(In)vulnerable Domains in Multilingualism 902729657X, 9789027296573

Table of contents :
(In)vulnerable Domains in Multilingualism
Editorial page
Title page
LCC page
Table of contents
Introduction
References
The DP, a vulnerable domain?
Conclusion
Notes
References
Child and adult acquisition of word order in the Italian DP
Discussion and conclusions
Notes
Appendix
References
Null Subjects and optional infinitives in Basque
Conclusion
Notes
References
The acquisition of subjects in bilingual children
Conclusion
Notes
References
Approaches to bilingual acquisition data
Notes
References
Factors accounting for code-mixing in an early developing bilingual
Conclusion
Note
References
Syllable final consonants in Spanish and German monolingual and bilingual acquisition
Conclusion
Notes
Appendix
References
Interrogative elements as subordinators in Turkish
Abbreviations
Notes
References
Child L2 acquisition
Final considerations
Notes
References
The Verb–Object parameter in simultaneous and successive acquisition of bilingualism
Conclusions
Notes
References
Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP?
Conclusion
Notes
References
Name index
Subject index

Citation preview

(In)vulnerable Domains in Multilingualism

Hamburg Studies in Multilingualism The Hamburg Studies in Multilingualism (HSM) publishes research from colloquia on linguistic aspects of multilingualism organized by the University of Hamburg.

Editors Jürgen M. Meisel Jochen Rehbein Conxita Lléo University of Hamburg

Volume 1 (In)vulnerable Domains in Multilingualism Edited by Natascha Müller

(In)vulnerable Domains in Multilingualism Edited by

Natascha Müller University of Hamburg

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 (In)vulnerable Domains in Multilingualism / edited by Natascha Müller. p. cm. (Hamburg Studies in Multilingualism, issn 1571–4934 ; v. 1) Includes bibliographical references and indexes. P115.2.I58 2003 404’.2--dc21 isbn 90 272 1921 4 (Eur.) / 1–58811–373–6 (US) (Hb; alk. paper)

2003041811

© 2003 – 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

< R/ RE E FF

"intro"> "kup"> "ber"> "eze"> "hin"> "gaw"> "deu"> "lle"> "her"> "kat">

Table of contents

Introduction Natascha Müller The DP, a vulnerable domain? Evidence from the acquisition of French Tanja Kupisch

vii

1

Child and adult acquisition of word order in the Italian DP Petra Bernardini

41

Null Subjects and optional infinitives in Basque Maria-José Ezeizabarrena

83

The acquisition of subjects in bilingual children: Pronoun use in Portuguese-German children Marc-Olivier Hinzelin

107

Approaches to bilingual acquisition data Ira Gawlitzek-Maiwald

139

Factors accounting for code-mixing in an early developing bilingual Margaret Deuchar and Rachel Muntz

161

Syllable final consonants in Spanish and German monolingual and bilingual acquisition 191 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo Interrogative elements as subordinators in Turkish: Aspects of Turkish-German bilingual children’s language use Annette Herkenrath, Birsel Karakoç and Jochen Rehbein Child L2 acquisition: An insider account Mary Aizawa Kato

221

271

< /R/TREARGET E FF

"moh"> "roe"> "ni"> "si"> "toc">

vi

Table of contents

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism Anja Möhring and Jürgen M. Meisel

295

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? Tom Roeper

335

Name index

361

Subject index

367



Introduction Natascha Müller Universität Hamburg

Research in language acquisition has shown that children develop particular grammatical phenomena much later than others and that some grammatical domains are prone to error in acquisition in the sense that children will deduce systems for these domains which do not correspond to the target system. These so called vulnerable grammatical domains differ from robust ones which are acquired quickly and in an error-free fashion. Although most researchers agree on the existence of vulnerable and invulnerable domains in language acquisition, it is still a matter of debate as to how to characterize and explain (in)vulnerablity. One possibility is to assume that vulnerable domains differ from invulnerable domains in requiring much more complex syntactic, phonological, or semantic analysis (cf. Jakubowicz, Nash, Rigaut & Gérard 1999, Müller, Cantone, Kupisch & Schmitz 2002, Platzack 2001, Rizzi 1993/94, Roeper 1996, Sorace 2000, Wexler 1998 among others). The following contributions take the perspective of multilingualism on the issue of vulnerability. It is of course possible that vulnerability does not only depend on the grammatical phenomenon but also on the language acquisition type. Thus, it may be the case that a domain which is invulnerable in first language acquisition turns out to be problematic for second language learners. The interesting question is how learners deal with such problematic domains and what this might reveal for language acquisition theory in general. Or, a bilingual child, who disposes of resources from more than one language, might acquire a vulnerable grammatical phenomenon in a way which characterizes it more aptly as an invulnerable phenomenon. The child might have used resources from the respective other language. Studies of individuals with more than one language demonstrate the complexity of the language system in general, which is made up of distinct interacting components. They also might help to reveal how these components, the computational system and the general cognitive system for example,

viii Introduction

function and interact and how issues like complexity and vulnerability are related to these components. Tanja Kupisch studies the structure of the DP in the acquisition of French by one monolingual and two French-German bilingual children, one of them clearly unbalanced with French as her weaker language. The three longitudinal studies reveal that the DP domain is not vulnerable. This is shown to be the case with respect to determiner omissions, a phenomenon which is characteristic of a developmental stage in the acquisition of French. Kupisch argues that the frequency of determiners, consistency in their syntactic distribution and transparency of grammatical encoding help the child to quickly abandon the stage in French during which obligatory determiners are being omitted. This can be observed even in the bilingual child whose French develops very slowly compared to other children acquiring French. The conclusion reached by the author is that vulnerability is not a question of the interface character (the interface level connecting internal grammar with other cognitive systems) alone. Rather, other factors, such as inconsistency in the syntactic distribution of grammatical elements seem to contribute to the vulnerability of a particular syntactic domain. Petra Bernardini also investigates the DP. In contrast to Kupisch, however, she looks at the acquisition of word order within DP in Italian by two SwedishItalian bilinguals, one of them unbalanced with Italian as the weaker language, the other one with Italian as the stronger language. The grammatical phenomenon investigated is the positioning of possessives and attributive adjectives. The data of the bilinguals are compared with Italian monolingual children and with Swedish adult learners of Italian. The general result is that the bilingual child with Italian as the stronger language resembles monolingual Italian children, while the bilingual child with Italian as the weaker language can be compared with adult second language learners of Italian. Interestingly, the problematic constructions for the bilingual child with Italian as the weaker language are those which have variable word-order of the adjective in the target-system. One of these word orders is found in the Swedish target. Bernardini concludes that “ambiguous input” in the sense of Müller & Hulk (2001) might be the reason for the difficulties observed in bilingual first language acquisition. In contrast to what Kupisch found for determiner omissions, language imbalance can be used to predict the language which causes difficulties for the child in the domain of word order within DP. Maria-José Ezeizabarrena investigates the occurrence of Basque root infinitives in Spanish-Basque bilingual children from the perspective of overt/

Introduction

non-overt subjects which occur in these structures, comparing the data to those from a monolingual Basque child. Interestingly, the author comes to the conclusion that the bilingual children do not differ, neither qualitatively nor quantitatively, from the monolingual child with respect to the use of root infinitives. Root infinitives are analyzed as null auxiliary structures, thus similar licensing mechanisms apply for null subjects in finite clauses and in root infinitives. The occurrence of root infinitives in child language corresponds to a point in development during which the children process complex verb forms. Marc-Olivier Hinzelin studies the acquisition of the null-subject parameter in two bilingual Portuguese-German children. The null-subject parameter has to be set to different values in the two languages: Portuguese is a null-subject language, whereas German is not. The author has chosen this parameter in order to show that the two children have separate grammatical systems from early on and that there is no cross-linguistic influence in this grammatical domain. Besides the omission of subjects, Hinzelin also studies the use of personal pronouns and the acquisition of finiteness in the two languages: Interestingly, the acquisition of finiteness marks the onset of target-like (pronominal) subject use, a relation one would also predict from grammatical theory. Hinzelin concludes that the null-subject property seems to be an invulnerable grammatical domain in early bilingualism. Ira Gawlitzek-Maiwald focusses on the investigation of utterances from two German-English bilingual children containing elements of both languages. These mixed utterances have a booster function or the function of “pooling of resources” in the sense that the slower language may profit from the weaker language in bilingual first language acquisition. The author finds that mixing is highly systematic and temporary in the children studied; more interestingly, it occurs in a developmental period in which the children have “structural gaps” in the slower language. In other words, mixing is a strategy followed by the bilingual child in order to cope with the asynchronic development in the two languages and to make use of her/his bilingualism in a positive way. Margaret Deuchar and Rachel Muntz are concerned with mixed utterances in an English-Spanish bilingual child. They study three specific factors which have been proposed in the literature to influence the children’s early mixing. The first factor is the type of interaction between the adult and the bilingual child. Adults may respond with more monolingual or more bilingual strategies when the bilingual child mixes the two languages and this behavior might influence the child’s mixing rate. The authors found no positive correlation between the adult’s interaction strategies and the child’s mixing rate. The next

ix

x

Introduction

factor is related to the nature of the adult’s input. One might expect that the child adjusts the proportion of code-mixing to that in the adult input. Again, the authors found no clear statistical relation between adult and child mixing. The last factor studied is the child’s relative proficiency in each language and its relation to the child’s frequency of mixing in each language. Here the prediction would be that the child mixes more in the non-dominant than in the dominant language. The subject studied by Deuchar and Muntz could be classified according to the criteria used as “balanced”. But even for the recording session where one could argue for a slight dominance, no correlation is found between the child’s dominance and the mixing rate. The authors suggest that selfawareness might be an explanatory factor in the sense that the child mixes more in the Spanish than in the English context because she becomes aware of the fact that her Spanish-speaking interlocutors can all understand English, whereas not all English-speaking interlocutors also understand Spanish. Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo study the acquisition of codas by German and Spanish monolingual and bilingual children. The two target-languages differ with respect to the syllable structure. In general, one can say that Spanish has much less complex syllable structure than German. Both languages contain syllables with codas, but German allows much more different consonants and consonant clusters at the end of syllables as compared with Spanish, which has strong restrictions on the type of elements allowed in codas. The monolingual data show that Germanspeaking children acquire codas earlier than their Spanish-speaking peers. This result is argued to be due to the greater frequency of codas in German than in Spanish and to the importance of codas in the phonology of German (displaying a high percentage of monosyllabic words for which truncation — a general pattern of language acquisition — would result in a violation of binarity, a prosodic requirement). With respect to the bilingual data, the authors argue in favor of early separation of two phonological systems. At the same time, they observe cross-linguistic influence in the sense of Gawlitzek-Maiwald (1996). They observe that the bilingual children acquire codas as early as monolingual children in German, but earlier as monolingual children in Spanish. The observation is interpreted as one language (German) having a bootstrapping effect for the other language (Spanish). The papers so far have concentrated on a relatively early age span in bilingual first language acquisition. The paper by Herkenrath, Karakoç and Rehbein looks at bilingual children at a more advanced age, between four and thirteen years.

Introduction

Annette Herkenrath, Birsel Karakoç and Jochen Rehbein investigate connectivity in five Turkish-German bilingual children. The bilingual data are compared with monolingual Turkish data. The comparison is the basis for the hypothesis that the languages in the bilinguals are influencing each other, in the sense that an innovative type of connectivity used in Turkish is due to the German influence. The paper addresses the type of influence involved in this process. Apart from the completely innovative part, the bilinguals use an option in Turkish which is only marginally available in Turkish to a much greater extent when they connect finite embedded interrogative structures to a matrix construction: The bilinguals display a greater tendency to move wh-elements to a construction-initial position and to postpone wh-constructions to the matrix construction, even in cases when they play a rhematic role. In other words, a peripheral property of Turkish was turned into a core feature of the Turkish as used by the bilingual children. The paper addresses the question whether the results of the study on the bilingual children can be interpreted as an emerging new variety of Turkish developing in Germany or in Western Europe. The summarized papers have focused on early bilingualism from birth. The papers by Mary Kato and Anja Möhring and Jürgen Meisel compare first and early child/adult second language acquisition. Mary Kato’s paper is a description and explanation of the bilingualism of the author, her L1 being Japanese, the L2 acquired during early childhood Brazilian Portuguese (BP) and the L2 acquired during adulthood and in nonnatural setting English. Kato looks at all three languages from a competence and a proficiency perspective. Interestingly, the language the author is most proficient in is BP, the L2. Although proficiency has deteriorated in Japanese and some aspects have fossilized, the author still is able to make grammaticality judgements in quite complex syntactic constructions. In other words, the competence has remained unharmed by fossilization. The question Kato poses for the acquisition of BP is whether she had direct or indirect access to UG. It is argued that for the head directionality parameter, direct learning of the functional lexicon allowed to set the parameter to the correct value in BP. With respect to the null-subject property, for which Japanese and BP are in a subset relation, the Japanese grammar represented the author’s first hypothesis about the grammar of BP, which was revised later during acquisition. Notwithstanding, Kato defines some areas in BP for which she is less proficient as compared to native-speakers who have acquired BP from birth: these involve automaticity in the gender agreement domain and prosodic aspects. In the case of Kato’s L2

xi

xii

Introduction

English, acquired after the critical age and without natural input, some of the properties of the specialized (language) module like automaticity and speed are lacking. The hypothesis put forward for the differences between BP and English is related to critical age, i.e. how early the L2 learner has started to acquire the L2. If there are critical (and different) ages to set prosodic parameters, phonological parameters, and syntactic parameters, one may explain why there are learners who are syntactically and phonologically competent, but not prosodically competent. If these critical ages follow the order given above and if this order is invariant, one may explain why there will be no L2 learner who is prosodically and phonologically competent but not syntactically competent. Kato suggests that the base to learning her L2 English might be her first L2, BP, in the sense that a peripheral property of her BP was turned into a core feature of English. Anja Möhring and Jürgen Meisel compare simultaneous and successive acquisition of two languages with respect to the head directionality parameter, OV — VO in particular. They study German-French bilingual children and L2 learners of German who have a Romance language as L1. The paper also investigates the cluster of phenomena related to the setting of the OV–VO parameter, adverb placement and particle constructions. They show that the parameter is set correctly at different values very early in the bilingual children and they find that the surface phenomena related to the setting of the parameter in grammatical theory are developmentally related in child grammar. As for adult L2 acquisition, they observe that the different surface phenomena are not related and that the learners seem to learn the placement of verbs and objects gradually and separately for different syntactic contexts (main vs. subordinate clauses). They suggest that L2 learners cannot set parameters to different values as opposed to the ones needed for the L1, since L2 acquisition can be characterized by the loss of access to the parametrized principles of UG. As a consequence, the L2 learners have to use other learning mechanisms like inductive learning in order to learn the OV property of the German target system. The differences of simultaneous and successive acquisition of two languages are ultimately explained as a result of different age of onset of acquisition. The papers thus far have all investigated individuals who acquire/have acquired more than one language. The paper by Thomas Roeper studies monolingual child and adult grammar under the perspective of a theory of multiple grammars where each module involves a different economy metric. More specifically, it is argued that economic factors have to be measured independently in each

Introduction xiii

module of grammar. Simplicity in one module generates complexity in another, e.g. syntactic simplicity can generate semantic complexity and vice versa. If particular modules of grammar (and particular speech registers, lexical classes etc.) make different choices with respect to simplicity, and if children prefer default choices early in development, more than one grammar is involved in every speaker and therefore every speaker accesses multiple grammars. On the basis of pied-piping, where the [+wh] feature to be checked at the CP level is “buried”, it is argued that the notion of Seek Sublabel is part of the economy system, in that economy requires minimal depth for Seek Sublable. The requirement of minimal depth for Seek Sublable is investigated with respect to different grammatical phenomena in child grammar, verb raising/do-insertion, wh-fronting and deletion of Agr. The analysis presented provides indirect support for Chomsky’s (2000) claim that Agr does not project an independent node but is subsumed under TP. The book is a collection of research contributions which have been presented at two conferences held at the Research Center for the study of Multilingualism of Hamburg University, founded by the Deutsche Forschungsgemeinschaft (German Science Foundation) in July 1999. I would like to thank the reviewers for their valuable comments: Peter Auer, Aurora Bel, Thomas Berg, Ute Bohnacker, Piero Bottari, Maria Teresa Guasti, Stefanie Haberzettl, Cornelia Hamann, Aafke Hulk, Regina Köppe, Elizabeth Lanza, Pilar Larrañaga, Elisabeth Loebel, Elena Nicoladis, Johanne Paradis, Teresa Parodi, Andrew Radford, Katrin Schmitz, Carla Soares, Jeanine Treffers-Daller, Elisabeth van der Linden, Fred Weerman. Thanks to Tanja Kupisch and Martin Hilpert for helping me with the final version of the manuscripts. Enjoy your trip!!

References Chomsky, N. 2000. “Derivation by Phase”. In: Ken Hale. A Life in Language, M. Kenstowicz (ed.), 1–52, Cambridge Mass.: MIT Press. Gawlitzek-Maiwald, I. and Tracy, R. 1996. “Bilingual bootstrapping”. Linguistics, 34, 901–926. Müller, N. and Hulk, A. 2001. “Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages”. Bilingualism: Language and Cognition 4(1), 1–21. Jakubowicz, C., Nash, L., Rigaut, C. and Gérard, C.-L. 1999. “Determiners and clitic pronouns in French-speaking children with SLI”. Language Acquisition 7, 113–160.



xiv Introduction

Müller, N., Cantone, K., Kupisch, T. and Schmitz, K. 2002. “Zum Spracheneinfluss im bilingualen Erstspracherwerb: Italienisch-Deutsch”. Linguistische Berichte 190, 157–206. Platzack, Ch. 2001. “The vulnerable C-domain”. Brain and Language 77, 364–377. Rizzi, L. 1993/94. “Some notes on linguistic theory and language development: the case of root infinitives”. Language Acquisition 3(4), 371–393. Roeper, T. 1996. “The role of merger theory and formal features in acquisition”. In Generative Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations and Crosslinguistic Comparisons, H. Clahsen (ed.), 415–449. Amsterdam/Philadelphia: John Benjamins. Sorace, A. 2000. “Differential effects of attrition in the L1 syntax of near-native L2 speakers”. In Proceedings of the 24th Annual Boston University Conference on Language Development, C. Howell, S. Fish and T. Keith-Lucas (eds), 719–725. Somerville, MA: Cascadilla Press. Wexler, K. 1998. “Very early parameter setting and the unique checking constraint: a new explanation of the optional infinitive stage”. Lingua 106, 23–79.



The DP, a vulnerable domain? Evidence from the acquisition of French* Tanja Kupisch Research Center on Multilingualism, University of Hamburg

Recent studies have shown that the acquisition of determiners, and particularly articles, appears unproblematic for children whose target language is French. Examining the acquisition of C-related phenomena in Swedish and German, Platzack (2001) argues that the CP represents a vulnerable domain, showing that grammatical operations related to the CP, unlike operations independent of this domain, are acquired comparatively late and that associated elements tend to be omitted frequently. According to the Configurational Hypothesis (Giorgi and Longobardi 1991) the structure of DP parallels clause structure. As a consequence, the DP could be expected to constitute a vulnerable domain as well. In the present contribution I will analyze longitudinal data from three children acquiring French: a monolingual and two GermanFrench bilingual children, one of them unbalanced with French as her weaker language. It will be shown that at least in French the DP appears to be an invulnerable domain. I will discuss this finding from a cross-linguistic perspective, arguing that frequency of determiners and consistency in their syntactic distribution facilitates and thus accelerates the acquisition process.

1.

Introduction

The notion of vulnerable domains has been used quite frequently in recent studies of language acquisition. A general picture emerging from various studies based on different languages is that particular grammatical phenomena are acquired relatively early while the emergence of others is rather late. Certain grammatical elements are rarely produced in a target-deviant fashion while others are prone to errors. This generalization seems to hold across different groups of language learners. In particular, the domains that are problematic for early first language learners, monolingual or bilingual (balanced or unbalanced),1 second language

2

Tanja Kupisch

learners, and children with language impairment, often coincide. Even crosslinguistic parallels can be drawn. Hence, grammatical systems are unstable in particular domains. Although the notion of vulnerability had been present in the literature for some time, it was a recent paper by Platzack (2001) titled The Vulnerable C-domain that popularized the use of this expression in order to refer to such unstable areas. Following Rizzi (1997) the author assumes that the purpose of the C-domain is to link the information present at IP and VP to the discourse. The C-domain is taken to consist of two parts: ForceP, facing outwards, and FinP, facing inwards. ForceP is a sentence type projection indicating whether a sentence is interrogative, imperative, or declarative. FinP has a head with the feature [finite] and licenses tense and mood in IP. Finiteness anchors a proposition in the real world. It is this interface character that is assumed to render the CP vulnerable.2 The idea of ‘anchoring’ is not unique to generative theories. For example, cognitive semantic approaches assume that tense markers have the function of locating a process to extralinguistic reality by expressing relations such as ‘prior’ or ‘simultaneous to the moment of speaking’. “[W]hereas a simple verb (e.g. fall) merely names a type of process, a finite clause (It fell; It may fall; It would fall) profiles a process instance and situates it with respect to the time and immediate reality of the speech event.” (Langacker 1991: 321)

In Langacker’s approach, tense is assumed to effect the grounding of finite clauses. The notion of ground is used for the speech events, the participants, and the time and place of speaking. Grounding elements have the function of locating abstract concepts with respect to the ground. Other grounding elements besides tense morphemes are articles, demonstratives, and certain quantifiers. These effect the grounding of nominals in languages such as English, French, and German. “Whereas a single noun (e.g. cat) merely means a ‘type’ of thing, a full nominal (this cat; some cat; any cat; a certain cat) designates an ‘instance’ of that type and gives some indication of whether and how the speech-act participants have succeeded in establishing mental contact with that particular instance.” (Langacker 1991: 321)

Cognitive semantic theorists like Langacker draw a parallel between tense markers and determiners which consists in their grounding function, i.e. in their function to relate abstract types of processes and objects to the speech event. Let us now return to generative theory. In this field parallels between

The DP, a vulnerable domain?

nouns and verbs, or more precisely, between noun phrases and clauses have motivated the suggestion that the noun phrase is headed by the functional category D. This so-called DP-Hypothesis originally goes back to Szabolcsi (1983/84) and Abney (1987) and has been adopted in most contemporary work in the domain of generative grammar. The idea that DP-structure parallels clause structure has become known as the Configurational Hypothesis (Giorgi and Longobardi 1991: 2) and it is present in a number of generative works on nominal structure (e.g. Valois 1991, Siloni 1997). It is assumed that both CP and DP consist of three layers. The rightmost layer contains lexical projections or, in cases of movement, traces of these. This is the domain in which thetaroles are assigned. The middle field can be characterized as the functional layer. Here, grammatical relations, e.g. noun-adjective agreement in the case of the nominal domain and Subject–Verb agreement in the case of the clause, are expressed. The leftmost layer has the function of connecting the proposition to the extralinguistic world. According to recent work, finiteness is checked in the domain of C (Rizzi 1997), and definiteness in D (Longobardi 1994). The idea can be formalized as in (1) for the CP and for the DP. DP/CP

(1)

D¢/C¢

D D0/C0

FP Spec

F¢ F0

NP/VP N¢/V¢

Spec N0/V0

XP

Platzack (2001) compares C-related and solely I-related phenomena in Swedish in the performance of early L1 learners, children with specific language impairment (SLI), L2 learners, patients with Broca’s aphasia, and an adult control group. The author finds that none of the speakers has problems with syntactic properties that are solely related to the I-domain, but that all except for the adult control group have problems with phenomena related to the C-domain.

3

4

Tanja Kupisch

He adduces evidence from acquisition studies in German where the same generalizations hold (Clahsen et al. 1996, Hamann et al. 1999, among others). Further evidence for the vulnerability of the C-domain has been provided by Hulk and Müller (2000) and Müller and Hulk (2001) in bilingual GermanItalian and Dutch-French children. Given their configurational parallels and the assumption that the CP represents a vulnerable domain, one should expect the DP to be vulnerable as well. Recall that the reason for which the CP is considered vulnerable is its interface character. As will be shown in Section 2, determiners encode both semantic and pragmatic information in addition to syntactic features. Hence, they are very likely to be vulnerable if location at the interface is taken to be a crucial factor. This article is organized in the following way. Section 2 provides a brief introduction to the use of articles, dealing with grammatical, semantic and pragmatic aspects. In Section 3 I will review cross-sectional and longitudinal studies dealing with the acquisition of determiners in French, concentrating on the empirical results. Section 4 presents the results of three longitudinal studies on the emergence and regular use of determiners in French based on data from one monolingual and two bilingual children, one of them unbalanced with French as her weaker language. The results will be discussed in Section 5 from the perspective of several theoretical approaches. A conclusion will be provided in Section 6.

2. The use of articles in adult French French has prenominal definite and indefinite articles. Articles are marked for singular and plural, and for feminine and masculine gender in the singular. Furthermore, French makes use of the partitive article. While the plural form of the indefinite article, des, is used to refer to an indefinite number of countable objects, the partitive denotes an indefinite amount of uncountable objects (names of substances such as sucre ‘sugar’, or abstract nouns such as amitié ‘friendship’). An overview is provided in Table 1. The plural of the indefinite article and the two forms of the partitive article are very similar in form and function. They are derived from the preposition de ‘of ’ and the definite article and refer to a nonspecific number of countable objects or a nonspecific amount of something uncountable. The use of French articles furthermore involves some phonological rules. The final vowel of the

The DP, a vulnerable domain?

Table 1.Articles in French singular masculine

plural feminine

both genders

def.

indef.

part.

def.

indef.

part.

le chat

un chat

du pain

la fleur

une fleur de la farine

def.

indef.

les chats/ des chats/ filles filles

definite singular article is dropped if followed by a noun with an initial vowel, as in l’hôtel ‘the hotel’, l’escalier ‘the stairs’, l’habitude ‘the habit’, l’explication ‘the explanation’. Indefinite singular and plural articles, if they precede nouns with an initial vowel, are subject to liaison, i.e. the word final sound of the determiner fuses with the first vowel of the following noun. Thus, les hôtels is pronounced as [le zotεl].3 French nouns are obligatorily preceded by an article when they occur in argument positions, as in (2). Articles are not used before proper names (cf. (3a) vs. (3b)) or before nouns in predicative contexts, as in (4a).4 The contrast between (4a) and (4b) illustrates that an article does not render the sentence ungrammatical. Rather, the nominal obtains a specific reading, i.e. it is presupposed that the physician is known to the hearer. Occasionally articles are also absent from constructions, often idiomatic, in which the noun complement is introduced by a preposition. Such cases are illustrated in (5). (2) a.

J’ai vu une chatte. ‘I saw a cat.’ b. *J’ai vue chatte. ‘I saw cat.’

(3) a.

J’ai rencontré Marie. ‘I met Mary.’ b. *J’ai rencontré la Marie. ‘I met the Mary.’

(4) a.

Mon voisin est physicien. ‘My neighbour is a physicist.’ b. Mon voisin est le physicien. ‘My neighbour is the physicist.’

5

6

Tanja Kupisch

(5) a.

Regarde la pauvre fait lui faire un bisou là avec amour et tendresse. ‘Look the poor one makes him give a kiss there with love and tenderness.’ b. Elle va lui transformer ses carottes en patates. ‘She will turn him his carrots into potatoes.’

In contemporary works within generative grammar articles are assumed to be generated in the DP-layer, specifically, under the head D, the position in which the referential features of nominals are checked. That is, D is associated with a feature [±definite]. In the presence of a definite article, D is specified by the feature [+definite], as in (6b), while it is specified by the feature [−definite] in the presence of an indefinite article, as in (6a). (6) a. Yesterday I saw a cat and a dog playing in the garden. b. Suddenly, the cat ran away.

According to Abney (1987), determiners take an NP or its extended projection as their complement. Furthermore the determiner is supposed to assign argumenthood to the nominal expressions it occurs with. The Universal Determiner Requirement (Longobardi 1994) states that nominals in argument position must involve a Det position. Prenominal elements such as articles, demonstratives and quantifiers are often treated as representatives of the class of determiners. Their complementary distribution has motivated the suggestion that they all occupy D0, the topmost functional projection of the nominal domain. Giusti (1997) argues that this complementary distribution is not universal, taking this as a starting point in arguing that the elements do not constitute a homogeneous category. She claims that only articles are purely functional elements that originate in D0 as extended heads of the noun phrase, while demonstratives and quantifiers are lexical elements. Demonstratives are generated in specifier positions lower in the syntactic tree than the article and are moved to SpecDP to check their referential features. Quantifiers are subdivided into two groups: They are adjectives when preceded by other determiners as in Italian, e.g., i due ragazzi ‘the two boys’, and they are lexical heads when they select a full DP as their complement, as illustrated in (7).

The DP, a vulnerable domain?

QP

(7) Spec

Q¢ Q0

DP

tous ‘all

les garçons the boys’

From a distributional perspective, Giusti’s (1991, 1997) arguments in favor of an independent projection QP above DP are highly plausible, although her claim that “contrary to other determiners, articles are not inserted for semantic, but rather for grammatical (morpho-syntactic) reasons” (ibid. 104) might be considered problematic. Giusti mainly draws on examples where the definite article appears before nouns that do not have specific reference. However, there is more to article use than the specific vs. nonspecific distinction. As early as 1973 Brown observed that “the semantic rules governing [article] use can only be stated in dyadic terms, the dyad being that of the speaker and listener” (ibid. 341). According to Brown, the referential aspects of article use are determined not only by semantic but also by pragmatic factors. Apart from deciding whether a referent is specific or nonspecific, the speaker has to take into account whether or not the listener is familiar with the referent. To give an example, a speaker uttering (6a) chooses the indefinite article, even though the referent is specific for her/him, because the referent is presupposed to be unknown to the hearer. Furthermore, the definite article can be used in prototypical and generic statements, as in (8), although no particular referent is instantiated. In such cases count nouns obtain a homogenized reading. (8) a.

Les chats sont des mammifères. (generic reading) ‘Cats are mammals.’ b. J’adore les chats. (prototype reading) ‘I love cats.’

In generative theory, definite articles in cases such as (8a,
b) are also referred to as “expletive” articles. Such articles do not make the DPs refer to any particular referent. Similarly, articles before proper names, e.g. in German or in some Italian dialects, are referred to as expletives, because proper names are said to be intrinsically referential in these languages (e.g. Penner and Weissenborn 1996: 165).5 In both cases the article does not contribute to the establishment of a referent. It is said that such definite articles “cannot be associated with an

7

8

Tanja Kupisch

overt semantic content, and their presence can only be accounted for in syntactic terms” (e.g. Bottari et al. 1993/94: 289).6 In the present study I will depart from the assumption that all of UG is available to the child from the outset in linguistic development. The child’s task in the acquisition process consists of the acquisition of the forms and functions in his/her target language that are associated with particular syntactic categories, such as D.

3. Studies on the omission of articles in early child speech The studies summarized in this section are concerned with acquisition data from French. They have all shown that articles start to be used at a comparatively early age and that target-like use is attained within a relatively short period of time in this language. Chierchia et al. (1999, 2000) have examined article omission in 16 monolingual children speaking English, Swedish, Italian, and French. They defined three stages in the acquisition process: (i) the bare noun stage (henceforward BN-stage), in which all nouns occur bare, (ii) the free variation stage, characterized by an optional use of articles, (iii) the target-stage, in which articles are regularly provided in obligatory contexts. Their cross-linguistic investigation revealed that a target-like use of articles is rather late in children acquiring English and Swedish as compared to children acquiring French and Italian. Moreover, the free variation stage is clearly shorter in the latter group. Regarding the four French children they noted that article omission practically ceased with a mean length of utterance (MLU) of between 3 and 3.5.7 For Philippe they observed that during the entire period examined, covering the age span between 2;1 and 2;6, omission is never above 8%. It should be added that none of the children were older than 2;6, i.e. in all of them target-like use was attained before that age. Granfeldt (2000) compared the acquisition of articles in three bilingual Swedish-French L1 learners to that of adult Swedish learners of French. The group of bilingual children differed with regard to the point at which the 90% criterion of article use in obligatory contexts was reached but the author showed that this was due to individual differences in their linguistic development. Mimi, whose stronger language until the age of 2;6 was French, reached this point at age 2;2 (MLU 3.2). Anne, also with French as her stronger language, reached the 90% criterion only at the age of 3;7 (MLU 3.6), but she has been

The DP, a vulnerable domain?

shown to be a comparatively slow learner in previous studies (cf. Schlyter 1994). John, with French as his weaker language, seems to be an exception. 90% of use in obligatory contexts was reached only at the age of 3;9 (MLU 4.3) (ibid. 272), but Granfeldt himself, who does not apply the 90% criterion, suggests that Jean acquires articles between the age of 2;6 and 2;9, reporting that “from 2;6 onwards he makes very few omissions even in Det-Adj-N contexts.” (ibid. 275) Granfeldt’s group of L2 learners consisted of four adult Swedish learners of French with time of exposure to French ranging between three and eight months. He found that omission errors were very limited, most of the time below 10%, even at the early stages. Omissions typically occurred in contexts in which their first language, Swedish, requires the noun to be bare. For instance, Martin produced jouer trompette ‘play trumpet’ instead of jouer de la trompette probably due to the Swedish translational equivalent spela trumpet. Other omissions were found before indefinite plural nouns, i.e. contexts in which Swedish nouns occur bare but in which French requires the use of the indefinite article des. Jacubowicz et al. (1998) have looked at the elicited production and comprehension8 of definite articles, nominative and accusative clitics and reflexive pronouns in a group of 13 French-speaking children with SLI and 20 normally developing children. The former group ranged in age between 5;7 and 13 years. The latter were aged between 5;6 and 5;11. With all expression types except articles there was a significant difference between the two groups in the accuracy rates attained. While normally developing children were at high levels of accuracy for all tested types, children in the SLI group reached a significantly higher accuracy rate in their use of articles than in their use of clitics and reflexives. Both normally developing children and all but one of the SLI children employed articles to 100% in obligatory contexts. Jacubowicz (1999) compared the acquisition of functional categories hosting tense morphemes, determiners, and clitics by normally developing children and children with SLI. She conducted a study with 12 three year-olds, 12 four year-olds, a control group of ten adults and 28 children with SLI ranging in age from 5;6 to 13 years. The results for clitics and determiners replicated those obtained by Jacubowicz et al. (1998), revealing that accusative clitics were subject to impairment, while determiners tended to be used in obligatory contexts. Both groups of normally developing children provided determiners more than 95% of the time. In the SLI group 16 children (ranging in age between eight and ten years) displayed accuracy rates beyond 95%, 12 children (mean age 6;10) had accuracy rates beyond 80%. The high accuracy

9

10

Tanja Kupisch

rates in determiner use by normally developing children in this study as well as in Jacubowicz et al. (1998) are expected because the children were at least three years of age. On the other hand, it is surprising that even the SLI children in both studies hardly ever omitted any determiners. Berkele (1983) has provided a detailed description of the emergence of articles in the bilingual German-French child Caroline. Caroline already uses various article forms (les, la, le, un) at the age of 1;10. At first, bare nouns prevail; between the age of 2;4 and 2;7 their number decreases with the use of definite articles increasing. The number of BNs decreases to 0% at the age of 2;7,20 (morpheme based MLU 3.2). Interestingly, Berkele shows that the article is typically omitted before mass nouns encoding nonspecific reference, as in c’est chocolat ‘it’s chocolat’. That is, Caroline treats these objects like nonindividualized mass nouns as they are in German. Recall that a similar observation was made by Granfeldt for adult Swedish learners of French. Koehn (1994) studied the acquisition of number morphology within the DP in the bilingual German-French child Ivar. The author points to an interesting difference in the grammatical encoding of plurality in German and French, noting that “in French Ivar begins by marking number on articles while in German first systematic markings occur on nouns” (ibid. 45). This might be taken to imply that due to form syncretism the article is not a reliable plural marker in German. For instance, the form die ‘the.FEM’ is not used to mark plural nouns exclusively, but it is also the form representing the feminine singular nominative and accusative articles. In German, therefore, the noun ending is a more reliable plural marker than the article. The opposite holds for French. Here, plural marking on nouns is mostly not audible. On the other hand, the form of the plural article is used solely to refer to plural referents.9 Consequently, the article is a reliable plural marker, while the noun ending is not. One might suspect that this grammatical function of les and des accelerates the use of plural articles or articles in general.

4. The use of articles by a monolingual and two bilingual children acquiring French 4.1 The data For the present investigation three French corpora were used. The monolingual corpus, Grégoire, available in CHILDES (Child Language Data Exchange

The DP, a vulnerable domain?

System, Mac Whinney and Snow 1985), was collected by Christian Champaud, researcher at the Centre National de la Recherche Scientifique (C. N. R. S.). Grégoire grew up in Paris, France. He has two elder brothers. The recordings, lasting between 30 minutes and an hour, were conducted at the child’s home in natural play situations. Besides the investigator, the mother was always present during the recordings. Sometimes the father, grandmother, or one or both brothers were also involved. All available transcripts, covering the age span between 1;9,18 and 2;5,27 (10 recordings), were analyzed. The bilingual data originate from the research project “Early bilingualism in childhood: Comparing Italian/German and French/German”, directed by Natascha Müller, University of Hamburg. Both Alexander and Céline grew up in Hamburg, Germany. Alexander’s father is German, his mother is French and speaks German very well. The opposite case is met in Céline’s family, in which the mother is German and the father is French. Both children have an elder brother. The bilingual children were also studied longitudinally. They were videotaped twice monthly, sometimes with longer breaks due to visits to France. The interviewers, one French-speaking, one German-speaking, visited each child at home and played with her/him for approximately half an hour in each language. While the French investigator was playing and speaking with the child, the German investigator operated the video equipment, and vice versa. Thus, the child always interacted with either of the two interlocutors. Sometimes the person recording was spontaneously involved in the interaction. The age period examined for Alexander covers the age span between 2;2,6 and 2;11,20 (17 recordings). A comparison of MLU values and language use in his two languages points to a slight dominance of Alexander’s French as opposed to his German (see Figure 1). I will, however, not discuss the issue of language dominance in Alexander, because the present study is limited to the investigation of French, where his linguistic development is comparable to that of monolingual French children. The investigation period for Céline covers the age span between 2;1,14 and 4;0,19 (43 recordings). It was necessary to examine a longer period of time, because Céline starts using French productively only by age three. Céline differs from Alexander and Grégoire in that her French develops slower (see the MLU in Figure 2), and is produced with lower frequency (see the number of utterances in Figure 3). Thus her French appears to be comparatively “weak”. The weaker language issue has been addressed extensively in the works of Schlyter and colleagues (Schlyter 1993, 1994, Schlyter and Bernardini 2001, Schlyter and Håkansson 1994). The weaker language is acquired more slowly

11

Tanja Kupisch

5 4,5 4 3,5 3 2,5 2 1,5 1 0,5 0 2; 2, 2; 6 2, 2 2; 0 3, 2 2; 4 4, 2 2; 0 5, 25 2; 6, 8 2; 7, 2; 6 7, 2 2; 7 8, 1 2; 2 8, 2 2; 8 9, 1 2; 8 10 2; ,2 10 ,2 2; 3 11 2; ,6 11 ,2 0

French German

Figure 1.MLU values of Alexander (French vs. German). 6 5 4 Céline Alexander Grégoire

3 2 1

1

3

4

6

3;

3;

9 2;

3;

7 2;

0

6 2;

3;

4 2;

3;

3 2;

10

1 2;

2;

0 2;

9 1;

10

0 1;

12

Figure 2.MLU values of Grégoire, Alexander and Céline in French.

and produced with lower frequency, not only with respect to other children acquiring the same language (cf. Figures 2 and 3), but also compared to the stronger language. The remaining part of this section will be concerned with a comparison of Céline’s French and her German. The criteria to determine language proficiency used here are MLU, number of utterances per session, number of verb types, as well as language mixings (cf. also Genesee, Nicoladis and Paradis 1995). Figures 4–6 suggest that Céline’s French is weaker than her German in terms of syntactic complexity, frequency, and with respect to the lexicon. A noticeable increase in the number of utterances by age 2;11 suggests that she starts to use French more productively from that moment, while an increase in the complexity of her utterances (reflected in higher MLU values) is observed only by age 3;3.

The DP, a vulnerable domain?

1

3

4

6

3;

3;

9

3;

7

2;

0

6

2;

3;

4

2;

10

3

2;

3;

1

2;

2;

0

2;

10

2;

Céline Alexander Grégoire

1;

1;

9

500 450 400 350 300 250 200 150 100 50 0

Figure 3.Number of utterances per session (Grégoire, Alexander and Céline in French). 6 5 4 3

Céline (Fr.) Céline (Ger.)

2 1

3;5

3;3 3;4

3;0 3;1

2;1 0 2;1 1

2;7 2;8

2;5 2;6

2;3 2;4

2;0 2;1

0

Figure 4.MLU values of Céline.

Another typical characteristic of the weaker language is the unidirectionality of mixing. That is, elements from the stronger language are mixed into the weaker language but not vice versa. Figure 7 illustrates the portion of entirely German utterances (i.e. intra-discourse mixings), and intra-utterance mixings (i.e. utterances containing elements from both languages), and entirely French utterances in Céline when addressing the French interlocutor. There is a clear difference in the amount mixings10 in French, as compared to German (Figure 8). While the amount of mixed utterances in German is maximally 2.3% but mostly 0%, the share of non-French utterances in the French recording part amounts to more than 90% in several recordings, particularly before age 2;7. Thus, another typical characteristic of the weaker language is met. Moreover, it

13

Tanja Kupisch

500 450 400 350 300 250

French

200

German

150 100 50

1

3

4

5

3;

3;

3;

8 2;

0

7 2;

3;

6 2;

3;

5 2;

11

4 2;

10

3 2;

2;

1 2;

2;

0 2;

0

Figure 5.Number of utterances per session in Céline. 90 80 70 60 50 40 30 20 10 0

French German

2;0 ,9 2;1 ,6 2;3 ,15 2;4 ,13 2;5 ,24 2;6 ,21 2;8 ,2 2;8 ,29 2;1 0,5 2;1 1 2;1 ,3 1,2 9 3;0 ,27 3;3 ,12 3;4 ,9 3;5 ,15 3;6 ,12

14

Figure 6.Verb types in Céline (from Cordes 2002).

has been shown that Céline makes use of particular communicative strategies in her weaker language. Her speech contains a large number of speech patterns, as e.g. the pattern c’est ART N ‘that’s ART N’ (Kupisch 2001). To summarize the previous section, Céline clearly shows an unbalanced development in her two languages. Her French evolves slow both compared to other children and to her German. The comparison of MLU values and the number of utterances in Alexander and Grégoire suggests that Alexander’s development is similar to that of monolingual French children.

