge University Press. Sperber, D. & Wilson, D. 1995. Relevance: Communication and Cognition. 2nd edn. Oxford: Blackwell. Tamburelli, M. 2005. The importance of paradigm formation in bilingual acquisition: Evidence from Italian. University College London Working Papers in Linguistics 17: 27–57. Tomasello, M. 2003. Constructing a Language. Cambridge MA: Harvard University Press. Tsimpli, I.-M. 1996. The Prefunctional Stage of Language Acquisition. New York NY: Garland.
Index of subjects
A Acceleration, 185, 229, 269, 285, 287, 289, 290 Accusative, 33, 35–8, 40, 41, 48, 213, 238, 241, 245, 272 see also unaccusative Adjunction, 18, 165 Adult language, 2, 14, 17, 20, 28, 40, 59, 62, 77, 79, 162 Anaphor/a, 9, 122, 235, 237, 240, 241, 244–47, 250, 261–65 anaphoric expression, 123 anaphoric element, 235, 243, 246, 263 anaphoric relation, 244, 246 anaphoric binding, 246, 247, 261, 263, 264 Aphasics, 163 Article/s, 211–13, 215, 220, 221, 224, 226, 228, 229, 235, 273, 285, 286 see also indefinite/definite article B Balanced/unbalanced bilinguals/bilingualism, 9, 169, 269, 270, 289, 290 Bare nouns, 220, 228, 229, 273 Basque, 7, 8, 183, 185, 186, 188, 189, 190–207, 295 Belfast English, 66, 109 Bilingual, 1, 6, 8, 9, 10, 42, 59, 161–63, 166, 168, 169, 178–80, 183–6, 201–3, 205, 209–211, 214, 220, 221, 224, 225, 228–34, 269–71, 273–7, 279–85, 287, 289–91, 295, 296 bilingual first language acquisition, 6, 7, 8, 11, 161–3, 210, 269, 271, 275, 292
Binding, 8, 9, 74, 119, 121–4, 133, 134, 235–7, 240, 243–8, 257, 261, 263–5, 296 binding conditions, 244, 245 binding principles, 8, 74, 236, 244, 247 binding theory, 9, 74, 236, 244, 266 C Case/s, 3, 30–57, 187, 189, 213, 238, 240, 241, 244, 245, 263, 272 case assignment, 35, 119 case checking, 35 case licensing, 31, 32, 35, 43, 44, 48, 54, 57, 189 case marking, 32, 33, 35, 47, 57, 75 case–chain, 3, 35, 40, 45, 47, 57 see also inherent cases, structural case, lexical case Catalan, 45, 162, 184, 202, 220 Clitic/s, 5, 7, 9, 17, 28, 62, 119, 122, 123, 126, 132, 133, 166, 167, 173, 174, 238–47, 250–52, 254–8, 261, 262, 272–4, 277, 282, 287, 288, 291 clitic subject, 7, 166, 167, 174 cliticization, 2, 240, 241 see also enclitic, non-clitic, object-clitic, reflexive clitic Competence, 1, 13, 14, 31, 42, 57, 74, 81, 231, 232, 297 Construction/s, 2, 3, 13–28, 37, 41, 48, 61–3, 65–7, 69, 70–2, 78–84, 107, 110, 161, 162, 164, 166, 167, 169–180, 183, 185, 186, 189, 192,
194–6, 198, 200, 202, 210, 241, 247, 252, 258, 261, 272, 295, 296 see also dislocated constructions, item-based constructions, V2 constructions Contact phenomena, 8, 209–11, 230, 231 Continuity, 61, 62, 76, 79, 119, 295, 297 continuity approach, 62, 63 continuity hypothesis, 5, 107, 115 Conversational implicatures, 137, 146, 153 CP, 3, 4, 6, 7, 61, 62, 65, 69, 76, 79, 161, 163, 164, 166–8, 185, 190, 191 CP domain, 65, 66, 69, 70, 76, 79, 83, 296 see also Split-CP and SpecCP Cross-linguistic influence, 6–8, 162, 163, 166, 183, 185, 186, 201–3, 209–11, 220–24, 227, 228, 230, 231, 270, 271, 280, 282, 283, 285, 286, 289, 290 Crossover, 122, 123, 125, 133 Cue/s, 75, 76, 78, 82–4, 94, 115 D Danish, 66 Dative, 33, 37–9, 41, 48, 213 Definite article, 89 Delay/ed, 1, 8, 9, 74, 75, 79, 80, 82–5, 108, 227, 228, 235, 236, 259, 262, 269, 270, 271, 274, 275, 277, 279, 281–3, 286–91, 296 Delay of Principle B Effect, 236, 259