The DP, a vulnerable domain?

100% 80% 60% 40%

intra-utterance mixings German French

20%

2; 0, 9 2; 1, 14 2; 4, 13 2; 6, 7 2; 8, 2 2; 9, 20 2; 11 ,3 3; 0, 13 3; 3, 12 3; 4, 23 3; 6, 12

0%

Figure 7.Language mixings in Céline’s French (from Cordes 2002). 100% 80% 60% 40%

intra-utterance mixings French German

20%

2; 0, 9 2; 1, 14 2; 4, 13 2; 6, 7 2; 8, 2 2; 9, 20 2; 11 ,3 3; 0, 13 3; 3, 12 3; 4, 23 3; 6, 12

0%

Figure 8.Language mixings in Céline’s German (from Cordes 2002).

4.2 The study This section presents an empirical investigation of the emergence and use of articles in obligatory contexts in the three children introduced above. Determiners other than articles were counted as well, for two reasons. First, current theories, according to which articles and other determiners are associated with one and the same syntactic position, predict a simultaneous appearance, or at least near simultaneous appearance of these elements. More importantly, calculation of omission rates solely on the basis of article-noun sequences and BNs is not justified because it is impossible to tell whether the omitted element

15

Tanja Kupisch

would have been an article or any other kind of determiner. Nor can we determine the exact status of presyntactic devices, which have also been included in the study. Following Chierchia et al. (1999) I will use the term “BN-stage” when referring to the period prior to the emergence of determiners; the use of determiners is considered target-like when the omission rate in obligatory contexts is below 10%. Grégoire had passed the BN-stage when he was recorded for the first time (1;9,18). Until the age of 2;0,5 (MLU 2.38), most prenominal elements represent monosyllabic placeholders (henceforward MPHs) in the form of [6], [e], and [a] (cf. Bottari et al. 1993/94 for a definition of MPHs). Omission rates range between 55% and 75%. From the age of 2;1,25 (MLU 2.69) there was a noticeable decrease in the use of MPHs, and the omission rate dropped to 34%. The use of determiners other than articles strongly increased after 2;3 (MLU 2.76), thus at least two months after the emergence of unreduced article forms. Determiner omission was below 10% from the age of 2;5,1. Grégoire’s MLU had then increased to 3.96. Figure 9 illustrates the absolute number of contexts for determiner use and omission. Articles and other determiners were subsumed under the category Det+N; MPHs and phonologically reduced forms of determiners were categorized as “proto-determiners”. Similar to Grégoire, Alexander had already passed the BN-stage when he was recorded for the first time. In the first recording at the age of 2;2,6 (MLU 2.33) he already used various article forms (un, une, le, la). A target-like use is attained by the age of 2;4,20 (MLU 2.74), when determiner omissions were below 10% for the first time. The forms comprised the two plural articles des and les, as well as other determiners (e.g. deux ‘two’, cette ‘these’, beaucoup de ‘a 200 180 160 140 120 100 80 60 40 20 0 2;5 ,1 2;5 ,13 2;5 ,27

2;3 ,0

BNs proto-determiners Det+N

2;0 ,5 2;1 ,25

1;9 ,18 1;9 ,28 1;1 0,2 0 1;1 1,2 2

16

Figure 9.Absolute number of BNs, proto-determiners and DPs in Grégoire.

The DP, a vulnerable domain?

140 120 100 80

BNs Det+N

60 40 20

2;11,20

2;11,6

2;10,23

2;10,2

2;9,18

2;8,28

2;8,12

2;7,27

2;7,6

2;6,8

2;5,25

2;4,6

2;4,20

2;3,24

2;2,27

2;2,6

2;2,20

0

Figure 10.Absolute number of BNs and DPs in Alexander.

lot of ’, tout ‘all’). The results are shown in Figure 10. Note that Alexander, unlike Grégoire, does not use any proto-determiners, possibly due to the fact that he was recorded at a later age. Céline’s linguistic development with regard to the DP can be subdivided into three stages. Up to approximately age three the number of French utterances (mean 28 per recording) and the number of French DPs (mean 3.8 per recording) are very low. The mean MLU is just below 2.0. In the second stage (age 2;11,29–3;1,10) there is a sudden rise in the number of French utterances (mean 126 per recording) and French DPs (mean 21 per recording), the MLU remaining around two. This sudden enlargement of the (nominal) lexicon with no corresponding enlargement in structure (cf. the MLU) was probably due to the presence of Céline’s French speaking grandmother during that time. In the third stage, the mean number of DPs did not increase noticeably, but there was a strong increase in the MLU, corresponding with an increasing use of lexical verbs (cf. Figure 6). In particular, while the occurrence of nominals in the first two stages was restricted to isolated DPs and presentationals like c’est un _ ‘that’s a _’, nouns were increasingly used as arguments of lexical verbs in the third stage. Furthermore, at 3;4,9 (MLU 3.57) Céline started to use determiners other than articles: numerals like deux ‘two’, possessives like ma ‘my’, sa ‘her/ his/its’, ses ‘their’, interrogative pronouns like quel ‘which’ and indefinite pronouns like chaque ‘each’. This significant change in her development

17

Tanja Kupisch

Table 2.Developmental stages in the production of DPs by Céline

age span duration no. of recordings no. of DPs mean no. of DPs/session mean no. of utterances/ session mean MLU

First stage

Second stage

Third stage

2;0,9–2;11,15 0;11,6 20 76 3.8 28 (min. = 1/ max. = 119) 1.67

2;11,29–3;1,10 0;1,11 4 84 21 126 (min. = 82/ max. = 188) 2

3;3,12–4;0,9 0;9,17 19 409 21.5 116 (min. = 57/ max = 172) 3.1

coincides with her return from a long stay in France during the summer vacation. Table 2 illustrates the development sketched above. The first recording was conducted at the age of 2;0,9 (MLU 1.5). Single instances of determined nouns, un bête ‘an animal’ and une girafe ‘a girafe’, occurred in the third and fourth recording. At the age of 2;3,15 (MLU 1.62) Céline used 9 determiners (4 different types) and article use seems to be productive, but their number decreases to values between zero and four in subsequent sessions, in which she hardly produces any French nouns at all. It is thus impossible to say whether determiner use is really productive before the recordings at age 2;7,19 and 2;8,2, in which 13 and 12 determiners respectively are produced. Figure 11 shows the absolute number of DPs as compared BNs in Céline. As mentioned before, first language learners typically pass through a BN-stage, followed by a stage of inconsistent determiner use, before the target50 45 40 35 30 25 20 15 10 5 0 2; 0, 9 2; 1, 14 2; 4, 13 2; 6, 7 2; 8, 2 2; 9, 20 2; 11 ,3 3; 0, 13 3; 3, 12 3; 4, 23 3; 6, 12 3; 8, 0 3; 9, 18 3; 11 ,1 4; 0, 19

18

Figure 11.BNs and DPs in Céline.

BNs Det+N

The DP, a vulnerable domain?

stage is reached (cf. Chierchia 1999). Céline follows this pattern, but her development is delayed with respect to other children. Since the amount of BNs never exceeds the number of 7 (cf. Figure 11), BN- and free variation stage appear less pronounced. While Grégoire passes through a phase with a high number of BNs, which decrease with age (cf. Figure 9), Céline produces very few BNs, and these do not cluster at a particular stage. The use of determiners by Céline can be considered target-like from the age of 3;0,27. The highest MLU value that had been reached until that moment is 2.69, a value even lower than that of the other children when approaching the adult grammar. The omission rate had been 0% twice before (age 2;8,2 and 2;9,20), but higher omission rates in subsequent sessions (often due to omissions before singular count nouns) suggest that the target-stage has not been reached by then.11 Omission rates above 10% in recordings after age 3;0,27, by contrast, are either due to very low absolute numbers (age 3;1,10 and 3;9,18) or to a few omissions before mass nouns (age 3;11,15 and 4;0,19). These results show that in terms of MLU Céline patterns similar to other children acquiring the DP in French, only is her development delayed with regard to age.12 One could argue, however, that a delay with respect to age is not the only point in which Céline differs from the other children. First, the number of BNs that she produced before reaching the target-stage is very low compared to other children. It seems as if she converged to the adult grammar without really passing the other two stages. This could be due to language influence. Gawlitzek-Maiwald and Tracy (1996) and Schlyter and Bernardini (2001) proposed that early mixings in the language of bilingual children might 100 90 80 70

Céline

60

Alexander

50

Grégoire

40 30 20 10

7

11

6

3;

3;

4

3;

9 10

3

3;

3;

1

3;

3;

0

7 2;

3;

6 2;

3;

4 2;

9 10

3 2;

2;

1 2;

2;

0 2;

1; 9 1; 10

0

Figure 12.Relative number of determiner omission in Céline, Alexander, and Grégoire.

19

20

Tanja Kupisch

represent a temporary strategy of syntactic bootstrapping. More precisely, the child uses his/her expertise in one language to solve problems in the other language (cf. also Müller et al. 2002). It is intuitively plausible to assume that Céline’s German has a booster function as well, only that bootstrapping does not result in mixing in her case. In German, Céline’s use of determiners was target-like from the age of 2;4,13. Thus, when the child started to produce French nouns more frequently, she already knew that they are typically preceded by articles in German and could easily have discovered that this applies to French as well. This line of reasoning would explain the almost complete absence of BNs from her French. On the other hand, it would also predict a much earlier convergence to the target-stage in French. Instead, Céline keeps omitting determiners in French after the age of 2;4,13, even before singular count nouns. The latter would be unexpected if there was such a positive influence from German. Another point in which Céline differs from the other children is that there are four recordings with omission rates above 10% during the period previously defined as the target-stage. Only twice (age 3;1,10 and 3;9,18) these can be traced back to low numbers. In the other two recordings (age 3;11,15 and 4;0,19) they are clearly due to a few omissions before mass nouns, as in (9), contexts in which the article would be obligatorily absent in German under a non-specific reading. Thus, there seems to be language influence, but at a later stage. (9) a.

et sucre (Céline 3;11,15) ‘and sugar’ b. eh ben avec mayo e ketchup (Céline 3;11,15) ‘er, with mayonnaise and ketchup’ c. mais il va café d’dans (Céline 4;0,19) ‘but it goes coffee in’ d. là/p’tit café (Céline 4;0,19) ‘there/small coffee’ e. c’est thé (Céline 4;0,19) ‘that’s tea’

Nonetheless, it appears that the above cases constitute exceptions rather than the rule. First, Céline produces a low number of mixed DPs as compared to entirely French DPs (mixed DPs constitute 30% before, and 8.7% after age 2;11,15). Second, besides the utterances in (9), there are other contexts in which German, unlike French, does not allow the use of a determiner. Such contexts are illustrated in (10a–d) together with the German equivalent (the examples are taken from the corpus).

The DP, a vulnerable domain?

(10) a.

[looking at animals] des sominix [= animaux] (Céline 3;0,13) ‘(indef.pl) animals’ German: ‘Tiere’ b. c’est quoi du foin? (Céline 3;5,15) ‘it is what (part) hay?’ German: ‘Was ist (das,) Heu?’ c. ça c’est — du café pour la barbie (Céline 3;5,29) ‘that that’s — (part) coffee for the barbie’ German: ‘Das (da) das ist — Kaffee für die Barbie.’ d. oui toutes les couleurs (Céline 3;0,27) ‘yes all the colors’ German: ‘Ja alle Farben.’

In such cases of potential transfer, one would expect target-deviant utterances if there was influence from German. A quantification of these contexts in the Céline-corpus is illustrated in Table 3, which provides the number of possible contexts for transfer and the number of correct article uses.13 Table 3.Use of articles with nouns which in German do not require articles Contexts

No. of contexts

Correct

Target-deviant

Target-deviant (%)

a.indef. plural b.generic c.mass d.others

8 1 26 2

8 1 19 2

0 0 7 0

0% 0% 26.9% 0%

Total

37

30

7

18.1%

The table reveals that articles are regularly provided in 81.9% of all contexts. Only with mass nouns a slight tendency to omit the article was noted.14 Although most of them were correctly preceded by the partitive article, as in (10c), a small number of 7 cases occurred bare. On the whole, the numbers suggest that Céline develops a grammatical system for French determiners independently of her dominant language. Furthermore, mass nouns constitute less than 7% of the total number of nouns in the adult input to the child.15 As shown above, Céline’s omission of determiners with mass nouns is not consistent; it is more manifest after than before the target-stage is reached, and it does not appear to have a noticeable effect on the acquisition process.

21

22

Tanja Kupisch

To conclude, the results obtained in the present investigation, including the results of previous studies, permit the following generalizations. Most French children start to use articles and other determiners in a target-like fashion from between the ages of 2;2 and 2;7 with an MLU ranging between 2.7 and 3.5. This generalization holds for monolingual children and also for some of the bilinguals, as e.g. Alexander, Mimi (Granfeldt 2000), and Caroline (Berkele 1983). Another group of children does not match the above pattern with respect to age, but nonetheless with respect to MLU. This seems to be the case, e.g., for the bilingual Swedish-French girl Anne (Granfeldt 2000), and for Céline. These two children do not have any problems in acquiring the DP. Their development is delayed only with respect to age. As a result, the French DP does not appear vulnerable. It is not even so in unbalanced children whose dominant language could potentially motivate the production of target-deviant BNs in certain contexts. Céline’s case shows that the developmental pattern displayed by children whose linguistic development is delayed with respect to monolingual children may differ from that typical of monolingual children. It also seems as if some bilingual children and adult learners of French tend to produce BNs due to language contact with a Germanic language. However, neither language contact nor delay prevent the children from attaining a target-like use of articles with an MLU comparable to that of monolingual children.

5. Discussion The results obtained here and in previous studies provide several pieces of evidence in favor of the invulnerability of the syntactic category DP in French. The question then is, what are the properties that make the French DP easy to acquire? In trying to answer this question I will discuss acquisition theories dealing with the omission of functional categories in early child speech in Sections 5.1 to 5.5. The basic idea characterizing generative approaches to language acquisition is the assumption of the so-called language faculty, a component of the human mind dedicated to language. The language faculty has an initial state, which, under the influence of the environment, undergoes changes until it reaches the final state, or target grammar. The properties attributed to UG have not been constant within the past decades, especially if one compares the Principles and Parameters Framework with Minimalism. There is still no unified account of

The DP, a vulnerable domain?

what constitutes UG and the nature of the parameters. Given the diversity of approaches, deviating to a greater or lesser extent, I cannot be exhaustive in the following presentation and will mainly focus on studies that have dealt with the acquisition of determiners. 5.1 Theories in the Principles and Parameters Framework Within the Principles and Parameters Framework two main opposing views can be distinguished. Supporters of the Full Competence or Continuity Hypothesis suggest that child grammar is identical to adult grammar at any moment in acquisition. Opponents of continuity, on the other hand, deny that all structure is available from the outset of the acquisition process. There are a number of hypotheses differing with regard to what elements are assumed to be absent from child grammar. Many scholars have proposed the lack of functional categories as opposed to lexical categories. A possible explanation for the initial absence of functional categories is the Maturation Hypothesis according to which functional categories emerge by some kind of biological maturation approximately at the age of two. Radford (1990) cites numerous examples of BNs produced by English children below the age of two. He says that theta-theory, X¢-theory, and phrasal projections of lexical categories are available to these children, but functional categories are not. Radford suggests that the appearance of the functional category D is due to genetically programmed maturation (1990:218). The maturation account has often been criticized because of contradictory evidence from languages other than English. Especially in languages with rich morphology children can be shown to use functional morphology, bound and unbound, from very early on in acquisition (see e.g. Guasti 1993/94). Also, under the assumption that it is functional categories that mature, cross-linguistic variation cannot be accounted for, because it would predict the simultaneous emergence of a given functional category in all languages. Supporters of strong continuity suggest that child grammar is identical to adult grammar at any moment in acquisition. While defenders of maturation have to account for the presence of functional categories before maturation, defenders of continuity have to account for their initial absence. In the following I will discuss some of the proposals couched in terms of continuity. Apart from Chierchia et al. (1999, 2000), all approaches discussed presuppose the Universal Determiner Requirement, according to which nominals in argument position must involve a Det position. This requirement is met differently in individual languages, i.e. languages differ with respect to the extent to which they allow BNs.

23

24

Tanja Kupisch

In French, as in Italian, the Det requirement tends to be satisfied at PF through insertion of an overt determiner.16 According to Hoekstra et al. (1997), the D position is universally projected in the adult language whether or not a noun is preceded by an overt determiner. The same holds for child language. That is, D is syntactically active and can be spelled out by determiners even in very early stages of acquisition. The occurrence of BNs in child language is attributed to the underspecification of the morphological feature “number” that is encoded by determiners. For instance, if D is specified with the feature [+singular], it must be filled by a determiner encoding this feature, as in (11a). If, on the other hand, D does not have any number feature, it can remain empty, as illustrated by (11b). The child can represent nominals either as in (11a), or as in (11b), whereas the adult has only access to (11a). DP

(11) a.

DP

b.

D [+Sg]

NP

D

NP

the

N

N

cat

cat

Chierchia et al. (2000) note that the approach fails to explain the developmental sequences observed in longitudinal studies of article omission (BN-stage, free variation stage, target-stage). In order to capture the total omission in the beginning, Hoekstra et al. would have to assume not only an optional but an obligatory underspecification of number. But then, how does the change from the first stage to the free variation stage come about? Furthermore, the authors claim that the underspecification of number also accounts for the initial absence of subject pronouns and verbal inflection. Yet, these elements often emerge at different points in the acquisition process. This could be dealt with by assuming that some elements encode more features than others and are thus more “complex”. Such an explanation in terms of complexity would also explain cross-linguistic variation. However, it is contradictory to the observation made by Koehn (1994) with regard to the encoding of number features. To recall, Koehn noted that bilingual German-French children tend to encode number features in German on nouns before using articles. Thus, the point when children realize the necessity of number specification does not necessarily correspond to the onset of article use.

The DP, a vulnerable domain?

Unlike Hoekstra et al., Penner and Weissenborn (1996), claim that the acquisition of determiners may be delayed because it involves morphological information that is initially not accessible. Based on acquisition data from German, the authors argue that the earliest articles have to be analyzed as “impostors” that have neither morphological nor semantic content. The first “real” determiners that children use are expletive definite articles, i.e. articles that occur in generic nominals (e.g., die Milch ‘the milk’), situative and discourse unica (e.g., die Sonne ‘the sun’), and proper names (e.g., die Maria, ‘(the) Mary’). Given the lack of semantic content, their occurrence makes the child realize the syntactic need to overtly spell out D. In other words, such elements trigger the Det requirement.17 Penner and Weissenborn’s approach is empirically substantiated, but looking at longitudinal data of German, Eisenbeiss (2000: 45) observes just the contrary, namely, that children start to use expletive articles only when the use of “semantically motivated” articles is almost or already target-like. Moreover, it is debatable whether the above DPs are really semantically empty. Secondly, proper names are not preceded by articles in all languages and varieties of German. In fact, such nouns remain bare e.g. in Standard German, French and some Southern Italian dialects.18 Furthermore, studies dealing with referential aspects of articles use have shown that generic utterances scarcely occur in child speech before the age of three (e.g. Serratrice 2000, Kupisch 2000). Of course, under the assumption that the child needs very little experience and a single trigger is enough for the child to acquire certain elements, their argumentation is plausible. It is clear that input alone cannot explain acquisition, nevertheless, its relevance should not be underestimated. If it did play a role, this would mean that the number of potentially available triggers is substantially reduced in some languages or varieties. Also, the theory would predict an earlier emergence of determiners in children acquiring e.g. a Southern German or Northern Italian dialect than in children acquiring a Standard German dialect or French, contrary to fact. An alternative solution to the initial absence of certain elements from child speech is provided by the Lexical Learning Hypothesis, also referred to as weak continuity. The basic idea underlying lexical learning, e.g., as found in the work of Clahsen et al. (1996), is that UG, in particular X¢ theory, is available to the child. That is, if there is an X, this X projects to XP. However, the tree need not be completely projected, i.e. some functional projections, e.g. CP, may be missing. The tree structure expands gradually under the guidance of input and UG principles. The addition of functional layers to the existing tree structure is

25

26

Tanja Kupisch

motivated by the discovery of grammatical properties encoded by the heads of functional projections. Clahsen et al. have shown how their theory applies to German acquisition data. Initially, determiners are optional and the possessive suffix -s is omitted. Determiners and adjectives occur in complementary distribution, show a high percentage of agreement errors, and determiners are not inflected for case. As soon as the possessive marker -s is acquired, determiners are used target-like and can be freely combined with adjectives. These findings have been confirmed by other studies on the DP (e.g. Eisenbeiss 2000, Granfeldt 2000). One advantage of this approach is that, unlike the previously discussed accounts, it allows for cross-linguistic variation because the discovery of initially missing elements may depend on the accessibility or salience of certain functional categories in terms of frequency, regularity of appearance, stress patterns etc. On the other hand, the initial absence of D also implies the unavailability of referential features. Studies of the referential use of articles have shown that children’s first articles typically occur in the function of namings and in deictic functions. In choosing between definite and indefinite articles in these functions children make hardly any errors (Brown 1973, Maratsos 1974, KarmiloffSmith 1979, among others). It is possible to argue, as supporters of lexical learning do, that the first determiners occur in formulaic expressions. But then, one should also consider whether or not formulaic expressions play a role in the instantiation of new functional layers. Another explanation for article omission in early child speech is found in Chierchia et al. (1999, 2000). As mentioned in Section 2, the authors looked at 16 monolingual children acquiring French, Italian, English and German, and found that the omission pattern differed with respect to the target language. The cross-linguistic differences were couched in terms of the Nominal Mapping Parameter (NMP) (cf. Chierchia 1998a,
b). According to the NMP the existence of a BN-grammar in child language, though deviating from the adult grammar, still conforms to UG because the projection of D is not universal but depends on parametric variation. The basic idea of the NMP is that all nouns in the Romance languages have the status of predicates and D must be projected for a nominal to be able to occur in argument position. The opposite case is met in languages such as Chinese, where all nouns are arguments and D is never projected. Bare nominals in Germanic languages are licensed with mass nouns and with indefinite plural nouns (usually with an existential or generic reading). In such contexts D is not projected.

The DP, a vulnerable domain?

Cross-linguistic variation in the acquisition data is explained as follows. At the initial state, the parameter is set to the “Chinese” value. All children assume that nominals in their language are arguments, D does not have to be projected, nouns occur bare. The BN-stage is attested in all languages, but its length differs depending on the target language. In Italian and French it is particularly short. The discovery of plural markings and determiners then leads children to reset their parameter to the “Germanic” parameter, which results in a free variation of BNs and DPs. Children whose target language is a Romance one very soon realize that argument nominals are always preceded by an article, and consequently reset their parameter to the target setting, while “[t]heir Germanic peers may continue to display non-adult performance, although their setting of the NMP is correct, since they have still to deal with the lexical classification of nouns.” (Chierchia et al. 2000: 31). That is, children in the latter group have to decide for each noun separately whether it is mass or count. The approach is intuitively plausible and allows for cross-linguistic variation, but the following points should be noted. First of all, Romance and Germanic are not consistent families. This has also been shown in studies on the acquisition of determiners. For example, acquisition data from Icelandic and Swedish (cf. Bohnacker 1997, 1999) suggest that determiners are acquired very early in these languages, as opposed to, e.g. English.19 Furthermore, the contexts for German, as for English BNs, are strictly defined. Thus, German does not have “free-variation” but child grammar does. Another point that should be added concerns the assumption that plural markings on nouns, besides determiners, trigger the passage from one stage to another. Accordingly, it should be disadvantageous for French children that plural marking on nouns is mostly not phonetically realized, contrary to fact.20 5.2 Acquisition theories in the Minimalist Program Within the Minimalist Program UG consists of formal features and operations. The child’s task in the acquisition process consists in constructing lexical items, discovering the formal features that are grammaticalized in the target language, and refining the computational system. Based on the observation that determiners tend to be preserved by children acquiring French, while accusative clitics are subject to omission (cf. Section 2), Jacubowicz (1999) suggests that patterns of omission or impairment vs. relative preservation of functional categories can be explained in terms of the complexity of the syntactic computation. In particular, complexity depends on the properties of functional elements in the language a

27

28

Tanja Kupisch

(normal or language impaired) child is acquiring. Two complexity criteria were proposed, both of which predict that the syntactic computation is less complex for determiners than for clitics. Firstly, determiners must be present in every sentence. Secondly, they are canonically merged in the lexical domain with a predicate. Neither criterion holds for clitics, which have to be present in some sentences only, and which are merged in the functional domain (see Jacubowicz et al. 1998 for the analysis of clitics adopted in the framework). The assumption that the complexity of syntactic operations has an impact on the acquisition process is very plausible, but one might object to the claim that D, a category of the “less complex” type, is only syntactically necessary, as opposed to clitics which are also semantically necessary. The author does point out that determiners are not devoid of semantic content, but she distinguishes them from functional categories that are only projected when required by semantic-conceptual considerations (ibid. 6). Firstly, this assumption is not in accord with the distinction, common in generative approaches, between substantive, or referential and expletive articles. Admittedly, it might be easier to have a concept in mind when hearing him than when hearing the, but, as shown in Section 2, referential determiners have semantic and pragmatic functions besides their syntactic one. Only expletive articles can be argued to have a purely syntactic motivation. But even disregarding these points it is questionable what, under these theoretical considerations, would be the status of determiners in other languages, such as German, a language which obligatorily requires BNs in some contexts. German BNs fulfill the same functions as in French are expressed by DPs, but this does not make the use of determiners in the former language more “semantically motivated” than in the latter. In fact, both languages have means to express the same semantic distinctions. Speakers of both languages must choose between definite and indefinite articles, but only in one of the languages a speaker additionally has the option of not using any form at all. Lastly, in sentences such as Jean je l’ai vu hier ‘Jean I him have seen yesterday’ the clitic may be said to have an emphatic function, while the semantic content is conveyed by Jean. 5.3 Prosodic accounts According to Demuth (1994), Gerken (1996), Lleó (1998), Lleó and Demuth (1999) and others, the omission of determiners is due to prosodic factors. Their hypothesis is compatible with the idea of strong continuity, explaining why determiners are omitted despite an underlying grammar according to which D

The DP, a vulnerable domain?

is projected. Common to all prosodic approaches is the assumption that weak syllables are more likely to be omitted than strong syllables. A crucial notion is that of “extrametric syllables”, i.e. syllables that are not immediately dominated by a foot but rather by the prosodic word. Such syllables are found both within single words, as illustrated in (12a), but also within sequences of articles and nouns, as in (12b). (12) prosodic word level

PW

foot level

F

syllable level

w

s

w

a.

ba-

na-

na

b.

the

ba-

by

It has been proposed that prosodic structure conforms to four principles. One of these principles, the “exhaustivity-constraint”, states that each unit in the prosodic hierarchy is dominated by the immediately higher unit. Thus, a syllable cannot be immediately dominated by a prosodic word, unless the constraint is violated, as e.g. by the first syllables ba- and the in the examples (12a,
b). The above researchers have observed that exactly those syllables that violate the exhaustivity-constraint are more frequently omitted in child language. In terms of Optimality Theory (OT), this would mean that the exhaustivity-constraint is ranked very high in child language.21 The prosodic truncation approach has been criticized for its alleged inability to account for cross-linguistic variation. In other words, the theory makes similar predictions for French as for Italian, German and English, contrary to empirical facts. As Lleó (1998) and Lleó and Demuth (1999) argue, this criticism is not justified. Based on data from German and Spanish they show that Spanish displays a larger number of extrametric syllables than German does. Thus, a Spanish child, having more exposure to extrametric syllables, very soon realizes that his/her target-language requires their production. They suggest that “a possible explanation for these different developments in the two language groups lies in the cross-linguistic differences in the frequency of multisyllabic words in the target-language.” (Lléo and Demuth ibid. 410). Yet, the account fails to explain why in one and the same individual learner omission of extrametric syllables is abandoned earlier word-internally, like in (12a), than in free standing syllables, as in (12b) (see e.g. Bottari et al. 1993/94). Again,

29

30

Tanja Kupisch

this could be traced back to the input: while children usually do not hear nana, they hear combinations of baby and elements other than articles, as e.g. one/my/ your baby. But then, one has to account for the fact that numerals, possessive pronouns and articles are not acquired simultaneously. Hence, article omission in child language cannot be due to prosodic factors alone, although prosody may play a role. 5.4 Semantic or pragmatic motivations: An alternative account? An alternative to assuming that determiner omission can be explained by deficiencies in the grammatical or prosodic domains is to assume that the early omission of determiners is due to ignorance of the relevant semantic or pragmatic principles. For instance, it might be hypothesized that children initially have difficulties in choosing an article according to the contextual requirements. They would have problems discovering the anaphoric relationship that exists between the italicized DPs in (6), repeated here as (13). (13) Yesterday I saw a cat and a dog playing in the garden. Suddenly, the cat ran away.

For several reasons this hypothesis should be discarded. As mentioned above, in the speech of children around the age of two articles are mostly used in the naming function and in the deictic function. The former function requires the use of an indefinite article, the latter requires a definite one but it is hard to uncover clear cases of target-deviant article use. If children had problems in discovering semantic or pragmatic principles, one would expect to find cases of article misuse. Of course, one may argue in favor of a prior stage in which articles are omitted due to an avoidance strategy, but again, it remains open why there is cross-linguistic variation, at least among languages as closely related as French, Italian and German. Lastly, some studies have observed that semantically motivated, referential uses of articles tend to occur earlier in the development than expletive articles (cf. Eisenbeiss 2000: 45). 5.5 The invulnerability of the French DP Presupposing the existence of an innate language capacity and a UG that provides the child with some abstract principles that guide the acquisition process, and further presupposing that the child has to learn the language specific forms by which syntactic categories are expressed, why not assume that

The DP, a vulnerable domain?

the phonetic realization of a syntactic category — in this case D — and its regular use in obligatory contexts is triggered by the elements that most frequently and most typically occupy this position? In other words, I want to propose that it is determiners and in particular articles (including the “expletive” ones but not exclusively these) that makes the child realize that his/her target-language necessitates their (first) use. It is suggested that their frequency and the consistency in their distribution in French facilitates acquisition. An argument along these lines might not be in the original spirit of UG, i.e. the hypothesis that children learn their languages despite impoverished input and thus must have some device that helps them. While it is clear that rote memorization and imitation of input alone cannot explain acquisition, crosslinguistic variation in the emergence of particular grammatical elements suggests that its role should not be underestimated. In the following I will compare the distribution of BNs in French, representative of a language in which BNs are very infrequent, with German, a language in which the DP should turn out to be more vulnerable because BNs are more frequent and their distribution is less transparent. 5.5.1 The distribution of BNs in French and German A number of studies have shown that children acquiring a Romance language start to use articles earlier than children acquiring a Germanic language such as German or English (e.g. Chierchia et al. 1999, Lleó and Demuth 1999, Kupisch 2000, among others). I suggest that the early emergence of articles in French is closely related to their distribution in the target language. This suggestion will be clarified by means of a comparison with German. The obligatoriness of D should be discovered early in French because determiners are frequent and their distribution is consistent and thus highly predictable. At first sight, this generalization seems to apply equally to German and French since in both languages determiners are frequent (compared to other functional categories) and they always precede nouns (indirectly if the noun is also preceded by an adjective). However, German children frequently encounter nouns that are not preceded by determiners, while this is rarely the case for French children. Figure 13 shows the relative number of bare nouns in nominal expressions in the two languages. The data is representative of the speech of monolingual adult speakers of French and German when talking to children. The percentages are based on a number of 2524 nouns in German and 2396 in French. Nominals have been categorized according to three classes: nouns preceded by articles (Art+N), nouns preceded by determiners other than

31

32

Tanja Kupisch

90% 80% 70% 60% 50%

French German

40% 30% 20% 10% 0% Det+N

Art+N

BNs

Figure 13.BNs vs. DPs in adult French vs. adult German.

articles including quantifiers (Det+N), and BNs. Proper names have been excluded from the evaluation because their number may vary according to play situation and degree of familiarity between child and interlocutor. Furthermore, cross-linguistic differences in the amount of proper names are unexpected. The figure shows that less nominals are preceded by determiners (articles and others) in German than in French (83% vs. 95%). The occurrence of BNs in French is restricted to a small number of contexts. Most of them (43 items) occurred in negated utterances, like (14a).22 A comparatively large number (37 items) was found as complements of prepositions, as in (14b), and in idiomatic expressions (26 items), the latter mostly with abstract nouns and exclusively in object positions, as in (14c). It should be added that many of the expressions among the group of BNs before prepositions appear to be idiomatic, as well. (14) a.

mais le docteur il a pas de bonbons à distribuer ‘but the doctor has no candy to give away’ b. mettre par terre ‘put on the ground‘, une prison pour vaches ‘a prison for cows’, nager sans bouée ‘swim without life belt’, un truc à manches langues ‘a thing with long sleeves’, une nouvelle poussette avec siège éjectable ‘a new baby buggy with ejector seat’ c. faire chaud ‘be warm [weather]’, faire attention ‘pay attention’, avoir envie ‘like to do’, avoir faim ‘be hungry’, avoir soif ‘be thirsty’, avoir raison ‘be right’, avoir peur ‘be afraid’, avoir besoin ‘need’

The remaining number of BNs (14 items) comprise nouns in predicative positions, vocatives, exclamations, and repetitions of child utterances.

The DP, a vulnerable domain?

Let us turn to German. In German, BNs are not only more frequent, but they are also found in a variety of different contexts. First, similar to French, BNs are frequent in idiomatic expressions and after prepositions (cf. (15a,
b)) (in the latter case they are often idiomatic, like in French), and they also occur in vocatives, exclamations, and predicative contexts. However, unlike in French, their distribution is not restricted to the object position, but they can occur in subject position as well if their reading is generic (cf. (15c)), or existential, if the noun is emphasized, as in (15d). (15) a.

mir ist egal wo du das reintust, hauptsache das is da nachher drin ‘I don’t care where you put this, what matters is that it is in there afterwards’ b. du gehst nach hause, ißt fisch als beilage und fällst in ohnmacht ‘you go home, eat fish as a side dish and faint’ c. piraten sind dumm ‘pirates are stupid’ d. POMMES willst du essen? ‘It is French fries you want to eat?’ (a–d from adult-child conversation)

Summarizing the previous part, taking aside the number of proper names, a very small portion (5%) of French nouns occurs bare. French is exceptional in this respect, and differs even from other Romance languages such as Italian, in which the occurrence of BNs is also lexically restricted, but which are assumed — in contemporary generative theories (e.g. Longobardi 1994) — to obligatorily project the D position. A study comparable to the one above has in fact shown that the percentage of BNs amounts to 13% in Italian. French thus constitutes a language in which BNs are extremely rare. A child exposed to French hardly encounters any bare nouns. I assume that this enables the emergence and target-like use of determiners at a very early age. The suggestions made here are compatible with UG. Under the assumption of strong continuity and an initial underspecification of certain categories or features, the proposal would explain why obligatoryness of D is triggered earlier in some languages than in others. Under the assumption of lexical learning it would explain why French children become aware of a D position that must be filled earlier than children acquiring other languages.

33

34

Tanja Kupisch

6. Conclusion In this paper I have shown that the French DP is not vulnerable, contrary to what might be expected if there were parallels in the syntactic configurations of DP and CP. Since vulnerable domains are typically defined as domains in which syntactic information is related to the discourse, one might conclude that the French DP is determined by syntactic principles exclusively and that it is unrelated to the discourse — an assumption that is indeed implied in some generative approaches. However, this assumption is in conflict with a number of acquisition studies showing that children make hardly any errors in choosing between definite and indefinite articles, a choice which requires the consideration of contextual requirements. Drawing on the results obtained from the present study, I suggest that an interface status alone is not sufficient to account for the vulnerability of a grammatical domain. Another possible conclusion, which is not incompatible with the previous suggestion, is that it is not DP but QP that constitutes the left periphery of the nominal domain. The claim that QP is a syntactic projection above DP is empirically motivated (cf. Giusti 1991, 1997), and the vulnerability of elements typically attributed to this category should be tested in further research. However, even if QP was vulnerable, this obviously does not affect the emergence of articles. An outcome of this study is that vulnerability is not a question of interface status alone. There must be other factors that contribute to the vulnerability of a particular syntactic domain. To answer the question of what these could be I considered the emergence of articles from a cross-linguistic perspective. The well-established fact that children acquiring French/Italian/Spanish start to use articles earlier than children acquiring English/German fails to be explained by a number of approaches. Articles are not omitted to the same extent in all languages. Why? What distinguishes languages such as English, Swedish, German, Italian, French, and Spanish? Cross-linguistic differences are often underestimated. One should consider that the information associated with a particular syntactic domain might differ from language to language, and that this information is not always equally accessible. I proposed that the crucial factor is a language specific one. Regarding distributional aspects, languages are transparent to a different degree, and some languages are more challenging than others. I suggested that the acquisition process with regard to the French DP is accelerated because the distribution of determiners in French is consistent and transparent. I assume that this makes the French DP invulnerable.



The DP, a vulnerable domain?

Notes * This work was funded in the framework of the SFB 538 Mehrsprachigkeit (Collaborative Research Center No. 538 Multilingualism by the Deutsche Forschungsgemeinschaft (DFG). The responsibility for the contents of this study lies with the author. I would like to thank the members of my research project, Natascha Müller, Katja Cantone, and Katrin Schmitz, as well as two anonymous reviewers for commenting on earlier versions of this paper. I am grateful to Julia Cordes for her contributions on the Céline-corpus, Sören Anders for his lastminute corrections, and in particular Tessa Say and Christian Koops for correcting my English and for their critical and stimulating comments and suggestions. 1. The two languages of a bilingual child may develop differently, in the way that the child is less proficient in one language than in the other. Such imbalance is not restricted to bilingual children who acquire two languages successively, but may also be found in the simultaneous acquisition of two languages (e.g. Schlyter 1993, 1994). In the following I will use the term “weaker language” to refer to cases of unbalanced bilingualism. 2. The underlying idea is that linguistic elements requiring only syntactic knowledge can be acquired more easily than elements requiring semantic and/or pragmatic knowledge as well. However, one might argue just the contrary, namely that is easier to acquire elements that are not only syntactically but also semantically and pragmatically motivated. The latter argumentation would be in keeping with the fact that lexical categories tend to be acquired earlier than functional categories. 3. There are some exceptions to this rule, e.g. les haricots [le ariko] ‘the beans’, les héros [le ero] ‘the heroes’. 4. This does not hold if the nominals are preceded by an adjective: (i) J’ai rencontrée la belle Marie. (ii)

Mon voisin est un bon physicien.

5. An anonymous reviewer correctly noted that it would be more appropriate to describe proper names as being inherently marked for definiteness rather than inherently referential. Proper names are first of all just names and they are referential only if used in referring expressions. Furthermore, languages differ with regard to whether or not they allow bare proper names as referential expressions. While German and Italian do allow them, the definite article is considered obligatory before proper names e.g. in Catalan, European Portuguese, and Modern Greek. 6. This assumption is not uncontroversial. According to Korzen (1998: 106), the definite article grammaticalizes the feature [+given], i.e. the existential presupposition of the designated entity, even if not used to refer endo- or exophorically. 7. Unless indicated otherwise, I refer to word-based MLU. 8. The investigation of determiners was limited to the production task. 9. This does not imply that the French plural article les is unifunctional. As Karmiloff-Smith (1979: 51) points out, it also encodes, although not primarily, “totality” in addition to plurality.

35

36

Tanja Kupisch

10. I will use the term “mixings” to cover both intra-discourse and intra-utterance mixings. 11. Relative numbers resulting from recording sessions in which less than five French noun tokens were produced were not taken into account. Such recordings were also excluded in Figure 12. 12. The duration of the period between first determiner use and target-like use was not compared for two reasons. First, due to a low number of nouns in many of Céline’s recordings, it is difficult to say when exactly the BN-stage ends. Second, in the other two children the BN-stage has not been captured at all. 13. The number of correct contexts includes mixed DPs. Imitations and utterances that were unclassifiable, e.g. because the referent could not be clearly determined as either singular or plural, have been excluded. 14. This statement only refers to mass nouns when used with nonspecific reference. Of course, there were cases in which mass nouns, such as e.g. pain ‘bread’, were used with a definite article because a particular bread or piece of bread was referred to. 15. This is a preliminary result of an input analysis in Kupisch (in prep.). Mass nouns constitute between 2.4% and 6.9% of all nouns used by an investigator in a recording session of 30 minutes. 16. Italian differs from French in allowing an empty D in lexically governed positions. The empty D is, however, optional (see e.g. Longobardi 1994, Chierchia 1998a,
b). 17. As pointed out before, it is debatable whether such articles can really be considered semantically empty. Furthermore, languages behave differently with regard to this “lack of semantic content”. For instance, the definite article is obligatory with discourse unica, proper names etc. in Modern Greek. This seems to imply that notions such as “semantic content” are rather language specific, which underlines the point made further below that languages differ with regard to the availability of triggers. 18. See, however, footnote 4. 19. It is possible that their early emergence in these languages is related to the fact that definite articles are represented by bound morphemes, although this is not what Bohnacker mainly argues for. 20. One might also raise the question of how to classify languages such as Finnish that have no article but case. Are Finnish nouns of the argument or of the predicate type? 21. In OT, constraints are ranked, and an optimal structure may conform to higher ranked constraints while violating lower ranked ones. 22. One could indeed argue that such nouns are not really “bare”, because pas de “none of” can be treated in a way similar to beaucoup de “a lot of”, indicating a quantity, which is zero in the former case. These cases were, however, included in the study since it is unclear how this construction is understood by the child.

The DP, a vulnerable domain?

References Abney, S. 1987. The English noun phrase in its sentential aspect. Ph.D. Dissertation, Cambridge, Mass.: MIT. Berkele, G. 1983. Die Entwicklung des Ausdrucks von Objektreferenz am Beispiel der Determinanten. Eine empirische Untersuchung zum Spracherwerb bilingualer Kinder (Französisch/Deutsch). MA Thesis, University of Hamburg. Bohnacker, U. 1997. “Determiner phrases and the debate on functional categories in early child language”. Language Acquisition 6: 49–90. Bohnacker, U. 1999. Icelandic plus English: Language differentiation and functional categories in a successively bilingual child. PhD Dissertation, University of Durham. Bottari, P., Cipriani, P. and Chilosi, A. M. 1993/94. “Protosyntactic devices in the acquisition of Italian free morphology”. Language Acquisition 3: 327–369. Bottari, P., Cipriani, P., Chilosi, A. M., and Pfanner, L. 1998. “The determiner system in a group of Italian children with SLI”. Language Acquisition 7: 285–298. Brown, R. 1973. A First Language: The Early Stages. Cambridge, Mass.: Harvard University Press. Chierchia, G. 1998a. “Plurality of mass nouns and the notion of ‘semantic parameter’”. In Events and Grammar, S. Rothstein (ed.), 53–103. Dordrecht: Kluwer. Chierchia, G. 1998b. “Reference to kinds across languages”. Natural Language Semantics 6: 339–405. Chierchia, G., Guasti, M. T. and Gualmini, A. 1999. “Early omission of articles and the syntax/semantics map”. Talk presented at GALA. Potsdam, September 1999. Chierchia, G., Guasti, M. T. and Gualmini, A. 2000. “Nouns and articles in child grammar and the syntax/semantics map.” Manuscript, University of Milan/University of Siena/ University of Maryland, College Park. Clahsen, H., Eisenbeiss, S. and Penke, M. 1996. “Lexical learning in early syntactic development.” In Generative Perspectives on Language Acquisition, H. Clahsen (ed.), 129–159. Amsterdam/Philadelphia: John Benjamins. Cordes, J. 2002. Zum unausgewogenen doppelten Erstspracherwerb eines deutsch-französisch aufwachsenden Kindes: Eine empirische Untersuchung. MA Thesis, University of Hamburg. Demuth, K. 1994. “On the underspecification of functional categories in early grammars”. In Heads, Projections and Learnability, B. Lust, M. Suner and J. Whitman (eds), 119–134. Hillsdale, NJ: Lawrence Erlbaum Associates. Eisenbeiss, S. 2000. “The acquisition of the DP in German child language”. In: Acquisition of Syntax. Issues in Comparative Developmental Linguistics, M.-A. Friedemann and L. Rizzi (eds), 26–62, London: Longman. Gawlitzek-Maiwald, I. and Tracy, R. 1996. “Bilingual bootstrapping”. Linguistics 34: 901–926. Genesee F., Nicoladis, E. and Paradis, J. 1995. “Language Differentiation in Early Bilingual Development”. Journal of Child Language 22: 611–631. Gerken, L. A. 1996. “Prosodic structure in young children’s language production”. Language 73: 113–144. Giorgi, A. and Longobardi, G. 1991. The Syntax of Noun Phrases. Cambridge: CUP. Giusti, G. 1991. “The categorial status of quantified nominals”, Linguistische Berichte 136: 438–454.

37

38

Tanja Kupisch

Giusti, G. 1997. “The categorial status of determiners”. In The New Comparative Syntax, L. Haegemann (ed.), 95–123. London: Longman. Granfeldt, J. 2000. “The acquisition of the determiner phrase in bilingual and second language French”. Bilingualism: Language and Cognition 3 (3): 263–280. Guasti, M. T. 1993/94. “Verb syntax in Italian child grammar: Finite and nonfinite verbs”. Language Acquisition 3: 1–40. Hamann, C., Penner, Z. and Lindner, K. 1999. “German impaired grammar. The clause structure revisited”. Language Acquisition 7: 193–245. Hoekstra, T., Hyams, N. and Becker, M. 1997. “The underspecification of number and the licensing of root infinitives”. In Proceedings of the 21st Annual Boston University Conference on Child Language Development, E. Hughes, M. Hughes and A. Greenhill (eds), 293–306. Somerville, Mass.: Cascadilla Press. Hulk, A. and Müller, N. 2000. “Cross-linguistic influence at the interface between syntax and pragmatics”. Bilingualism: Language and Cognition 3 (3): 227–244. Jakubowicz, C. 1999. “Functional categories in (ab)normal language acquisition”. Paper presented at GALA. Potsdam, September 1999. Jakubowicz, C., Nash, L., Rigaut, C. and Gérard, Ch.-L. 1998. “Determiners and clitic pronouns in French-speaking children with SLI”. Language Acquisition 7: 113–160. Karmiloff-Smith, A. 1979. A Functional Approach to Child Language: A Study of Determiners and Reference. Cambridge: CUP. Koehn, C. 1994. “The acquisition of gender and number morphology within NP”. In Bilingual first language acquisition: French and German grammatical development, J. M. Meisel (ed.), 29–52. Amsterdam/Philadelphia: John Benjamins. Korzen, I. 1998. “On nominal determination — with special reference to Italian and comparison with Danish”. In Nominal Determination, G. Hansen (ed.), 67–132. Copenhagen Studies in Language 21. Kupisch, T. 2000. Artikelauslassungen bei einem bilingual deutsch-italienischen Kind, MA Thesis, University of Hamburg. Kupisch, T. 2001, “The acquisition of the DP in French as the weaker language”. Working Papers in Multilingualism 32: 1–34. Hamburg: Sonderforschungsbereich 538/Universität Hamburg. Kupisch (in prep.). The acquisition of articles by bilingual German/Italian and German/French children. Dissertation, University of Hamburg. Langacker, R. 1991. Concept, Image, and Symbol. Berlin, New York: Mouton de Gruyter. Lleó, C. 1998. “Proto-articles in the acquisition of Spanish: Interface between phonology and morphology”. In Modelle der Flektion: 18. Jahrestagung der Deutschen Gesellschaft für Sprachwissenschaft, R. Fabri, A. Ortmann and T. Parodi (eds), 175–195. Niemeyer: Tübingen. Lleó, C. and Demuth, K. 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, E. Hughes, M. Hughes and A. Greenhill (eds), 407–418. Somerville, Mass.: Cascadilla Press. Longobardi, G. 1994. “Reference and proper names: A theory of N-movement in syntax and logical form”. Linguistic Inquiry 25: 609–665.



The DP, a vulnerable domain?

Mac Whinney, B. and Snow, K. 1985. “The child language data exchange system”. Journal of Child Language 12: 271–296. Maratsos, M. P. 1974. “Preschool children’s use of definite and indefinite articles”. Child Development 45: 446–455. Müller, N. and 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., Cantone, K. F., Kupisch, T. and Schmitz, K. 2002. “Zum Spracheneinfluss im bilingualen Erstspracherwerb: Italienisch-Deutsch”, Linguistische Berichte 190: 157–206. Penner, Z. and Weissenborn, J. 1996. “Strong continuity, parameter setting and the trigger hierarchy: On the acquisition of the DP in Bernese Swiss German and High German”. In Generative Perspectives on Language Acquisition, H. Clahsen (ed.), 161–200. Amsterdam/Philadelphia: John Benjamins. Platzack, C. 2001. “The vulnerable C-domain”. Brain and Language 77: 364–377. Radford, A. 1990. Syntactic Theory and the Acquisition of English. Blackwell: Oxford, England. Rizzi, L. 1997. “The Fine Structure of the Left Periphery”. In Elements of Grammar. Handbook in Generative Syntax, L. Haegeman (ed.), 281–337. Kluwer: Dordrecht. Schlyter, S. 1993. “The weaker language in bilingual Swedish-French children”. In Progression and Regression in Language, K. Hyltenstam and Å. Viberg (eds), 289–308. Cambridge, CUP. Schlyter, S. 1994. “Early morphology in Swedish as the weaker language in French-Swedish bilingual children”. Scandinavian Working Papers in Bilingualism 9: 67–86. Schlyter, S. and Håkansson, G. 1994. “Word order in Swedish as the first language, second language and weaker language”. Scandinavian Working Papers on Bilingualism 9: 49–67. Schlyter, S. and Bernardini Röst, P. 2001. “Climbing up the tree of the dominant language: On the role of code mixing for the development of the weaker language in young bilinguals”. Ms, Romanska Institutionen, University of Lund. Serratrice, L. 2000. The emergence of functional categories in bilingual first language acquisition. PhD Dissertation, University of Edinburgh. Siloni, T. 1997. Noun Phrases and Nominalizations. Dordrecht: Kluwer. Szabolsci, A. 1983/1984. “The possessor that ran away from home”. The Linguistic Review 3: 89–102. Valois, D. 1991. The Internal Syntax of DP. PhD Dissertation, University of California.

39



Child and adult acquisition of word order in the Italian DP* Petra Bernardini Lund University

It is generally agreed that children are able to acquire two languages to which they are exposed from birth as two L1s. Little attention has been drawn to cases when one language develops slower and is weaker in different aspects. A question under debate is whether the weaker language develops as an L2 rather than as an L1. The goal of this paper is to contribute to the understanding of the weaker language. This paper compares the acquisition of word order in the Italian DP by bilingual Italian/Swedish children with that of monolingual Italian children and Swedish adult second language learners of Italian. In this study there seems to exist a relationship between language dominance and the acquisition of word order in the DP in terms of N-raising. Non-target-like pre-nominal attributive adjectives and only pre-nominal possessives are observed in the Swedish-dominant bilingual child and the adults. The Italian-dominant child and the monolingual children position both types of adjectives target-like from the beginning. However, the preference for the pre-nominal position in the adults and the child with Italian as the weaker language occurs with different adjectives (obligatory post-nominal vs. variable adjectives). It is argued that transfer of the N-raising-parameter value in the adults and ambiguous input in the case of the child cause the preference for the pre-nominal adjectives.

1.

Introduction

There is general consensus in the field of language acquisition that children who are exposed to two languages from birth progress along an acquisition path analogous to that of monolingual acquisition in early childhood (= L1 acquisition). Children are able to differentiate their two systems without confusion of grammars. For example, in the various studies on bilingual French/German children’s acquisition in Meisel (1990; 1994), or Dutch/French in De Houwer

42

Petra Bernardini

(1990), French/English, in e.g., Genesee et al. (1995), Italian/English in Serratrice (1999), it is reported that the development of the two languages with regard to parameter setting is autonomous almost without any interaction. These studies indicate thus that L1 acquisition and the acquisition of two languages from birth (= 2L1 acquisition) are identical. On the other hand, in studies of adult second language acquisition (= L2 acquisition) it is quite commonly reported that the L1 and the L2 interact to different degrees (Eubank 1996, Vainikka and Young-Scholten 1996). There are even scholars who suggest that the initial stage of L2 is identical to the L1 of the learner (cf. Schwartz and Sprouse 1996) in terms of parameter settings. But although the simultaneous acquisition of two languages from birth is equal to monolingual acquisition of the two languages, in the sense that the languages develop separately with respect to parameter settings, there is evidence for that the two languages may have some influence on each other (Döpke 1998; Döpke 2000; Müller 1998; Müller and Hulk 2001) just as has been proposed to various degrees in L2 acquisition. Language interaction in 2L1 acquisition seems to depend on three factors. The first factor is the structural overlapping of the two languages. Müller and Hulk predict language influence to occur at interface levels in areas that are also problematic to monolingual children. This is predicted to occur although children set the parameters of their L1 to the correct value. Language interaction in 2L1 acquisition is also reported to occur when one language is dominant and the other is weaker, language contact seems to be more common when children speak in the weaker language. For example, intrasentential code-mixing is more frequent in the language of which the child has less command (see Petersen (1988), Nicoladis and Secco (1998), Genesee et al. (1995), Huss (1991); Bernardini and Schlyter (2002). Particularly in cases when one language develops faster, because of language dominance (Bernardini and Schlyter 2002), it seems as if the more developed language might provide the higher functional structure in intrasentential code-mixing. Language interaction in 2L1 acquisition seems also to occur when one language develops faster than the other in a particular domain. For ex. Gawlitzek-Maiwald (1996) and Tracy and Gawlitzek-Maiwald (2002) in their studies of the simultaneous acquisition of English and German, notice that mixed utterances consist of German IP and English VP. Their explanation is that in monolingual acquisition of German the IP develops faster than in monolingual acquisition of English. Thus, in cases where one language develops slower in a bilingual child

Child and adult acquisition of word order in the Italian DP

exposed to both languages from birth, production data show evidence for interaction between the two languages. This kind of interaction is similar to that found in L2 acquisition, where it is often reported that learners use their knowledge of L1 in their use of L2. In Schlyter (1993, 1994, 1995, 1999) it is suggested that bilingual children’s weaker language has a different acquisition path than that of an L1. In fact, it is suggested that the weaker language, when clearly weaker, is similar to L2 acquisition. We suggest that the DP-domain in the language pair Swedish/Italian is a suitable testing ground for language influence. In Swedish adjectives may only precede the noun while in Italian they are both pre- and post-nominal. (1) a.

Swedish en fin stor röd boll a nice big red ball

b.

Italian una bella grande palla rossa a nice big ball red

In this contribution I investigate the acquisition of noun/adjective word order in Italian by monolingual Italian children, bilingual Italian/Swedish children (Italian weaker/Italian stronger) and by adult learners with Swedish as a native language. Considering the above discussion on differences in language contact in 2L1 (+language differentiation, +language interaction) vs. L2 acquisition (−language differentiation, +language interaction), my main concern will be to discover whether there are differences between the bilingual children and adults in the positioning of adjectives in Italian. The inclusion of the monolingual children is necessary if we wish to differentiate L1 from L2 acquisition. According to the hypothesis “simultaneous bilingual acquisition = 2L1 acquisition” (Genesee et al. 1995, Meisel 1990, 1994), the bilingual children’s acquisition path should have more in common with that of the monolingual children. According to the hypothesis “weaker language = L2” (Schlyter 1993), the child who has Italian as the weaker language will have an acquisition path that has more in common with that of the adults, while the acquisition path of the child who has Italian as the stronger language should have more in common with that of the monolingual L1 (according to the hypothesis implied in “weaker language = L2, then stronger language = L1”). The paper is organised as follows: in Section 2 I introduce the DP systems of Italian and Swedish with respect to word order. In Section 3 I introduce the relevant previous studies of L1, 2L1 and L2 acquisition on the word order in the DP and on language interaction. In 4 predictions for the acquisition of the Italian DP are made. Section 5 presents the corpora and the method of analysis. Section 6 reports on the results of the acquisition of word order in the DP by

43

44

Petra Bernardini

monolingual Italian children (6.1.), the acquisition of the same phenomenon in adult Swedish learners of Italian as L2 (6.2.), and the bilingual Swedish/Italian children’s acquisition of word order in the Italian DP (6.3.). In 7 I discuss the findings of this work and offer some conclusions.

2. Word order in the DP 2.1 The N-raising Hypothesis The so-called N-raising hypothesis (Bernstein 1991, Cinque 1995) accounts in general for the difference in word order between the adjective and the noun in Romance and Germanic.1 In both Italian (as Romance languages in general) and Swedish (as in Germanic in general) the structural basic order is A-N. The N-A order is derived by “N-raising”, i.e. the N moves from NP to target an AgrP intermediate to NP and DP2. (2) a.

Swedish

b.

Italian DP

DP D en

D una

AgrP

AgrP macchinai

NP

NP

AdjP

N

stor

bil

AdjP

N

grossa

ti

The N-movement hypothesis is necessary to account for different word orders in the DP in a unidirectional branching framework (as in Kayne 1994) considering that stacked adjectives have a hierarchical order in relation to each other, which is also preserved when adjectives are post-nominal (Cinque 1995: 303): (3) a.

en fin stor röd boll a nice big red ball b. una bella grande palla rossa a nice big ball red

beautiful > big > red beautiful > big > red

In Italian (and French) some adjectives may be both pre- and post-nominal, depending on the interpretation:

Child and adult acquisition of word order in the Italian DP

(4) a.

un amico vecchio a friend old (an old/aged friend) (intersective/objective) b. un vecchio amico an old friend (a dear old friend) (non-intersective/subjective)

Laenzlinger (2000) suggests that the different interpretations of position-variable adjectives are triggered by quantification, subjective reading and by reduced morphological forms of the prenominalization of the adjective in Romance.2 The difference between pre- and post-nominal adjectives is accounted for, following Cinque (1995), Laenzlinger (2000) and others, assuming that adjectives are merged in the Specifier of a semantically relevant functional projection. This can be seen by their order in relation to each other and by the distance to the noun based on their semantic meaning: (5) [quantifOrdinal > Cardinal] > [speaker-orientedSubjective Comment > Evidential] > [internal phys. prop.Size > Lenght > Height > Speed > Depth > Width] > [measureWeight > Temperature > ?Wetness > Age] > [external physic. propertyShape > Color > Nationality/ Origin > Material]

That adjectives are post-nominal in Romance is thus seen as the result of N-raising above the lower adjectives. 2.2 The distribution of possessive adjectives in Italian and Swedish The first difference to note between the two languages is that in Italian but not in Swedish the possessive must be preceded by a determiner. (6) a.

Il mio libro./un mio libro. the my book /a my book b. min bok *-en./*en min bok. my book the /a my book ‘my book/one of my books’

That the Italian possessive form must be preceded by an article leads one to consider the possessive form as an adjective. According to Giorgi and Longobardi (1991) the possessive form of Italian is in SpecDP. For languages such as (standard) Swedish, English and French which do not allow an article together with the possessive form, these authors suggest that Pos is in the head

45

46

Petra Bernardini

of DP, thus being a determiner. Schoorlemmer (1998) proposes further that the possessive form (mio etc.) reflects a formal licensing property. The position in nominal syntax hosting the pre-nominal possessor would then be SpecPosP, a structural licensing position comparable to SpecIP in the sentence. Thus, the possessive adjective would be an XP. When it is pre-nominal, we observe (7a) and when it is post-nominal (7b): DP

(7) a. Spec

D¢ D0 il

PosP Pos¢

Spec mioi

Pos0

AgrP Spec

Agr¢ Agr0 canej ti

NP Spec

N tj

DP

b. Spec

D¢ D0 il

PosP Spec

Pos¢ Pos0

AgrP Spec

Agr¢ Agr0 canei

NP Spec

N

mio

ti

Child and adult acquisition of word order in the Italian DP

We will therefore account for the fact that in Italian but not in Swedish the article must be present with the possessive assuming that, in Swedish, Pos0 carries a [definite] feature that forces Pos0 move further to D0 (8). In Italian Pos0 does not carry [definite] and remains in Pos0 (9). (8) [DP [D La [Pos miaj [Agr macchinai [NPj [Ni]]]. (9) [DP Minj [Posj [Agr bili [NPj [Ni]]]]]

We can account for the possibility of post-nominal possessives in Italian in the same way as for the attributive adjective, by assuming that N raises overtly in the case where Pos is post-nominal (10): (10) [DP La [XP macchinai [Pos miaj [Agri [NP [PosjNi]]]].

The problem with (7) is that if the post-nominal possessive of this kind owes its position to overt N-raising, there is no explanation for the fact that in this position the possessive receives a different interpretation, to highlight the possessor, i.e. in a contrastive sense. Cardinaletti (1998) claims that although the pre- and post-nominal possessive adjectives have the same forms, and could be accounted for by movement, this would not be optional. “The existence of both [la sua casa/la casa sua] could suggest that possessive movement is optional in Italian. A closer look […] reveals that it is not.” (Cardinaletti 1998: 18). Cardinaletti analyses pre-nominal possessives as deficient and postnominal ones as strong, since the latter possessives (as strong pronominal forms in general, as opposed to deficient) may, for instance, be contrasted, while prenominal possessives cannot be contrasted (11):3 (11) a.

la casa sua (strong) / la casa sua e mia ‘the house his/her / the house his/her and my’ b. la sua casa (weak) / la mia *e sua casa ‘the his/her hose / the my and his/her house’

The strong/deficient opposition implies that the deficient possessives must move to SpecPosP (see (7)) while we suppose that strong possessives stay in SpecNP. Possessives may also be post-nominal without the presence of a determiner. This is assumed to be due to N-to-D-raising (Longobardi 1994). (12) a.

Giannii mio ti Gianni my b. Casai miai house my

47

48

Petra Bernardini

This kind of N-raising has also been proposed for the Swedish definite suffix (Delsing 1993): (13) a.

bili-en ti car-the b. en bil a car

3. The acquisition of word order in the DP 3.1 Comparing L1, 2L1 and L2 Up to this point, the L1 acquisition of word order in the Romance DP has not been studied exhaustively. In Antelmi’s (1997) case study of the Italian girl Camilla, the possessive appears first in the post-nominal position at the age 1;9. (14) tata mia/papà mio/mamma mia [from Antelmi:

88] nanny my /daddy my /mommy my

At the age of 1;6 the possessive occurs only in isolation. At the age of 1;9 the possessive is post-nominal. It is produced without a determiner until 2;0. At the age of 2;1 the possessive is also pre-nominal, still without the article. At the age of 2;2 the possessive co-occurs with the determiner in both the pre- and the post-nominal position. Thus, there is a discrepancy of 3 months between the first occurrences of the post-nominal and the pre-nominal possessives. The determiner becomes productive one month after the appearance of the prenominal possessive. Antelmi did not observe any errors in the positioning of possessives. She did not study attributive adjectives. Granfeldt (2000b) studied the appearance of attributive adjectives in the French child Gregoire and noticed that the first attributive adjectives were prenominal. These seemed to be the only ones from age 2;0.5 (file G5) to age 2;5.1 (G8), whereas the post-nominal attributive adjectives made their appearance only at age 2;7 (G9). Bohnacker (1997) shows that the early productivity (i.e. before the age of 2;0), of the so-called definite suffix -en/et (‘-the’) suggests that N-raising takes place very early in Swedish. She adopts the same analysis of the definite suffix, an N-to-D approach, to the Scandinavian DP as Delsing (1993). This suggests that N-raising is easy for monolingual children. Volterra and Taeschner (1978) studied, among other things, the acquisition

Child and adult acquisition of word order in the Italian DP

of attributive adjectives in two Italian/German girls who acquired these two languages from birth. Attributive adjectives appeared earlier in German (age 2;3) than in Italian (age 2;9). The child Lisa had some errors in placing adjectives, which cannot be attributed to language dominance since they appeared in both directions (cf. (15) and (16)). (15) Lisa‘ will nur Schuhe dunkelbraun (Lisa 3;6) V. and T. (1978: 322) ‘Lisa wants also shoes darkbrown’ (16) Quel‘ bianco pecora (Lisa 3;6) (V. and T., ibid) ‘that white sheep’

Paradis and Crago (2002) studied the acquisition of the DP in French by normal L1 (three 3-year olds and 7-year olds), SLI and L2 children. These data included recordings of spontaneous speech. They found that the L2 pre-school children, who had had two years of exposure to French prior to the study (English L1) did not produce any post-nominal adjectives while the L1 and SLI children did. The younger L1 children made one positioning error for each position (Total N of adj with nouns = 67 pre-nominal, 31 post-nominal). The older L1 children produced 110 pre-nominal adjectives and 11 post-nominal ones without errors. The SLI children produced 64 pre-nominal adjectives without errors and 12 post-nominal ones of which 2 should have been prenominal. The greatest difference among these learners appears thus to be that the L2 children did not produce any post-nominal adjectives at all. Wetter (1996) noted transfer from Italian to German in Italian pre-school children learning German as a L2. These children made errors with color adjectives, which are obligatory post-nominal in Italian. Granfeldt (2000a,
b) compared the acquisition of DP of bilingual SwedishFrench children and adult Swedes learning French as a L2. The bilingual children used pre-nominal attributive adjective from first occurrence to the end of their third year of life. One child had a post-nominal adjective before the age of three. In contrast to the children, the adult L2 learners did produce postnominal adjectives from the first recordings. They also made some errors in positioning, although very few. In the pre-nominal position, shown in (17), there are, in total, 11 errors in the recordings of the seven learners. (17) le‘ allemagne couple [Henry 1, French L2, from Granfeldt: 98] ‘the German couple’

Parodi, Schwartz and Clahsen (1997) studied the acquisition of the German DP by adult L2 learners with different L1s who had been living in Germany for

49

50

Petra Bernardini

several years at the first recording. The four learners Ana, Bongiovanni, Lina and Bruno, who had Romance (Italian and Spanish) languages as L1, positioned the adjective post-nominally in German (which is not target-like). In the beginning of the study (approximately the first 5 months) they positioned the adjective post-nominally in German in about one third of the cases. The frequency decreased over time (see Table 1). Table 1.The positioning of adjectives in German by L2 learners with Romance L1 (Parodi et al. 1997) Learner

Months in Germany (after Parodi % post-nominal adj. in Geret al.: (18, Table 1) man (after Parodi et al.: 21, Table 2)

Ana

5 24

25 0

Bongiov.

1.5–4 20

37.5 20

Bruno

1.5–3 6–16 25

13 0 9.1

Lina

2–4 10.5 17

28.1 26.6 0

Parodi et al. suggested that the positioning errors were due to influence from their Romance L1. This was confirmed, they believed, by the fact that the Korean and Turkish learners (languages with only pre-nominal adjectives) did not position the adjectives post-nominally. Parodi et al. also studied the omissions of articles and found that the learners with Korean and Turkish as L1 omitted the article much more frequently in German than did the learners with Romance as L1. Parodi et al. suggested that this difference also indicates to transfer as an explanation, since Korean and Turkish do not have obligatory determiners, while Romance languages do. In this sense, the latter learners would transfer their knowledge of obligatory determiners from their L1 to their L2. In this task too, the Romance learners had approximately 30% errors (omissions) in the beginning, with a decrease in the following recordings. Previous research on the acquisition of the DP in Romance and Germanic languages gives some support to the idea that monolingual children might have

Child and adult acquisition of word order in the Italian DP

more ease in (some kind of) N-raising than L2 learners (cf. the early productivity of the post-nominal definite suffix in Swedish (Bohnacker 1997) and the post-nominal possessive in Italian (Antelmi 1997). The two bilingual children in Volterra and Taeschner’s study showed some influence in both directions. Children learning French as L2 did not use post-nominal adjectives in the study of Paradis and Crago (2002). Romance children as well as adults learning German produced post-nominal adjectives in German (e.g., Wetter (1996), Parodi et al. 1997). Swedish adult L2 learners of French pre-nominalised obligatorily post-nominal adjectives in French (Granfeldt 2000a). 3.2 The transferability of word order in the DP The idea that will be explored in our analysis is that ambiguous input may lead learners to an erroneous word order in those cases where the target language offers two possibilities of which one is shared with the other language. Whether this is independent of language-external factors, such as language dominance will be shown in the data. The idea stems from Müller (1998) and Müller and Hulk (2001) who investigated bilingual (German plus another language) children’s acquisition of subordinate clauses in German (Müller 1998) and object drop in bilingual Germanic/Romance acquisition (Müller and Hulk). Müller and Hulk (2001) suggest that indirect influence should occur in the domains of syntax that are vulnerable. The CP domain would be one, because Platzack (2001) has suggested that it is vulnerable in L1/L2 acquisition and language attrition. The suspected cause of its vulnerable character is the fact that the CP is a window to other cognitive domains and an interface level to pragmatics. The shared idea of Müller (1998) and Müller and Hulk (2001) is that indirect cross-linguistic influence occurs when a syntactic construction in one of the two languages of the bilingual child allows for more than one grammatical analysis from the perspective of the child grammar, when the other language contains positive evidence for one of these possible analyses. Müller and Hulk (2001) compared the frequency and persistency of object drop in Dutch/French, German/French and German/ Italian bilingual children who were exposed to both languages from birth with monolingual children of the respective languages. They found that object drop was more common among the bilingual children than among the monolingual Romance children. Yip and Matthews (2000) studied the acquisition of Cantonese and English in a child who had been exposed to these two languages from birth. They

51

52

Petra Bernardini

studied the transfer from Cantonese of wh-in-situ interrogatives, null objects and pre-nominal relatives into English. Comparing the presence of these phenomena in English with the development of MLU (in words) in the two languages, they concluded that this type of transfer was due to language dominance, since the MLU of Cantonese was generally higher than that of English. In the case of object drop in English by the bilingual child, they also considered input ambiguity as an explanation. Another study that seems to speak in favour of indirect cross-linguistic influence, in the sense of M&H, in combination with language dominance in bilingual L1 acquisition, is Schlyter and Håkansson (1994), who studied the acquisition of the V2 parameter in Swedish by monolingual L1 children, bilingual 2L1 children (French as the other language), and L2 children. The bilingual children who had Swedish as a stronger language had the same quantitative proportions of SV/VS as the L1 monolingual children, (about 65/35) while the ones with Swedish as the weaker language had the same proportions as the child L2 learners (about 90/10) (see Schlyter and Håkansson 1994: 58).

4. Predictions for the acquisition of word order in the Italian DP As far as N-raising is concerned, we expect to find a transfer-effect in the child who has Italian as a weaker language if the Hypothesis “weaker language = L2” is correct. This is motivated by the fact that L2 learners seem to transfer parameter values. If the hypothesis is incorrect, influence should still be observed in the bilingual children. This is motivated by Müller’s (1998) suggestion that influence is unidirectional and is predicted to occur in the direction from a language were a particular construction is non ambiguous to a language were the same construction is ambiguous. We predict the positioning of adjectives constitutes such an area. Italian has two possibilities for positioning the adjective, of which one, the pre-nominal position, is shared with Swedish. Accordingly, the Italian monolingual children should place adjectives mainly correctly and mainly post-nominally. The adult L2 learners on the other hand are expected to pre-nominalise adjectives that in Italian are obligatorily post-nominal, since they are expected to transfer the Swedish value of the N-raising parameter. The bilingual children are expected to show influence from Swedish because they receive ambiguous input (i.e. both pre- and post-nominal adjectives in Italian and only pre-nominal in Swedish) which favours the pre-

Child and adult acquisition of word order in the Italian DP

nominal position in Italian. In addition, the Swedish/Italian bilingual children are expected to differ group internally, as a consequence of the dominance of one of their languages. The child for whom Swedish is dominant is expected to behave as an L2 learner, according to the Weaker Language = L2 Hypothesis. The child for whom Italian is dominant is expected to behave just as a monolingual child, according to the same hypothesis. However, both children are also expected to have some influence from Swedish since the positioning of adjectives is an ambiguous input area. The difference between the two children must therefore be revealed in adjectives that are obligatorily post-nominal and those that are variable. The child with Italian as the weaker language should have problems with the obligatory post-nominal adjectives (as an L2 learner) and the variable ones (as a 2L1 child). In general, what has hitherto been predicted about the attributive adjective should also apply to the possessive adjective, but evidence for influence from Swedish can best be determined in cases where the learner uses the pre-nominal possessive to contrast, which is non-target-like in Italian. We expect the bilingual children to show influence from Swedish in preferring the prenominal position, as a consequence of the ambiguous input and the adults to contrast the pre-nominal position. If the weaker language is like L2, then we expect to find examples of contrastive use of the pre-nominal possessive when Italian is the weaker language.

5. Corpora and comparative criteria 5.1 The acquisition corpora 5.1.1 The monolingual Italian children The monolingual Italian longitudinal child data that I will analyse are taken from the CHILDES database (see Cipriani et al. 1993 and Bottari et al. 1993/ 1994). I have taken into account four of the children from this corpus. These children were recorded in their homes while talking in spontaneous interaction with an Italian adult, more or less once a month from approximately the age of two to three. Information about the children’s ages in the files is in Table 2. These children grew up in the central-northern part of Italy, and were considered by Cipriani et al. to be normal for their age as judged by MLU development.

53

54

Petra Bernardini

Table 2.The monolingual Italian children’s ages at the recordings (from CHILDES) Martina

Raffaello

Viola

Rosa

File

Age

File

Age

File

Age

File

Age

02 03 04 05 06 07 08 09 11 12 13 14

1;7,18 1;8,2 1;8,17 1;9,1 1;10,29 1;11,2 1;11,20 2;1,12 2;3,1 2;3,22 2;4,13 2;5,21

01 02 03 04 05 06 07 08 09 10 11 12 13 14 15

1;7.7 – 1;10,20 1;11,25 2;00,10 2;00,28 2;1,15 2;3,14 2;4,29 2;5,13 2;6,13 2;7,14 2;8,15 2;9,06 2;11,14

01 02 03 04 05 06 07 08 09 10

1;11,16 2;0,17 2;1,02 2;1,17 2;4,2 2;5,1 2;5,15 2;7,20 2;8,16 2;10,16

01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21

1;7,13 1;9,11 1;10,8 1;11,24 – 2;01,14 2;01,29 2;2,23 2;3,29 2;4,23 2;5,25 2;6,29 2;7,26 2;9,4 2;9,24 2;10,14 2;11,12 2;11,23 3;0,24 3;1,29 3;3,23

Legend: Age, Year;months,days, – = there is no information on the age of the child in the file.

The monolingual children’s overall language development was divided into four phases by Cipriani et al., following the criteria of MLU (in words) and the type of utterances the children produced. For example, verbless utterances were characterised as typical of early ages, while subordination was typical of later ages. According to these criteria, the four children’s language development is displayed in Table 3. In order to compare the monolingual children with the other corpora of the present study, the phases of Cipriani et al. where they have correlated age and MLU (see Cipriani et al.:41, 46, 54, and 62) are shown in Table 3.4 In phase 2, Cipriani et al. note that various determiners and modifiers (demonstratives, possessives, attributive adjectives and numerals) become more frequent, quoting the following examples from the children of our study (Cipriani et al. (1993: 55 (40)): (18) a. Rosa: b. Martina:

questa sua mamma. tutti i topolini.

‘this his/her/its mother’ ‘all the mice’

Child and adult acquisition of word order in the Italian DP

Table 3.Correlation of age and MLU in the monolingual children Phase 1: pre-syntactic

Raffaello: Rosa:

MLU

Age

1.2–1.6 1.5–1.6

1;7–1;11 1;9–2;2

MLU

Age

1.6–2.1 1.9–2.4 1.6–2.4 2.2–1.9

1;8–1;11 2;0–2;3 2;2–2;5 2;0–2;5

Phase 2: primitive syntax

Martina: Raffaello: Rosa: Viola:

Phase 3: completedness of the nuclear phrase

Martina: Raffaello: Rosa: Viola:

MLU

Age

2.1–2.6 2.4–3.0 2.4–2.9 1.9–2.9

2;1–2;3 2;4–2;6 2;7–2;9 2;5–2;7

Phase 4: consolidation and generalisation of the rules in complex structural combinations

Martina: Raffaello: Rosa: Viola:

c. e. i. k.

Raffaello: Martina: Rosa: Rosa:

MLU

Age

2.6–2.6 3.0–3.0 2.9–3.0 2.9–1.8

2;3–2;7 2;8–3;0 2;10–2;11 2;6–2;10

l’orso cattivo. un bello vestito. tutti i calli! (cavalli) mucchine piccine queste.

‘the evil bear’ ‘a beautiful dress’ ‘all the horses’ ‘cows.small small these’

In phase 2 the MLU is 1.6–2.4, as can be seen in Table 3. This means that the comparison with the bilingual children should concern a similar MLU span. 5.1.2 The adult Swedish L2 learners The L2 data, also longitudinal, consist of a Swedish middle aged couple, Karl and Stina, who were recorded in Italian during one year’s stay in Italy. Karl

55

56

Petra Bernardini

Table 4.The adult learners Adult learner

Rec.

Months spent in Italy before each recording

Stina Karl

5 5

0, 3, 6, 9, 12 0, 3, 6, 9, 12

Legend: 0 months in Italy, after an Italian language course in Sweden before leaving for Italy.

worked for a Swedish company in the northern part of Italy, and Stina, his wife, accompanied him. Five conversations between the couple and an Italianspeaking interviewer were recorded during the year, with approximately three months between the recordings. The length of each recording varied from fifteen to thirty-five minutes. Karl and Stina had also received some formal Italian language instruction before and during their stay in Italy. Information about Karl and Stina is shown in Table 4 (after Bardel 2000: 76, Table 1). Bardel studied Karl and Stina’s acquisition of negation in Italian as compared to the development of finiteness. The two learners were at approximately the same level throughout the period studied, although Stina was more careful and did not speak if she didn’t think it was right. Karl was the opposite and spoke without worrying about grammatical errors. These two learners had several occurrences of postverbal negation during the whole period studied (which is not target-like in Italian but in Swedish), even in the last recordings. This was interpreted as influence from Swedish. 5.1.3 The bilingual Swedish/Italian children The bilingual Swedish/Italian corpus consists of longitudinally collected spontaneous speech of two bilingual children who were both growing up in bilingual families. Lina, a girl, grew up in Italy, in the same area as the monolingual children (see above), with a Swedish mother and an Italian father. Lukas, a boy, grew up in Sweden with an Italian mother and a Swedish father. His mother is from the northern part of Italy.5 The two children were audio-recorded at least once a month from age 1;8 and 2;0, respectively, and the data collection stopped at age 3;7 for both. The children were recorded during a session of approximately one hour of interaction, thirty minutes with a native Italian speaker and thirty minutes with a native Swedish speaker (in Lukas’ case it was however 45 minutes in Italian and 20 in Swedish). Transcriptions of one such session per month were made according to the CHAT format (MacWhinney and Snow, 1985). Both the transcriptions and the tapes are used for the analysis. Pertinent

Child and adult acquisition of word order in the Italian DP

Table 5.The bilingual children – Lina Age

It. utt.

It. words

It. MLU

Sw. utt.

Sw. words

Sw. MLU

1;8 1;9 1;11 2;2 2;2,16 2;4,23 2;7 2;8,21 2;9,19 2;10,5 2;10,24 2;11,11 3;6 3;7

80 83 149 175 111 28 154 251 203 329 483 267 282 272

144 120 274 351 225 65 329 607 492 879 1284 651 796 788

1.8 1.5 1.8 2.0 2.0 2.3 2.1 2.4 2.4 2.7 2.7 2.4 2.8 2.9

64 24 13 67 12 48 17 22 8 16 32 15 1 1

82 26 16 84 12 57 19 23 11 19 42 26 2 1

1.3 1.1 1.2 1.3 1.0 1.2 1.1 1.0 1.4 1.2 1.3 1.7 – –

2910

7105

338

420

Legend: Age, year; month, day; Italian utt(erances), total amount of Italian utterances produced by the child during the recording with the Italian interlocutor; Italian words, total amount by the child during the recording with the Italian interlocutor; Italian MLU, MLU of the Italian utterances and words produced by the child during the recording with the Italian interlocutor; Swedish MLU, MLU of the Swedish utterances and words produced by the child during the recording with the Swedish interlocutor, etc.

information on the recordings is shown in Table 5 for Lina and 6 for Lukas. In Bernardini (2001) Lina’s Italian was considered to be stronger than her Swedish while the opposite was true of Lukas, based on MLU development and on the amount of utterances and words produced during the interaction with the interlocutor of each language (cf. willingness to speak, as a criterion for language dominance (Schlyter 1993, 1994, 1999 f. ex.)). In Tables 4 and 5 the two criteria concerning this balance are shown. Tables 5 and 6 show that the utterances in Italian are far more than those produced in Swedish by Lina, while the opposite is observed for Lukas. Italian develops for both children, although for Lukas this language is not developing as much as his Swedish. For Lina the situation is the opposite. Her Italian develops normally (see Table 3; Cipriani et al. (1993); Antelmi (1997) for the normal development of Italian) while her Swedish is not as productive, with a rather low MLU. In Figure 1 we see the difference between the two children in MLU development in Italian. As is clear in Figure 1, Lina’s Italian MLU is higher at every age point than Lukas’, approximately by one word per utterance. However, given the possibility that Lukas’ language development is simply slower than Lina’s, I also compare

57

Petra Bernardini

Table 6.The bilingual children – Lukas Age

It. utt.

It. words

It. MLU

Sw. utt.

Sw. words

Sw. MLU

2;0 2;1 2;2 2;3 2;4 2;5 2;7 2;7,21 2;8,14 2;11 3;1,15 3;3,15 3;4,15 3;6,13 3;7,4

133 32 71 66 75 111 43 29 22 151 89 246 75 56 118

144 35 80 97 149 162 73 64 36 223 123 589 149 62 229

1.1 1.1 1.1 1.5 1.4 1.5 1.7 2.2 1.6 1.5 1.4 2.3 2.0 1.1 2.0

191 53 249 214 109 288 416 124 338 263 292 304 328 39 92

361 88 568 480 338 903 1215 344 1133 828 1044 1297 1319 169 388

1.9 1.7 2.3 2.3 3.1 3.1 2.9 2.8 3.4 3.1 3.6 4.2 4.02 4.3 4.2

1243

1992

3300

9307

Legend: See Table 5.

MLU

58

3.5 3 2.5 2 1.5 1 0.5 0

Lina Lukas

1,

8

2,

1

2,

4 7 2,

,2

1 1 2,

0,

5 1 3,

,1

5 6 3,

,1

3

Age

Figure 1.Italian MLU in Lina and Lukas.

the two languages of each child, as shown in Figures 2a and 2b. First Figure 2b shows that Lukas’ Swedish develops normally. His language development is normal, i.e. his L1 is developing with normal speed of acquisition for his age. Second, Figure 2b shows that the other language of this child, Italian, is not developping normally according to age, i.e. his MLU does not improve from 2 to 3 between the ages 2;0 and 3;7. In fact there is a discrepancy in MLU between

3.5 3 2.5 2 1.5 1 0.5 0

MLU I

6 3;

10

,2

4

19 2;

2;

9,

7 2;

16 2, 2;

1;

1;

11

MLU S

8

MLU

Child and adult acquisition of word order in the Italian DP

Age

Figure 2a.MLU S(wedish) and I(talian) in Lina. 5

MLU

4 3

MLU S

2

MLU I

1 0 3;7

3;4

3;1

2;8,14

2;7

2;4

2;2

2;0

Age

Figure 2b.MLU S(wedish) and I(talian) in Lukas.

Swedish and Italian, by approximately one word per utterance in all the recordings in the age period 2;0–3;7. The same is observed for Lina in Figure 2a. Recall that the noun phrase expands in Phase 2 in the monolingual children, when their MLU is above 1.5. Thus, the first files of Lukas are not relevant for the comparison with the monolingual children. There are also recordings at later ages when his MLU is below 1.6, but this might be due to unwillingness to speak. Since the children speak less in their weaker language (as shown in Figure 2), this could be reflected in the MLU. If the child hardly says anything but yes and no, the MLU will be very low. Another criterion for language balance in this study is therefore upperbound (the longest utterance produced by the child in each recording). Figure 3 shows this for each of the bilingual children.

59

10 8 6 4 2 0

UBS

3;

6

4 ,2

2;

10

7 2;

19

3 2;

9,

2 2;

2;

9

UBI

1;

UB

Petra Bernardini

Age

Figure 3a.U(pper)B(ound) in S(wedish) and I(talian) in Lina. 20 15 UB

60

UBS 10 UBI 5 0 2;

0

2;

3

2;

7

2;

11

3;

4

Age

Figure 3b.U(pper)B(ound) in S(wedish) and I(talian) in Lukas.