Index of subjects Determiner omission, 214, 220–30, 269, 284–6 Dislocated constituents, construction/s, 17, 21, 22, 28, 172 see also left dislocation Dominant/dominance, 8, 185, 201, 202, 209–11, 215, 216, 218, 220, 228, 230–32, 270, 276, 295, 296 dominant language, 211, 218, 230, 269, 270, 276 see also language dominance DP/s, 7, 35, 36, 39–41, 43, 44, 47, 54, 57, 62, 63–5, 67, 69, 76, 80, 82, 114, 166, 167, 173, 174, 190–95, 198, 209, 212, 213, 228, 229, 231, 263, 271–3, 287, 288 DP subject, 65, 67, 69, 70, 80, 81, 166, 173, 174 DuFDE, 6, 161, 163, 168 Dutch, 68, 75, 77, 90, 92, 112, 134, 162, 220, 236, 237, 241–3, 245, 250, 282 E Economy, 4, 61–3, 73–84, 297 see also structural economy Embedded clause/s, 4, 5, 64, 66, 67, 69, 79–81, 87–96, 100, 105–110, 115 embedded contexts, 78, 80, 84, 88, 103, 105, 108 embedded questions, 64, 66, 69, 92 embedded V2, 4, 64, 94 Emphasis, 16, 185, 191, 198, 200, 203, 211, 295 Enclitics, 100 Exclamatives, 64, 66 Expletive, 108, 189, 272 F Finiteness, 83, 164, 168, 169, 175, 177, 178 Focus, 33, 65–7, 81, 129, 185, 186, 188–192, 195–203, 239 French, 1, 2, 6–9, 13–18, 28, 42, 45, 55, 81, 161–79, 209–31, 237, 242, 243, 250, 262, 269–91, 295, 296
Frequency, 4, 8, 14, 21, 61, 63, 70, 73, 79, 81–4, 88, 97, 103, 201, 211, 213, 214, 228, 231, 272, 282 see also input frequency Full Clause Hypothesis, 42, 43, 54 G Genitive, 33, 35–8, 40, 41, 48, 184, 213 German/Germanic, 1, 6, 7, 8, 9, 21, 42, 45, 55, 61–3, 67, 68, 70, 77, 80, 81, 87, 88, 90, 92, 103, 107, 161–79, 191, 209–31, 269–89, 295–7 see also Swiss German Given information, 65, 66, 75, 80, 82 Greek, 146, 262 Grice, 137, 138, 139, 141, 142 H Hebrew, 5, 6, 45, 137, 138, 147, 148, 150, 155, 241, 295 I Icelandic, 66, 108, 236 Illocutionary force, 65, 76 Indefinite article, 148, 155, 212, 220 Information structure, 18, 33, 62, 65, 69, 73, 75, 81, 183, 188–92, 194 Inherent Case/s, 37, 39–41, 48, 50, 53, 54, 245 Input, 1, 2, 4, 13, 14, 24, 42, 61–3, 73, 76–9, 81–4, 87, 88, 93–8, 102–4, 107, 109, 115, 126, 127, 132, 162, 177–9, 210, 229, 230, 271, 297 input frequency, 4, 61, 63, 79, 82–4 Interface/s, 10, 162, 163, 178, 182, 183, 185, 186, 201 Inversion, 77, 92, 163, 164, 166, 167, 175, 204 see also subject-verb inversion IP, 6, 61, 62, 65, 161, 163, 165, 185, 190, 191, 202, 210 IP domain, 4, 61, 67, 76, 84, 185, 189, 190