The upperbound of Italian and Swedish is even more uneven than that of MLU in both the children. Most importantly, we see that at the end of the period studied, the upperbound of Italian is higher than it was in the beginning of the period, even though the MLU is low (cf. Figure 2b). 5.2 Method for analysing the word order in the DP In order to analyse the position of the adjectives in the four monolingual children I used the transcriptions available in the CHILDES database, searching them through with the tools FREQ and COMBO of the CLAN program (see MacWhinney 2000). Lists of every child’s complete vocabulary in the files were made by FREQ. Subsequently, everything that could possibly be an adjective was included in a search file; COMBO could search for the context in which the adjectives occurred in every file. I also had access to the audio recordings of the Swedish adults and the bilingual children. In the children’s case, it was necessary to take into account forms

Child and adult acquisition of word order in the Italian DP

that did not correspond to the adult norm, such as un atto for ‘un altro’ (another), or butto for ‘brutto’ (ugly), bio for ‘mio’ (my). In other words, forms that are reduced or alternated due to typical child pronunciation. In the case of the adults and the bilingual children, Italian adjectives combined with Swedish nouns, and vice versa, were excluded. Adjectives were considered as modifiers that require their NP to be preceded by a determiner (i.e. article or demonstrative) in argument position. This criterion includes indefinite adjectives (such as altro) but not demonstratives or quantifiers (as quello/questo ‘that/this’ tutti ‘all’ or tanti ‘many’). This criterion for sorting out demonstratives and numerals is motivated by the fact that they may be considered more determiner-like. The article may not co-occur with the demonstratives, numerals, and the quantifier tanti, and follows the quantifier tutti. It was noted whether the adjectives occurred pre- or post-nominally, and whether they were correctly positioned. The latter judgement was based on the context and by listening to intonation in the recordings. For the monolingual children the context in the transcriptions was the only source of information. The same procedure as described above was carried out with the possessive adjective. The identification of non-target-like production of adjectives (attributive and possessive) is based on position and type of adjective. As we have seen in Section 2, there are attributive adjectives that may occur in both positions. To decide whether or not these were target-like, the intended meaning was judged in terms of the context. Finally, repetitions of the interlocutors’ previous utterances are not counted.

6. Child and adult acquisition of word order in the Italian DP 6.1 The monolingual Italian children In total, the four monolingual children as a group produced 54 possessive adjectives and 78 attributive adjectives in all sixty files taken together (see Table 7), which divided by four makes 13.5 possessive adjectives/child and 19.5 attributive adjectives/child. The results from these children’s positioning of possessive adjectives will be treated in 6.1.1 and their attributive adjectives in 6.1.2. 6.1.1 The monolingual Italian children’s positioning of possessive adjectives Both the pre-nominal and the post-nominal positions are observed in the monolingual children’s data. All possessive adjectives are positioned correctly, as is shown in Table 7. Observe that the pre-nominal position is more frequent than the

61

62

Petra Bernardini

post-nominal position: 42 cases in contrast to 12. This is not surprising, since the latter should only be used in contrastive sense, except in the case of casa mia. Although the pre-nominal position of the possessive adjective dominates in the monolingual children, the first position found in each of the monolingual datasets is the post-nominal one, as is shown in Table 8.6 Table 7.The monolingual children’s total amount of possesssive adjectives in the two positions Child

total Pos

Pos N

*Pos N

N Pos

*N Pos

Martina Raffaello Rosa Viola

12 5 35 2

10 4 27 1

– – – –

2 1 8 1

– – – –

54

42

–

12

–

Legend: *, incorrect in this position

Table 8.First appearance of possessive adjectives in the monolingual children

Mart Raffaello Rosa Viola

Pos N

N Pos

2;4,13 2;5,25 2;5,25 2;8,16

2;3 2;5,13 2;0 2;1,02

For all four children, the first occurrence of the post-nominal possessive adjective is of the type casa mia, but also with kinship nouns. With these nouns the possessive is generally pre-nominal (see Cardinaletti (1998: 41), not preceded by the determiner (mia nonna/mio fratello ‘my grandmother/my brother). (19) a.

Martina: sì, sì ah, bene do ado [/] do a casa mia, eh! ‘yes, yes, well I’m going to home my’ [Mart11, 2;3] Mother: ci sei a casa tua, aspetta … ‘you are at home [e.g. home your], wait’ b. Raffaello: babbo mio. ‘my father’ Mother: cosa babbo tuo? ‘what do you mean, your father?’ Raffaello: no babbo Stefano. ‘not Stefano’s father’ c. Rosa: nonna m [//] mia. ‘grandmother mine’ d. Viola: babbo mio. ‘daddy mine’

Child and adult acquisition of word order in the Italian DP

Since there is no indication that a determiner is missing the examples in (3) might be considered cases of the type casa mia or Gianni mio which need not be preceded by the article.7 The same cases are in fact the first ones to appear in Antelmi’s (1997) longitudinal case study of a monolingual Italian girl’s L1 acquisition. They occur during the age span 1;9–2;2 (see Antelmi 1997: 88), as I have already noted in Section 3, repeated here as (20). Antelmi interpreted such constructions as if the copula was missing (papà (è) mio ‘daddy (is) mine’). (20) tata mia/papà mio/mamma mia [from Antelmi 1997: 88] ‘“nanny” my/daddy my/mommy my’

However, since, for example, the mother in (19b) interprets it as a unit, it is quite possible that constructions such as (19) are cases of the type casa mia/babbo mio. The post-nominal possessive without determiner is replaced, or at least outperformed in frequency by the pre-nominal possessive (see in particular Rosa, who has most possessives), which becomes more productive than the post-nominal possessive after the first files with only the post-nominal articleless possessive (see Table 9). The determiner is omitted ungrammatically with both the pre-nominal and the post-nominal possessive (which is rarer in all children). In about a third to half of the time, the determiner is present in obligatory contexts with the possessive (except Viola who hardly produces any possessive adjectives whatsoever, see Table 9, column 2). In none of the children is there a 100% of determiners in obligatory contexts with the possessive adjective. This is the case for both the pre-nominal and post-nominal position. The post-nominal possessive does not have many occurrences where the determiner is obligatory. Consider that the definite article is productive quite late in all the children (see Cipriani et al. (1993: 142–143, Figures 21–24). It is only in Phase 3 that article omission becomes less frequent (from 0–40%). This might suggest that there is no DP structure in these children’s noun syntax. However, given the fact that they have post-nominal possessives very early and that the article is present in at least half of the pre-nominal cases, this cannot be the case. The N must have raised somewhere. To suppose that it has raised to D is out of the question, since determiners are omitted so often, both with the pre-nominal possessive (Table 9) and generally (see Cipriani et al.: 142–143, Figures 21–26). It is generally known that Italian monolingual (see Pizzuto and Caselli 1992; Chini 1997) and bilingual children (Müller 1990; Kupisch et al. 2001) acquire

63

64

Petra Bernardini

Table 9.The monolingual children’s possessive contexts Child

Pos N

N Pos Casamia/ babbo mio

File, Age

other

Martina 11, 2;3 13 2;4

2(2casa mia) 9(6+D,3−D)

Raffaello 10, 2;5 11, 2;6 12, 2;7 13, 2;8

1(-D) 1(+D) 2(2–D)

2 (babbo mio, camera mia)

Rosa 05, appr. 2;0 11, 2;5,25 12, 2;6,29 13, 2;7,26 14, 2;9,4 15, 2;9,24 16, 2;10,14 17, 2;11,12 18, 2;11,23 19, 3;0,14 20, 3;1,29 21, 3;3,23

1 2 12(4+D,8−D) 4(4−D) 3(1+D,2−D) 5(4+D,1−D) 1(+D) 1(-D) 2(1+D,1−D) 13(6+D,7−D)

1(nonna mia) 1(casa mia) 2(tappa tua, mamma mia) 1 (a mucche mia) 2(2casa sua) 1(-Dposto tuo)

1(nonna mia) 1(babbo mio) 1(a maetta mia) 2(1casa mia,1 mamma 1(-D(a)mico mio) mia) 2(2casa tua)

Viola 03, 2;1,2 09, 2;8

1(babbo mio) 2(2+D)

Legend: _amico mio = omission of determiner+amico mio; (a)mico mio = the ‘a’ is not pronounced by the child; DPosN = possessive+noun sequence preceded by a noun in obligatory context; _PosN = omitted determiner in obligatory context with possessive noun sequence; DNPos = noun+possessive sequence preceded by a determiner in obligatory context; _NPos = omitted determiner in obligatory context with noun possessive sequence. If it is not marked with ‘_’ the determiner is not obligatory.

gender and number quite rapidly. This is thought to be due to the fact that Italian has clear morpho-phonological cues for gender in the noun-ending. Therefore it is reasonable to assume that the Italian monolingual children raise

Child and adult acquisition of word order in the Italian DP

the N to NumP overtly, thereby deriving the post-nominal possessives. The reason why they do not fill D0 might be that they do not yet master the rule that arguments are DPs and not NPs (Stowell 1991). 6.1.2 The monolingual Italian children’s positioning of attributive adjectives In two of the monolingual Italian children studied here, the attributive adjective first makes its appearance pre-nominally (Rosa and Viola). In one child, Martina, the first adjectives that occur with a noun are positioned both pre- and postnominally. In Raffaello, the first occurrence is post-nominal. The age and position of the first appearance of the attributive adjectives are shown in Table 10. Table 10.The position of the first occurences of attributive adjectives in the monolingual children Child

AN

NA

Martina Raffaello Rosa Viola

1;11,2 –* 2;0 2;5,15

1;11,2 2;3,14 1;9;11 2;7,20

*In the Raffaello files the A N order never occurs

The fact that the adjectives appearing earliest are positioned differently in the four children is of course difficult to interpret without examining which type of adjectives are found first. In Martina’s first file with attributive adjectives there are 4 post-nominal adjectives (lungo ‘long’ 2, grosso grosso ‘very big/fat’ 2) and 1 pre-nominal (grande ‘big’). All of her adjectives are of the kind that may be pre- and post-nominal, except grosso grosso ‘very big/fat’. In Raffaello’s first file there are 5 post-nominal adjectives (4 cattivo ‘evil’, 1 grosso), of which both types are position-variable; Rosa and Viola have one occurrence each of altro ‘(an)other’, which is obligatory pre-nominal. The fact that the children have both pre-and post-nominal adjectives in their first files of appearance points to a need to consider the children’s overall production of attributive adjectives. This is shown in Table 11 (see also Table A, Appendix). In Table 11 it seems that the pre-nominal position predominates slightly over the post-nominal one, occurring 42 and 36 respectively. However, it should be noted that in the pre-nominal position, the adjective altro is by far the most common, as we see in Table 12. We can thus read from Table 12 that if 39 pre-nominal adjectives out of a

65

66

Petra Bernardini

Table 11.The position of attributive adjectives in the monolingual Italian children Child

A, total no.

AN

NA

Martina Raffaello Rosa Viola

18 16 39 5

13 – 26 3

5 16 13 2

Total

78

42

36

Table 12.Pre-nominal adjectives in the monolinual children Child

AN

altro N

Martina Raffaello Rosa Viola

12 – 24 6

10 – 24 5

Total

42

39

total of 42 pre-nominal adjectives consist of the indefinite adjective altro, there remains only 3 pre-nominal variable adjectives in the whole Italian monolingual corpus that I have investigated here. The fact that the variable adjectives are positioned preferably in the postnominal position tells us that the monolingual Italian children do not have any significant problems with N-raising. Considering that the kind of N-raising we are talking about is when N raises to an AgrP intermediate to NP and DP, it is not a problem to suppose that this AgrP is projected, even though these Italian children have high percentages of article omission (see Cipriani et al.). 6.2 The adults 6.2.1 The adults’ positioning of the possessive adjective The adult learners produce in total 25 possessive adjectives. As can be seen in Table 13, they are always pre-nominal. They make two errors in putting the possessive adjective to the left of the noun when it is used in a contrastive way, as in (21) and (22) (21) Stina3: eeh Karls genitori e miei genitori anche. ‘Karl’s parents and my parents too’

Child and adult acquisition of word order in the Italian DP

(22) Karl2: sì # una macchina, eeh, no [/] no mia macchina me# ‘yes (pause) a car, eeh, no no MY car but (pause) di ditta la macchina. ‘of company the car’ Table 13.The position of possessive adjectives in the adult learners Adult

total no. Pos

Pos N

*Pos N

N Pos

*N Pos

Stina Karl

8 17

7 16

1 1

– –

– –

The total absence of the possessive adjective in its post-nominal position in combination with the ungrammatical contrastive use of the pre-nominal possessive in Italian indicates that Swedish is influencing Italian. 6.2.2 The adults’ positioning of the attributive adjective The adults differ from the monolingual children in that they produce more attributive adjectives (see Table 14), although there are fewer files in the adult corpus (also less recording time, see 5.3.): the four monolingual children produce altogether 78 attributive adjectives. The adults produce both pre- and post-nominal adjectives, 82 pre-nominal and 31 post-nominal ones (see Table B, Appendix for occurrences in each file). There is one difference between the two adults: Stina does not make any error in putting the adjective pre-nominally, but only post-nominally (although only once, against 10 correct cases) as in (23). (23) Interviewer: Modena # ci sei stata? ‘Modena (pause) there have (you)been? Stina2: sì, sì piacio # è un città bella. ‘yes, yes like (pause) (it)is a town beautiful

Table 14.The adults’ positioning of attributive adjectives Adult

A, total no.

AN

*A N

NA

*N A

Stina Karl

30 73

19 42

– 11

10 20

1 –

Total

103

61

11

30

1

67

68

Petra Bernardini

Karl, on the other hand makes errors in both ways, those in pre-nominal position predominating, as in (24). (24) a.

Karl3: un altra problema è un uomo un molto importante uomo in Svezia. ‘another problem is a man a very important man in Sweden’ b. Karl4: stupido # eh più facile lavoro con svedese gente. ‘stupid, (pause) much easier work with Swedish people’ c. Karl4: spesiale pasta, speciale. ‘spesial pasta, special’ (he is correcting the pronunciation)

Hence, Karl does make errors both in the position of the complement of the AP as in (24a) and with adjectives that can only occur in the post-nominal position (as in (24b) and (24c)). In Table B (Appendix) the types of adjectives and their positioning in a developmental perspective in Karl and Stina are shown. 6.2.3 Summary Are there any differences between the adult learners’ and the monolingual children’s use of possessive adjectives and attributive adjectives? Yes indeed, with respect to possessives. All the Italian monolingual children begin with post-nominal possessives and use them throughout the period studied. The adults never use a single post-nominal possessive. However, they use prenominal possessives to contrast the possessor. This is not grammatical in Italian and was never done by the monolingual Italian children As far as the positioning of the attributive adjectives is concerned, there is one major difference as well. The adults position obligatorily post-nominal adjectives pre-nominally. This difference, we have predicted, may be due to influence from Swedish. As far as variable adjectives are concerned, there are a few errors (see Table B, Appendix). 6.3 The bilingual Italian/Swedish children

The bilingual Italian/Swedish children’s positioning of Italian possessive adjectives In contrast to the monolingual children, the bilingual children initially use the possessive adjective in its pre-nominal position, as is shown in Table 15. In this respect the bilingual children are a homogeneous group. They differ, however, in another respect, which is the total production of possessive adjectives. This is shown in Table 16. Lina, whose Italian is stronger than her Swedish (Table 5a, Figures 2a and 3a), has a total amount of possessive adjec6.3.1

Child and adult acquisition of word order in the Italian DP

Table 15.First appearance of Italian possessive adjectives in the bilingual children

Lina Lukas

Pos N

N Pos

2;8,21 2;4

2;9,19 –

tives that is distributed in a similar way to the monolingual group with regard to the position of the element: 47 pre-nominal possessives and 19 post-nominal ones (the monolinguals have, both as a group and as individuals, approximately the same distribution (see Table 19). Table 16.The bilingual children’s total amount of Italian possessives in the two positions Child

total no. Pos

Pos N

*Pos N

N Pos

*N Pos

Lina Lukas

66 9

47 8

– –

19 –

– –

* = incorrect in this position

Lukas (Italian as the weaker language, Table 6b and Figures 2b and 3b) differs from his bilingual peer and the monolingual children in that he has no post-nominal possessive adjectives in all his 15 files. 6.3.2 The bilingual children’s positioning of attributive adjectives The two bilingual children differ regarding the first appearance of attributive adjectives. Both positions are observed at age 2;8,21 in Lina, whereas the prenominal one is observed as early as 2;1 in Lukas. The post-nominal adjective is not observed before age 3;3,15 in Lukas, although his Italian MLU is comparable to the monolingual children’s when their modifiers are productive (see Table 17). Table 17.The position of the first occurences of attributive adjectives in the bilingual children Child

AN

NA

Lina Lukas

2;8,21 2;1

2;8,21 3;3,15

69

70

Petra Bernardini

Their production of attributive adjectives is quite low, comparable to the monolingual child Viola (5 in total, see Table 18). This is shown in Table 14. Table 18.The position of attributive adjectives in the bilingual children Child

A, total no.

AN

*A N

NA

N *A

Lina Lukas

6 8

3 2

– 4

3 2

– –

Total

14

5

4

5

–

Whereas Lina does not make any errors in the positioning of attributive adjectives, Lukas’ errors concern those adjectives that can be in both positions, but which when used contrastively, should be post-nominal, as in (25) and (26). (25) Mother: un cane, lo facciamo qua, un occhio, un naso, le orecchie. ‘a dog, we make it here, an eye, a nose, the ears’ Lukas (2;1): grosso cane. ‘BIG dog’ Mother: ma lui ha la bocca più grossa [eh]? but he has the mouth more big [“has he not”]? ‘but he has a bigger mouth’ (26) Lukas (2;3): jamen, questa # piccolina bocca. [Swedish Yes but], this small mouth’ Mother: [uhm], ha la bocca più piccola quella. ‘[“yes”], that one has a bigger mouth’

In (25) the variable adjective grosso is pronounced with emphasis and cannot be in a subjective reading, since he cannot have an opinion of the dog, or a subjective way of looking at it while it still has to be drawn. The dog is not yet. Therefore grosso must be objective, describing an internal physical property of the dog. In (26) it is clear from the mother’s answer that what is meant is a contrastive use of piccolina. However, the error Lukas makes in (2;1–2;3) occur when he has an MLU below 1.6, and is thus not comparable with Phase 2, where the monolingual children begin to use modifiers productively; see Table C (Appendix A: the developmental perspective on the types of adjectives). One year later (see Table C, Appendix) he has post-nominal attributive adjectives, also used contrastively, as in (27) and (28). (27) Lukas (3;3,14): il trattore grosso riesce. ‘the tractor big can make it’

Child and adult acquisition of word order in the Italian DP

(28) Lukas: che, non è la pappa che non è, le mani pulite. (3;3,14) that, not is the food that not is (dirty), the hands clean ‘it is not the food that is not clean but you have to have clean hands’

The number of pre-nominal attributive adjectives produced by the monolingual children diminished substantially when the indefinite adjective altro was excluded. This check was also carried out on the bilingual children (see Table 19). Table 19.The pre-nominal adjectives in the bilingual children Child

total A N

altro N

Lina Lukas

3 6

3 2

Total

9

5

In fact three of Lina’s and two of Lukas’ pre-nominal adjectives are represented by altro. Considering that Lukas has six pre-nominal attributive adjectives of which four are erroneously put in this position there remains no correct pre-nominal attributive adjectives that can occur in both positions in his data. Lina’s number of pre-nominal attributive adjectives reduces to three, and these three are all of the kind that can occur in both positions, e.g. (29). (29) Lina (2;10,24): ora devo fare una casa bella! ‘now I have to make a BEAUTIFUL house!’ (She is drawing, and has not yet succeeded in making a beautiful house)

Thus, the bilingual children differ from each other in the positioning of the attributive adjective. One of them, Lina, whose Italian is much stronger than her Swedish, behaves as the monolingual children, whereas the other child, Lukas, whose Italian is weaker than his Swedish, behaves differently from all the other children. First, he begins to produce attributive adjectives earlier than the other children. But, second, he fails to put them in the post-nominal position, at first. At a late age, however, he also has post-nominal adjectives.

Summary of the bilingual children’s acquisition of possessive and attributive adjectives The pattern that emerges from the study of the bilingual children is clear. One child behaves more as a monolingual child and the other more as an L2 learner. This is most striking with regard to the positioning of the possessive. The child 6.3.3

71

72

Petra Bernardini

Lukas has only pre-nominal possessives, exactly as the adults. The other child, Lina, has both pre-and post-nominal possessives from the beginning and uses the type casa mia that the adults and Lukas do not use. Lukas does not use prenominal possessives in a contrastive way, although this is what the adults do. Regarding the attributive adjectives, one thing is clear: Lina does not make errors with obligatory post-nominal adjectives, and neither does Lukas. While Lina does not make any errors with variable adjectives, just as the monolinguals do not, Lukas does so quite regularly, just like Karl (adult learner). The adults make errors both with obligatorily post-nominal adjectives and variable adjectives, but Lukas only makes errors in the pre-nominalisation of contrastively used variable adjectives.

7. Discussion and conclusions In this paper I have compared the acquisition of word order in the Italian DP of three types of learners: monolingual L1 children, bilingual Swedish/Italian children, and adult L2 learners. Specifically, I looked at adjective/noun order, adopting the N-raising hypothesis of Bernstein (1991), Cinque (1995), which states that the underlying structure of the word order of Germanic and Romance languages is the same, i.e. A N, and that the N A order of Romance languages is due to overt movement of the N. I also looked at the order of the possessive and the N, since this element can also be both pre- and post-nominal in Italian, but only pre-nominal in Swedish. One reason for studying word order in the DP was that this appears to be an ambiguos input domain, since possessive and attributive adjectives in Italian are both pre- and post-nominal, but only pre-nominal in Swedish. Müller (1998) has suggested that such domains are subject to influence. The word order in the DP has also been shown to be the target of transfer in L2 acquisition where Germanic is L1 and Romance is L2 and vice versa (see Parodi et al. 1997 and Granfeldt 2000a,
b). My research question was whether bilingual children would perform as L1 monolingual Italian children or as adult Swedish L2 learners of Italian. My results show that the monolingual Italian children begin with the postnominal possessive and later use the element in both positions without errors. Results are similar for the attributive adjectives, except that both word orders were found in the beginning. However, possessives were mainly pre-nominal and attributive adjectives were mainly post-nominal in the monolingual data.

Child and adult acquisition of word order in the Italian DP

My interpretation of these findings was that the monolingual children have overt N-raising to at least NumP, but no ‘active’ upper part of the DP structure. This is supported by three facts: 1. possessives were post-nominal without articles; 2. they were pre-nominal with variable article omission; 3. adjectives were mainly post-nominal. Comparing different learner groups in their ability to position the Italian adjective correctly with respect to the noun, it was found that when only N-raising was the measure, there seemed to be more similarity between monolingual L1 children and the 2L1 Italian/Swedish child with Italian as the stronger language than there was between this latter child and the 2L1 Italian/ Swedish child with Italian as the weaker language, who performed similarly to the two adult L2 learners of Italian. When the fact that Italian has variable adjectives was considered, it appeared that the child with Italian as the weaker language did not make any errors with adjectives that are obligatorily post-nominal in Italian, as the adults did. Nor did he stress the pre-nominal possessive as the adults did, even though he had only pre-nominal possessives, while the monolingual children had them in both positions. He did, however make some errors with adjectives that are variable with respect to position in Italian. These findings suggest that the child with Italian as the weaker language was influenced by ambiguous input, as monolingual and bilingual children have been predicted to be by Müller (1998) and Müller and Hulk (2001). It remains, therefore, to be seen if the difference between Italian as the weaker language and monolingual children/children with Italian as stronger language is just a difference between monolingual and bilingual acquisition. This calls for more studies on the acquisition of word order in the Romance DP. The adults had post-nominal adjectives in the beginning and made errors in pre-nominalising some of them. The errors concerned obligatorily postnominal adjectives. This suggests that the adults were about to reset their N-raising parameter from the Swedish value to the Italian value. The observed difference between the adults and the children could maybe also be accounted for in an alternative approach to N-raising. Alexiadou (2001) assumes that postnominal adjectives are generated in a predicative phrase below N. This could explain that the monolingual children have mainly post-nominal attributive adjectives, while the adults have mainly pre-nominal adjectives.8 The adults have access to higher functional categories and the children do not.9 How would this perspective account for the differences between the weaker language and monolingual acquisition? This has to be considered in future research.

73



74

Petra Bernardini

Notes * The research presented in this work is supported by a grant from the former Swedish Council for Research in the Humanities and Social Sciences (HSFR) to Suzanne Schlyter (F0686/1998). I am very grateful to Camilla Bardel for giving me access to her data on L2. Thanks to Anne Börjesson, Verner Egerland, Jonas Granfeldt and Suzanne Schlyter for reading and commenting on previous versions of this paper. I also gratefully acknowledge the comments of an anonymous reviewer. Responsibility for any errors remains with the author. 1. For example Cinque (1995); Giorgi and Longobardi (1994) for Italian; Giusti (1993; 1994; 1999) for mainly Italian, Rumanian and the Scandinavian languages, Dimitrova-Vulchanova and Giusti (1998) for Balkan languages; Longobardi (1994, 1996) for Italian in particular, (2000) for a typological view on different kinds of N-movement; Bernstein (1991, 1993) for Romance in general; Laenzlinger (2000) for French and English, Valois (1991) for French and Vallon; and Zamparelli (1995) for Italian and English. See however Alexiadou (2001), Lamarche (1991), Delsing (1993) for a different view. Delsing (1993) accounts for other word order differences with N-raising (e.g. Swedish post-nominal enclitic article -en/et ‘-the.uter/neuter’), but accounts for the difference in the word order between the noun and the adjective in Romance and Germanic differently. 2. The “liason” in gros_avions is an example of a weak morphological form. In Italian some position-variable adjectives are reduced morphologically when pre-nominal, as un bel vestito ‘a nice dress’ vs. un vestito bello ‘a dress nice’ (i.e. a dress that is nice, in comparison to another dress) Nespor (1988) observes some position-variable adjectives are interpreted as quantifiers in the pre-nominal position: (i) a. b.

Ho visto macchine diverse. (I) have seen cars different Ho visto diverse macchine. (I) have seen many cars

3. Cardinaletti analyses possessives in the same way as personal pronouns were analysed in Cardinaletti (1994) and in Cardinaletti & Starke (1999). 4. I have translated the names of the phases from Italian. 5. For more information about the bilingual corpus, see Bernardini (2001). 6. In the files Mart(ina)11, Raf(faello)10, Rosa10, Viola03. 7. In the Italian CHILDES corpus an omitted word is usually indicated with ‘0w’ and this has not been done in these cases. 8. See fn. 1. 9. See Aldridge et al. (1997) for a discussion of different approaches to N-raising and their consequences for interpreting acquisition data.

Child and adult acquisition of word order in the Italian DP

Appendix Table A.The monolingual children’s adjective contexts Child

AN

*A N

NA

File, Age

oblig.

Variable

oblig.

oblig

Variable

altro

bello, grande Piu/molto grosso-grosso, blu

Piu/molto grosso-grosso, blu

bello, grande

1

2

2

Mar. 03, 1;8 07, 1;11,2 08, 1;11,20 11, 2;3,1 12, 2;3,22 13, 2;4,13 14, 2;5,21

1 2 2 2 2

1 1

1 2

Raf. 10, 2;5,13 11, 2;6,13 12, 2;7,14 13, 2;8,15 14, 2;9,06 15,2;11,14

1 3 3

4 1 1 1 2 1

75

76

Petra Bernardini

Ros. 02, 1;9,11 05, 2;0 10, 2;4,29 11, 2;5,29 12, 2;6,29 13, 2;7,26 14, 2;9,4 15, 2;9,24 16, 2;10,14 17, 2;11,12 18, 2;11,23 19, 3;0,24 20, 3;1,29 21, 3;3,23

2 2 1 4

5 2 1 4 5 1

1 2

1 1 1 1 2 1

1 1

Vio. 03, 2;1,02 05, 2;4,2 07, 2;5,15 08, 2;7,20 09, 2;8,16

1 1 2 1 2

Legend: *, not targetlike in this position

1

Child and adult acquisition of word order in the Italian DP

Table B.The adult Swedish L2 learners’ adjectives in the files AN

AN

*A N

NA

NA

Oblig

variable

Oblig

oblig

variable

Altro/ primo

bello, grande

grosso-grosso, blu grosso-grosso, blu bello, grande Piu/Molto+Aoblig Piu/Molto+Aoblig / / variabile variabile

Stin. 1 2 3 4 5

1 5 4 2

1 4

4+1* 3

1, +1*

1

Kar. 1 2 3 4 5

1 4, +1* 9 7,+2* 4

2 3 12 3

1 1 16 3

4 9 3

2

*, non targetlike in this position

Table C.The bilingual children’s adjective contexts AN

*A N

NA

oblig

variable

oblig

oblig

variable

altro

bello, grande

Piu/molto grosso-grosso, blu

Piu/molto grosso-grosso, blu

bello, grande

Lina LII13 LII14 LII15 LII20

2

1

1 1 1

77

78

Petra Bernardini

Lukas LUI02 LUI03 LUI04 LUI09 LUI16 LUI18 LUI21

1* 1* 1* 1 1*

2

1 1

Legend: *, non-targetlike in this position

References Alexiadou, A. 2001. “Adjective syntax and noun raising: word order asymmetries in the DP as the result of adjective distribution”. Studia Linguistica 2001: 217–248. Aldridge, M., Borsley, R. D., Clack, S., Creunant, G. and Jones, B. M. 1997. ‘The acquisition of noun phrases in Welsh’, In A. Sorace, C. Heycock and R. Shillcock (eds), Proceedings of the GALA ‘97: Conference on Language Acquisition, Human Communication Research Centre, University of Edinburgh, 6–9. Antelmi, D. 1997. La prima grammatica dell’italiano: indagine longitudinale sulla morfosintassi italiana. Bologna: Il Mulino. Bardel, C. 2000. La negazione nell’italiano degli svedesi: sequenze acquisizionali e influssi translinguistici. PhD Dissertation, Études romanes de Lund, 61, Lund. Lund University Press. Bernardini, P. 2001. Lo squilibrio nell’acquisizione di due lingue nell’infanza: indagine longitudinale sulla sintassi nominale. Licentiatavhandling, Perles (Petites Études Romanes de Lund), Extra Seriem, Commentaires et Communications, 10: Lund University. Bernardini, P. and Schlyter, S. 2002. “Growing syntactic structure and code-mixing in the weaker language”. Paper presented at IASCL/SRCLD 2002: July 16–21, Madison, Wisconsin. Bernstein, J. 1991. DP’s in French and Walloon: Evidence for parametric variation in nominal head movement. Probus 3 (2): 101–126. Bernstein, J. 1993. Topics in the syntax of nominal structure across Romance. PhD Dissertation, The City University of New York. Bohnacker, U. 1997. “Determiner phrases and the debate on functional categories in early child language”. Language Acquisition 6: 49–90. Bottari, P., Cipriani, P., Chilosi, M. 1993/1994. “Morphosyntactic devices in the acquisition of Italian free morphology”. Language Acquisition 3: 327–369. Cardinaletti, A. 1994. “On the internal structure of pronominal DPs”. The Linguistic Review 11: 195–219. Cardinaletti, A. 1998. “On the Deficient/Strong Opposition in Possessive Systems”. In Possessors, Predicates and Movements in the DP. [Linguistik Aktuell 22], A. Alexiadou and C. Wilder (eds), 17–54. Amsterdam/Philadephia: John Benjamins.

Child and adult acquisition of word order in the Italian DP

Cardinaletti, A. and Starke, M. 1999. “The typology of structural deficiency: A Case study of the three classes of pronouns”. In Clitics in the Languages of Europe, van Riemsdijk (ed), 145–233. Berlin: Mouton de Gruyter. Chini, M. 1997. Genere Grammaticale e Acquisizione: Aspetti della morfologia nominale in Italiano L2. Milan: Francoangeli. Cinque, G. 1995. Italian syntax and Universal Grammar. Cambridge: Cambridge University Press. Cipriani, P, Chilosi, A.-M., Bottari, P. and Pfanner, L. 1993. L’acquisizione della morfosintassi: fasi e processi. Padova: Unipress. De Houwer, A. 1990. The Acquisition of Two Languages from Birth: A Case Study. Cambridge: Cambridge University Press. Delsing, L.-O. 1993. The Internal Structure of Noun Phrases in the Scandinavian Languages. A Comparative Study. PhD thesis, Lund, Department of Scandinavian Languages, Lund University. Dimitrova-Vulchanovica, M, and Giusti, G. 1998. “Fragments of Balkan Nominal Structure”. In Possessors, Predicates and Movements in the DP. [Linguistik Aktuell 22], A. Alexiadou and C. Wilder (eds), 333–360. Amsterdam/Philadephia: John Benjamins. Döpke, S. 1998. “Competing language structures: The acquisition of verb placement by bilingual German-English children”. Journal of Child Language 25: 555–584. Döpke, S. (ed). 2000. Cross-linguistic Structures in Simultaneous Bilingualism. Studies in Bilingualism 21. Amsterdam/Philadephia, John Benjamins. Eubank, L. 1996. “Negation in early German-English interlanguage: More valueless features in the L2 initial state”. Second Language Research 12: 73–106. Gawlitzek, I. and Tracy, R. 1996. “Bilingual boostrapping”. Linguistics 34: 901–926. Genesee, F., Nicoladis, E. and Paradis, J. 1995. “Language differentiation in early bilingual development”. Journal of Child Language 22: 611–631. Giorgi, A. and Longobardi, G. 1991. The syntax of noun phrases. Cambridge: Cambridge University Press. Giusti, G. 1993. La sintassi dei determinanti. Padova: Unipress. Giusti, G. 1994. “Enclitic articles and double definiteness: a comparative analysis of nominal structure in Romance and Germanic”. The Linguistic Review 11: 241–255. Giusti, G. 1999. The functional structure of noun phrases: A bare phrase structure approach. University of Venice Working Papers in Linguistics 9 (1–2). Granfeldt, J. 2000a. Le développement morphosyntaxic du syntagme nominal chez des enfants et des adultes: approche génerativiste. Litentiatavhandling, Perles (Petites Études Romanes de Lund), Extra Seriem, Commentaires & Communications. Lund, Romanska Institutionen, Lund University. Granfeldt, J. 2000b. “The acquisition of the determiner phrase in bilingual and second language French”. Bilingualism: Language and Cognition 3(3): 263–280. Huss, L. M. 1991. Simultan tvåspråkighet i svensk-finsk kontext. Uppsala, Acta Universitatis Upsaliensis, Studia Uralica Upsaliensia 21. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press. Kupisch, T., Müller, N. and Cantone, K. F. 2002. “Gender in monolingual and bilingual first language acquisition: Comparing Italian and French”. Lingue e Linguaggio 1: 107–149.

79

80

Petra Bernardini

Lamarche, J. 1991. “Nombre grammatical et placement de l’adjectif en français et en anglais”, Revue québécoise de linguistique théorique et appliquée 10 (2): 19–42. Laenzlinger, C. 2000. “French adjective ordering: Perspectives on DP-internal movement types”. In GG@G (Generative Grammar in Geneva) 1: 55–104. Longobardi, G. 1994. “Proper nouns and reference: A theory of N-movement in Syntax and Logical Form”. Linguistic Inquiry 25(4): 609–665. Longobardi, G. 1996. “The syntax of N-raising: A minimalist theory”. OTS working papers: 1–55. Longobardi, G. 2001. “The structure of DPs: Some principles, parameters, and problems”. In The Handbook of Contemporary Syntactic Theory. Blackwell Handbooks in Linguistics, M. R. Baltin and C. Collins (eds), 562–603. Oxford: Blackwell. MacWhinney, B. 2000. The CHILDES project: tools for analyzing talk. 3rd edition, Mahwah, NJ: Lawrence Erlbaum. MacWhinney, B. and Snow, K. 1985. “The CHILDES project”. Journal of child language 12: 413–44. Meisel, J. M. 1990. (ed.) Two First Languages — Early grammatical development in bilingual children. Dordrecht: Foris Publications. Meisel, J. 1994. Bilingual First Language Acquisition: French and German grammatical development. Amsterdam: John Benjamins. Müller, N. 1998. “Transfer in bilingual first language acquisition”. Bilingualism: Language and Cognition 1(3): 151–171. Müller, N. and Hulk, A. 2001. “Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages.” Bilingualism: Language and Cognition. 4(1): 1–21. Nespor, M. 1988. “Il Sintagma Aggettivale”. In Grande grammatica di consultazione, Vol 1: I sintagmi nominale e preposizionale, L. Renzi (ed), 425–442. Bologna: Il Mulino. Nicoladis, E. and Secco, G. 1998. “The role of translation equivalents in bilingual family’s code-switching”. Proceedings of the 22nd annual BUCLD. 576–585. Paradis, J. and Crago, M. 2002. “Comparing DP in L2 and SLI child French”. Paper presented at IASCL/SRCLD 2002, 16–21 July, Madison, Wisconsin. Paradis, J. and F. Genesee. 1997. “On continuity and the emergence of functional categories in bilingual first-language acquisition”. Language acquisition 6: 91–124. Parodi, T., Schwartz, B. D. and Clahsen, H. 1997. “On the L2 acquisition of the morphosyntax of German nominals”. Essex research reports in linguistics 15, University of Essex. Petersen, J. 1988. “Word-internal code-switching constraints in a bilingual child’s grammar”. Linguistics 26: 479–493. Pizzuto, E. and Caselli, M. 1992. “The acquisition of Italian morphology: implications for models of language development”. Journal of Child Language 19: 491–557. Platzack, C. 2001 “The Vulnerable C-domain”. Brain and Language 77: 364–377. Serratrice, L. 1999. The emergence of functional categories in bilingual first language acquisition. PhD dissertation, University of Edinburgh. Schlyter, S. 1993. “The weaker language in bilingual Swedish-French children”. In Progression and regression in Language, K. Hyltenstam and Å. Viberg (eds), Cambridge: Cambridge University Press.



Child and adult acquisition of word order in the Italian DP

Schlyter, S. 1994. “Early morphology in Swedish as the weaker language in French-Swedish bilingual children”. Scandinavian Working Papers in Bilingualism 9: 67–86, University of Stockholm. Schlyter, S. and Håkansson, G. 1994. “Word Order in Swedish as the First Language, Second Language and Weaker Language”. In Scandinavian Working papers on Bilingualism, K. Hyltenstam (ed), 49–67. Stockholm University, Center for Research on Bilingualism, Stockholm. Schlyter, S. 1995. “Morphoprosodic schemata in bilingual acquisition”. In The development of morphological systematicity in a crosslinguistic perspective, K. Maroldt and H. Pirshwa (eds), 67–86, Tu˝bingen, Narr. Schlyter, S. 1999. “Evolution des types de mélanges entre deux et quatre ans chez quelques enfants bilbingues”. In Alternances des langues et constructions de savoir, cahiers du français contemporain 5, V. Castellotti and D. Moore (eds), Fontenay Saint-Cloud: ENS éditions. Schoorlemmer, M. 1998. “Possessors, articles and definiteness”. In Possessors, Predicates and Movements in the DP. [Linguistic Aktuell 22], A. Alexiadou and C. Wilder (eds), 55–87, Amsterdam/Philadephia: John Benjamins. Schwartz, B. D. and Sprouse, R. 1996. “L2 cognitive states and the full transfer/full access model”. Second Language Research 12: 40–72. Stowell, T. 1991. “Determiners in NP and DP”. In Views on Phrase Structure, K. Leffel and Bouchard (eds), 37–56. Dordrecht: Kluwer Academic. Tracy, R. and Gawlitzek-Maiwald, I. 2002. “The strength of the weak: Asynchronies in the simultaneous acquisition of German and English”. Paper presented at IASCL/SRCLD 2002, July 16–21, Madison, Wisconsin. Vainikka, A. and Young-Sholten, M. 1996. “Gradual development of L2 phrase structure”. Second Language Research 12(1): 7–39. Valois, D. 1991. The internal syntax of DP. PhD dissertation, UCLA. Wetter, M. 1996. Tra una e due lingue materne: L’acquisizione del tedesco in bambini italofoni, germanofoni e bilingui. Bulzoni Editore: Rome. Volterra, V. and Taeschner, T. 1978. “The acquisition and development of language by bilingual children”. Journal of Child Language 5: 311–326. Yip V., Matthews S. J. 2000. “Syntactic transfer in a Cantonese-English bilingual child”. Bilingualism: Language and Cognition 3(3): 193–208. Zamparelli, R. 1995. Layers in the Determiner Phrase. PhD dissertation, Rochester, New York, Department of Linguistics, The College Arts and Sciences, University of Rochester.

81



Null Subjects and optional infinitives in Basque* Maria-José Ezeizabarrena University of the Basque Country

Child root infinitives are considered to be underspecified structures with frequent null subjects. This paper deals with the study of the overt and nonovert subjects of the structures of root infinitives found in early speech production in Basque language, in order to determine whether there is a direct relationship between the different properties of the subjects and the presence of verb inflection in child speech. No underspecification has been observed during the root infinitive stage in this language where root infinitives are frequent also in the adult language. Finally, child Basque root infinitives have been interpreted as null auxiliary (auxiliary or modal) finite verbs and the target-deviant infinitivals have been explained as consequence of the less restrictive nature of the mechanisms that license root infinitives in the target language.

1.

Introduction

Several researchers who have studied the acquisition of Basque have observed that, at the initial stage, children’s language production does not manifest any evidence for the presence of functional categories in their grammar (see Meisel (ed.) 1994). At the beginning, neither verb inflection, nor morphological case or determiners are found during the initial recordings of longitudinal corpora. In this sense, early Basque is not very different from early English: nouns, demonstratives, adverbs and uninflected verbs with different tense and person references constitute the initial material for speech production in both languages. At a given moment, shortly after the two-word-stage, some morphology begins to appear: nouns bear case marking and verbs start to show inflectional marking. The well known hypothesis of the initial absence of functional categories of Radford (1986) was proposed as an explanation for these features of early

84

Maria-José Ezeizabarrena

production, and adopted in the study of early Basque (Barreña 1994, 1995, Ezeizabarrena 1994; Elosegi 1998; Larrañaga 1994, 2000). The gradual acquisition of functional categories (Guilfoyle and Noonan 1992) was the hypothesis we defended in a previous work (Ezeizabarrena 1996) in order to explain the gradual appearance of single and double agreement marking in verb inflection. In that work we interpreted the scarcity of different agreement markers during the earliest period, together with the frequent root non-finite forms, as a possible consequence of the lack of functional categories, following the approach of the split Infl hypothesis (Pollock 1989, Chomsky 1989). In the generative framework, where parameter setting is related to functional categories, an assumption such as the initial absence of functional categories predicts the impossibility of setting the pro-drop parameter, at least before the acquisition/activation of the corresponding functional category I or T. This assumption is not in contradiction with other facts observed in earlier stages of acquiring non pro-drop languages, such as English and Dutch, where non-finite verb inflection correlates in many cases with the absence of overt subjects (Hyams 1996). This author states that the null subject (NS) phenomenon and the early optionality of root infinitives are both derived from the underspecification of functional categories such as I (root infinitives, null Case and deictic interpretation) and Det (determinerless DPs, no scrambling and deictic familiar interpretation). Moreover, the significantly higher number of NS with infinitives than with finite verbs observed in three non pro-drop languages (French, Dutch and German) demonstrates that it is the Case Filter which blocks the presence of subjects in the absence of I (Krämer 1993). In her approach, the Case Filter is applied in both adult and child language. A similar parallelism between the development of DP and IP has also been found in pro-drop languages like Spanish and Catalan (Grinstead 1996, 1998). The existence of an early stage in child Spanish and Catalan, where only NS have been found, has been explained by this author in terms of the Case Filter. In Grinstead’s approach it is the lack of tense and number inflection which prevents nominative Case from being checked, and therefore, he concludes that early null subjects in the two pro-drop languages are Case unmarked PRO. Such a proposal does not appear to be confirmed very robustly by the data, as suggested by Aguado and Pine (2002) and Bel (2001b). These authors find quite a stable rate of null subjects in the first two years but no initial stage without overt subjects in the acquisition of either language. Bel concludes that the prodrop parameter is set to the correct value in Spanish from very early on. Even if the language produced by the child appears to be initially under-

Null Subjects and optional infinitives in Basque

specified at some points, such as I (root infinitives) and Det (NS, determinerless NP), more general aspects of the grammar, such as the Case Filter and the prodrop parameter, are available to the child at this age. The observation of such facts has lead researchers to suggest that the difference between child and adult grammar is a result of differences between the pragmatic system of children and that of adults (Hyams 1996). 1.1 Optional infinitives cross-linguistically The existence of a root infinitive or optional infinitive stage where children produce non-finite sentences in place of finite verbs has been documented for different languages in the literature on language acquisition, as the examples in (1) indicate. (1) Root infinitives in different languages1 German a. Thorstn das hab-en [Poeppel and Wexler 1993 (Wexler 1994)] Thorstn that have-inf Spanish b. quema-r (Bel 2001a) burn-inf c. tapa-r el pie (Bel 2001a) cover the foot d. e(s)te coche (a)pa(r)ca-(r) (M 2;07) (Ezeizabarrena 1997) this car park-inf e. ese pone-(r) (J 3;11) (Ezeizabarrena 1997) this put-inf Catalan f. sorti-r (Bel 2001a) come out-inf g. un altre pinta-r (Bel 2001a) another paint-inf French h. pas manger la poupée [Pierce 1992 (Wexler 1994)] neg eat the doll i. pas tomber bébé [Pierce 1992 (Wexler 1994)] neg fall baby

85

86

Maria-José Ezeizabarrena

Basque j. Ho(r)i amatau that-A turn off k. hemen-dik pasa here-from go through

(M 2;01) (Ezeizabarrena 1997) (J 2;07) (Ezeizabarrena 1997)

Non-finite forms other than infinitives may also appear in main clauses (2). In this sense, English -ing forms without their auxiliary have been also considered as instances of optional infinitives (OI) in the literature (Wexler 1998). (2) Present participles (Gerunds) in Catalan and Spanish a. això recollint, mama (Bel 2001a) this pick-ger mummy b. du(r)miendo (Bel 2001a, Ezeizabarrena 1997) sleep-ger

The fact that such structures show different rates and properties cross-linguistically has been related to different properties of languages, such as their pro-drop nature (Wexler 1994, 1998; Sano 1998), the morphological marking of plurality, and the semantic properties of infinitivals (Hoekstra and Hyams 1998). As Wexler (1998) and Hoekstra and Hyams (1998) suggest, the root or optional infinitive stage is not a stage where finite verb forms are absent, or where the child does not know the relevant grammatical principles such as movement, checking, and so on. It is a stage where root infinitives are possible grammatical sentences for children in contexts where they are not so in the adult language. In order to avoid misunderstandings, I consider a terminological clarification necessary. The term Optional Infinitive (or Root Infinitive in Rizzi’s terms) has been proposed by Wexler (1994) in relation to a developmental stage, the OI-stage, and refers to the non-finite forms produced by children — frequently, but not systematically — in contexts where a finite form is required. In this paper, we distinguish between optional and root infinitive. Optional Infinitive (OI) will be used to refer to root infinitives used in contexts where they are in free distribution with finite forms, e.g. grammatical root infinitives, whilst we refer to root infinitives used target-deviantly by the children as *RI. 1.2 Optional infinitives in Basque Following the prediction that if pro-drop, then non-OI language (Wexler 1994, 1998), children acquiring Basque, a pro-drop language,2 are not expected to go

Null Subjects and optional infinitives in Basque

through an optional infinitive stage. However, data of monolingual and bilingual children have shown that such structures (OIs and *RIs) also exist in some pro-drop languages (Barreña 1995, Ezeizabarrena 1997, 2001; Liceras, Valenzuela and Díaz 1999). We assume that this frequency can be related to properties of particular languages, in the sense that some languages license OI in the adult language in contexts where others do not. Thus, OI appear to be more frequent in adult German and Dutch than in French (Krämer 1993), and more frequent in adult Basque than in Spanish (Ezeizabarrena 2002). The aim of this paper is to present data from some early production of finite and non-finite verb forms of a pro-drop language, Basque, in order to discuss the relationship between NS and the absence of verb inflection in this language. Data of one bilingual child, Mikel, will be analyzed throughout the different sections. In some sections his data will be compared to data of two more children, one monolingual and one bilingual. Different results will be presented in order to argue that the application/non-application of the Case Filter does not offer any satisfactory explanation for the high rate of NS with non-finite verbs. In this sense, our data seem to confirm the early application of the Case Filter as proposed by Krämer (1993), even if Basque data show considerable differences with respect to the languages she analyzed: French, German and Dutch. Moreover, a second proposal made by this researcher, the Bare VP Hypothesis complemented by the Null Modal Hypothesis also appears to be compatible with the Basque data. The present work is organized as follows: in Section 2 the target language is described. Section 3 deals with the child database. After the presentation of the corpus in 3.1, the development of root infinitives will be presented in 3.2. Section 3.3 includes the study of overt and null subjects with finite and nonfinite verbs, followed by 3.4 where rates of root infinitives are put together with rates of inflected auxiliaries in order to show the inverse development of both kinds of verbs. Conclusions are summarized in Section 4.

2. Target system: Basque 2.1 Null arguments Basque is a language with a very rich agreement system, a high variety of syntactic orders,3 and frequent null arguments. The verbal inflection shows agreement with arguments bearing ergative, absolutive, and dative Case, so that

87

88

Maria-José Ezeizabarrena

finite verbs agree with their one — absolutive —, two — ergative and absolutive, or absolutive and dative — or three — ergative, absolutive and dative — arguments (3a) obligatorily.4 The fact that these three arguments can be overt or null (in parenthesis in (3)) has been related to rich morphology, in the sense of the pro-drop parameter formulated by Rizzi (1986). It is noteworthy that most verbs only show analytic structure in this language where only about twenty verbs can be inflected. Thus, agreement marking is generally realized on the auxiliary. There are two auxiliaries, the transitive one for accusative and unergative verbs, and the intransitive one for unaccusative verbs. (3) a.

(zu-k) (lagun-a-k) (ni-ri) aurkeztu you-e friend-det-apl I-d introduce-pf Ø-d-i-zki-da-zu a3-pres-root-pl-d1s-e2s ‘you have introduced (some) friends to me’ b. (zu-re lagun-a-Ø) etorr-i Ø-d-a you-g friend-det-a come-pf a3s-pres-root ‘your friend has come’ c. (zu-re etxe-a-Ø) (ni-ri) gusta-tzen Ø-z-a-i-t you-g house-det-a I-d like-imp a3s-pres-root-dative-d1s ‘I like your house’

2.2 Rich and uniform agreement system The Basque agreement system is also uniform in the sense of Jaeggli and Safir (1989) where the agreement markers are different for the three persons, and where (at least)5 six different inflected forms, three for singular and three for plural are found in the whole paradigm.6 Moreover, Basque verb inflection shows a partial relation of homophony of Agr to its corresponding personal pronouns (4). Notice that there is no third person pronoun in Basque, so that it is the demonstrative which operates in its place. (4) Pronouns 1s ni 2s zu 3s (hura) 1pl gu 2pl zuek 3pl (haiek)

Absolutive agr. Verb forms n-aiz ‘I am’ z-ara ‘you are’ Ø-da ‘(that) is’ g-ara ‘we are’ z-are-te ‘you are’ Ø-dira ‘(these) are’

Ergative agr. Verb forms du-t ‘I have (it)’ du-zu ‘you have (it)’ du-Ø ‘(that) has (it)’ du-gu ‘we have (it)’ du-zue ‘you have (it)’ du-te ‘(these) have (it)’

Dative agr. Verb forms di-zu-t ‘I have it to you di-da-zu ‘You have it to me’ di-o-t ‘I have it to him’ di-gu-zu ‘you have it to us’ di-zue-te ‘they have it to you’ di-e-te ‘they have it to them’

Null Subjects and optional infinitives in Basque

As seen in (4) Basque verb morphology shows quite consistent homophony with first and second person pronouns and therefore Basque belongs to the group of languages with [+pronominal] Agr in the sense used by Kato (1999). As she predicted for NSL, Basque has no weak pronouns, which are supposed to be in complementary distribution with uniform agreement paradigms. In such languages, to which Spanish and Basque belong, strong pronouns are supposed to be in a higher position than SpecIP, namely in the SpecΣ.7 In Kato’s proposal, where pro is assumed not to exist, strong pronouns in NSL are cases of doubling of the subject agreement itself. Each verb has compound (participle + auxiliary) forms but only some very frequently used verbs have simple forms. Basque shows transitive edun ‘to have’ and intransitive izan ‘to be’ auxiliaries. Transitive (eraman ‘bring’) and unergative (lo egin ‘sleep’) verbs take the transitive auxiliary, and unaccusative (joan ‘go’) verbs take the intransitive one. Person agreement markers are located in the verb inflection in present tense as follows: A-root-D-E; participles bear the aspect morphemes. (5) Analytic forms in Basque: Participle Auxiliary Verb Root+Aspect a. bota-Ø Ø-du-t ‘I have thrown (it)’ throw-pf a3s-root-e1s b. bota-Ø Ø-di-zki-zu-t ‘I have thrown them to you’ throw-pf a3s-root-pl-d2s-e1s c. bota-TZEN Ø-du-t ‘I usually throw (it)’ throw-ipf a3s-root-e1s d. etorr-i n-aiz ‘I have come’ come-pf a1s-root

Although Agr appears to be the licenser of empty categories in Basque, there is empirical evidence that argues for the need for a different proposal in order to explain the similarities found between finite and infinitive sentences not found in other languages. For instance, Null referential arguments are found in nonfinite constructions such as nominalizations, where null arguments are in free distribution with overt arguments; this fact excludes the possibility of PRO in these contexts. Moreover, different tests such as word order, case marking, cross over effects, etc., have shown that finite and non-finite sentences behave alike in many respects, except in the presence of overt agreement markers (Zabala and Odriozola 1996, Elordieta 2001a,
b).

89

90

Maria-José Ezeizabarrena

2.3 Some remarks on adult optional infinitives As suggested in the introduction (1.2), optional infinitives are grammatical root infinitives. Some examples are listed in (6). There are contexts where the auxiliary can be omitted in Basque. Omission takes place in contexts such as imperatives very frequently (6a), as well as in some narrative contexts (6d). Elliptical responses to modal or egin ‘do’ containing questions are frequent as well (6b,
c). Examples in (6) can be considered bare VPs with null auxiliary, or null modal, similar to the grammatical Dutch infinitives quoted by Krämer 1993.8 (6) a.

etorri (zaitez)! come (a2pl-root ‘come!’ b. zer egin duzu? ‘what have you done?’ -berandu etorri late come ‘to come late/I have come late’ c. zer nahi duzu? ‘what do you want? -asko jakin a lot know ‘to know a lot’ d. atzo etxera iritsi (nintzen), eta inor ez yesterday home-to arrive (a1s-root-past) and nobody ‘Yesterday, I arrived home and nobody was there’

Basque optional infinitives9 do not appear to have the semantic restrictions observed in Dutch, where only eventive verbs can appear as grammatical infinitives (Wijnen 1998). Eventive (6a,
b,
d) and non-eventive (6c), are both possible OI in adult Basque.

3. Optional infinitives in early Basque As we have seen above, OI can appear in adult language, which means that root infinitives are also grammatical in the target language. Therefore, in what follows we will distinguish between grammatical root infinitives (OIs) and deviantly used root infinitives (*RIs). Some examples of *RIs in different languages have been listed in (1) and other non-finite forms used target-

Null Subjects and optional infinitives in Basque

deviantly have been also reported in child speech (2). For the present work we only consider the infinitive/past participle forms (6), ignoring other non-finite forms such as the future/irrealis -ko participle (etorri-ko ‘will come’) or the imperfect -tzen (etortzen ‘usually come’) (5c). 3.1 The corpus The data we will analyze in most detail in this paper are taken from one bilingual child, Mikel, acquiring Basque and Spanish simultaneously at home. His father, a monolingual Spanish speaker speaks Spanish and his mother, who is bilingual, speaks only Basque to her children. In some sections, Mikel’s data are compared to data of another bilingual child Jurgi, and a monolingual Basque speaking child Oitz. The three children were videotaped twice a month over a twoyear period, in spontaneous speech with their parents or their caregivers in 30’ sessions for each language. All the material was collected in the HEGEHJ-BUSDE long-term project developed jointly between the University of the Basque Country in Vitoria-Gasteiz and the University of Hamburg in Germany. 3.2 OIs and *RIs A considerable number of root infinitives have been found in the corpus of bilingual children (Ezeizabarrena 1997, 2001, 2002) and monolingual children (Ezeizabarrena and Barreña 2002). Following Wexler (1994, 1998) *RIs are not expected in early Basque, a pro-drop language. However, the presence of root infinitives in adult language extends the possibility of their appearance to early speech too. It is interesting at this point to determine whether child root infinitives correspond to adult use. Although their rate decreases gradually, the rate of target-deviantly used root infinitives (*RI) does not disappear at age 4. The question arises whether they are performance errors, or due to some production strategy based on the omission of the auxiliary. In earlier work we found that even if the null modal hypothesis were the explanation for some of the scarce *RIs in Spanish, performance could account for the rest. However, the considerably high number of occurrences in Basque and their gradual decrease calls for a more adequate explanation (Ezeizabarrena 2001, 2002). In what follows we will show that it is the hypothesis of the production strategy that holds true. Root infinitives do exist in the target language in several contexts where the auxiliary is omitted, as has been mentioned before.10 Therefore, it is not

91

92

Maria-José Ezeizabarrena

surprising that grammatical infinitive forms also appear in the child language, in imperative sentences (7a) or in elliptical responses (7b). (7) a.

hartu Ana! pick up Ana ‘pick (it) up, Ana!’ b. zer nahi duzu? ‘what do you want (to do)?’ -lotu ‘tie (it)’

(M 1;09)

(M 1;11)

However, child root infinitives are not all restricted to the contexts mentioned in (6) and (7). They have been produced sometimes to refer to ongoing, past or future events, without any previous discursive context that would allow their use in the adult language. (8) a.

bota(ko dut) hemen throw here ‘(I will) throw here’ b. bota (dut) throw ‘(I have) throw(n)’ c. nor-k ipini (du)? who-E put ‘Who (did) put (it)?’

(M 2;01)

(J 2;01)

(O 2;04)

A first attempt to investigate this phenomenon of the optional finiteness in Basque was carried out in a study of the development of infinitive rates in relation to the acquisition of agreement and tense (Ezeizabarrena 1997). In addition to the data of the two bilingual children studied in that work, we will present some data of the monolingual Oitz studied by Barreña (1995). Both bilingual children, Mikel and Jurgi, as well as the monolingual Oitz, go through 4 different stages in the development of verb finiteness (Table 1). Such stages have been delimited based on the criterion of age of productive marking of each feature -Agr-S, Agr-O and T- in the verbal inflection. This means that each stage starts at the moment where one of the two possibilities takes place: a) production of the same inflected verb with at least two different, and in both cases, context-adequate morphemes, or b) production of the same (non-zero) affix attached to two different verbs (Meisel 1994, Ezeizabarrena 1996). Following such productivity criterion, Stage I corresponds to the period where

Null Subjects and optional infinitives in Basque

productive marking cannot yet be confirmed in the data; Stage II includes the time after the beginning of contrastive agreement marking in the verb, where only one overt marking can be reported (Subject agreement). Stage III corresponds to the period where inflected verbs show overt marking with more than one argument. Stage IV corresponds to the last period of the study, after the appearance of the past forms. Table 1.Age of production of productive verb inflection in Basque11

Stage I II III IV

First Verbs agreement with 1 arg. agreement with 2 arg. T (Past)

Bilingual

Bilingual

Monolingual

Mikel

Jurgi

Oitz

1;07–1;11 2;00–2;03 2;04–2;09 2;10- 4;00

1;11–2;07 2;08–3;00 3;01–3;03 3;04–4;01

1;06–2;02 2;03 2;04–2;06 2;07–3;00

Previous work on the study of the acquisition of Basque and Spanish by bilingual children has shown the following results for both languages: a. the acquisition of subject agreement is not followed by the disappearance of target-deviant infinitives. Neither does the acquisition of agreement with the second argument have consequences for their disappearance, although their frequency of use decreases in both stages II and III. b. target-deviant infinitives decrease during the acquisition process, but so do target-like non-finite verbs. Table 2 indicates a quite parallel development of Inf (target-like and targetdeviant infinitives) and *RIs in the three corpora, though in general, root infinitives are more frequently produced by the monolingual child than by the bilingual children Mikel and Jurgi, both with the total Inf and with the targetdeviant *RIs. However, rates of infinitives decrease gradually in the three corpora from the first to the fourth stage (M 48.5% to 16.2%; J 60.2% to 7.7% and O 81.6% to 12.8%). Target-deviantly used infinitives also show a decreasing tendency from stage I to stage IV (M 31.6% to 3.2%, J 20.5% to 1.9% and O 38.6% to 0.5%). Such a continuous decrease has not been found in the Spanish corpus of M and J, where their number is much lower, and where non-finite forms are generally very much restricted in the target-language (Ezeizabarrena 2001, 2002). At stage III *RI-rates are lower than 15% in every case, and after acquiring T their average does not reach 4%. Table 2 shows some differences

93

94

Maria-José Ezeizabarrena

Table 2.Target-like and target-deviantly used root infinitives in Basque (1;07–4;01) Mikel (biling.)

Jurgi (biling.)

Oitz (monol.)

Stage

Inf /V

*RI/Verbs

Inf/V

*RI/Verbs

Inf/V

*RI/Verbs

I II III IV

48.5% 33.7% 14.3% 16.2%

31.6% 15.2% 2.6% 3.2%

60.2% 37.4% 11.4% 7.7%

20.5% 10% 3.1% 1.9%

81.6% 54.8% 41.7% 12.8%

38.6% 22.5% 13.1% 0.5%

Total

20.1% 6.4% (464/2309 (148/2309 tokens) tokens)

15.7% 4.4% (199/1267 (56/1267 tokens) tokens)

34.6% 11.8% (560/1614 (191/1614 tokens) tokens)

between the bilingual children on the one side and the monolingual on the other: *RIs undergo a significant decrease at stage II, and at III their average remains below 4% in Mikel’s and Jurgi’s corpora. In the case of the monolingual child, the decrease takes place more gradually during stages I to III, while the minimum average of *RIs is delayed to stage IV. Even if the acquisition of T which marks the beginning of stage IV is followed by some decrease of *RIs in Jurgi and Oitz, the data in Table 2 leads to the conclusion that T is not responsible for the development of *RIs, which have decreased considerably from stage I to III. 3.3 Root infinitives and NS Null subjects are much more frequent with nonfinite than with finite verbs in the acquisition of non pro-drop languages (Krämer 1993, Hyams 1996, Rasetti 2000). Therefore NS with infinitive verbs have been generally considered as PRO, and NS with finite verbs as pro. In this section we will analyze data of Mikel in more detail than in the previous sections, in order to study his null and overt subjects in more depth. For Figure 1 and Figure 2 Mikel’s data have been revised up to age 3;06 in order to account more accurately for the subjects produced with finite and non-finite verbs. Some slight differences have been included in this new inventory, with respect to previous work (Ezeizabarrena 1997, 2002) referred to in Section 3.1, 3.2 and 3.4; among others, dative marked subjects of double marked unaccusative verbs such as psych-verbs have now been counted as subjects, instead of the absolutive arguments of these verbs in earlier corpora; moreover, coordinated structures with the same subject have been counted as one finite sentence with overt subject, some dubious examples have been reconsidered, etc.

Null Subjects and optional infinitives in Basque

%

100 90 80 70 60 50 40 30 20 10 0

1;08 1;09 1;10 1;11 2;00 2;01 2;02 2;03 2;04 2;05 2;06 2;07 2;08 2;09 2;10 2;11 3;00 3;01 3;02 3;03 3;04 3;05 3;06

NS/Vfin

NS/*RI

%NS/Inf (imperatives excluded)

Figure 1.NS in Basque. Mikel.

As Figure 1 shows, null subjects are very frequent in the corpus of Mikel with both finite and non-finite verbs, as well as with *RIs. However, with the exception of the first recording at age 1;08, where no subject has been realized with any of the 6 verbal forms, subjects appear in all recordings where verbs are reported. Notice that there are no *RIs in the recordings of ages 2;05, 2;06, 2;08, 2;09, and 3;03, hence the blanks in Figure 1 at these points. Mikel’s data show that the rate of overt subject production remains quite constant from the beginning, even before age 2;00, namely the point at which we considered I to be undoubtedly active (stage II). NS with finite verbs show an average of 65.4% (range 40–84%); NS with infinitives are more frequent as their average of 90.4% indicates (range 50–100%) and NS with *RIs are in between with an average of 83.7% (range 33–100%). Moreover, if we group the data of Figure 1 in the four stages of Table 1, we can see more clearly that the frequency of explicit subjects with finite verbs does not vary very much in the whole corpus: it remains quite constant from age 2 to 4 with an average of 34.6% (range 24.3–39.3%), in Mikel, even at the earliest stages (I and II), where object agreement markers are not yet productive. Figure 2 includes rates of overt production of subjects with the three kinds of verbs: finite, infinitive, and *RIs during the four stages. Rates of overt production of subjects tend to increase throughout the whole period of study in the case of finite verbs. On the contrary, the production of NS with Inf shows a lower and a more irregular development. Though rates of *RIs tend to increase during the studied period till 3;06, the total number of root infinitives seems to follow a different course. NS are more frequent with *RIs than with grammatical infinitives, though less than with finite verbs. By stage IV the course of finite verbs and of *RIs show a parallel trajectory. Data from Figure 1 and Figure 2 indicate that child root non-finite verbs behave very similarly to finite forms, and that *RIs, with their 83.7% NS

95

Maria-José Ezeizabarrena

%

96

100 90 80 70 60 50 40 30 20 10 0

Stage I % S/Vf

Stage II % S/*RI

Stage III

Stage IV till 3,06

%S/Inf (imperatives excluded)

Figure 2.Overt subjects in Basque. Mikel (1;08–3;06).

average, appear to be closer to finite forms (65.3% NS) than is the whole group of infinitives (88.1%). If our interpretation is correct, the null modal hypothesis proposed for languages such as German and Dutch appears to be compatible with our data. However, Basque *RIs can have not only modal interpretation, but also reference to present or past events, as is the case in the adult OI mentioned in Section 2.3. If they are finite structures with omitted inflection, the null element must not be restricted to modals as it is not in the target language. For the same reason, neither should the eventivity constraint work in child *RIs, as data have confirmed (Ezeizabarrena 2002). Our results do not correspond to what was found in other pro-drop languages such as Spanish and Catalan by Grinstead (1996, 1998). Grinstead found an initial stage where subjects are absent, followed by a second stage after the acquisition of the specification of I, where the rate of overt subjects oscillates around 20–30%. Paradis (2001) comes to a similar conclusion. All the data mentioned in this Section 3.3 are compatible with the idea that case assignment works in child language from the beginning and with the idea that *RIs are finite verbs with null inflection. Following Krämer’s null modal hypothesis as explanation for root infinitives in languages like Dutch and German, we think that such forms correspond to a more general phenomenon, in which null modal structures are one of the several possible manifestations. More generally, we conclude that root infinitives are null auxiliary (modal, auxiliary) structures not only in Basque but also in other languages, such as French, German, Dutch, and Spanish. See Boser et al. (1992).12 The underspecification hypothesis suggested by Hyams (1996) and Hoekstra and Hyams (1998) for other languages, such as English, is the next point for consideration in this section. In order to test whether subjects of *RIs are underspecified in Basque, revised data of Mikel have been put in Table 3. In this table, overt subjects have been classified into pronouns, demonstratives,

Null Subjects and optional infinitives in Basque

interrogatives, and two kinds of DPs. In Basque the article is not a word but an affix (-a, in standard Basque), added to the last word of the noun phrase (3). No noun or NP has been found without this determiner. However such a suffix has a rather peculiar nature, in the sense that it is not a marker for specificity, as are articles in other languages. Therefore, we have separated one word DPs from other DPs in order to see more clearly whether the frequency of evident DPs varies through the different kinds of clauses. Table 3 shows that no difference exists with regard to subject specificity between finite, infinitive, and *RIs in Mikel’s data. Pronouns and demonstratives together constitute about a half of the total number of subjects in every case (59.1% with finite, 48% with infinitive and 42.7% with *RIs). Table 3.Subjects with finite, infinitive and *RI verbs in Basque. Mikel (1;07–3;06). Total S Pronoun Demonst Interrog

One word DP

More complex DP

Finite V

434

97 (22.3%)

160 (36.8%)

14 (3.2%)

133 (30.6%)

30 (6.9%)

Infinitive *RI

21

6 (28.5%)

3 (14.2%)

1 (4.7%)

7 (33.3%)

4 (19.0%)

Total Inf

25

8 (32%)

4 (16%)

1 (4%)

8 (32%)

4 (16%)

It is noteworthy, however, that personal pronouns start to appear in sentences containing a verb after the acquisition of AgrS at 2;0, and most of them (26 in the whole corpus) correspond to imperative sentences like (9c). Such pronouns have not been included in Table 3. First pronouns with finite verbs appear at 2;1 in Mikel’s corpus. The existence of pronouns has also been reported earlier, at 1;9 in verbless sentences (Larrañaga 1994). (9) a.

ama ezto(t) guro ni(k) mum neg-aux-e1s want I-e ‘Mum, I don’t want’ b. ni tu(s)tau in n-a(z) I-a frighten do a1s-be ‘I’ve frightened’ c. zu-(k) ipini! you-e put put it on!

(M 2;01)

(M 2;01)

(M 2;00)

97

98

Maria-José Ezeizabarrena

Moreover, rates of one word DPs are very similar in the three groups (about 30%) in contrast to infinitives, which show a lower rate (14%). With regard to complex DPs, they are more frequent with *RIs (19%) and infinitives (12.9%) than with finite verbs. This distribution of subjects with both kinds of verbs shows that underspecification does not account for *RIs in Basque, and neither does it account for the rest of root non-finite verbs in this language. 3.4 Basque *RI verbs as null auxiliary structures Throughout this paper, we assume that omission can be a very useful strategy for language acquisition, a characteristic that has been observed in different linguistic areas. In phonology, deletion appears as one of the processes (deletion, reduplication, epenthesis) available to the child in order to adapt language production to articulatory and cognitive limitations.13 In verb morphology, omission seems to be much more frequent than are errors. Mistakes in person marking are very rare in languages with rich morphology, as has been widely mentioned in the literature (Guasti 1993/1994, Serrat and Aparici 2001). In syntax, deletion takes place much more frequently than other processes do, for instance reduplication, as in the case of overtly realized traces of moved wh-elements.14 As mentioned in Section 2, auxiliaries and modals can be omitted in adult Basque. Krämer (1993) and Bel (2001a) have also shown that root non-finite sentences are grammatical in several Germanic and Romance languages. Hence, it is not surprising that such structures appear in child language, at least in languages that license root non-finite verbs.15 Due to the fact that every language constrains optional infinitives in a particular way, different OI rates are expected cross-linguistically in adult and child language. For instance, Dutch and German infinitives seem to be eventive verbs while this is not necessarily the case of English bare forms (Wijnen 1998, Hoekstra and Hyams 1998). As Hoekstra and Hyams (1989) suggest, children appear to be very sensitive to the morphosyntactic distinctions in the target systems, though their grammar has less restrictive licensing conditions. At this point, it is interesting to note that bilingual children have shown a rate of root infinitive usage which corresponds to that of monolingual use in each of the respective languages. This fact has been taken to indicate that bilingual children are very sensitive to the optionality of finite/non-finite verb forms, and that they do not transfer the rules of finite/non-finite optionality from one language to the other (Ezeizabarrena 1997, 2002, Hulk and Müller 2000).

Null Subjects and optional infinitives in Basque

In our opinion, it is the higher rate of appearance of infinitives in root sentences, rather than the category itself, that really distinguishes child language from adult language. If root infinitives are auxiliary-omitted finite structures, one can assume that this is also true of *RIs, as both types of verbs show a similar development. In what follows we present a further set of data in this discussion on Basque root infinitives: the increase of transitive finite verbs parallel to the decrease of root infinitive verbs. In this respect, Mikel behaves similarly to the other bilingual child, Jurgi (Table 4). Initially, at Stage I, the period where infinitive forms (grammatical OI and *RI together) make up about 50% of the verbs, the presence of transitive inflected verbs is very low (7.2% in M and 10.3% in J), in contrast with Stage IV where the rate of infinitives is less than 17% and the rate of transitive verbs is nearly 50% of the finite verbs. Production rates of transitive finite verbs seem to stabilize at Stage III (around 35–40%), namely, after the productive marking of overt agreement with two arguments. Interest in the development of transitive inflected forms is a result of their ability to bear two or more overt agreement markers. In contrast, intransitive double marked auxiliaries are much less frequent, as they generally bear one agreement marking (Absolutive) and only sometimes the Dative marking. Table 4.Rates of transitive inflected verbs in Basque Mikel

Jurgi

Stage

Inf/V

Vfin Vtr/Vitr

% Vtr/Vfin

Inf /V

Vtr/Vitr

% Vtr/Vfin

I II III IV

48.5% 33.7% 14.3% 16.2%

5/64 45/153 146/215 600/617

7.2% 22.7% 40.4% 49.3%

60.2% 37.4% 11.4% 7.7%

3/26 23/88 78/148 281/387

10.3% 23.4% 34.5% 42.1%

Total

20.2%

796 /1049

43.1%

15.7%

385/649

36.1%

As shown in (5), information about person and number agreement (with one, two or three arguments), tense and mood is codified in the inflected verb form in Basque. All this complexity does not appear in child language production simultaneously (see Table 1). If our interpretation is correct, children start producing complex inflected forms very gradually: inflected forms agree only with the subject initially (absolutive and ergative), and later with one or two objects. At age 4, not every inflected form of the paradigm had been reported

99

100 Maria-José Ezeizabarrena

yet.16 During a period of about a year, where agreement is available in order to check D features, it appears to be the case that the concrete inflected verb form is not yet accessible to children at production level.17 The question whether target-deviant root infinitive forms are instances of performance errors or of some other mechanism arises again at this point in the argument. We suggest that such forms correspond to a developmental stage in Basque, where feature checking is available to the child, but their morphological realization is not. If our interpretation is correct, our data show that children overuse an option also available in the adult language such as the omission of auxiliary verbs during the time where complex verb forms are being processed. In the whole corpus, inflected forms are more frequent with intransitive than with transitive verbs (Table 4). This could correspond to some error-avoiding strategy, especially in the pre-AgrO stage (Stages I and II), that is at the stage where the verb agrees only with its absolutive and ergative marked subject arguments. Stage III marks a quantitative and qualitative change in the development of *RIs: their rate stabilizes and the production of transitive auxiliaries does so too. At Stage III agreement with two arguments starts to appear regularly, although this does not mean that double marked inflected forms are always produced in a target-like way: on the one hand, inflected forms are not produced in every required context, though animate objects seem to be more salient than inanimate for object agreement, in the sense that agreement with nonperson objects is sometimes lacking, contrary to what happens with animate objects. On the other hand, not every inflected form with double overt marking corresponds to a target-like verb use. Neologisms have been found only very seldom, but a quite systematically target-deviant selection of verb paradigms has been reported (Ezeizabarrena 1996, Fernández and Ezeizabarrena, in press). Error-avoiding strategies have been suggested in several works on first and second language acquisition.18 Meisel, Clahsen and Pienemann (1981) distinguish between different kinds of simplification observed in L2 learning: (a) elaborative simplification, which serves the purpose of bringing the learner’s interim grammar closer to the target norm; (b) restrictive simplification, which reduces the grammar to make it easier to handle, and (c) the conformative simplification in which the learner simplifies his grammar by conforming to the model system and eliminating any non-standard forms. Simplification of some kind appears to be compatible with Mikel’s and Jurgi’s *RIs, though more research is needed on this topic.

Null Subjects and optional infinitives in Basque

Moreover, a similar auxiliary-omission strategy has been also found in the speech of Basque agrammatics. Basque aphasics produce a high rate of root non-finite forms, as Laka and Erriondo (2000) and Laka and Ezeizabarrena (2001) have shown. See also Centeno (1996) for Spanish. However, aphasics seem to create new forms more frequently than children do. Some consequences for production and perception can be observed from the early production of root infinitive forms: a. children produce well-formed words (participles, infinitives, roots), but not incomplete words such as stems or neologisms, which shows that they distinguish between possible and impossible words in their target grammars. b. infinitives of auxiliary verbs as izan ‘be’ are not frequent, contrary to what is found in African-American English (Labov 1995). This indicates that the production of root non-finite forms is not a result of a “substitution” process, but of the omission of light elements (auxiliaries). c. children are (over)using an OI-rule that exists in their grammar, rather than applying a universal rule, or transferring rules from one language to another (bilinguals have 20% infinitives in Basque root sentences but only 3% in Spanish). d. speakers avoid overly complex verb forms in a stage when they are processing different kinds of features. They prefer to drop the auxiliary rather than to produce crashing structures, in the sense of Chomsky (1995).

4. Conclusion The study of root infinitives in early Basque has confirmed the presence of targetdeviantly used infinitival structures in root sentences. Basque, a language with very rich and uniform verb morphology appears to be a good candidate for the early setting of the NS parameter to its correct value (Jaeggly and Safir 1989, Kato 1999). This seems to be the case in our data, as the early presence of null subjects indicates. Their rate of occurrence, which remains quite constant throughout the study (around 30%), shows a very early adult-like setting of the pro-drop parameter. Overt subjects have shown a very constant behaviour during the study period regardless of the presence or absence of overt agreement features of the verb, contrary to what has been found in English and other languages (Hyams 1996, Hoekstra and Hyams 1998, Grinstead 1995). Although no non-finite structures other than infinitives have been studied in the present paper, data of early Basque are compatible with the hypothesis of the existence of similar

101



102 Maria-José Ezeizabarrena

licensing mechanisms for null subjects in finite clauses and *RIs, although their behaviour in other non-finite clauses remains unclear, even in the adult language (Elordieta 2001a,
b). In any case, subjects with *RIs in Basque are not underspecified structures. Moreover, the early presence of strong pronouns as doubling structures in a higher position than IP confirms the existence of such structures at age 2;0 or earlier. *RIs in Basque have been interpreted as null auxiliary structures, as suggested for other languages (Krämer 1993, Kato 1999). Contrary to Wexler’s, our data disconfirm the relation between *RIs and the pro-drop parameter (Wexler 1994, 1998). We conclude that child *RIs are the consequence of the overuse of licensing mechanisms of OI existing in the target language. Our prediction is that rates of *RI production in a particular language will be proportional to the rates of licensing of OI in this language. For instance, Child *RIs can be constrained by eventivity, if such a property constrains adult OI in this language, but not necessarily. Neither does morphological uniformity seem to be directly related to the absence of *RIs. Finally, omission is not due to performance errors or to the absence of functional categories. Following the Continuity Hypothesis, we concluded that the checking domain and its corresponding mechanisms are active and probably target-like at age 2;0, but the child still needs to learn the paradigm, namely the overt realisation of abstract features.

Notes * The research reported in this paper has been partially supported by the Ministry of Education and Culture (PB98–0239), The Basque Government (PI-1999–18) and the University of the Basque Country (9/UPV00033.130–13614/2001). I am grateful to Tony Duffy for his corrections of style. I am also very grateful to Natascha Müller, Tanja Kupisch and to the two anonymous reviewers for their helpful suggestions. 1. Abbreviations: a: Absolutive case; d: Dative case; e: Ergative case; g: genitive case; ger: gerunds; Inf: infinitive (target-like and target-deviant); ipf: imperfect; J: Jurgi; M: Mikel; O: Oitz; oi: Optional Infinitive; pf: perfect; pl: plural; *RI: target-deviant root infinitive; s: singular; ( ): non pronounced speech segments; ( ; , ): age in years; months, days. 2. See Eguzkitza (1987) and Ortiz de Urbina (1989) among others. 3. See Elordieta (2001a) for a detailed discussion on this topic. 4. Dative agreement is not obligatory in some Eastern varieties. See Fernández and Ezeizabarrena (in press) and references listed there.

Null Subjects and optional infinitives in Basque 103

5. We ignore here the forms of the familiar 2nd singular. 6. There are some defective verbs only inflected for the third person: dirudi ‘it seems’. 7. In Kato’s approach strong pronouns are supposed to be in the same “default” Case than Topic DPs. We will not go in depth in this topic. 8. Krämer (1993: 210) lists some examples like (21b), produced as expressing an intention which will immediatelly be followed by an action. (21) b.

Even koffie zetten just coffee make-inf

9. Verb forms listed in (6) correspond to what is called infinitive in other languages: the verb form selected for dictionary entries (e.g. Nola esaten da etorri inglesez? ‘How do you say to come in English?’), and as complement of modal verbs (etorri nahi dut ‘I want to come’) is the same as the past participle in Basque. Actually, such non-finite forms are participles/ infinitives. 10. The parentheses contain the phonetically non-produced segments. 11. Symbols: V: Total of finite and non-finite verb forms in Basque including infinitives/past participles (bota ‘throw/thrown’), present participles (bota-tzen ‘throwing’), and future participles (bota-ko ‘will throw’). Inf
: target-like and target-deviantly used infinitives in Basque. It represents -ø, -n, -tu, -i marked infinitive/participle verb forms like bota ‘throw/thrown’. *RI: target-deviantly used root infinitives. 12. In contexts containing a non-finite verb form and no overt tense auxiliary [….]. The features of this empty pronominal category are recoverable under the general licensing conditions for null pronominals (Boser et al. 1992: 56, quoted by Bel 1997). 13. See Lleó (1997) for a review of different explanatory proposals on child speech. 14. See Crain and Lillo-Martin (1999) for some examples of English and German. 15. Auxiliary deletion is less frequent in Romance languages than in German, Dutch and Basque. 16. As the marking of 3rd person singular for ergative and absolutive is zero, and agreement with three arguments where E and A both are 1st or 2nd person does not exist, Basque inflected forms cannot bear more than two overt person morphemes. Threefold marked forms only exist in case of agreement with ergative and dative marked 1st and 2nd persons, and absolutive 3 pl: (i) eman-Ø Ø-di-zki-zu-t give-pf a3s-root-pl-d2s-e1s ‘I have given them to you’ Such forms are not reported in child Basque before 3;11 in Mikel’s case and neither are they in Peru’s and Jurgi’s corpora (Barreña 1995, Ezeizabarrena 1996). 17. We assume that it is accessible for comprehension. 18. See Celce-Murcia (1977) and Liceras, Valenzuela and Díaz (1999).

104 Maria-José Ezeizabarrena

References Aguado, J. and Pine, J. M. 2002. “There is no evidence for a ‘no overt subject’ stage in early child Spanish: A note on Grinstead (2000).” Journal of Child Language 29: 865–874. Almgren, M, Barreña, A., Ezeizabarrena, M. J., Idiazabal I. and Mac Whinney, B. (eds). 2001. Research on Child Language Acquisition. Proceedings of the 8th Conference of the IASCL. Somerville: Cascadilla. Barreña, A. 1994. “Sobre la adquisición de la categoría funcional COMP por niños vascos”. In La adquisición del vasco y del castellano en niños bilingües, J. M. Meisel (ed.), 231–284. Frankfurt am Main: Vervuert. Barreña, A. 1995. Gramatikaren jabekuntza-garapena eta haur euskaldunak. Bilbao: University of the Basque Country. Bel, A. 1997. “On root infinitives in Catalan, Spanish and English”. Paper presented at ISAPL ’97, Porto, 25–27 June. Bel, A. 2001a. “The projection of aspect: A key in the acquisition of finiteness?”. In Research on Child Language Acquisition. Proceedings of the 8th Conference of the IASCL, M. Almgren et al. (eds), 1297–1313. Somerville: Cascadilla. Bel, A. 2001b. “Sintonía entre la gramática infantil y la adulta: a propósito de los sujetos en castellano y catalán”. Paper presented at III Encuentro Internacional sobre la Adquisición de las Lenguas del Estado. Málaga, September 2001. Boser, K, Lust, B., Santelmann, L., Withman, J. 1992. “The syntax of CP and V2 in early child German: The strong continuity hypothesis”. In Papers from the Tenth Annual Meeting, NELS, K. Broderick (ed.), 51–65. Amherst, MA: GLSA. Celce-Murcia, M. 1977. “Phonological factors in vocabulary acquisition: A case study of a two-year-old English-French bilingual”. Working Papers in Bilingualism 13: 27–41. Centeno, J. G. 1996. Use of Verb Inflections in the Oral Expression of Agrammatic Spanishspeaking Aphasics. Ann Arbor: UMI Dissertation Services, Bell & Howell. Chomsky, N. 1989. “Some notes on economy of derivation and representation”. MIT Working Papers in Linguistics 10: 43–75. Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Clahsen, H. (ed.). 1996. Generative Perspectives on Language Acquisition. Amsterdam: Benjamins. Crain, S. and Lillo-Martin, D. 1999. An Introduction to Linguistic Theory and Language Acquisition. Malden, MA: Blackwell. Eguzkitza, A. 1987. Topics on the Syntax of Basque and Romance. Indiana: Indiana University Linguistics Club. Elordieta, A. 2001a. Verb Movement and Constituent Permutation in Basque. Utrecht: Landelijke Onderzoekschool Taalwetenschap. Elordieta, A. 2001b. ‘’The role of verbal agreement’’. In On Case and Agreement. Kasu eta komunztaduraren gainean, B. Fernández and P. Albizu (eds), 113–132. Bilbao: University of the Basque Country. Elosegi, K. 1998. Kasu eta preposizioen jabekuntza-garapena haur elebidun batengan. Bilbao: University of the Basque Country.