see also Spec-IP Italian, 45, 107, 122, 123, 138, 146, 150, 155, 162, 184, 187, 214, 220, 237, 274, 282 L Language development, 8, 9, 13, 14, 18, 25, 28, 61, 75, 109, 156, 214, 235–7, 249, 262, 275, 291 see also Linguistic development Language dominance, 8, 185, 209–211, 215, 228, 230–32, 270, 276 Language norm, 28, 270, 275 Language transfer, 167, 170, 172, 179 Left periphery, 2, 6, 7, 161, 163–6, 169, 171–9, 183, 186, 188, 190, 191, 202, 296 Left-dislocation, 13–15, 17, 18, 21 Lexical Case/s, 3, 38–41, 47, 50–54 Lexical Learning Hypothesis, 77 Lexical subject, 2, 13–16, 18, 21–22, 24, 26, 28 Lexicon, 9, 14, 178, 185, 231, 241, 243, 245, 250 Linguistic development, 222, 226, 230, 235, 262 M Mixed language utterances, 8, 209, 210, 216–20, 230–32 MLU, 9, 19, 20, 46, 168, 191, 192, 215, 216, 222–30, 248, 249, 269, 275–86, 289–91 N Narrow focus, 186, 188, 189, 192, 195, 197–203 Negation, 37, 41, 48, 61–84, 92, 95, 141, 196, 235 Nominative, 33–6, 40, 41, 48, 51, 52, 54, 57, 119, 189, 213, 240, 244, 246 Non-clitic subjects, 166, 167, 174 Non-pronominal subject/s, 107, 108, 110, 111, 114, 115, 118
Index of subjects see also pronominal subject/s Non-subject-initial declaratives, 61–3, 66, 68, 70, 72–4, 77, 81–4 Norwegian, 2, 3, 61–84, 92, 117, 162, 250, 266, 295 Null subject/s, 28, 75, 80, 162 Number, 3, 17, 21, 31, 32, 34, 35, 39, 40, 43–9, 51, 55–7, 164, 213, 238, 240, 241, 244, 245, 263 O Object clitic/s, 9, 262, 272–4, 277, 282, 287, 291 Object shift, 62, 77, 78, 80, 82–5, 239 Old and Middle English, 5, 67, 110, 112, 115 Omission, 162, 213, 214, 220–26, 228–30, 235, 269, 271, 272, 274, 280–88, 291 see also rate of omission, determiner omission P Parameter/s, 75, 87, 88, 106, 109, 110, 115, 119, 161, 162, 184, 212, 230, 231, 241, 245, 295 Performance, 2, 3, 7, 13, 14, 18, 25–9, 40, 49, 53, 54, 57, 74, 75, 121, 171, 231, 232, 246–50, 253–65, 290, 296, 297 Plural noun, 3, 44, 45, 48, 49, 51, 53, 54, 56–7, 148 Postverbal, 5, 96, 119–21, 122, 123, 125, 126, 128–35, 164, 167, 172, 192, 193, 196 Pragmatic, 4, 5, 7, 8, 10, 17, 18, 61, 63, 66, 73–75, 79, 80–82, 84, 88, 108, 109, 115, 123, 131, 137–9, 141–8, 153–6, 162, 163, 183–6, 188–91, 196, 201–3, 212, 220, 231, 235, 236, 271, 295, 297 Preposition, 3, 16, 17, 33, 35, 38, 39, 41, 48, 51, 90, 199, 235 Presupposition, 108, 143, 185, 188, 190, 192, 199
Preverbal, 5, 8, 114, 115, 119, 120, 122, 123, 125, 126, 128, 129, 131, 132, 135, 167, 183, 192, 195, 196, 202, 251 Principles and Parameters, 161, 295 Pronominal subject, 3, 4, 61–3, 65, 67, 69–72, 80, 84, 107, 108, 110, 111, 114, 115, 118, 197 see also non-pronominal subject Pronoun, 2, 3, 7, 9, 10, 13, 15–18, 20, 21, 23, 25–8, 35, 47, 54, 55, 62, 67, 69, 74, 78–80, 82, 91, 100, 107, 108, 111–113, 121–123, 125, 133, 161, 174–6, 178, 179, 188, 192–5, 199, 221, 235–8, 239, 240, 244–7, 250–7, 260–3, 265, 271–4, 277, 287, 288, 291 see also reflexive pronoun, strong pronoun Q Quantifier Raising, 5, 121, 122, 125, 133 R Reflexive, 9, 74, 236–8, 240–47, 250–57, 260–63 reflexive clitic, 238, 240, 241, 243, 245–47, 250–52, 255–7, 261 reflexive pronoun, 240, 245, 254, 260, 262 Russian, 3, 21, 31–57, 74, 75, 236, 295 Russian Case, 32, 34, 35, 56 S Scalar implicature, 6, 137–9, 141, 142, 145–7, 150, 152, 153, 155 Scandinavian, 62, 239, 267, 294 Scrambling, 75, 77, 238 Serbo-Croatian, 8, 9, 235–8, 247, 260, 262, 295 Spanish, 2, 5, 7, 8, 11, 45, 58, 119, 121–3, 125, 126, 132, 133, 162, 163, 183–94, 197, 198, 200–3, 205, 206,