Null Subjects and optional infinitives in Basque 105

Ezeizabarrena, M. J. 1994. “Primeras formas de concordancia en euskera”. In La adquisición del vasco y del castellano en niños bilingües, J. M. Meisel (ed.), 131–230. Frankfurt am Main: Vervuert. Ezeizabarrena, M. J. 1996. Adquisición de la morfología verbal en euskera y castellano por niños bilingües. Bilbao: University of the Basque Country. Ezeizabarrena, M. J. 1997. “Infinitivos verbales en el lenguaje de niños bilingües”. In Views on the Acquisition and Use of a Second Language. Proceedings of EUROSLA ’97, L. Díaz, and C. Pérez (eds), 177–190. Barcelona: Pompeu Fabra. Ezeizabarrena, M. J. 2001. “Non-finite root sentences: A language specific or a child language specific option?”. Working Papers in Multilingualism 26, Vol. 1: 54–72. Ezeizabarrena, M. J. 2002. “Root infinitives in two pro-drop languages”. In The Acquisition of Spanish Morphosyntax: The L1/L2 Connection. A. T. Pérez-Leroux and J. M. Liceras (eds), 33–64. Dordrecht: Kluwer. Ezeizabarrena, M. J. and Barreña, A. 2002. “Does language dominance influence the use of root infinitives?” Paper presented at the IASCL-SRCLD Congress. Madison, 17–21 July. Fernández, B. and Ezeizabarrena, M. J. in press. Itsasaldeko solezismoa, datiboaren lekualdatzearen argipean. Euskaltzaindia. Grinstead, J. 1996. “Morphological underspecification and overt subjects in child Catalan and Spanish”. In Perspectives on Spanish Linguistics, J. Gutiérrez-Rexach and L. SilvaVillar (eds), Vol.1, 55–67. UCLA. Grinstead, J. 1998. Subjects, Sentential Negation and Imperatives in Child Spanish and Catalan. PhD Dissertation, University of California. Guasti, M. T. 1993/1994. “Verb Syntax in Italian Child Grammar: Finite and Nonfinite Verbs”. Language Acquisition 3 (1): 1–40. Guilfoyle, E. and Noonan, M. 1992. “Functional categories and language acquisition”. The Canadian Journal of Linguistics 37: 241–272. Hoekstra, T. and Hyams, N. 1998. “Aspects of root infinitives”. Lingua 106: 81–112. Hyams, N. 1996. “The underspecification of functional categories in early grammar”. In Generative Perspectives on Language Acquisition, H. Clahsen (ed.), 91–128. Amsterdam: Benjamins. Hulk, A. and Müller, N. 2000. “Bilingual first language acquisition at the interface between syntax and pragmatics”. Bilingualism: Language and Cognition, 3 (3): 227–244. Jaeggli, O. and Safir, K. 1989. “The null subject parameter and parametric theory”. In The Null Subject Parameter, O. Jaeggli and K. Safir (eds), 1–44. Dordrecht: Kluwer. Kato, M. 1999. “Strong and weak pronominals in the null subject parameter”. Probus 11: 1–37. Krämer, I. 1993. “The licensing of Subjects in early child language”. In Papers on Case and Agreement II, Cambridge: MIT Working Papers in Linguistics 19, Ph. Collins (ed.), 197–212. Labov, W. 1995. “The case of the missing copula: the interpretation of Zeroes in AfricanAmerican English”. In Language I, L. R. Gleitman and M. Liberman (eds), 25–54. Cambridge: Bradford. Laka, I. and Erriondo, L. 2000. “Aphasia in Basque”. Journal of Neurolinguistics 14: 133–157. Laka, I. and Ezeizabarrena, M. J. 2001. “Exploring the Regression Hypothesis: A Comparison of Data from Acquisition and Agrammatism in Basque”. Poster presented at the workshop: Bilingualism and Brain Plasticity. Trieste 22–25 March.



106 Maria-José Ezeizabarrena

Larrañaga, M.P. 1994. “La evolución del caso en euskera y castellano”. In La adquisición del vasco y del castellano en niños bilingües, J.M. Meisel (ed.), 113–150. Frankfurt am Main: Vervuert. Larrañaga, M. P. 2000. Ergative Sprachen, akkusative Sprachen, der Erwerb des Kasus bei bilingualen Kindern. Frankfurt am Main: Vervuert. Liceras, J., Valenzuela E. and Díaz, L. 1999. “L1/L2 Spanish grammars and the pragmatic deficit hypothesis”. Second Language Research 15 (2): 161–190. Lleó, C. 1997. La adquisición de la fonología de la primera lengua y de las lenguas extranjeras. Madrid: Visor. 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: Benjamins. Meisel, J. M. (ed.). 1994. La adquisición del vasco y del español por niños bilingües. Frankfurt: Vervuert/Iberoamericana. Meisel, J. M., Clahsen, H. and Pienemann, M. 1981. “On determining developmental stages in natural second language acquisition”. Studies in Second Language Acquisition 3: 109–135. Ortiz de Urbina, J. 1989. Some Parameters in the Grammar of Basque. Dordrecht: Foris. Paradis, J. 2001. “The use of subjects by a Spanish-English bilingual child: crosslinguistic interference or the influence of the input?”. Working Papers in Multilingualism 26, Vol. 1: 67–78. Pollock, J. Y. 1989. “Verb movement, universal grammar and the structure of IP”. Linguistic Inquiry 20: 365–424. Radford, A. 1986. “Small children’s Small Clauses”. Bangor Research Papers in Linguistics 43: 51–76. Rasetti, L. 2000. “Null subjects and root infinitives in the child grammar of French”. In The Acquisition of Syntax, M-A. Friedemann and L. Rizzi (eds), 236–268. London: Longman. Rizzi, L. 1986. “Null objects in Italian and the theory of pro”. Linguistic Inquiry 17: 501–557. Sano, T. 1998. Roots in Language Acquisition. A Comparative Study of Japanese and European Languages. UCLA. Serrat, E. and Aparici, M. 2001. “Morphological errors in early language acquisition: Evidence from Catalan an Spanish”. In Research on Child Language Acquisition. Proceedings of the 8th Conference of the IASCL, Almgren et al. (eds), 1260–1277. Somerville: Cascadilla. Wexler, K. 1994. “Optional infinitives, head movement and the economy of derivations.” In Verb Movement, D. Lightfoot and N. Hornstein (eds), 305–350. Cambridge: Cambridge University Press. Wexler, K. 1998. “Very early parameter setting and the unique checking constraint: A new explanation of the optional infinitive stage”. Lingua 106: 23–79. Wijnen, F. 1998. “The temporal interpretation of Dutch children’s root infinitivals: The effect of eventivity”. First Language 18: 379–402. Zabala, I. and Odriozola, J. C. 1996. “On the relation between DP and TP. The structure of Basque Infinitivals”. Catalan Working Papers in Linguistics 5/2: 231–281.



The acquisition of subjects in bilingual children Pronoun use in Portuguese-German children* Marc-Olivier Hinzelin Research Center on Multilingualism, University of Hamburg

1.

Introduction

In the study of bilingual children the differentiation of languages is regarded as a major issue by researchers as well as by the general public. Do bilingual children develop two different grammars or do they (for a certain period) create a fused grammar? Meisel (2001: 15–16) summarizes the different hypotheses about grammatical differentiation in bilingual development as follows: (1) Fusion Hypothesis (2) Differentiation Hypothesis a. Interdependent Development Hypothesis b. Autonomous Development Hypothesis

It has been shown for different language combinations that there is no fusion of grammatical systems (e.g. Genesee 1989, Meisel 1989, De Houwer 1990 and Genesee, Nicoladis and Paradis 1995). The remaining question within the Differentiation Hypothesis is whether one language may influence the development of the other (Interdependent Development Hypothesis) or whether the development of each language is comparable to the development of monolinguals (Autonomous Development Hypothesis). The Autonomous Development Hypothesis for bilingual language acquisition (cf. Meisel 2001: 16) predicts that there is no influence, once the grammatical systems have been separated, from one language onto the other in the grammatical development of the two languages and that both systems are differentiated as soon as children

108 Marc-Olivier Hinzelin

have access to grammatical knowledge. In this work, I want to argue in favor of the Autonomous Development Hypothesis, i.e. bilingual children show no sign of cross-linguistic influence and develop two autonomous grammars (cf. Paradis and Genesee 1996, 1997, Meisel 2001). In the same direction point the results of the comparison between the acquisition of subjects and of Verb– Object order in Hinzelin and Möhring (2001), the comparison of the development of Verb–Object order in 2L1 and L2 acquisition by Möhring and Meisel (2003), and the study on the acquisition of subjects in a Catalan-English bilingual by Juan-Garau and Pérez-Vidal (2000). Comparisons of both languages of the bilingual child show that different structures emerge early and are acquired easily (see Meisel 1989 where he analyzes word order phenomena and Subject–Verb agreement). In my work this will be shown on the basis of pronouns, lexical subjects and finiteness. I will report on the acquisition of subjects by two Portuguese-German bilingual children, comparing their development in the two languages. The study of the acquisition of subjects can shed some light on the question of influence in bilingual acquisition because, in its grammatical basis, it depends on the setting of the pro-drop or Null Subject Parameter (NSP). The notion of a parameter allows us to focus on the task faced by the bilingual child who has to choose between the two different values of the parameter, if the two languages differ in this respect. According to the Interdependent Development Hypothesis, the bilingual German-Portuguese children may be expected to experience difficulty with setting the NSP due to the fact that the parameter is set to different values in each of their two languages. In Portuguese the subject pronoun is generally omitted — the verb ending allowing for its identification — whereas German requires the presence of a (pronominal) subject. The possible influence of one language onto the other could have two directions in this case: On the one hand the use of subject pronouns in German — a nonnull subject language — spoken by the bilingual German-Portuguese children may be less frequent or delayed, as compared to monolingual German children, due to the influence of Portuguese — a null subject language. The extreme case would be a setting of the parameter to the [+pro-drop] value. On the other hand we may expect a high frequency of subject use in Portuguese due to the influence of German or even a setting of the parameter to the German value. In this paper, I will focus on the acquisition of subject pronouns and of finite verb forms and their productive use by children. The different subject types will be studied qualitatively and quantitatively and will be compared with the rate of subject omission. Finally, I will look at verb placement errors in German, a V2

The acquisition of subjects in bilingual children 109

language, and the use of inverted subjects in Portuguese — free subject inversion being a property attributed to null subject languages. All these phenomena are related to finiteness or the NSP, whose connection in acquisition I want to show in this work as formulated in the following starting hypothesis. Although German and Portuguese show a conflicting parameter setting, the bilingual children (Portuguese-German) exhibit no difficulties in setting the Null Subject Parameter to the correct value in each of their two languages. The setting of the NSP is closely linked to the appearance and productive use of personal pronouns in subject position, the acquisition of finiteness, and the systematic use of subjects in non-null subject languages (cf. Section 3). In a feature-based approach (cf. Section 4), this development should be reflected by the instantiation of the φ-features (especially the [person] feature) and — in non-null subject languages — also of the [EPP] (= Extended Projection Principle) feature in the child’s grammar.1

In Section 2, I describe the corpus analyzed in this study. In Section 3, I address theoretical issues concerning the Null Subject Parameter. In Section 4, general acquisitional questions are raised, and the results based on the development in Portuguese and German are presented in 4.1 and 4.2 respectively. Conclusions are drawn in Section 5.

2. The data This study examines longitudinal data from two Portuguese-German bilingual children, Luís and Daniel. The children were video-recorded once a fortnight in naturalistic play situations for about half an hour in each language (for methodological details established for the DUFDE project and followed here see Köppe 1994). I have analyzed every second recording (i.e. approximately one per month) from 1;07,212 to 3;03,10 (for Daniel) and from 2;01,20 to 3;05,23 (for Luís). This sums up to a total of 10,570 utterances (Daniel: 6,092 utterances (German recordings: 3,956, Portuguese recordings: 2,136), Luís: 4,478 utterances (German recordings: 2,717, Portuguese recordings: 1,761). In this work, only utterances containing a verb have been analyzed. Pronouns are only considered when they occur together with a verb. Only subject omissions and realizations with morphologically finite verb forms (except imperatives) and stems are considered in Figures 3 to 6 and 9 to 12 because imperatives and nonfinite forms (infinitives and participles) normally do not require a subject.3 Both children are growing up in middle class families in Hamburg, Germany.

110

Marc-Olivier Hinzelin

The parents observe the principle une personne — une langue. The native language of the mothers is Portuguese. Luís, who acquires European Portuguese, is trilingual, his father, a Colombian, speaks Spanish with him. He acquires German in a home day care setting (with a Tagesmutter), later on (age 3) contact with other German children in a home day care intensifies his exposure to German. Through friends of the family, Luís is in contact with all three languages as well. During the first recordings, Luís often speaks Portuguese with the German interlocutor. Daniel, learning Brazilian Portuguese, has a German father. His mother sometimes speaks German to him as well. Around age 2 a visit of his grandmother strengthens his Portuguese. A younger brother is born when Daniel is around two and a half years old. The development of the MLU is given in Figures 1 and 2 (Portuguese) and 7 and 8 (German) for both children. One learner from each variety of Portuguese (European and Brazilian) was selected because of the differences in subject use between these two. The Brazilian learner is more likely to be influenced by German because Brazilian Portuguese is more similar to German in its use of subject pronouns.

3. The Null Subject Parameter (NSP) 3.1 Starting hypotheses The analysis presented here is based on Chomsky’s (1995, 2000) Minimalist Program, Cardinaletti and Starke’s (1999) pronoun classification and the formulation of the Null Subject Parameter by Kato (1999). Earlier accounts (after the initial works by Perlmutter 1971, Chomsky 1981, 1982, and Rizzi 1986), for example by Jaeggli and Safir (1989), related inflected verb forms to null subjects. This connection is interesting for my study because the acquisition of verbal morphology can be directly observed in the data. Therefore, I assume here that the acquisition of the inflectional paradigm of a given language by the child is closely related to the use of (pronominal) subjects. Cardinaletti and Starke (1999) distinguish between two pronoun classes: strong and deficient. They further subdivide the deficient pronouns into weak and clitics. “Clitic pronouns are deficient with respect to weak pronouns which are in turn deficient with respect to strong pronouns, both distributionally, morphologically, semantically and prosodically.” (Cardinaletti and Starke 1999: 216). Agreement affixes are not counted as deficient pronouns. Kato (1999) suggests an initial split between two classes (strong and weak). Kato then

The acquisition of subjects in bilingual children

argues, following Everett (1996), for three subclasses of weak pronouns: free pronouns, clitics and pronominal affixes. They are in allomorphic relation to each other and in complementary distribution. Kato (1999) further proposes the elimination of the empty category pro and argues that only languages with a [+pronominal] inflectional morphology offer the possibility of null referential subjects. The inflectional morphemes are [+interpretable] and have to be integrated into the system of weak pronouns (“weak” is referring here to Kato’s classification not to Cardinaletti and Starke’s). The productive use of personal pronouns in subject position is taken here in combination with the acquisition of finiteness as a prerequisite for the setting of the NSP. I suppose that the acquisition of the inflectional paradigm of the verb is directly related to the use of pronominal subjects. The grammatical basis for both is the instantiation of the [person] feature in the child’s grammar. At the same time, the [EPP] feature is instantiated in non-null subject languages (i.e. the NSP is set to the [−pro-drop] value) and subject use becomes systematic. 3.2 The Null Subject Parameter (NSP) in Portuguese and German 3.2.1 Differences between (European) Portuguese and German In the Principles and Parameters approach, a given parameter is associated with various surface phenomena — the so-called clustering effect. Table 1 illustrates the Null Subject Parameter and its clustering effects in European Portuguese (EP) and German. There is no general agreement about the phenomena linked to the NSP.4 The three properties taken into consideration here are chosen because they are the most likely ones to cluster and also to occur in child language. This study focuses on the referential null subject property. Expletives and subject inversion will be considered but, as I will show, there is no evidence for their clustering at the time of parameter setting (see Section 4). Languages that allow referential null subjects like (European) Portuguese license free subject inversion but normally do not have overt expletives.5 German, like all the non-null subject languages, does not permit null subjects, neither referential nor expletive ones. But two exceptions have to be considered in German: The topic-drop phenomenon exists in German, i.e. the omission of the first constituent of a sentence (e.g. a subject, a fronted object or adverb). These null elements are discourse-licensed. The second phenomenon is the obligatory null expletive in impersonal passive constructions like heute wird getanzt (‘Today (there) will be dancing.’) or er verspricht, dass getanzt wird (‘He promises that (there) will be dancing.’). In these constructions, with the

111

112

Marc-Olivier Hinzelin

first position in the sentence (SpecCP) occupied by another element or the presence of a complementizer in COMP like dass (‘that’), the use of the expletive es would render the sentence ungrammatical (*heute wird es getanzt, *er verspricht, dass es getanzt wird). Table 1.Clustering of properties in European Portuguese and German Property

European Portuguese

German

[+pro-drop]

[+pro-drop]

[−pro-drop]

i. Referential null canto subject ‘(I) sing.’ comemos o bolo ‘(We) eat the cake.’

*(ich) sing(e) ‘I sing.’ *(wir) essen den Kuchen ‘We eat the cake.’

ii. Null expletive chove ‘(it) rains’ pronoun parece que as crianças comeram o bolo ‘(It) seems that the children ate the cake.’

*(es) regnet ‘it rains’ *(es) scheint, dass die Kinder den Kuchen gegessen haben ‘It seems that the children ate the cake.’

iii. Free subject comeram o bolo as crianças! *haben den Kuchen gegessen die inversion in have eaten the cake the children! Kinder simple sentences ‘The children have eaten the cake!’ die Kinder haben den Kuchen gegessen! the children have the cake eaten! ‘The children have eaten the cake!’

3.2.2 Differences between European and Brazilian Portuguese The two main Portuguese varieties (Brazilian and European) differ in their use of personal pronouns. In Brazilian Portuguese (BP) the subject pronoun is used very frequently (cf. Kato 1999, 2001, Noll 1999:58–64) and not only in cases of contrast, emphasis, etc. Kato (1999: 4) even states that the “[…] Modern Brazilian Portuguese core grammar of the younger generation has lost the null referential subject […]” but “[…] it retains the null expletive […], may have the arbitrary pro […], the null bound pronoun […], and the ‘anaphoric’ pro […].” Duarte (1993) shows the frequency change in the use of null subjects by comparing plays between 1845 and 1992. The use of null subjects drops from around 80% in 1845 to around 25% in 1992 (Duarte 1993: 112). Duarte (1993) relates this finding to the change in the pronominal paradigm. In view of the differences in pronoun use between the two main Portuguese varieties some additional comments are warranted (see Table 3 for the paradigm):

The acquisition of subjects in bilingual children

In standard Brazilian varieties (e.g. São Paulo, Rio de Janeiro) the pronoun você (together with the short form cê) is used instead of tu and the expression a gente is frequently preferred to nós, both agree with the verb in 3ps (= 3rd person singular; cf. Cunha and Cintra 2000:293–294, 298, Noll 1999:58–64). Therefore, these verb forms (2ps and 1pp) are not used at all or are less frequently used.6 The 2pp pronoun (as well as this verb form) is not in general use any longer (cf. Cunha and Cintra 2000: 287). Instead the pronoun vocês, which agrees with 3pp, is used. Thus we normally encounter only three different verb forms in BP: 1ps, 3ps and 3pp. These changes in the pronominal system of BP (compared to earlier stages of BP and to contemporary EP) and the resulting weakening of agreement are treated extensively by Galves (1993, 1997). Another difference concerns post-verbal subjects in “free inversion” constructions: they are marginal in contemporary BP compared to their more frequent use in 18th century BP; this change has been attributed to the high frequency of personal pronoun use and is thus reflecting the parametric change (in progress) of the NSP in BP (cf. Andrade Berlinck 2000 and Kato 2000). Kato (2000) relates the restricted VS order in BP to its partial pro-drop nature; both are due to a change in the nature of the agreement system, namely the creation of a “free weak pronominal paradigm”.

4. The setting of the Null Subject Parameter by the Portuguese-German bilingual children In this section, I deal with phenomena that I regard as decisive for the setting of the NSP because of their tight relationship to its grammatical basis, i.e. their connection to the INFL node. Nevertheless they are not part of the NSP itself (see below): – – – –

productive (and first) use of subject pronouns, systematic (and target-like) use of subjects, acquisition of finiteness, and verb movement (in Portuguese to I, in German via I to C).

The emergence and productive use of personal pronouns and the acquisition of finiteness are closely linked because the features on finite verbs ([person], [number] and [EPP]) have to be checked in the INFL node.7 Without the presence of a subject (in non-null subject languages) and without the movement of the verb, the derivation would crash because feature checking would

113

114

Marc-Olivier Hinzelin

not be possible (see next paragraph). The Null Subject Parameter can only be fixed and the [EPP] feature instantiated when the phenomena mentioned above are present. This holds true for both languages, German and Portuguese. Under minimalist assumptions the child has to select from the universal set of features those that occur in his/her language or languages. The child has to decide whether the feature on a certain element is interpretable or uninterpretable in the target language. If an uninterpretable feature is chosen, then it is necessary to realize an interpretable feature of the same kind on another element in order to delete the uninterpretable one before interpretation at LF. In non-null subject languages [person] and [number] features are uninterpretable on the verb and have to be deleted. Therefore the presence of pronouns, which bear the respective interpretable features, is obligatory and the pronouns have to be acquired at the same time or earlier.8 (Cf. Atkinson 2000, Radford 2000, Say 2001.) I assume that the [EPP] feature is instantiated at the same time as the [person] feature since they are affiliated with the same functional category. The personal pronoun as an element that bears [person] per se is clearly indicating this acquisitional step. A verb only bears [person] once two different grammatical persons are differentiated. Early productive 3ps verb endings do not require the presence of a [person] feature because there is no contrast to other endings. Therefore the acquisition of finiteness (differentiation of two grammatical persons) and the productive use of personal pronouns indicate all together the instantiation of the [person] feature and the setting of the NSP (instantiation of the [EPP] feature) in both languages. It is assumed here that, before the setting of the parameter, the use and omission of “subjects” is not constrained by the syntax (cf. Meisel 1994). Early “subjects” are mostly proper nouns or lexical “subjects”; pronouns do not occur at all or are extremely scarce (not productive). This finding indicates that the nature of the elements in subject position changes in connection with the acquisition and productive use of finite verb morphology. Therefore early subject omissions are neither instances of pro nor is the inflectional morphology [+pronominal]. On the contrary, these “subjects” are not subjects in the syntactic sense, that is, they are not governed by the same grammatical principles as in the adult grammar (cf. Meisel 1994: 94). Only when personal pronouns are used productively, finiteness has been acquired and the IP is present, it is possible to speak of true syntactic subjects (also in the case of lexical subjects and proper nouns). The language use during this stage is pre-syntactical or pre-grammatical; children have a “protolanguage” (Bickerton 1990) and rely on semantic or pragmatic principles (cf. Meisel 2001: 25–26). The division

The acquisition of subjects in bilingual children

into two acquisitional stages (see next paragraph) reflects these assumptions. A different view has been taken by Hyams (1986). She argues that at the outset of development an American child has the Null Subject Parameter set to the [+null] value like in Italian or Chinese. Hyams (1986: 64) states: “[E]arly English differs from adult English in that the former is a ‘pro-drop’ language.” Contrary to Hyams, Bloom (1990), Valian (1991) and Valian, Hoeffner and Aubry (1996) argue for a performance account of early missing subjects. Valian et al. (1996: 154) state that “even very young English-speaking children understand that English requires subjects; children’s inconsistent use of subjects reflects processing limitations, or a performance deficit.” Later, the effective working memory increases and this allows the inclusion of subjects and other obligatory elements by the child with greater consistency. The acquisition of finiteness is one important step in language acquisition and is associated with syntactic phenomena like verb movement (cf. Meisel 1990, 1994, and Meisel and Müller 1992). As far as language acquisition is concerned, I define finiteness according to Meisel (1994: 109) as agreement involving more than one grammatical person and the presence of the IP. The productive use of a verb form is defined as follows: “A form is regarded as being used productively when it appears with more than one verb (type) and in several subsequent recordings.” (Meisel 1994: 125) Pronoun use is defined here as being productive when pronouns appear in several subsequent recordings and when they are used for more than one grammatical person. According to this definition, I divide grammatical development into two stages. The beginning of the second stage, i.e. the productive use of pronouns in subject position, coincides to a large extent with the acquisition of finiteness (cf. also Bittner 2000: 216 for German). During stage I, there are only very few elements in subject position, mainly lexical elements and proper nouns. The lexical “subjects” are mostly not “full” DPs but bare nouns (cf. Tables 5 and 10 for Portuguese and German respectively). This corroborates my assumption of a pregrammatical phase. During this phase there is no productive use of an article, there are no productive pronouns and the verbs are not finite. Since the early lexical “subjects” are no D projections it is also not likely that strong pronouns, which are DPs, appear. When personal pronouns appear at all, they are very rare and their use is far from being systematic, i.e. they do not seem to be governed by the grammar and their use is not predictable. Therefore one cannot speak of productive use. The second stage (stage II) starts with the first systematic and productive appearance of personal pronouns. In general, the use of all kinds of subjects rises considerably (see Sections 4.1 and 4.2).

115

116

Marc-Olivier Hinzelin

Furthermore, the various clustered phenomena mentioned in Table 1 should appear within a short period of time after the setting of the parameter as predicted by parameter theory with respect to acquisition (cf. Meisel 1995). The data contain no evidence for this assumption. The expletive in German and the free subject inversion in Portuguese (the only overt phenomenon) appear only very late (as compared with the setting of the NSP) and are rarely used by the children (cf. 4.1.2 and 4.2.2 for the findings and possible explanations). 4.1 Portuguese 4.1.1 Finiteness and Personal Pronouns Finiteness, according to the definition given above, is acquired when there is productive use of the 3rd person singular (-a in the 1st conjugation) and the 1st person singular (-o in all conjugations).9 The presence of the IP projection in the children’s sentence structures can be shown by the use of modals with infinitives. Table 2 shows the verb endings of regular verbs (-ar, 1st conjugation class), Table 3 gives the pronominal paradigm. Table 2.Verb endings of regular verbs in Portuguese (-ar, 1st conjugation class) falar ‘speak’

1st person

2nd person

3rd person

Singular Plural

falo falamos

falas (falais)

fala falam

Table 3.Subject pronouns in Portuguese

Singular Plural

1st person

2nd person

3rd person

eu nós/a gente (3ps)

tu/você; cê (agrees with 3ps) (vós)/vocês (3pp)

ele; ela eles; elas

Table 4 indicates the two children’s development of verb forms and pronouns. During stage I Luís produces 45% of his sentences with “subjects” (62 elements in subject position out of a total of 139 utterances containing a verb — excluding imperatives, infinitives, participles, gerunds, imitations and unclear cases = 62/139) of which 55% (34/62) are proper nouns and 26% (16) lexical “subjects” but only two (3%) personal pronouns (cf. Table 5 and Figure 6; there are ten instances of demonstrative or relative pronouns (= 16%)). At 2;10,01, at the beginning of stage II, pronoun use rises to 49%

The acquisition of subjects in bilingual children

(31/63) of occurrences in one recording (with 63 subjects (= 56%) in 113 analyzed utterances). Daniel’s stage I lasts until 2;08,28. There are only five (19%) “subjects” (5/27) during this stage, of which four are lexical “subjects” and one is a proper noun. Pronouns start to be used for the first time at 2;08,28 and indicate the beginning of stage II (cf. Table 5 and Figure 4). The assumption of this acquisitional step is corroborated by the rise of the MLU at this age (cf. Table 4 and Figures 1 and 2). Daniel’s 3rd and 1st person agreement are used productively at the same age of 2;08,28. (Finiteness is thus acquired, according to the above definition, at 2;08,28.) Luís’ acquisition of verbal morphology occurs at approximately the same age: 3rd person agreement is used productively at 2;03,16, 1st person markings at 2;08,26. (Thus, at 2;08,26 finiteness is acquired.) With the acquisition of finiteness, personal pronouns start to be used systematically. Daniel uses personal pronouns for the first time at 2;08,28; there are no prior occurrences. Before this age, he uses only proper nouns and lexical subjects in subject position. Systematic use of pronouns by Luís also starts at 2;08,26 — before this age only two instances of 3rd person pronouns are found. This can be interpreted as a corroboration of Kato’s (1999) view that the inflectional morphemes are weak pronouns because strong pronouns and inflectional morphemes emerge simultaneously and indicate the instantiation of the [person] feature in the children’s grammar. During stage I, eight agreement errors occur in Daniel’s data (in seven he uses a 3ps form for 1ps reference) and also eight in Luis’ data (six 3ps for 1ps). For monolingual comparison data see Clark (1985) on Romance in general, Faria (1993) on EP and Valian and Eisenberg (1996), Oliveira (1999), Simões (1999, 2000), Figueira (2000), and Oliveira (2001) on BP. Examples of verbs (and subjects): Stage I: (1) (é) auto (is car ‘That is a car.’

(Daniel, 1;09,02)

(2) (o) (ca:o) (quebô) (= o carro quebrou) (the (car (broke ‘The car broke.’

(Daniel, 2;04,26)

117

118

Marc-Olivier Hinzelin

Table 4.Verb forms and pronouns in Portuguese (Age, MLU, Form) 1st verb form 1st produc- 1st pronouns Productive 2 productive (3ps) tive verb (later used use of pro- verb forms form (3ps) pronouns) nouns (3ps & 1ps) (finiteness)

3 productive verb forms (3ps, 1ps & 2ps)

Daniel 1;09,02 1.28 tá, é ‘is, is’

2;08,28 1.89

2;08,28 1.89 eu, você/cê (ele, ela)

2;08,28 1.89

2;08,28 1.89

–

Luís

2;03,16 1.63

2;05,12 2.73 ela (eu, tu, ele)

2;08,26 2.93

2;08,26 2.93

3;00,22 2.91

2;01,20 1.13 dá, ápi (= cabe/abre) ‘gives, suits/ opens’

Table 5.Subjects in Portuguese — DPs and Pronouns Child NOR (AU)

Subject omissions (% of AU)

Personal pro- Demonstrative Lexical subnouns and other jects (% of AU) pronouns (BN/DPs/U) (% of AU) (% of AU)

Proper nouns (% of AU)

Daniel 12 (27)

22

0

0

1

(81%)

(0%)

(0%)

4 (3/1/0) (15%)

Luís 7 (139)

77

2

10

34

(55%)

(1%)

(7%)

16 (13/2/1) (12%)

Daniel 4 (221)

132

73

9

8

1

(60%)

(33%)

(4%)

(4%)

(0.5%)

Luís 5 (402)

168

114

52

35

33

(42%)

(28%)

(13%)

(9%)

(8%)

Stage I

(4%)

(24%)

Stage II

NOR = number of analyzed recordings, AU = analyzed utterances, BN = bare nouns, U = unclear.

The acquisition of subjects in bilingual children

5,5 5 4,5 4 3,5 3 2,5 2 1,5 1 1 0,1 3;0 0 1,0 2;1 3 0,0 2;1 8 8,2 2;0 5 7,2 2;0 8 6,2 2;0 0 5,3 2;0 6 4,2 2;0 9 3,2 2;0 0 3,0 2;0 6 1,2 2;0 3 0,2 2;0 7 1,1 1;1 0 0,2 1;1 2 9,0 1;0 4 8,0 1;0

Figure 1.Daniel – Development of the MLU in Portuguese. 5,5 5 4,5 4 3,5 3 2,5 2 1,5 1 3 5,2 3;0 9 4,1 3;0 4 3,1 3;0 9 2,1 3;0 2 0,2 3;0 5 1,2 2;1 9 0,2 2;1 1 0,0 2;1 6 8,2 2;0 7 7,2 2;0 0 6,1 2;0 2 5,1 2;0 4 4,1 2;0 6 3,1 2;0 7 2,1 2;0 0 1,2 2;0

Figure 2.Luís – Development of the MLU in Portuguese.

(3) ão tá (= cão está) dog is ‘There is the dog.’

(Luís, 2;03,16)

(4) tá itschi anda ozinho (= ? Luís anda sozinho)  Luís goes alone ‘Luís does it himself.’

(Luís, 2;07,27)

Stage II: (5) eu toco I touch-1ps ‘I touch it.’

(Daniel, 2;08,28)

(6) ’cê viu (’cê = você) you saw-3ps-pps ‘You saw it.’

(Daniel, 2;08,28)

119

120 Marc-Olivier Hinzelin

(7) eu que:o voltar (que:o = quero) I want-1ps to-go-back ‘I want to go back.’

(Daniel, 2;10,03)

(8) eu vou [s]irar benny ([s]irar = tirar) I will-1ps [pull Benny (= his cat) ‘I will pull Benny (= his cat).’

(Daniel, 2;11,00)

(9) ·imÒ eu pinto  I paint-1ps ‘I am painting.’

(Luís, 2;08,26)

(10) e ê e ele ainda não vai imboa (imboa = embora)  and  he yet not goes-3ps away ‘And he does not leave yet.’

(Luís, 2;10,01)

(11) eu tenho outro I have-1ps other ‘I have got another one.’

(Luís, 2;08,26)

(12) eu vai bucar um papel (bucar = buscar) I will-1ps look-for a paper ‘I will look for a piece of paper.’

(Luís, 2;08,26)

The examples of stage I show the predominance of “3ps” verb forms (mostly present tense, but also perfect (pps: pretérito perfeito simples) and imperfect; Daniel: 26 out of 27 utterances containing an analyzed verb (i.e. excluding imperatives, infinitives, participles, gerunds, imitations and unclear cases); Luís: 139 out of 139 utterances) and the use of lexical “subjects” (if used at all). The examples of stage II demonstrate 1ps verb forms (5 and 9), the use of subject pronouns (5–12) and the use of modals (7) and auxiliaries (8 and 12) clearly indicating the activation of a functional projection (IP; see also Table 6). Summarizing, one can say that the acquisition of finiteness signals the onset of target-like subject use, as illustrated in Figures 3 to 6. This corroborates the postulated relation between finiteness and pronominal subject use. All figures show that the use of pronominal subjects starts with the acquisition of finiteness10 and that the rate of null subjects and of phonetically realized subjects is about 50%. 4.1.2 Post-verbal subjects Post-verbal subjects are frequently found with unaccusative verbs but free subject inversion (or Romance inversion), which implies that a constituent (generally the object) is placed between verb and inverted subject, is rarely found. Only one clear example was found (13); two other examples were copula constructions.

The acquisition of subjects in bilingual children

Table 6.Developmental sequence of finite verb forms (productive use) in Portuguese Da 2;8

Lu 2;3–2;8

-a -e é tá

3ps 3ps 3ps 3ps

2;08,28 * 2;08,28 2;08,28

2;03,16 2;06,10 2;08,26 2;05,12

-iu -ou

3ps pps 3ps pps

* 2;08,28

2;07,27 2;10,01

Da 2;8

Lu 2;8–3;0

-o sou/estou

1ps 1ps

2;08,28 *

2;08,26 *

-s (EP)

2ps

–

3;00,22

tem (BP)/há (EP)

3ps

2;10,03

*

-am -em são/estão

3pp 3pp 3pp

* * *

* * *

-mos

1pp

*

*

Da = Daniel, Lu = Luís; * = form not found in the data.

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1 0,1 3;0 0 1,0 2;1 3 0,0 2;1 8 8,2 2;0 5 7,2 2;0 8 6,2 2;0 0 5,3 2;0 6 4,2 2;0 9 3,2 2;0 0 3,0 2;0 6 1,2 2;0 3 0,2 2;0

% subject omissions

% pronominal subjects

Figure 3.Daniel – Pronominal subject use in Portuguese.

(13) tem bugallá abi luí (= tem ? lápis Luís) has-3ps ? pencil Luís ‘Luís has got the pencil.’

(Luís, 2;08,26)

(14) não é carro polítia aquêle (polítia = polícia) not is-3ps car police that ‘That one is not a police car.’

(Luís, 2;08,26)

121

122 Marc-Olivier Hinzelin

70 60 50 40 30 20 10 0 1 ,1 00 3; 0 ,0 11 2; 3 ,0 10 2; 8 ,2 08 2; 5 ,2 07 2; 8 ,2 06 2; 0 ,3 05 2; 6 ,2 04 2; 9 ,2 03 2; 0 ,0 03 2; 6 ,2 01 2; 3 ,2

00 2;

personal pronouns

other pronouns

proper nouns

lexical subjects

Figure 4.Daniel – Types of subjects in Portuguese. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 2 0,2 3;0 5 1,2 2;1 9 0,2 2;1 1 0,0 2;1 6 8,2 2;0 7 7,2 2;0 0 6,1 2;0 2 5,1 2;0 4 4,1 2;0 6 3,1 2;0 7 2,1 2;0 0 1,2 2;0

% subject omissions

% pronominal subjects

Figure 5.Luís – Pronominal subject use in Portuguese.

(15) é aí meu carro is-3ps there my car ‘My car is there.’

(Daniel, 3;00,11)

That there are no examples of free inversion in Daniel’s data is probably due to its marginality in contemporary BP (see Section 3.2.2). As far as EP is concerned this construction exists in adult speech (as focused subject construction, cf. Costa 2001) but it could be expected to be rare in child speech because of its complexity.11

The acquisition of subjects in bilingual children

70 60 50 40 30 20 10 0 2

,2

5

,2

proper nouns

00

3;

9

1

,2

11

2;

10

other pronouns

2;

6

,0

10

2;

7

,2

08

2;

0

,2

07

2;

2

,1

06

2;

4

,1

05

2;

,1

04

2;

personal pronouns

lexical subjects

Figure 6.Luís – Types of subjects in Portuguese.

4.2 German 4.2.1 Finiteness and personal pronouns In German the endings to be acquired (in order to establish finiteness according to the definition above) are -t for 3rd person and zero ending or -e for 1st person with main verbs (the 2nd person emerges later, see footnote 9; for the endings see Table 7; the pronominal paradigm is given in Table 8). Another indicator for finiteness is the V2 placement. Table 7.Verb endings of regular verbs in German sagen ‘say’

1st person

2nd person

3rd person

Singular Plural

sage/-0 sagen

sagst sagt

sagt sagen

Daniel uses the 3rd person -t ending productively at age 2;04,26. The -0 ending with 1st person reference is also produced at 2;04,26, but there is only one use per recording over three consecutive recordings and the stem is also used for 3rd person reference. I therefore consider finiteness to be acquired at the age of 2;07,25 when the use of the 1st person becomes more regular and forms with the -e ending begin to be used, as well. At the same age at 3;02,19, Luís’ 3rd and 1st person agreement show up. Thus I consider finiteness to be acquired at this age. During stage I (and before the acquisition of finiteness), Daniel uses 27% “subjects” (23 elements in subject position out of a total of 85 utterances

123

124 Marc-Olivier Hinzelin

Table 8.Subject pronouns in German

Singular Plural

1st person

2nd person

3rd person

ich wir

du ihr

er; sie; es/der; die; das/man ‘one’ sie/die

containing a verb — excluding imperatives, infinitives, participles, gerunds, imitations and unclear cases = 23/85), of them five (22%) are pronominal forms (9 lexical “subjects” (= 39%), 3 proper nouns (= 13%) and 6 interrogative pronouns (= 26%)). At 2;07,25 he uses eight (89%) subject pronouns (8/9; with 9 subjects (= 39%), one instance of topic-drop (= 4%) and 13 subject omissions (= 57%) in 23 analyzed utterances) and then the number of pronouns increases considerably (and the number of subject omissions that are not due to topic-drop decreases (below 10%)). This indicates the beginning of stage II (cf. Table 10 and Figure 10). Luís’ stage I lasts until 2;10,29, but there is a transitional phase between 2;10,29 and 3;02,19 during which he uses personal pronouns only sporadically (cf. Table 10 and Figure 12). Agreement errors in German show up almost exclusively during stage I. The sixteen (55%) instances (16/29; with 29 subjects in 46 analyzed utterances: 8 lexical subjects (= 28%), 5 proper nouns (= 17%)) of pronouns found in Luís before finiteness is acquired, occur exclusively during the transitional phase of five and a half months.12 In this phase he already uses pronouns productively but not very frequently; he does not make productive use of two different verb forms yet. The development in German is shown in Table 9, the acquisition process is also reflected by the MLU development (cf. Figures 7 and 8). For comparison data see Clahsen (1982), Mills (1985), Hamann (1996), Bittner (2000) for monolingual German and Meisel (1990, 1994) for bilingual German-French acquisition. Examples of verbs (and subjects): Stage I: (16) is da is there ‘It is there.’

(Daniel, 1;10,20)

(17) mann (sitz) bus man (sit bus ‘The man is sitting in the bus.’

(Daniel, 2;03,00)

The acquisition of subjects in bilingual children

Table 9.Verb forms and pronouns in German (Age, MLU, Form) 1st verb form 1st produc- 1st pronouns Productive 2 productive (3ps) tive verb (later used use of pro- verb forms form (3ps) pronouns) nouns (3ps & 1ps) (finiteness)

3 productive verb forms (3ps, 1ps & 2ps)

Daniel 1;10,20 1.3 will, is ‘wants, is’

2;04,26 1.68

1;10,20 2;07,25 1.3 2.63 ich (das, du, er, der, die, man, die (pl.), es, wir, sie)

2;07,25 2.63

2;11,00 3.45

Luís

3;02,19 1.67

2;10,29 2;10,29 2.37 2.37 ich, man (das, du, die, der, die (pl.), es, wir)

3;02,19 1.67

3;04,19 1.75

2;08,26 1.56 geht ‘goes’

5,5 5 4,5 4 3,5 3 2,5 2 1,5 1 1 0,1 3;0 8 1,2 2;1 0 1,0 2;1 3 0,0 2;1 5 7,2 2;0 8 6,2 2;0 0 5,3 2;0 6 4,2 2;0 9 3,2 2;0 0 3,0 2;0 6 1,2 2;0 3 0,2 2;0 7 1,1 1;1 0 0,2 1;1 2 9,0 1;0 4 8,0 1;0

Figure 7.Daniel – Development of the MLU in German.

(18) do geht (= ? doch geht) sure goes ‘Sure, it works.’

(Luís, 2;08,26)

(19) lui ol (= ? Luís holt/holen) Luís get ‘Luís gets it.’

(Luís, 2;08,26)

125

126 Marc-Olivier Hinzelin

Table 10.Subjects in German — DPs and Pronouns Child NOR (AU)

Subject omissions [topic-drop/ no t. d.] (% of AU)

Personal pronouns (% of AU)

Demonstrative and other pronouns (% of AU)

Lexical Proper nouns subjects (% of AU) (BN/DPs/U) (% of AU)

Daniel 12 (85)

62 [7/55] (73%) ([8%/65%])

5

6

3

(6%)

(7%)

9 (7/1/1) (11%)

Luís 9 (5)

4 [3/1] (80%) ([60%/20%])

0

0

0

1 (20%)

Daniel 9 (553)

86 [37/49] (16%) ([7%/9%])

379

29

45

4

(69%)

(5%)

(8%)

(1%)

Luís 7 (159)

44 [11/33] (28%) ([7%/21%])

87

2

17

7

(55%)

(1%)

(11%)

(4%)

Stage I

(4%)

Stage II

NOR = number of analyzed recordings, AU = analyzed utterances, BN = bare nouns, U = unclear.

Stage II: (20) ich komme I come-1ps ‘I am coming.’

(Daniel, 2;07,25)

(21) er will spiel’n autos he wants to-play cars ‘He wants to play cars.’

(Daniel, 2;07,25)

(22)  hab ich gefund’n topic-drop have-1ps I found ‘I have found that.’

(Daniel, 2;07,25)

The acquisition of subjects in bilingual children 127

5,5 5 4,5 4 3,5 3 2,5 2 1,5 1 3 5,2 3;0 9 4,1 3;0 4 3,1 3;0 9 2,1 3;0 2 0,2 3;0 5 1,2 2;1 9 0,2 2;1 1 0,0 2;1 6 8,2 2;0 7 7,2 2;0 0 6,1 2;0 2 5,1 2;0 4 4,1 2;0 6 3,1 2;0 7 2,1 2;0 0 1,2 2;0

Figure 8.Luís – Development of the MLU in German.

(23) siehst du? see-2ps you? ‘Do you see it?’

(Daniel, 2;07,25)

(24) ich mach da[x] (= das) I do-1ps this ‘I will do this.’

(Luís, 2;10,29)

(25) gehst du? go-2ps you? ‘Are you leaving?’

(Luís, 3;02,19)

(26)   (möchte) ich geh’n topic-drop (?) (want-1ps I go ‘I want to go (to bed).’

(Luís, 3;02,19)

(27) wollen wi eine eisenbahn machen mit das? want-1pp we a train make with that ‘Do you want to make a train with me out of that?’

(Luís, 3;05,23)

The modals in (21) and (27) indicate the use of a functional projection like IP or CP because they are base-generated in IP. The examples (22) and (26) probably constitute the first evidence for a CP structure with the first constituent being dropped (topic-drop of the object in SpecCP), the verb in second position (V2) and the subject inverted (in SpecIP). This confirms the acquisition of finiteness with verb movement to C (see also Table 11, comparison data from Meisel 1994 (French-German) is given in order to evaluate the linguistic development of the children studied here). Figures 9 and 11 clearly show that the use of pronominal subjects in both

128 Marc-Olivier Hinzelin

Table 11.Developmental sequence of finite verb forms (productive use) in German13

-t is’ -0 -0

3ps 3ps 3ps mod 1ps mod

Da 2;0–2;7

Lu 2;11–3;2 C 1;11–2;4

Iv 2;0–2;3

P 2;7–2;10

2;04 2;00 2;07 2;07

3;02 2;11 3;05 3;02

1;11 2;1 2;4 2;5

2;0 2;0 2;3 2;3

2;9 2;7 2;10 2;8

Da 2;7–3;2

Lu 3;2–3;5

C 2;5–2;8

Iv 2;5–2;8

P 2;11–3;3

-0 -e bin

1ps 1ps 1ps

2;07 2;07 2;11

3;02 3;02 *

2;5 2,10 *

2;7 2;8 2;7

2;10 2;9 3;3

-st bist

2ps 2ps

2;11 3;02

3;04 *

2,4 *

2;5 2;8

2;11 *

-en sind

3pp 3pp

3;00 3;01

3;03 3;05

2;5 2;6

2;8 2;8

2;11 3;10

-en sind

1pp 1pp

3;02 *

3;05 *

2;9 2;8

2;6 *

3;3 3;11

Da = Daniel, Lu = Luís; C = Caroline, Iv = Ivar, P = Pierre; mod = modal; * = form not found in the data.

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0 3,1 3;0 27 2, 3;0 00 2, 3;0 20 1, 3;0 11 0, 3;0 28 1, 2;1 00 1, 2;1 03 0, 2;1 25 7, 2;0 28 6, 2;0 30 5, 2;0 26 4, 2;0 29 3, 2;0 00 3, 2;0 26 1, 2;0 23 0, 2;0 17 1, 1;1

% subject omissions without t. d.

% pronominal subjects and t. d.

Figure 9.Daniel – Pronominal subject use (including topic-drop) in German.

children before the acquisition of finiteness is at chance rather than guided by the grammar. As mentioned at the beginning of this section, there is no systematic pronoun use during this stage and pronouns do not have the properties of subjects typical of the adult grammar. The pronouns that are used are very scarce and not productive. After the instantiation of the [person] and the [EPP] feature in the children’s grammar, that is, after the acquisition of

The acquisition of subjects in bilingual children 129

70 60 50 40 30 20 10 0 0 ,1 03 3; ,27 02 3; ,00 02 3; ,20 01 3; ,11 00 3; ,28 11 2; ,00 11 2; ,03 10 2; ,25 07 2; ,28 06 2; ,30 05 2; ,26 04 2; ,29 03 2; ,00 03 2; ,26 01 2; ,23 00 2; ,17 11 1;

personal pronouns

other pronouns

proper nouns

lexical subjects

Figure 10.Daniel – Types of subjects in German. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 3 5,2 3;0 19 4, 3;0 14 3, 3;0 19 2, 3;0 22 0, 3;0 25 1, 2;1 29 0, 2;1 01 0, 2;1 26 8, 2;0

% subject omissions without t.d.

% pronominal subjects and t. d.

Figure 11.Luís – Pronominal subject use (including topic-drop) in German.

finiteness and at the onset of productive pronoun use, pronominal subject use rises, reaching over 90% for Daniel from the age of 2;10,03 onwards and around 80% for Luís from 3;02,19 onwards. The high frequency of pronominal subject use indicates that, at this point, the NSP is set to the correct value in German. In the figures instances of topic-drop, permitted by the German target grammar, are counted together with the use of lexical pronouns. Thus the percentage of subject omissions only contains cases violating the target grammar. (The remaining cases of subject omission could be due to performance constraints as mentioned at the beginning of this section.)

130 Marc-Olivier Hinzelin

70 60 50 40 30 20 10 0 3 ,2 05 3; 9 ,1 04 3; 4 ,1 03 3; 9 ,1 02 3; 2 ,2 00 3; 5 ,2 11 2; 9 ,2 10 2; 1 ,0 10 2; 6 ,2

08 2;

personal pronouns

other pronouns

proper nouns

lexical subjects

Figure 12.Luís – Types of subjects in German.

4.2.2 Expletives After the setting of the NSP the other phenomenon, besides the overt referential (pronominal) subject, expected to show up is the use of expletives. Their emergence indicates for Hyams (1986) the end of the optional subject stage. But the expletive pronoun is used only later; maybe due to maturational constraints or to the recording situation, e.g. there is no use of a meteorological verb that would imply the use of an expletive. Moreover clear expletive constructions are rarely encountered in the data. Sometimes it is difficult to decide whether a specific use of das (‘that’, with the variants des/’s) is used in an expletive way or with a general, not further determined reference. The first clear case of an expletive es (‘it’) occurs at the age of 3;03,10 with Daniel: (28) es fehlt noch ein’ hier there misses still one here ‘There is still one missing here.’

(Daniel, 3;03,10)

Until the end of the period covered in this study (until 3;05,23) there is no clear case of expletive es in Luís’ data. Juan-Garau and Pérez-Vidal (2000: 185) report as well the absence of expletives up to the beginning of 3;4 in the English of the bilingual Catalan-English child Andreu studied. (Subjects are already used overtly in 93.8% of Andreu’s clause productions in English at 3;2 (Juan-Garau and Pérez-Vidal 2000: 186–187).) 4.2.3 Verb placement errors Verb placement errors could provide evidence against verb movement in German, i.e. the presence of the functional projections IP and CP.

The acquisition of subjects in bilingual children

Table 12.Verb placement errors in German Daniel

Luís

[−V2]

[−Vend]

[−V2]

[−Vend]

Stage I Stage II

3 18

13 6

2 13

2 2

Total Total (all)

21 40

19

15 19

4

There are some verb placement errors that are not very frequent and are also attested in monolinguals (cf. Mills 1985:158–167). All realizations of morphologically finite forms that are not V2 and of non-finite forms that are not verb-end are counted as verb placement errors. V1 realizations that are due to topic-drop (missing subject, object or adverb) were not considered. The errors reported in Table 12 do not constitute evidence against verb movement. Most utterances show correct word order and the finite verb does not stay in final position but moves to a V3 or V1 position as examples (29) and (30) illustrate: (29) [iç] de mach ein ein eiðebahn [I ? make a a railway ‘I make a railway.’

(Daniel, 2;11,28)

(30) muß die hier bleibn must that here stay ‘That one must stay here.’

(Daniel, 2;11,28)

It should be noted that there is neither quantitative nor qualitative evidence against verb movement at the time when pronouns are used productively and finiteness is acquired.

5. Conclusion The difference in the percentages of subject use in the two languages reflects the setting of the Null Subject Parameter to distinctive values in German and in Portuguese by the bilingual children. Subject use becomes systematic (in the languages that require a pronominal subject). Personal pronouns in subject position are used productively at 2;08 (Portuguese – Daniel and Luís) and at 2;07 (German – Daniel) and 2;10 (German – Luís) respectively. The data show

131



132

Marc-Olivier Hinzelin

that finiteness is acquired at 2;08 (Portuguese – Daniel and Luís) and at 2;07 (German – Daniel) and 3;02 (German – Luís) respectively. The setting of the NSP is closely linked to the appearance and productive use of personal pronouns in subject position and to the acquisition of finiteness occurring in the same developmental period. The [person] and the [EPP] features are now instantiated in the child’s grammar. The systematic use of subjects in the non-null subject language German reflects this grammatical development. Bilingual children exhibit no difficulty in setting the Null Subject Parameter to its respective value in each of their two languages, i.e. in German to [−prodrop] and in Portuguese to [+pro-drop]. The frequency of pronoun use differs in the two languages reflecting the properties of the target grammar and their application. (Cf. the similar results obtained by Juan-Garau and Pérez-Vidal (2000) for a Catalan-English bilingual child.) Concerning pronominal subject use, I have not found any instances of influence of one language on the other. These findings support my assumption that bilingual children develop the two language systems autonomously. On the basis of the analysis presented here, I conclude that bilingual children are able to keep their grammatical systems separate and exhibit no difficulties in setting a given parameter to contrasting values in each of their two languages. This confirms the Autonomous Development Hypothesis.

Notes * This study has been carried out as part of the research project “Simultaneous and successive acquisition of bilingualism”, directed by Jürgen M. Meisel. This project is one of currently 13 funded by the Deutsche Forschungsgemeinschaft (German Science Foundation. within the framework of the SFB 538 Mehrsprachigkeit (Collaborative Research Center on Multilingualism), established at the University of Hamburg. Besides the author of the present paper, the researchers currently working in this research team are Jürgen M. Meisel (principal investigator), Matthias Bonnesen, María Pilar Larrañaga and Anja Möhring. I would like to thank the team as well as Gisella Ferraresi, Maria L. Goldbach, Regina Köppe, Esther Rinke and Tessa Say for discussions and comments on earlier drafts of this paper. Furthermore I would like to thank two anonymous reviewers for their input to this paper. Thanks to Margaret Kehoe and Lori Grosland for correcting my English. The responsibility for the contents of this study lies with the author. 1. As the reader will notice in Section 3 and 4, the starting hypothesis refers to both German and Portuguese. In Portuguese the productive use of personal pronouns, although they are not obligatory, together with the acquisition of finiteness indicates the instantiation of the

The acquisition of subjects in bilingual children

[person] feature because the pronouns and verbal affixes both bear this feature. As I will show in Section 4, the setting of the NSP and its value depend on this acquisitional step. 2. Age in (years;months,days). 3. In Portuguese, a subject (lexical subject or overt pronoun) that bears nominative Case may precede (inflected) infinitive forms (infinitivo pessoal/flexionado). 4. Chomsky (1981: 240) enumerates five: (i) missing subject, (ii) free inversion in simple sentences, (iii) “long wh-movement” of subject, (iv) empty resumptive pronouns in embedded clause, (v) apparent violations of the *[that-t] filter. All the phenomena clustering with the referential null subject property have been questioned in the literature and counterexamples have been found (cf. Gilligan 1987, Wanner 1993, Haider 1994, Nagy and Heap 1998, and Heap 2000, among others). The clustering has been established empirically, there is no theoretical relation between all of the phenomena. 5. There are, however, European Portuguese varieties that exhibit overt expletive pronouns, e.g. ele chove ‘it rains’ (cf. Spitzer 1917, Cunha and Cintra 2000: 284). 6. In Daniel’s speech (his mother is from the city of Itabuna, state of Bahia, in the NorthEast of Brazil), tu is never encountered. Nevertheless tu is used in this region (and others) together with a 3ps verb form, e.g. tu fala. I am grateful to Diana Rodrigues de Carvalho for pointing this out to me. 7. As one anonymous reviewer pointed out, there are different points of view and the theory is still developing: “[i]n the framework of Chomsky (1995), followed by the author, the EPP is presented as a strong D-feature of the I head that has to be eliminated during the derivation. Subsequent works describe EPP as a feature of a functional category that requires the presence of a referential-like expression in its local domain (Chomsky [2000]) or as a property of a feature that a functional head bears and that is required to trigger movement (cf. Pesetsky [and Torrego 2001]).” Another new view on the [EPP] feature is proposed by Nash and Rouveret (2002). Note that in Kato’s (1999) account (elimination of pro) the inflectional morphemes are integrated into the system of weak pronouns, i.e. the endings of finite verb forms are pronouns, therefore strong pronouns and inflectional morphemes should emerge simultaneously. 8. The features are also deleted in the case of full DPs in subject position, these bearing [number] and default 3rd [person] features. 9. The first person singular is chosen here because it emerges in the children studied in both languages before the second person. 10. However, in her study on monolingual EP acquisition Soares (2002:140–154) found that the 1ps/2ps/3ps contrast (finiteness) emerged before pronominal subjects. 11. Hereby I mean complexity in terms of pragmatic complexity to express focus, not structural complexity. Costa (1998, 2000) proposes that post-verbal subjects in EP are in SpecVP (i.e. in situ). (Cf. also the discussion of the VOS order in Spanish in Zubizarreta 1998: 125–127 and the book on the syntax of Subject–Verb inversion in Portuguese by Ambar 1992.) The question of the subject position in free inversion constructions is also addressed by, among others, Longobardi (2000) and Roberge (1986).

133

134 Marc-Olivier Hinzelin

12. There are two instances of the pronoun ich (‘I’) together with an infinitive at age 2;10,01. In counts only pronouns occurring with a morphologically finite verb form are considered. 13. The data of the Portuguese-German children is presented in comparison with the data of French-German children in Meisel (1994: 98).

References Ambar, M. M. 1992. Para uma sintaxe da inversão sujeito-verbo em português. Lisboa: Edições Colibri. Andrade Berlinck, R. de 2000. “Brazilian Portuguese VS order: A diachronic analysis”. In Brazilian Portuguese and the Null Subject Parameter, M. A. Kato and E. V. Negrão (eds), 175–194. Frankfurt am Main: Vervuert. Atkinson, M. 2000. Minimalist Visions. Ms., University of Essex. Bickerton, D. 1990. Syntactic Development: The Brain Just Does It. Ms., University of Hawaii at Manoa. Bittner, D. 2000. “Are there correlations in the acquisition of definite pronouns and finite verbs in German?”. In Papers on Language Change and Language Acquisition, A. Alexiadou, N. Fuhrhop, U. Kleinhenz and P. Law (eds), 199–217. ZAS Papers in Linguistics (Berlin) 15. Bloom, P. 1990. “Subjectless sentences in child language”. Linguistic Inquiry 21: 491–504. Cardinaletti, A. and 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. Chomsky, N. 1981 (1982). Lectures on Government and Binding. The Pisa Lectures. 2nd revised ed. 1982 (1st ed. 1981). Dordrecht: Foris. 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. Chomsky, N. 2000. “Minimalist inquiries: The framework”. In Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik. R. Martin, D. Michaels and J. Uriagereka (eds), 89–155. Cambridge, MA: The MIT Press. Clahsen, H. 1982. Spracherwerb in der Kindheit. Eine Untersuchung zur Entwicklung der Syntax bei Kleinkindern. Tübingen: Narr. Clark, E. V. 1985. “The acquisition of Romance, with special reference to French”. In The Crosslinguistic Study of Language Acquisition. Volume 1: The Data. D. I. Slobin (ed.), 687–782. Hillsdale, New Jersey: Erlbaum. Costa, J. M. M. da 1998. Word Order Variation. A Constrained-based Approach. The Hague: Holland Academic Graphics. Costa, J. 2000. “Focus in situ: Evidence from Portuguese”. Probus 12: 187–228. Costa, J. 2001. “Focus in European Portuguese: Issues for language acquisition”. In Proceedings of the Colloquium on Structure, Acquisition, and Change of Grammars: Phonological and Syntactic Aspects. Volume I. University of Hamburg, 27th–29th October 2000, K. F. Cantone and M.-O. Hinzelin (eds), 37–46. Working Papers in Multilingualism Series B 26, Hamburg: Sonderforschungsbereich 538/Universität Hamburg.

The acquisition of subjects in bilingual children

Cunha, C. and Cintra, L. F. L. 2000. Nova gramática do português contemporâneo. 16.a edição. Lisboa: Edições João Sá da Costa. De Houwer, A. 1990. The Acquisition of Two Languages from Birth: A Case Study. Cambridge: Cambridge University Press. Duarte, M. E. L. 1993. “Do pronome nulo ao pronome pleno: A trajetória do sujeito no português do Brasil”. In Português brasileiro. Uma viagem diacrônica. Homenagem a Fernando Tarallo, I. Roberts and M. A. Kato (eds), 107–128. Campinas: Editora da Universidade Estadual de Campinas Unicamp. Everett, D. L. 1996. Why There are no Clitics. An Alternative Perspective on Pronominal Allomorphy. [Dallas, Tex.]: The Summer Institute of Linguistics and The University of Texas at Arlington. Faria, I. H. 1993. “A aquisição da noção de ‘agente’ e a produção de sujeitos sintácticos por crianças portuguesas até aos dois anos e meio”. Revista Internacional de Língua Portuguesa (RILP) 10: 16–50. Figueira, R. A. 2000. “L’acquisition du paradigme verbal du portugais. Les multiples directions des fautes”. Cahiers d’Acquisition et de Pathologie du Langage (CALaP, Université René Descartes, Paris V) 20: 45–64. Galves, C. C. 1993. “O enfraquecimento da concordância no português brasileiro”. In Português brasileiro. Uma viagem diacrônica. Homenagem a Fernando Tarallo, I. Roberts and M. A. Kato (eds), 387–408. Campinas: Editora da Universidade Estadual de Campinas Unicamp. Galves, C. 1997. “La syntaxe pronominale du portugais brésilien et la typologie des pronoms”. In Les pronoms. Morphologie, syntaxe et typologie, A. Zribi-Hertz (ed.), 11–34. Saint-Denis: Presses Universitaires de Vincennes (Université Paris VIII). Genesee, F. 1989. “Early bilingual development: One language or two?”. Journal of Child Language 16: 161–179. Genesee, F., Nicoladis, E. and Paradis, J. 1995. “Language differentiation in early bilingual development”. Journal of Child Language 22: 611–631. Gilligan, G. M. 1987. A Cross-linguistic Approach to the pro-drop Parameter. Dissertation, University of Southern California. Haider, H. 1994. “(Un-)heimliche Subjekte — Anmerkungen zur Pro-drop Causa, im Anschluß an die Lektüre von Osvaldo Jaeggli and Kenneth J. Safir, eds., The Null Subject Parameter”. Linguistische Berichte 153: 372–385. Hamann, C. 1996. “Null Arguments in German Child Language”. Language Acquisition 5: 155–208. Heap, D. 2000. La variation grammaticale en géolinguistique: Les pronoms sujet en roman central. München: Lincom Europa. Hinzelin, M.-O. and Möhring, A. 2001. “The acquisition of subjects and verb–object order in bilingual children”. In Proceedings of the Colloquium on Structure, Acquisition, and Change of Grammars: Phonological and Syntactic Aspects. Volume II. University of Hamburg, 27th–29th October 2000, K. F. Cantone and M.-O. Hinzelin (eds), 1–25. Working Papers in Multilingualism Series B 27, Hamburg: Sonderforschungsbereich 538/Universität Hamburg. Hyams, N. M. 1986. “The AG/PRO parameter in early grammars”. In Language Acquisition and the Theory of Parameters, N. M. Hyams, 63–109. Dordrecht: Reidel.

135

136 Marc-Olivier Hinzelin

Jaeggli, O. and Safir, K. J. 1989. “The null subject parameter and parametric theory”. In The Null Subject Parameter, O. Jaeggli and K. J. Safir (eds), 1–44. Dordrecht: Kluwer. Juan-Garau, M. and Pérez-Vidal, C. 2000. “Subject realization in the syntactic development of a bilingual child”. Bilingualism: Language and Cognition 3 (3): 173–191. Kato, M. A. 1999. “Strong and weak pronominals in the null subject parameter”. Probus 11: 1–37. Kato, M. A. 2000. “The partial pro-drop nature and the restricted VS order in Brazilian Portuguese”. In Brazilian Portuguese and the Null Subject Parameter, M. A. Kato and E. V. Negrão (eds), 207–240. Frankfurt am Main: Vervuert. Kato, M. A. 2001. “Strong and weak pronominals in the history of Brazilian Portuguese grammar”. In Proceedings of the Colloquium on Structure, Acquisition, and Change of Grammars: Phonological and Syntactic Aspects. Volume II. University of Hamburg, 27th–29th October 2000, K. F. Cantone and M.-O. Hinzelin (eds), 26–37. Working Papers in Multilingualism Series B 27, Hamburg: Sonderforschungsbereich 538/Universität Hamburg. Köppe, R. 1994. “The DUFDE project”. In Bilingual First Language Acquisition. French and German Grammatical Development, J. M. Meisel (ed.), 15–27. Amsterdam: Benjamins. Longobardi, G. 2000. “‘Postverbal’ subjects and the Mapping Hypothesis”. Linguistic Inquiry 31 (4): 691–702. Meisel, J. M. 1989. “Early differentiation of languages in bilingual children”. In Bilingualism across the Lifespan. Aspects of Acquisition, Maturity, and Loss, K. Hyltenstam and L. K. Obler (eds), 13–40. Cambridge: Cambridge University Press. 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: Benjamins. Meisel, J. M. 1995. “Parameters in acquisition”. In The Handbook of Child Language, P. Fletcher and B. MacWhinney (eds), 10–35. Oxford: Blackwell. Meisel, J. M. 2001. “The simultaneous acquisition of two first languages: Early differentiation and subsequent development of grammars”. In Trends in Bilingual Acquisition, J. Cenoz and F. Genesee (eds), 11–41. Amsterdam: Benjamins. Meisel, J. M. and 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 Placement. Functional Categories and V2 Phenomena in Language Acquisition. J. M. Meisel (ed.), 109–138. Dordrecht: Kluwer. Mills, A.E. 1985. “The acquisition of German”. In The Crosslinguistic Study of Language Acquisition. Volume 1: The Data, D.I. Slobin (ed.), 141–254. Hillsdale, New Jersey: Erlbaum. Möhring, A. and Meisel, J. M. 2003. “The verb–object parameter in simultaneous and successive acquisition of bilingualism”. In this volume. Nagy, N. and Heap, D. 1998. “Francoprovençal null subjects and constraint interaction”. In CLS 34: The Panels. M. C. Gruber, D. Higgins, K. S. Olson and T. Wysocki (eds), 151–166. Chicago: Chicago Linguistic Society; 34.2. http://english-1.unh.edu/nagy/ papers/fp.cls.html



The acquisition of subjects in bilingual children

Nash, L. and Rouveret, A. 2002. “Cliticization as unselective attract”. Catalan Journal of Linguistics 1 (to appear). Noll, V. 1999. Das brasilianische Portugiesisch. Herausbildung und Kontraste. Heidelberg: Winter. Oliveira, M. de 1999. “O parâmetro do sujeito nulo na aquisição da linguagem”. Cadernos de Estudos Lingüísticos (Campinas) 36: 131–145. Oliveira, I. 2001. “A aquisição do sintagma flexional no português brasileiro”. Letras de Hoje (Porto Alegre) 36 (3): 675–681. Paradis, J. and Genesee, F. 1996. “Syntactic acquisition in bilingual children. Autonomous or interdependent?”. Studies in Second Language Acquisition 18 (1): 1–25. Paradis, J. and Genesee, F. 1997. “On continuity and the emergence of functional categories in bilingual first-language acquisition”. Language Acquisition 6 (2): 91–124. Perlmutter, D. M. 1971. “A typological difference among languages”. In Deep and Surface Structure Constraints in Syntax, D. M. Perlmutter, 99–122. New York: Holt, Rinehart and Winston. Pesetsky, D. and Torrego, E. 2001. “T-to-C movement: Causes and consequences”. In Ken Hale. A Life in Language, M. Kenstowicz (ed.), 355–426. Cambridge: The MIT Press. Radford, A. 2000. Children in Search of Perfection: Towards a Perfect Acquisition Model. Ms., University of Essex. Rizzi, L. 1986. “Null objects in Italian and the theory of pro”. Linguistic Inquiry 17: 501–557. Roberge, Y. 1986. “Subject doubling, free inversion, and null argument languages”. The Canadian Journal of Linguistics 31: 55–79. Say, T. 2001. Feature Acquisition in Bilingual Child Language Development. Working Papers in Multilingualism Series B 18. Hamburg: Sonderforschungsbereich 538/Universität Hamburg. Simões, L. J. 1999. “Sujeito nulo na aquisição do português do Brasil”. Cadernos de Estudos Lingüísticos (Campinas) 36: 105–130. Simões, L. J. 2000. “Null subjects in Brazilian Portuguese: Developmental data from a case study”. In Brazilian Portuguese and the Null Subject Parameter, M. A. Kato and E. V. Negrão (eds), 75–103. Frankfurt am Main: Vervuert. Soares, C. 2002. “The acquisition of D, T and C: Additional evidence”. In The Process of Language Acquisition, I. Lasser (ed.), 135–163. Frankfurt am Main: Peter Lang. Spitzer, L. 1917. “‘es’ im Portugiesischen”. Zeitschrift für romanische Philologie 38: 713–718. Valian, V. 1991. “Syntactic subjects in the early speech of American and Italian children”. Cognition 40: 21–81. Valian, V. and Eisenberg, Z. 1996. “The development of syntactic subjects in Portuguesespeaking children”. Journal of Child Language 23: 103–128. Valian, V., Hoeffner, J. and Aubry, S. 1996. “Young children’s imitation of sentence subjects: evidence of processing limitations”. Developmental Psychology 32: 153–164. Wanner, D. 1993. “L’expression du sujet dans les langues romanes”. In Actes du XXe Congrès International de Linguistique et Philologie Romanes. Université de Zurich (6–11 avril 1992). Tome III, Section IV — Typologie des langues romanes, G. Hilty (ed.), 447–460. Tübingen: Francke. Zubizarreta, M. L. 1998. Prosody, Focus, and Word Order. Cambridge: The MIT Press.

137



Approaches to bilingual acquisition data* Ira Gawlitzek-Maiwald University of Mannheim They [i.e. multilingual children] speak everything because they hear everything all the time. MP (4;2.15)1

1.

Introduction

Language mixing is a topic of great interest to many researchers. In “Bilingual Bootstrapping” (Gawlitzek-Maiwald and Tracy 1996), we argued in favour of a certain type of language mixing as being a sign of language competence rather than a deficit. In the meantime, several authors have discussed our idea and made other suggestions relating to mixing and interlanguage influence. The aim of this paper is to clarify some ideas related to the concept of Bilingual Bootstrapping and, in comparison with these other approaches to pin down a number of relevant questions which play a crucial role in the analysis of bilingual data. My discussion is based on utterances of the bilingual children in the Tübingen project (cf. Tracy 1995, Gawlitzek-Maiwald and Tracy 1996, Gawlitzek-Maiwald 1997, Gawlitzek-Maiwald 2000).2 In the following section observations that led to the suggestion of Bilingual Bootstrapping will be summarised, then in Section 3 several recent analyses of bilingual data will be briefly introduced; in Section 4 some more data from two more children in the Tübingen project is presented before I discuss several puzzles related to the recent analyses and the data in Section 5, and finally suggest some (new) questions which might be fruitful for future research. 2. Why bilingual bootstrapping? The analysis of the bilingual child Hannah3 revealed that she separates her two languages, German and English in many utterances, cf. (1) to (5), but she also produces sentence internal mixing of a special and systematic kind, cf. (6) to (8).4

140 Ira Gawlitzek-Maiwald

(1) mama schuhe probieren mummy shoes try-on ‘mummy, try these shoes on’

H 2;1

(2) ich des essen I this eat ‘I want to eat this’

H 2;2

(3) mama do it up

H 2;1

(4) mama put [d6] knife in cup

H 2;25

(5) mama sing it please

H 2;2

(6) [IPgerm ich habe [VPengl gemade you much better]]6 I have … ‘I’ve made you much better’

H 2;4

(7) [IPgerm kannst du [VPengl move a bit]] ‘can you …’

H 2;8

(8) warte doch for ich bin fertig mit diesem puzzle wait particle for I am ready with this puzzle ‘wait until I am ready with this puzzle’

H 2;7

Although in some recordings Hannah produces more than 20% mixed utterances, the vast majority of her structures are clearly either German (69%) or English (18%).7 We do not take the mixing as an indicator of a fused system. While Hannah’s “monolingual” utterances develop just as they would in monolingual children of the respective languages, for approximately five months we observe that a subset of her mixed utterances show a crucial systematicity. We interpret this as the result of a strategy where the slower language profits from the faster developing one. We named this phenomenon Bilingual Bootstrapping and suggested a stronger (booster function across languages) and a weaker (pooling of resources) version. With respect to Hannah, arguments for the stronger version can be put forward. To name just one: Hannah acquires the English IP (although after the German IP) earlier than monolingual English children. For Adam and Laura, the children to be discussed in Section 4, the (weaker) pooling interpretation of the term is more appropriate. This, however, does not seem problematical as it has become clear during the last decade of research that children (monolingual as well as bilingual) might follow different strategies in their acquisitional processes. So, in principle there is no reason why this should not be possible with respect to Bilingual Bootstrapping as well. Different children will make use of it to different degrees.

Approaches to bilingual acquisition data

We conceive of Bilingual Bootstrapping as the observable effect of the two (or more) separately stored language systems being activated simultaneously in a bilingual individual. Furthermore, Bilingual Bootstrapping works domain specifically, i.e. we do not assume that the dominant language provides the phrase structure and the weaker language the lexicon, it is rather that one particular domain or construction develops quicker in one of the languages than in the other. This probably happens because there is more and/or clearer evidence available for a certain construction in the faster developing language than in the other. In the meantime, finer phonetic analyses have been pursued, and they also point to the fact that we are not confronted with fused systems but rather with coactivation (especially) during crucial developmental phases. Gut (2000a, 2000b) analyzes the same children as we did in Gawlitzek-Maiwald and Tracy (1996) and Gawlitzek-Maiwald (1997). She shows that Hannah, Adam, and Laura produce longer pauses before switches and in mixes than in monolingual utterances. She argues that these mixes cannot be equated with adult’s pauses, which help to structure discourse, and thus interprets the children’s pauses as “direct reflections of the processing load in her [Laura’s] speech production.” (Gut 2000b: 215). Crucially, in Hannah’s data these pauses occur in mixed utterances at the VP boundary. As illustrated in (6) and (7) above, in Hannah’s utterances this is often the switch site as the functional level IP is usually German, while the lexical projection VP is realized in English. What is special about the Bootstrapping mixes? They are defined by four characteristics: 1. This mixing correlates with, and probably even depends on, asynchronic developments of the two languages. This asynchrony in itself constitutes an argument against a fused system; if it were only one system, we would not observe an asynchrony. 2. The mixing is highly systematic. 3. It is a temporary phenomenon, and 4. it can be related to structural characteristics of the different language systems. Hannah’s German develops faster than her English with respect to simple clauses, i.e. we find German modals, auxiliaries, verbal inflection, and V2 movement while there are hardly any examples of these features/categories in English.8 (9) ich mache das noch — so I make this still  so ‘I’m still doing it like this’

H 2;4

141

142 Ira Gawlitzek-Maiwald

(10) du hast du den mund aufgemacht you have you the mouth open-made ‘you’ve opened your mouth’

H 2;4

(11) jetzt kannst du hause gehen now can you home go ‘now you can go home’

H 2;4

(12) I can’t walk

H 2;4

(13) where-is-the tractor

H 2;8

(14) mama what the colour

H 2;8

(15) he want(s) to have a firestation

H 2;9

In (9)–(11) the use of target-like agreement marking, auxiliaries and modals is illustrated. Although one of the first English modals is documented in the same recording, (12), other indicators for the presence of an English IP appear later, i.e. wh-questions — some still without verb or inversion — (13) and (14), and inflection of main verbs, (15). A few months later Hannah’s English is ahead of German with respect to infinitival constructions. While she already produces English to-infinitives, cf. (16), any type of German infinitival construction is rare; most frequent at that time are the aspectual infinitives such as (17). When she produces mixed infinitival constructions between 2;3 and 2;9, their structure is English filled with German or cognate elements, cf. (18) and (19). (16) I want to look at the puppets

H 2;3

(17) mama komm mit spazieren mummy come with go-for-a-walk ‘mummy, come with me for a walk’

H 2;3

(18) die mama helf mir [strap it in] the mummy help me … ‘mummy help me …’

H 2;3

(19) willst du auch eins zu look at want you too one to … ‘do you want one to look at’

H 2;9

Usually the rate of mixing in the individual corpora does not exceed 5%. However, we observe two peaks around 2;3 with more than 20% mixing and around 2;9 when mixing reaches 10% in a corpus. It is precisely at these peaks that a subset of her mixing is highly systematic: it affects particular structures

Approaches to bilingual acquisition data 143

and elements, thus helping to fill structural gaps in the slower language. Cf. (6) and (7) above where the IP is German and embeds an English VP and (8), (18), and (19) above where the English infinitival structures are filled with German lexical items. A purely pragmatic explanation could not account for this systematicity and the restriction to particular structural domains. As soon as the necessary layers and/or elements are acquired for the language which develops more slowly with respect to this domain, these mixed forms cease. Thus, in addition, this type of mixing is clearly a temporary phenomenon.

3. Alternative approaches to bilingual bootstrapping Analysing three bilingual English-German children in Australia, who are dominant in English, Döpke (1998, 2000) argues that morphological mixing is bi-directional while lexeme-based mixing is basically unidirectional. Döpke discusses in some detail German [S Aux V XP]-structures (some with finiteness marking on V) such as er kann nicht kommt rein (he can not comes in ‘he can’t come in’). She argues that these are due to the combined input of German and English which results in a higher cue strength for V XP structures. These structures are far from marginal as all children in her project produce them, one as often as 66% at one stage. Döpke’s assumption (2000: 99) is that these bilingual children have not yet acquired the hierarchical structure of their languages and thus rely on the combination of linear patterns, which are supported by one or both languages. These combinations might lead to ungrammatical structures as in the example er kann nicht kommt rein. However, alternative explanations in the line of Bilingual Bootstrapping are conceivable. These utterances can be analysed as mixed, i.e. they would consist of English structures with a head initial VP, which are filled with German lexical items. This leaves us to account for the inflection on some of the main verbs in these structures. These could, for example, be double markings of tense which have been documented for monolingual German as well as English children.9 Although Müller (1998) and Müller and Hulk (2001) do not discuss mixed utterances, it is fruitful to compare their analyses with the Bilingual Bootstrapping approach as some of the utterances we classified as mixed are not mixed on the lexical level (cf. (32), (34), (42) below), thus an alternative analysis cannot be ruled out from the onset. Müller (1998) analyses the acquisition of German word order in the

144 Ira Gawlitzek-Maiwald

bilingual (German/French) child Ivar. Ivar produces deviant word orders in German subordinate clauses, cf. (20). The same types of word order deviances are also found in some monolingual children, cf. (21). (20) wenn da kommt  andere schiffe … if there comes (*agr) other ships  ‘if there are other ships coming’ (21) will die meerjungfrau haben want the mermaid have daß du hast net die meerjungfrau gell that you have not the mermaid TAG ‘I want to have the mermaid so that you don’t have her’

Ivar 3;5.7

Benny 3;0.1910

Ivar’s acquisition of target-like CP takes a long time, longer than in monolingual children with deviant word order; Müller argues that this is a quantitative not a qualitative difference to monolingual acquisition. Müller interprets these findings as an instance of transfer from Ivar’s French. She assumes transfer to be a relief strategy which bilingual children might make use of if confronted with ambiguous input in one of their languages, e.g. SVO and SOV order in German subordinate clauses. If one of the options is supported by their other language, bilingual children might erroneously assume this structure for a while. Thus, transfer is supposed to be unidirectional, the target being always the ambiguous language. Müller and Hulk (2001) investigate the acquisition of object drop in three bilingual children (French/German, French/Dutch, Italian/German). In this study they concentrate on the children’s Romance languages. They find that much in the vein of Müller (1998), the bilingual children differ quantitatively but not qualitatively from monolingual children of the respective languages. The bilingual children drop more objects illegitimately than monolingual children of the same Romance languages. The children go through two phases of acquisition; the first is characterised by the gradual emergence of object clitics, frequent dropping of objects, and the absence of a C-system. During the second phase these features eventually disappear and approach target-like levels. However, the children observed by Müller and Hulk develop both their languages in parallel, not as the children in the Tübingen project in an asynchronic way. Müller and Hulk follow a claim by Roeper (1999), who argues that all children are multilingual, namely children who grow up with one mother tongue are internally so, and children with two or more mother tongues are

Approaches to bilingual acquisition data

externally so. External bilingualism slows the acquisition down as the mapping of Minimal Default Grammars onto target grammars takes longer because the children have to sort out more conflicting evidence in the input of two or more languages. Just as Gawlitzek-Maiwald and Tracy (1996) did with respect to Bilingual Bootstrapping, Müller and Hulk claim that crosslinguistic influence is related to the grammatical phenomena discussed and not to language dominance. Müller and Hulk maintain that dominance, as defined by the mean length of utterance, does not predict the direction of the influence, but rather that these bilingual children are object to a domain-specific cross-linguistic influence, which the authors name “mapping induced influence”, i.e. the children have more difficulties than monolingual children mapping pragmatic onto syntactic principles. This is supposed to be particularly so for the C-system. According to Müller and Hulk the structures in both languages develop in parallel. Schlyter and Bernardini (2000) rely on the metaphor of ivy for bilingual acquisition. They claim that their approach is based on Bilingual Bootstrapping to a large extent. Analysing code mixing data from one Swedish-Italian and three Swedish-French bilinguals, the authors propose three types of child bilingualism which will exhibit different contact phenomena: – –

–

For balanced bilinguals, bi-directional switching is expected. For slightly unbalanced bilinguals, unidirectional switching is expected where during a “weak period” functional morphemes of the dominant language are combined with lexical categories from the weak one. Finally, heavily unbalanced bilinguals are expected to produce mixing where the functional projections come from the strong and lexical projections from the weak language.

According to Schlyter and Bernardini, the weak language will eventually “climb up” the syntactic structures of the strong language in ivy fashion. In their analysis, the acquisition of CP is crucial. As soon as a child acquires the CP also in the weaker language the “Ivy mixing” ceases. However, there is a crucial difference between Bilingual Bootstrapping and “Ivy mixing”. Bilingual Bootstrapping applies specifically to certain domains whereas the authors seem to consider the Ivy Hypothesis to be a general strategy where the stronger language supplies the structure and lexical items are filled in from the weaker language. Supposedly this type of mixing only ceases once the CP is also acquired in the weaker language.