220, 233, 243, 250, 266, 295, 297 SpecCP, 7, 62, 163, 164, 167 Specific Language Impairment (SLI), 1, 9, 10, 163, 182, 249, 250, 262, 274, 287–91 Specificity, 5, 74, 137–9, 142–4, 147, 148, 153–5, 212 SpecIP, 62, 121, 163, 164, 167, 189, 190 Split-CP, 4, 61, 65, 76, 84, 88, 110, 161 Strong language, 8, 185, 209–11, 227, 229, 230, 269, 270, 285 Strong pronouns, 10, 272, 287, 288, 291 Structural Case, 35, 37, 41, 48, 49, 52, 53, 244, 246 Structural economy, 61, 76, 77, 84 Structure building, 61–3, 77–9 Subject clitic, 7, 17, 167, 173, 174, 272, 274 Subject position, 4, 5, 6, 36, 61–3, 66–8, 70, 73, 76, 80, 84, 119, 120, 122, 123, 137, 148, 149, 151–4, 164, 167, 194, 271, 272, 282, 287, 288, 291 Subject shift, 63, 66, 70, 76, 78–80, 82–4 Subject-object asymmetry, 5, 6, 137, 138, 147, 150–53, 155, 156, 274 Subject-verb inversion, 163, 164, 166, 167 Swedish, 68, 77, 83, 92, 163, 230 Swiss German, 2, 4, 80, 81, 87, 88, 90, 91, 93, 97, 100, 103, 104, 106, 114, 116, 117 Syntax-pragmatics interface, 162, 185, 186, 201 T Topic, 8, 10, 33, 65, 67, 81, 108, 110–12, 143, 144, 148, 153, 154, 162–4,
Index of subjects 184, 185, 188–91, 195–8, 200–202, 239, 271, 273 topicalized/topicalization/ topicality, 3, 61, 62, 79, 81, 108, 110, 115, 138, 166, 191, 271 Turkish, 75 U UG, 74, 76, 78, 297 Unaccusative, 7, 37, 183, 184, 186–9, 191–3, 198, 201–3, 205–7, 241, 296 Underspecification hypothesis, 32, 43, 55, 57 of functional categories, 44, 58, 85 of number, 3, 43, 45, 57 Universal quantifier/s, 5, 119, 120, 123, 124, 126 Usage-based theory, 2, 13, 15 V Verb movement, 3, 4, 61, 63, 64, 66, 68–70, 77–9, 81–3, 87–9, 92, 94–118, 297
Verb Projection Raising (VPR), 90, 91, 95, 103–6 Verb-final, 4, 87–90, 92–109, 112, 114, 115 Verb-placement error/s, 87, 92, 93, 98 Verb-second (V2), 3, 4, 6, 7, 61, 63–9, 76, 77, 79–84, 87–97, 102, 103, 108–15, 118, 161–4, 167, 169–80, 233 V2 constructions, 66, 69, 84, 161, 162, 164, 170, 172–80 V2 pattern, 90, 96, 102, 103, 108–10 see also embedded verb Vulnerability, 6, 7, 162, 163, 170–79, 228, 262, 289, 296 W Weak language, 9, 269, 270, 274, 280–91, 294 Weaker language, 211, 220, 228–32
Wh-complement, 89, 96, 99, 100–102, 108–10, 113, 118 wh-constituent, 64, 83, 101, 102, 109, 110 wh-phrase, 120, 166, 167, 188 wh-question, 17, 63, 65–70, 76, 77, 79, 81, 84, 107, 110, 114, 166, 167, 174, 188 wh-word, 64, 66, 81 Wide focus, 186, 188, 189, 191, 192, 196, 197, 199, 202, 203 Word order, 3, 4, 7, 8, 32, 33, 61–84, 87, 91, 95, 96, 106, 107, 109, 113, 115, 119, 164, 166, 167, 169, 171, 172, 176, 178, 179, 183, 185, 186, 188–91, 201–3, 238, 296 Y Yes/no-questions, 17, 65, 66, 68, 70, 82
In the series Language Acquisition and Language Disorders the following titles have been published thus far or are scheduled for publication: A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 46 Haznedar, Belma and Elena Gavruseva (eds.): Current Trends in Child Second Language Acquisition. A generative perspective. vi, 356 pp. + index. Expected July 2008 45 Guijarro-Fuentes, Pedro, María Pilar Larrañaga and John Clibbens (eds.): First Language Acquisition of Morphology and Syntax. Perspectives across languages and learners. 2008. vi, 302 pp. 44 Sekerina, Irina A., Eva M. Fernández and Harald Clahsen (eds.): Developmental Psycholinguistics. On-line methods in children’s language processing. 2008. xviii, 190 pp. 43 Savickienė, Ineta and Wolfgang U. Dressler (eds.): The Acquisition of Diminutives. A crosslinguistic perspective. 2007. vi, 352 pp. 42 Lefebvre, Claire, Lydia White and Christine Jourdan (eds.): L2 Acquisition and Creole Genesis. Dialogues. 2006. viii, 433 pp. 41 Torrens, Vincent and Linda Escobar (eds.): The Acquisition of Syntax in Romance Languages. 2006. viii, 422 pp. 40 Deen, Kamil Ud: The Acquisition of Swahili. 2005. xiv, 241 pp. 39 Unsworth, Sharon, Teresa Parodi, Antonella Sorace and Martha Young-Scholten (eds.): Paths of Development in L1 and L2 acquisition. In honor of Bonnie D. Schwartz. 2006. viii, 222 pp. 38 Franceschina, Florencia: Fossilized Second Language Grammars. The acquisition of grammatical gender. 2005. xxiv, 288 pp. 37 Montrul, Silvina A.: The Acquisition of Spanish. Morphosyntactic development in monolingual and bilingual L1 acquisition and adult L2 acquisition. 2004. xvi, 413 pp. 36 Bartke, Susanne and Julia Siegmüller (eds.): Williams Syndrome across Languages. 2004. xvi, 385 pp. 35 Sánchez, Liliana: Quechua-Spanish Bilingualism. Interference and convergence in functional categories. 2003. x, 189 pp. 34 Ota, Mitsuhiko: The Development of Prosodic Structure in Early Words. Continuity, divergence and change. 2003. xii, 224 pp. 33 Josefsson, Gunlög, Christer Platzack and Gisela Håkansson (eds.): The Acquisition of Swedish Grammar. 2004. vi, 315 pp. 32 Prévost, Philippe and Johanne Paradis (eds.): The Acquisition of French in Different Contexts. Focus on functional categories. 2004. viii, 384 pp. 31 Marinis, Theodoros: The Acquisition of the DP in Modern Greek. 2003. xiv, 261 pp. 30 Hout, Roeland van, Aafke Hulk, Folkert Kuiken and Richard J. Towell (eds.): The Lexicon– Syntax Interface in Second Language Acquisition. 2003. viii, 234 pp. 29 Fernández, Eva M.: Bilingual Sentence Processing. Relative clause attachment in English and Spanish. 2003. xx, 294 pp. 28 Shimron, Joseph (ed.): Language Processing and Acquisition in Languages of Semitic, Root-Based, Morphology. 2003. vi, 394 pp. 27 Salaberry, M. Rafael and Yasuhiro Shirai (eds.): The L2 Acquisition of Tense–Aspect Morphology. 2002. x, 489 pp. 26 Slabakova, Roumyana: Telicity in the Second Language. 2001. xii, 236 pp. 25 Carroll, Susanne E.: Input and Evidence. The raw material of second language acquisition. 2001. xviii, 461 pp. 24 Weissenborn, Jürgen and Barbara Höhle (eds.): Approaches to Bootstrapping. Phonological, lexical, syntactic and neurophysiological aspects of early language acquisition. Volume 2. 2001. viii, 337 pp. 23 Weissenborn, Jürgen and Barbara Höhle (eds.): Approaches to Bootstrapping. Phonological, lexical, syntactic and neurophysiological aspects of early language acquisition. Volume 1. 2001. xviii, 299 pp. 22 Schaeffer, Jeannette C.: The Acquisition of Direct Object Scrambling and Clitic Placement. Syntax and pragmatics. 2000. xii, 187 pp. 21 Herschensohn, Julia: The Second Time Around – Minimalism and L2 Acquisition. 2000. xiv, 287 pp. 20 Kanno, Kazue (ed.): The Acquisition of Japanese as a Second Language. 1999. xii, 180 pp. 19 Beck, Maria-Luise (ed.): Morphology and its Interfaces in Second Language Knowledge. 1998. x, 387 pp.
18 Klein, Elaine C. and Gita Martohardjono (eds.): The Development of Second Language Grammars. A generative approach. 1999. vi, 412 pp. 17 Archibald, John: Second Language Phonology. 1998. xii, 313 pp. 16 Hannahs, S.J. and Martha Young-Scholten (eds.): Focus on Phonological Acquisition. 1997. v, 289 pp. 15 Brinkmann, Ursula: The Locative Alternation in German. Its structure and acquisition. 1997. x, 289 pp. 14 Clahsen, Harald (ed.): Generative Perspectives on Language Acquisition. Empirical findings, theoretical considerations and crosslinguistic comparisons. 1996. xxviii, 499 pp. 13 Allen, Shanley: Aspects of Argument Structure Acquisition in Inuktitut. 1996. xvi, 244 pp. 12 Juffs, Alan: Learnability and the Lexicon. Theories and second language acquisition research. 1996. xi, 277 pp. 11 Yip, Virginia: Interlanguage and Learnability. From Chinese to English. 1995. xvi, 247 pp. 10 Lakshmanan, Usha: Universal Grammar in Child Second Language Acquisition. Null subjects and morphological uniformity. 1994. x, 162 pp. 9 Adone, Dany: The Acquisition of Mauritian Creole. 1994. xii, 167 pp. 8 Hoekstra, Teun and Bonnie D. Schwartz (eds.): Language Acquisition Studies in Generative Grammar. 1994. xii, 401 pp. 7 Meisel, Jürgen M. (ed.): Bilingual First Language Acquisition. French and German grammatical development. 1994. vi, 282 pp. 6 Thomas, Margaret: Knowledge of Reflexives in a Second Language. 1993. x, 234 pp. 5 Gass, Susan M. and Larry Selinker (eds.): Language Transfer in Language Learning. Revised edition. 1992. x, 236 pp. 4 Eckman, Fred R. (ed.): Confluence. Linguistics, L2 acquisition and speech pathology. 1993. xvi, 260 pp. 3 Eubank, Lynn (ed.): Point Counterpoint. Universal Grammar in the second language. 1991. x, 439 pp. 2 Huebner, Thom and Charles A. Ferguson (eds.): Cross Currents in Second Language Acquisition and Linguistic Theory. 1991. viii, 435 pp. 1 White, Lydia: Universal Grammar and Second Language Acquisition. 1989. xii, 198 pp.