145

146 Ira Gawlitzek-Maiwald

4. More evidence in favour of bilingual bootstrapping In this section, I am going to present some data from two more children of the Tübingen project in order to illustrate that the specific structural mixing which we characterised for Hannah as Bilingual Bootstrapping can also be found in other children. Adam and Laura are siblings. Their mother is German; their father English. Their parents follow the one-parent-one-language strategy. Adam and Laura were audio- or videotaped approximately fortnightly from 3;6 to 5;5 and from 2;5 to 4;2 respectively.11 The following examples illustrate that Adam and Laura are capable of producing purely German, cf. (22) to (25), and purely English utterances, cf. (26) to (29). As with Hannah, Laura’s and Adam’s acquisition of English and German is comparable to the acquisition of monolingual children of these languages. (22) ich kann nicht alleine machen I can not on-my-own make ‘I can’t do it on my own’

A 3;7

(23) des hat die laura gemacht this has the Laura made ‘Laura did it’

A 3;7

(24) ich hol ein anderes buch I fetch an other book

L 2;5

(25) des darf ich haben this may I have ‘I may have this’

L 2;5

(26) he doing up the engine

A 3;7

(27) make tea and milk

A 3;7

(28) WHAT baby dog doing

L 2;6

(29) I mix it now

L 2;9

In the following the mixed utterances will be discussed in some detail. Quantitative overviews and selected examples will serve to illustrate that Adam and Laura also produce the type of mixing we classify as Bilingual Bootstrapping. The graph in Figure 1 summarises the percentage of mixing in Adam’s corpora.12 Two aspects are striking: the rate of mixing depends to a considerable degree on the language of the interlocutor. In the German context, mixing never exceeds 5%. In the English context, we witness a general slow decrease of

Mixing in %

Approaches to bilingual acquisition data 147

14

14 12 10 8 6 4 2 0

11 9

8

8

6 5

3 0 1 3;7

3;10

4;2

3

2

1 English interlocutor German interlocutor

4;6

5;2

Age

Figure 1.Adam’s mixing rate.

the mixing rate from the beginning of recordings to the end.13 However, after the beginning of recordings, a second peak of mixing is reached with 9% around age 4;4. This peak coincides with the acquisition of the German CP. The following graph (30) summarises at what ages Adam acquires major structural layers in his two languages. Note again the asynchronic development of German and English, which is also illustrated by the examples (22)–(23) versus (26)–(27) above from the same recording date. Higher mixing rates occur after the German IP is acquired and co-occur with first CP precursors in German (3;6–3;10) and then again with the ultimate acquisition of the German CP (around 4;4). (30) Overview of Adam’s acquisition Mixing Germ. Precursors IP German CP acquired

3;6

3;10

Mixing Germ. CP acquired English IP

English CP

4;4

5;4

4;6

Criteria for the acquisition of a target-like IP in English are productive usage of auxiliaries, modals, and verbal inflection; for an English CP, the usage of inversion and wh-phrases in questions and subordinated clauses introduced by a complementizer. For a target-like German IP, the criteria are productive V2

148 Ira Gawlitzek-Maiwald

movement, Subject–Verb-agreement, and the occurrence of non-subject constituents in sentence initial position; for a German CP, a range of complementizer-introduced clauses with target-like word order and semantics. The examples in (31) to (34) illustrate mixes of the subset which is crucial for our discussion here. Although the percentage of the bootstrapping mixes in Adam’s data are smaller than in Hannah’s, they point in the same direction. Again we observe a structural “job sharing”. VPs are English or German, some with additional morphological mixing, while IPs are only German. (31) aber ich habe gethinke des ‘but I have thought this’ (32) I want a chimney builden I want a chimney build+inf (33) they didn’t got — they didn’t got hands zu drawing and they didn’t got feets zu zu — ‘they didn’t have hands for drawing and they didn’t have feet for for —’ (34) [wεn] she green is she gets sick ‘if she’s green she gets sick’

A 3;7 A 3;1014

A 4;4

A 4;6

On first sight (31) looks like a simple example of borrowing, but a closer look reveals that the structure of the VP is also English as the verb precedes its complement; the IP layer and verbal morphology in this example are German. This is the kind of analysis we have suggested in Gawlitzek-Maiwald and Tracy (1996). Meisel (2000) criticises it, however, his criticism is based on his assumptions that the German IP is also head final. Following Brandt et al. (1992), I assume that German main clauses are IPs with the head I preceding VP, thus in my analysis, finite verbs in German main clauses move to I, not to C, as Meisel (2000) assumes. The infinitival construction in (32) is highly complex: the main clause could be English or German, the VP is German, all lexical items are English, but want is subcategorised as German wollen, thus no to-infinitive is embedded and the bare infinitive of build is marked with the German infinitive ending -en. Adam produced (33) at a time when he had almost acquired English control constructions but produced only precursors to German zu- and no umzu-infinitives, thus he attempts to produce an English complement for zu, which does not really work and breaks off. In example (34), we witness German syntax of a subordinate clause with the verb in last position, this clause being introduced by an English lexeme which is cognate to a German complementizer.

Approaches to bilingual acquisition data 149

All these utterances illustrate that mixing can function as a relief strategy, in the sense of pooling the resources available from both languages. In the following paragraphs it will be argued that Laura, Adam’s younger sister, produces utterances which can be analysed in the same way. 24

25 21 20

17

15

13

10 5 0

6

6 3

2 2

2;5 2;7 2;9

7

1 2

3;1

Age

6 6

8 9 9

0

3;3

3;7

English interlocutor German interlocutor 4;1

Figure 2.Laura’s mixing rate.

The graph in Figure (2) shows that as with Adam, Laura’s mixing rate in the English context is high and rather low in the German context.15 Also the kind of mixed language used is different. When speaking mainly German, almost all influences from English can be classified as borrowings of single lexical items as in (35) and (36). When Laura interacts with an English speaking partner, we also find grey zone mixes as in (37), where English and German structure and/or the lexemes are similar, and structural and morphological mixes as in (38) to (43). (35) ich kann des pig baden I can this pig bathe ‘I can bathe this pig’

L 2;9

(36) puddle lane is da ‘puddle lane is there’

L 2;8

(37) [hi6 Is dfr6sI] ‘hier ist Dorothy’ or ‘here is Dorothy’

L 2;9

(38) nein du lookst des buch no you look(+agr) this book ‘no, you’re going to look at this book’

L 2;6

150 Ira Gawlitzek-Maiwald

(39) und ich will make a little one and I will ‘and I want to make a little one and I want’

L 2;9

(40) ich will spiel mit den TIEre I will play with the animals ‘I want to play with the animals’

L 2;9

(41) and then go we gell ‘and then we’re going, o.k.?’

L 3;2

(42) got you TOO a bicycle ‘do you also have a bicycle?’

L 3;2

(43) for dass des nich runterfällt for that this not down-falls ‘so that this does not fall down’

L 3;10

(38) is in fact an ambiguous example; as the English main verb has moved to second position, we do not know whether in D-structure it precedes or follows its complement, but even if it is “only” another instance of borrowing, it is noteworthy as the verb carries appropriate German agreement marking. (39) and (40) show the same syntactic structure, namely an English VP under a German IP. Note that the pronunciation of will is German for these two and all other occurrences of it in this sample. In (39) lexical items come from the same language as the different structural layers, while (40) consists only of German lexemes, but note that the verb appears only as a stem, as an English verb would. Still, this does not provide an argument for a fused system, as the vast majority of Laura’s utterances are either clearly German or English16 and follow the syntactic and morphological rules of the respective language. These morphological adaptations occur only in mixed utterances. In (41) and (42) the opposite happens; the utterances contain only English lexemes, however, they appear in structures which are clearly German. In (43), documented later, again a mixed infinitival construction occurs. An English infinitive complementizer introduces the utterance and is followed by a German finite structure, which is semantically appropriate. Apart from the high mixing in the first few recordings, two peaks of mixing can be identified around 2;9 and 3;0. Graph (44) sets these mixing peaks in relation to Laura’s acquisition of syntactic milestones in English and German. Again, it becomes clear that the mixing peaks occur at times when one of the two languages develops. During the first peak, Laura starts producing precursors to the English IP. During the second peak, the German CP is acquired.

Approaches to bilingual acquisition data

(44) Laura: Mixing

Mixing

Germ. Precursors IP Germ CP acquired precursors to EnglishIP

Germ. CP acquired

2;8

3;0

2;9

English & precursors IP of Engl. CP acquired

3;10

Note that in all examples discussed above the exact outcome of the mix depends on the construction concerned and the development of the languages concerned. Adam’s and Laura’s Bilingual Bootstrapping mixes differ from Hannah’s in that they are not as numerous, and for Adam and Laura we cannot argue that they acquire particular structures more quickly than the respective monolingual children. However, the weaker interpretation of Bilingual Bootstrapping, i.e. a pooling of resources across languages, can be traced in their data; namely that they manage to produce longer and more complex utterances by pooling their resources from both languages than they could just on the basis of their English or German.

5. Puzzles When working with bilingual data — in whatever theoretical framework — one is confronted with a number of conceptual puzzles. Those puzzles, which play a role in Bilingual Bootstrapping and the studies surveyed in Section 3, will be discussed here. The question of which of the two languages acquired is dominant in a child plays a role in the work of Schlyter and Bernardini (2000). Intuitively, it makes sense to talk about bilinguals being balanced or dominant in one of their languages, however, making the criteria for dominance explicit is a challenging task: What defines dominance? The majority of utterances, personal preference of the bilingual, proficiency — if so, how is it measured?, opportunities to speak? Not always is the definition of dominance made clear. Probably the most promising suggestion comes from Genesee et al. (1995), who use a number of criteria together to determine the possible dominance in bilingual speakers, i.e.

151

152

Ira Gawlitzek-Maiwald

MLU, longest utterance, percentage of multi-morphemic utterances, and number of word types in each language. However, according to most of these criteria Adam would not be classified as German dominant, which is clearly counter-intuitive. His and also Laura’s and Hannah’s dominance of German is established by a faster development of their German system. Thus I would claim that only a detailed system analysis and comparison reveals the dominant language. This might often correlate with quantitative differences, however, this need not be the case, as Adam illustrates.17 Many researchers — among them all authors discussed here — stress that bilinguals separate their languages from early on and that their language developments can be compared to monolingual acquisition of the same languages. If bilinguals produce deviant structures in their “monolingual” utterances, these fall within the range of deviant structures as we find them in monolingual children. On the one hand, this is a welcome finding because it stresses the normality of bilingualism, i.e. the fact that bilingualism cannot be made responsible for all deviances in language acquisition. On the other hand, it weakens the idea of transfer as proposed by Müller (1998) or the importance of cue strength as discussed by Döpke (1998). Why should we rely on the structures of the other language if children who do not acquire a second language produce the same range of (deviant) structures?18 Here is one more problem for the transfer approach in particular: The idea of ambiguities (or rather surface variation) in one language and the reinforcement of one of these works for the V2/VE variation in German subordinate clauses. However, it does not explain other constructions where we witness that children — for a limited time — find special mixing solutions, such as infinitival constructions. Real ambiguities exist for these in the target systems, i.e. different underlying structures, but the children mix against the expected direction of transfer: In English, surface sequences of the type [DP V to VP] are to be analysed as either raising structures (It seems to be raining again) or as control structures (She wants to win the race) depending on the main verb. Surface sequences of the type [DP V DP to VP] are either control structures with a ditransitive main verb (She persuaded her to take part) or exceptional case marking structures (ECM) (She expected him to come early). Thus to-infinitives in English realise three types of constructions: raising, control and ECM. In German on the other hand, zu-infinitives can only be raising or control structures. According to Müller and Hulk (2001), we would expect that children overgeneralize German structures because these are less ambiguous, however, the opposite is true. English infinitival constructions are developmentally ahead

Approaches to bilingual acquisition data

of German infinitival constructions. Thus the transfer approach is not borne out for these constructions in German/English bilingual children. However, Bilingual Bootstrapping helps to explain overgeneralizations and mixes in this domain, as the underlying syntactic structure of English infinitival constructions is easier, i.e. parallel to main clause IPs and CPs, while for German, there is disagreement among linguists as to the internal structure of German infinitival constructions (cf. Gawlitzek-Maiwald 1997: chap. 2). Another question which comes to mind is the following: What happens with structures where there is no real subset relation of the analyses of the languages concerned? Do children not show any evidence of inter-language influence and mixing in these areas? If they do, this is difficult to account for in the transfer approach. Bilingual Bootstrapping also faces a serious problem. If one takes the idea seriously that lexical items project syntactic structure, how then can German lexical items project an English structure or vice versa? At the moment, I have no conclusive answer to this. The last aspect that needs to be discussed in this section is the method of analysis of bilingual data. Genesee et al. (1995) strongly argue for a quantitative analysis as opposed to a qualitative one, based on the discussion of a few examples.19 In my opinion, both are necessary. This will be supported by two examples: Döpke’s (1988) analysis is mainly quantitatively oriented. In her tables, fairly large numbers of “residual” examples are not included. These are not further explained, and there are no examples provided. Thus the reader is left to wonder what the actual character and status of these utterances is. If the author had quoted a few of these utterances, the reader would not ponder whether these “residuals” might be either mixed or else highly relevant for the discussion in other ways. In Schlyter and Bernardini (2000) the “opposite” problem occurs. The authors mainly discuss individual examples, however, the reader neither knows enough about the children’s overall systems nor about the authors’ assumptions, so that some of the analysis remains unclear. For example, utterances such as (45) and (46) below are treated as instances of structural mixing, while (47) is taken to be an instance of borrowing (bold face marks the supposedly mixed elements). (45) so noch ein maison such again a house ‘once more such a house’

153

154

Ira Gawlitzek-Maiwald

(46) jag tycker om coquillage I think that mussel ‘I like mussel’ (47) on va manger des semlor one go eat (indefinite plural article) Swedish pastry ‘we’re going to eat pastry’ (cf. Schlyter and Bernardini 2000: 3f)

Schlyter and Bernardini argue that in (45) the child mixes the functional elements from his stronger language into the weaker language of the conversation. However, could it not be the case that the child defines the base language differently and borrows from French? Within the Bilingual Bootstrapping approach, this would not pose a problem as the base and dominant language need not be defined for the analysis, what matters is faster development with respect to a particular domain or construction. The analysis of (46) is a little more convincing as the authors state that it originates from a “period where his French grammar is not yet very developed” (Schlyter and Bernardini 2000: 3); when (47) is produced, apparently the boy’s French system includes inflection and complementizers. Strictly speaking, the Ivy Hypothesis can only be tested and differentiated from borrowing when there are at least two items of each language in an utterance and their syntax can be analyzed. Perhaps data such as the hypothetical example He comes kommt noch nicht (he comes comes yet not, ‘he doesn’t come yet’) might count as convincing evidence in favour of the Ivy hypothesis, as the double realization of the inflected verb might indicate that English indeed “climbs up” the German IP. These last observations really call for a system analysis which includes quantitative as well as qualitative information.

6. In place of a summary: New questions One theme came up in the discussion of different analyses of bilingual children’s language acquisition in Section 3, as well as in the puzzles in the last section, namely that bilingualism is not much different from monolingualism. Strictly speaking, monolingualism is a fiction (Tracy and Gawlitzek-Maiwald 2000, Tracy 2000). It is not only bilingual speakers who cope with several coexisting linguistic systems; this is also true for monolingual speakers who

Approaches to bilingual acquisition data

choose between different registers as needed, to name just one dimension of variation. Mixed utterances are just one more type of production where two or more structures are coactivated at the same time. Psycholinguistic research shows that the other structure/register/language is never turned off completely (cf. Grosjean 1995, Grosjean and Soares 1986). On the contrary, coactivation and competition are the norm. So instead of asking which elements are inserted or mixed into a host language it is more interesting to ask what structures of systems are coactivated, also in monolingual speakers, cf. (48) to (50). (48) Hohenschwanstein Hohenschwangau & Neuschwanstein (49) a.

was du machen? what you make(-fin) ‘What are you doing?’ b. willst du essen? want you eat ‘Do you want to eat?’ c. warum der auch ein keks ißt? why he too a biscuit eats ‘Why is he too eating a biscuit?’

(50) oh hab-ich ein PFLASter [h6d6] mami geSCHENK oh have-I a bandaid [?has] mummy given ‘Oh I have a bandaid that mummy has given to me’

adult B 2;11

B 2;11

B 2;1120

Julia 2;221

Example (48) is an adult slip of the tongue, more precisely a blend of the names of two famous Bavarian castles, i.e. two words are activated at the same time and blended into one expression. The utterances in (49) were produced by the monolingual boy Benny in one recording. They are all direct questions. Note the coexistence of different verb positions and target-like or deviant inflection. (50) is produced by a monolingual girl. Tracy (1991) argues that in the girl’s system the clearly articulated form hab-ich (have I) and the contracted form [h6d6] coexisted for some time, however, they fulfilled different functions. Eventually they were reanalysed by the child and converged to the target form covering all the formerly separate functions. These examples suggest that what Clyne (1987) calls “grey zones” and Müller and Hulk (2001) analyze as ambiguities support the development of coexisting mini-systems (even in monolingual children) which only eventually converge to adult-like structures. In the light of these observations and findings, it seems only natural that parallel structures strengthen each other more than

155



156 Ira Gawlitzek-Maiwald

divergent ones, and not only in bilinguals. This, for example, is also true for the development of German subordinate clauses, cf. Fritzenschaft et al. (1990) and Gawlitzek-Maiwald et al. (1992). Instead of asking whether bilingual children keep their languages apart, it now seems more interesting to ask what makes systems in language learners converge to one or two. After all, if this convergence fails, children do not develop adult-like systems and become language impaired. With respect to Bilingual Bootstrapping, we might raise the question again which Gawlitzek-Maiwald and Tracy asked some years ago: What exactly is it that children don’t know if they know exactly what to substitute from their other language?

Notes * I would like to thank friends and colleagues who read earlier versions of this paper for valuable comments, in particular: Natascha Müller, Rosemarie Tracy, Nichola Hayton, and two anonymous reviewers. I feel honoured to receive such detailed feedback. 1. This is the idea my monolingual German son had about multilingual language acquisition when I told him that I was working on this paper. 2. For a more detailed picture of the overall project, which was funded by the German Research Foundation (DFG, reference no. Tracy 238/1–3., cf. Fritzenschaft et al. (1990). The overall corpus consists of app. 120,000 utterances from six monolingual German and six bilingual German/English children. Here I am going to discuss the data of three bilingual children: Hannah (cf. Gawlitzek-Maiwald and Tracy 1996, Gawlitzek-Maiwald 1997, Tracy 1995), Adam (cf. Gawlitzek-Maiwald 1997, Tracy 1995), and Laura (cf. Gawlitzek-Maiwald 1997). 3. Hannah’s mother is English and her father German. The family changed from the oneparent-one-language strategy to English as family language, when Hannah was 1;7 as the parents thought that English received too little support. 4. Meisel (2000) also discusses the question of whether children who acquire two (or more) first languages simultaneously differentiate their languages from early on or whether a fused system might exist. After reviewing several studies, Meisel comes to the conclusion that “the Autonomous Development is to be preferred” (Meisel 2000: 27). 5. Note that the element [d6] is of ambiguous status. It could be either German or English, however, at this age Hannah systematically replaced target-like the with [d6], cf. Tracy (1995: 239). 6. Note that the participial prefix is German, however, the verbal stem and the word order within the VP are English. 7. These figures do not add up to 100% as two categories are omitted here: utterances which could belong to either language, like vocatives, and unidentifiable utterances. These rough

Approaches to bilingual acquisition data

figures are calculated across all coded data, i.e. a total of 5637 utterances produced between 2;1 and 4;3. The overall mixing rate calculated on all these utterances is 6%. For detailed tables please refer to Tracy (1995) or Gawlitzek-Maiwald (1997). 8. Paradis and Genesee (1997) also observe asynchronies in their bilingual French/English subjects. The children produce DPs at about the same time in both languages, while French IPs are produced considerably earlier and with higher frequency than English IPs. The overall rate of mixing in Paradis and Genesee’s data is 5.8% for one child and 6% for the other. This is similar to Hannah’s overall mixing rate. However, Paradis and Genesee do not find mixes with verbal inflection or auxiliaries. A possible reason for this could be that the children’s overall MLU never reached 2.0 in their sample, while Hannah’s MLU (in words) is 2.1 to 3.5 during the phase discussed. 9. Note that, as far as I can judge, Müller’s transfer approach (to be discussed below) could not account for Döpke’s data, as the head initial VP is not an target option in adult German. 10. For a detailed discussion of Benny’s subordinate clauses cf. Fritzenschaft et al. (1990) and Gawlitzek-Maiwald et al. (1992). 11. For more details on the data collection, analyses and the background of Adam and Laura, cf. Gawlitzek-Maiwald (1997 chap. 7). 12. Adam’s overall mixing rate — calculated for all coded protocols with 6167 utterances — is 6%. 13. Note that the mixing does not cease completely, however, we do not really expect this as Adam is a bilingual speaker who lives in a home where there is an activation potential for both languages. However, the type of Bilingual Bootstrapping mix ceases completely. 14. This utterance is repeated as (i) (i) my now\ I want a chimney builden now\ where Adam starts with a possessive pronoun and an adverb, then breaks off, repeats the previous utterance and adds the adverb at the end again, which is an English adverb position, just as the pre-subject position which he started off with. 15. For Laura, the overall mixing rate — calculated for all coded protocols with 5178 utterances — is 8%. Again, as for Adam, we do not expect the mixing to cease completely, cf. Footnote 13. 16. To name just one characteristic example: At age 2;9 Laura produces 212 utterances with verbs; 15% of these are mixed, 80% German, 5% English. 17. For a more detailed discussion of Adam’s language dominance cf. Tracy and GawlitzekMaiwald (in prep.). 18. See also the critical discussion of the transfer approach in Tracy (1998). 19. This controversy also plays a role in monolingual language acquisition research, cf. de Villiers (1992). 20. From Gawlitzek-Maiwald et al. (1992: 145). 21. From Tracy (1991: 221), cf. also the discussion on p. 423ff. Capitals indicate major stress.

157

158

Ira Gawlitzek-Maiwald

References Brandt, M., Reis, M., Rosengren, I. and Zimmermann, I. 1992: “Satztyp, Satzmodus und Illokution”. In Satz und Illokution 1, I. Rosengren (ed.), 1–90. Tübingen: Niemeyer. Clyne, M. 1987. “Constraints on Code-Switching: How Universal are they?” Linguistics 25: 739–764. de Villiers, J. 1992. “On the Acquisition of Functional Categories: A General Commentary”. In The Acquisition of Verb Placement: Functional Categories and V2 Phenomena in Language Acquisition, J. M. Meisel (ed.), 423–442. Dordrecht: Kluwer. Döpke, S. 1998. “Competing Language Structures: The Acquisition of Verb Placement by Bilingual German-English Children”. Journal of Child Language 25: 555–584. Döpke, S. 2000. “The Interplay Between Language-Specific Development and Crosslinguistic Influence”. In Crosslinguistic Structures in Simultaneous Bilingualism, S. Döpke (ed.), 79–103. Amsterdam/Philadelphia: John Benjamins. Fritzenschaft, A., Gawlitzek-Maiwald, I., Tracy, R. and Wagner, S. 1990. “Wege zur komplexen Syntax”. Zeitschrift für Sprachwissenschaft 9: 52–134. Gawlitzek-Maiwald, I. 1997. Der monolinguale und bilinguale Erwerb von Infinitivkonstruktionen: Ein Vergleich von Deutsch und Englisch. Tübingen: Niemeyer. Gawlitzek-Maiwald, I. 2000. “I want a chimney builden: The Acquisition of Infinitival Constructions in Bilingual Children”. In Crosslinguistic Structures in Simultaneous Bilingualism, S. Döpke (ed.), 123–148. Amsterdam/Philadelphia: John Benjamins. Gawlitzek-Maiwald, I., Tracy, R. and 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. M. Meisel (ed.), 139–180. Dordrecht: Kluwer. Gawlitzek-Maiwald, I. and Tracy, R. 1996. “Bilingual bootstrapping”. In Two languages: Studies in bilingual first and second language development, N. Müller (ed.). [Linguistics 34], 901–926. Genesee, F. Nicoladis, E. and Paradis, J. 1995. “Language Differentiation in Early Bilingual Development”. Journal of Child Language 22: 611–631. Grosjean, F. 1995. “A Psycholinguistic Approach to Code-Switching: The Recognition of Guest Words by Bilinguals”. In One Speaker, Two Languages: Cross-Linguistic Perspectives on Code-Switching. L. Milroy and P. Muysken (eds), 177–198. Cambridge: Cambridge University Press. Grosjean, F. and C. Soares. 1986. “Processing Mixed Language: Some Preliminary Findings”. In Language Processing in Bilinguals: Psycholinguistic and Neuropsychological Perspectives, J. Vaid (ed.), 145–179. Hillsdale, NJ: Lawrence Erlbaum. Gut, U. 2000a. Bilingual Acquisition of Intonation: A Study of Children Speaking German and English. Tübingen: Niemeyer. Gut, U. 2000b. “Cross-Linguistic Structures in the Acquisition of Intonational Phonology by German-English Bilingual Children”. In Crosslinguistic Structures in Simultaneous Bilingualism, S. Döpke (ed.), 201–225. Amsterdam/Philadelphia: John Benjamins. Meisel, J. M. 2000. “The Simultaneous Acquisition of Two First Languages: Early Differentiation and Subsequent Development of Grammars”. In Trends in Bilingual Acquisition, J. Cenoz and F. Genesee (eds), 11–41. Amsterdam/Philadelphia: John Benjamins.



Approaches to bilingual acquisition data 159

Müller, N. 1998. “Transfer in bilingual first language acquisition”. Bilingualism: Language and Cognition 1 (3): 151–171. Müller, N. and Hulk, A. 2001. “Crosslinguistic Influence in Bilingual Language Acquisition: Italian and French as Recipient Languages”. Bilingualism: Language and Cognition 4(1): 1–21. Paradis, J. and Genesee, F. 1997. “On Continuity and the Emergence of Functional Categories in Bilingual First-Language Acquisition”. Language and Acquisition: A Journal of Developmental Linguistics 6, 2: 91–124. Roeper, T. 1999. “Universal Bilingualism”. Bilingualism: Language and Cognition 2 (3): 169–186. Schlyter, S. and Bernardini, P. 2000. “Climbing up the Tree of the Dominant Language — on the Role of Code Mixing for the Development of the Weaker Language in Young Bilinguals”. Ms., University of Lund. Tracy, R. 1991. Sprachliche Strukturentwicklung: Linguistische und kognitionspsychologische Aspekte einer Theorie des Erstspracherwerbs. Tübingen: Narr. Tracy, R. 1995. “Child languages in contact: The simultaneous acquisition of two languages (English/German) in early childhood”. Habilitationsschrift, University of Tübingen. Tracy, R. 1998. “Transfer versus coexisting systems”. Bilingualism: Language and Cognition 1 (3): 185–186. Tracy, R. 2000. “Language Mixing as a Challenge for Linguistics”. In Crosslinguistic Structures in Simultaneous Bilingualism. S. Döpke (ed.), 11–36. Amsterdam/Philadelphia: John Benjamins. Tracy, R. and Gawlitzek-Maiwald, I. 2000. “Bilingualismus in der frühen Kindheit”. In Enzyklopädie der Psychologie 3: Sprachentwicklung, H. Grimm (ed.), 495–535. Göttingen: Hogrefe Verlag für Psychologie. Tracy, R. and Gawlitzek-Maiwald, I. (in prep.) “The Strength of the Weak: Asynchronies in the Simultaneous Acquisition of German and English”.



Factors accounting for code-mixing in an early developing bilingual* Margaret Deuchar and Rachel Muntz University of Wales, Bangor

1.

Introduction

Recent work has established that young developing bilinguals begin to use their two languages appropriately at a young age, generally before the age of two. However, as we shall see, the establishment of appropriate language use is a gradual process, and it is thought that its course may be influenced by various factors. In this paper we shall examine the impact of three specific types of factor which have been frequently discussed: the interaction between the adult and the child, the nature of the adult input, and the child’s relative proficiency in each language, or relative dominance. We shall consider which of these, if any, appears to influence the development of language choice in a case study of a child acquiring English and Spanish from birth.

2. Previous work on language choice in case study Deuchar and Quay (1999, 2000) established that Deuchar’s daughter, who was acquiring English and Spanish from birth, had some notion of appropriate language choice by age 1;7, when her MLU in both languages was 1.06. Audiovideo recordings in separate contexts (being addressed by an English-speaking interlocutor in one context and a Spanish-speaking interlocutor in the other) showed that her one-word utterances at ages 1;7 and 1;8 were more likely to be in English in the English context, and Spanish in the Spanish context. Table 1 is a summary of the results reported in Deuchar and Quay (2000: 102, Figure 6.3): the difference between the use of English and Spanish words in the two

162 Margaret Deuchar and Rachel Muntz

Table 1.Number of child’s one-word utterances at 1;7–1;8 Language of one-word utterances

English words

Spanish words

English context Spanish context

64 37

45 132

contexts was significant (chi square value 38.76, p < 0.001). English words tended to be selected to match the English context, whereas Spanish words tended to be selected to match the Spanish context. However, Deuchar and Quay (2000: 101) noted that “the pattern of matching the language to the context was stronger in the case of the Spanish context”. The proportion of language-appropriate words in the Spanish context was 78.11%, whereas in the English context it was 58.72%. We suggested that this preponderance of Spanish might be due to the fact that both the English-context and Spanish-context recordings were made in the home, where Spanish was normally spoken by both parents to the child. English was only spoken in the home in the presence of the monolingual grandmother, who was recorded interacting with the child during her weekly visit to the home. Thus the fact that Spanish was normally spoken in the home might have influenced the child’s productions even when she was being addressed in English. However, we left open the possibility that the communicative strategies used by the adults when interacting with the child might have an effect on the child’s language choice.

3. Parental Discourse Hypothesis (PDH) 3.1 Introduction to the PDH The PDH (labelled as such by Nicoladis and Genesee 1998) is due to Lanza (1992, 1997), who suggested that a developing bilingual’s language choice might be affected by the strategies used by her interlocutors in response to instances of “lexical mixing”, by which she means the production of content words in a language which is not that of the interlocutor, and which by adult standards is inappropriate for the context. On the basis of a study of Siri, a child acquiring Norwegian and English simultaneously, Lanza argued that the child’s greater use of mixing with her Norwegian-speaking father than with her Englishspeaking mother could be explained by the father’s greater willingness to accept utterances in the inappropriate language. When Siri used Norwegian utterances

Factors accounting for code-mixing in an early developing bilingual 163

with her mother, her mother “frequently feigned the role of a monolingual by questioning her child’s utterance” (Lanza 1992: 647). When she used English utterances with her father, however, he often continued the conversation without questioning what Siri had said. Lanza suggested that parental responses to language mixing could be classified according to five main strategies, which she placed on a continuum from the most monolingual (i.e. discouraging the child from using the inappropriate language) to the most bilingual (i.e. fully accepting the child’s use of the inappropriate language). Nicoladis and Genesee (1998) attempted to test the Parental Discourse Hypothesis in a quantitative analysis of data collected from five French-English developing bilinguals in Montreal, Quebec. Using Lanza’s continuum of strategies, they assigned scores ranging from 1 (for a maximally “monolingual” response) to 5 (for a maximally “bilingual” response) to all parental responses to the children’s instances of lexical mixing. They then calculated correlations between these scores and the children’s rates of mixing. A positive correlation would have been compatible with the PDH, since it would suggest that the more bilingually the parent responds, the more the child will be encouraged to mix. However, Nicoladis and Genesee found quite the opposite, a negative correlation. They also considered whether the parental responses had a more local effect on child mixing: to do this they examined whether or not the child continued to mix in the next turn, and whether this was linked to the type of strategy used by the parent. They discovered that the children tended to mix in the next turn whatever the strategy: this tendency was particularly marked following the most monolingual (“minimal grasp”) strategy, which is again the opposite of what the PDH would have predicted. Nicoladis and Genesee try to explain the difference between their conclusions and Lanza’s by suggesting that bilingual acquisition in a bilingual community such as in Montreal may proceed differently from in a monolingual community such as in Norway. However, Lanza (2001) argues that the quantitative approach which they took masks important developmental and interactional aspects which her own more qualitative study explored. Indeed, she reports that her own recalculation of her own data following the same methods as Nicoladis and Genesee led to a similar finding, that there was “no relationship between parental discourse strategy and mixing rate”, but she argues that “an indepth interactional and developmental analysis as done in Lanza (1997) has shown that that is not what is going on” (Lanza, 2001: 226). Two further studies arguing for a quantitative relation between child

164 Margaret Deuchar and Rachel Muntz

mixing and the adult input are those by Mishina (1999) and Juan-Garau and Pérez-Vidal (2001). Mishina compares the proportion of mixed utterances used by an English-speaking father with the proportion used by the Japanesespeaking mother to their child (aged 1;10–2;2) and then the proportion used by the child to each parent. She finds that the mother’s rate of mixing is significantly higher than the father’s in three out of four sessions, and that the child’s rate of mixing to the mother is significantly higher than to the father. However, no actual statistical correlation is performed. Correlation is also not used in the study by Juan-Garau and Pérez-Vidal (2001) of a Catalan-English developing bilingual, although the authors provide quantitative information about the extent of mixing by the child in Catalan and English contexts, as well as coding the parental responses using Lanza’s categories. The most dramatic result of the study is the abrupt decrease in mixing in the English context just before age three: the authors suggest this is related to a strange in parental strategies by the English-speaking father, and it would have been interesting to have this confirmed by the calculation of a correlation. 3.2 Testing the Parental Discourse Hypothesis in the case study 3.2.1 Rationale Despite Lanza’s reservations about a quantitative approach to the PDH, it seemed worth testing in relation to our own case study, especially in view of Nicoladis and Genesee’s idea that different results from their own might be obtained in a monolingual community. Our English-Spanish bilingual acquisition study was conducted in southern England, in a community where all interlocutors from outside the family addressed the child in English. 3.2.2 Language input in the case study During the child’s first year of life, the mother (a native speaker of English) spoke English to her when the two were alone together, but Spanish when the father (a native speaker of Spanish) was present. The mother and father always spoke Spanish together. When the child was about one year old, her mother began to speak Spanish to her even when the father was absent, so that Spanish became the home language. However, in the presence of monolingual Englishspeaking visitors to the home, both parents spoke English. The most frequent visitor was the maternal grandmother, a monolingual English speaker who visited one morning per week when the father was usually absent. The child attended a crèche regularly from age 0;4 where English was used by all present.

Factors accounting for code-mixing in an early developing bilingual 165

As this crèche was situated on the university campus where her mother worked, the child’s mother would speak English to her in this situation. A typical daily pattern of language use established by the mother is the use of Spanish over breakfast and while driving to the university. On entering the university crèche, the mother would switch to English. The child rather than the mother sometimes initiated this change to English on entering the crèche or to Spanish when entering the car on leaving the campus at the end of the day. Deuchar and Quay (2000: 6) estimated that the child heard English 71% of the time, and Spanish, 29% of the time from birth to age 1;0, and that in her second year, she was exposed to English 48% of the time, and to Spanish, 52% of the time. It is important to note that while Spanish was heard almost entirely from her parents, who also knew English, English was heard from a variety of interlocutors, all of whom (except her parents) were monolingual. This follows from her being brought up in a predominantly monolingual English-speaking country. 3.2.3 Data collection The data were collected in two main forms: diary records and audiovideo recordings. The data were collected in both a Spanish language context, defined as when the child was being addressed in Spanish, and in an English language context, when the child was being addressed in English. The mother kept daily diary records of her daughter from when the child was aged 1;2 until 2;10. These records were based on Spanish context interactions when the mother (or mother and father) was addressing the child in Spanish at home, and on English context interactions when the mother was addressing the child in English while they were on the university campus. Entries in the diary consisted of phonetic transcriptions of the child’s new utterances as they occurred. Audiovideo recordings were also made weekly in the two language contexts between the ages of 1;3 and 3;2. In the English context the child was addressed in English, mostly by the English-speaking grandmother (who spoke no Spanish), and sometimes also by the mother. In the Spanish context the child was addressed in Spanish, mostly by the Spanish-speaking father, and sometimes also by the mother. Most of the recordings were thirty minutes in duration and the two weekly recordings were usually made within two or three days of one another. A total of twenty recordings (for details, see Deuchar and Quay 2000: 20) were transcribed in the CHAT format and have been deposited with CHILDES (see MacWhinney 2000). They can be found at the CHILDES website (http:// childes.psy.cmu.edu).

166 Margaret Deuchar and Rachel Muntz

3.2.4 Data analysis In order to test the PDH in relation to our case, fourteen transcripts available on the website were used, and are listed in Table 2, together with the results of the analysis to be reported below. The transcripts selected consist of seven pairs, where each pair was based on one recording in an English context, the other in a Spanish context, both made within a few days of one another. The child’s age in these recordings ranges over a year, from 1;7 to almost 2;7. Whereas the analysis of language choice reported in Deuchar and Quay (2000) had been limited to data from ages 1;7 and 1;8, a wider range of recordings was chosen this time in order to gain a developmental perspective. For each transcript analysed, the total number of child utterances was calculated. Excluded from this total were any utterances that could not be unambiguously identified as either English or Spanish: this included proper names which did not have alternative English and Spanish versions. Also excluded were any utterances that were imitations of adult utterances produced within ten seconds before the child’s turn. The remaining utterances were classified as either belonging to the language of the context (English in the English-context and Spanish in the Spanish-context recordings) or as “mixed”. The term “mixed” was used for

Table 2.Child and adult code-mixing and adults’ average strategy scores Age of child

CHAT filename

Language context

Child’s rate of mixing (%)

Adult’s rate of mixing (%)

Adults’ average strategy score

Proportion of child responses which show continued mixing

1;7.5 1;7.8 1;9.5 1;9.9 1;11.4 1;11.7 2;0.26 2;1.4 2;1.30 2;2.5 2;5.2 2;5.5 2;6.2 2;6.21

870129eg 870201sf 870329sf 870402eg 870528eg 870531sf 870720sk 870728er 870823ej 870829sp 871126eg 871129sf 871226sf 880114eg

English Spanish Spanish English English Spanish Spanish English English Spanish English Spanish Spanish English

43.4 13.1 25.2 30.43 7.63 10.86 8.27 0 23.53 1.84 0.47 5.93 7.53 0

0.97 3.74 2.66 1.34 0.74 3.38 8.85 0 1.81 1.73 0 0.75 0 0

3.4 3.6 3.7 3.4 3.2 3.4 3.4 n/a 2.1 3.7 n/a 3.4 4 n/a

6/7 2/2 1/12 11/23 0/3 1/6 1/1 n/a 0/1 0/1 n/a 1/6 0/1 n/a

Factors accounting for code-mixing in an early developing bilingual 167

convenience to identify utterances which were either entirely in the noncontext language, or partly in English, partly in Spanish. (In our data there were almost no instances of grammatical mixing, so we did not distinguish between grammatical and lexical mixing.) For each transcript, the child’s rate of mixing was then calculated as a percentage of the total number of utterances after exclusions. Following Nicoladis and Genesee (1998), the adult responses following “mixed” utterances were then identified where they occurred, and were classified according to Lanza’s (1992, 1997) parental strategies, in order to determine whether the adult appeared to be “negotiating” a more monolingual or more bilingual context in Lanza’s terms. It was not always easy to classify the strategies unambiguously, as Lanza (1997: 271) also shows. Parental strategy scores were calculated in the same way as by Nicoladis and Genesee, so that a low score would indicate a more monolingual strategy, and a high score a more bilingual strategy. In order then to determine how effective the adult strategies were in achieving a return by the child to the language of the context, the child’s response, if any, to each adult strategy was coded as either showing continued code-mixing or not. 3.2.5 Results The results of the analyses are summarised in Table 2. The child’s rate of mixing in each transcript, as given in the fourth column of Table 2, is also shown in Figure 1. A glance at the Figure reveals several trends: (i) Mixing generally declines with age, with the exception of the English-context recording made at age 2;2, to be discussed below; (ii) Mixing is higher in the English than in the Spanish context in the first two pairs of recordings, but then the trend is reversed (with the exception of the recordings at age 2;2); (iii) After age two mixing is very low or non-existent in the English context (with the exception of age 2;2), but remains in the Spanish context. The overall decline in mixing is consistent with what has been reported in other studies (e.g. Köppe 1997), but its asymmetry, starting with a bias towards mixing in the English context, and changing to a bias towards mixing in the Spanish context, needs to be explained. The apparently anomalous amount of code-mixing in the English-context recording made when the child was 2;2 can be explained by the fact that this recording was made when the family was on holiday in the USA. The normal recording with the English-speaking grandmother could thus not be made, and instead the child was recorded during a visit to English-speaking relatives. In this recording the father and mother addressed the child and her three year-old monolingual cousin in English. However, the child would normally have been

168 Margaret Deuchar and Rachel Muntz

Rate of mixing (%)

50 45 40 35

English Spanish

30 25 20 15 10 5 0 1;7

1;9

1;11

2;1

2;2

2;5

2;6

Age of child

Figure 1.Child’s rate of mixing.

Rate of mixing (%)

50 45 40 35

English Spanish

30 25 20 15 10 5 0 1;7

1;9

1;11

2;1

2;2

2;5

2;6

Age of child

Figure 2.Child’s rate of mixing (simplified).

accustomed to hearing English from her parents only in the presence of English-speaking adults, and there was no monolingual adult present in this recording. With her father and mother as the only adults present, the situation may have appeared to the child to be similar to that familiar in the Spanishcontext recordings made at home. This may explain the unexpectedly large amount of Spanish produced by the child in this recording. Like Figure 1, Figure 2 shows the child’s mixing pattern over time, but the recordings made at age 2;2 have been omitted.

Factors accounting for code-mixing in an early developing bilingual 169

45 40 35 30

Child’s mixing rate (%) Average adult strategy score

25 20 15 10

2;6,2

2;5,5

2;2,5

2;1,30

2;0,26

1;11,7

1;9,9

1;9,5

1;7,8

1;7,5

0

1;11,4

5

Age of child

Figure 3.Average adult strategy score and child codemixing rates in all contexts.

The average adult scores for strategies in response to the child’s language mixing are shown in the penultimate column of Table 2. As can be seen, most scores are around the midpoint between a maximally bilingual and a maximally monolingual strategy, and there does not seem to have been a change over time. Correlations between the child’s rates of mixing and the adult strategies over the fourteen recordings were calculated, as well as separately within the English and Spanish contexts. (In the English context it should be noted that there are only four data points in the correlation calculation, because in three out of the seven recordings there were no examples of adult strategies because of a virtual or total absence of mixing by the child.). The overall relation between the child’s Table 3.Correlation between child’s code-mixing rates and adults’ average strategy scores Child’s mixing rate

Adults’ average strategy score (both contexts) Adults’ average strategy score (English context) Adults’ average strategy score (Spanish context)

n

Pearson’s r

p (2-tailed significance)

14

−0.315

0.345

4

−0.249

0.751

7

−0.206

0.658

Number of child responses

170 Margaret Deuchar and Rachel Muntz

35 30 25

Continued codemixing Revert to language of context

20 15 10 5 0 Minimal Expressed Adult grasp guess repetition

Move on

Codeswitch

Adult strategies

Figure 4.Child’s response to adults’ use of strategies.

mixing and the adult strategies is shown in Figure 3, and the results are summarized in Table 3. As the table shows, there was no significant overall correlation between the child’s mixing and the adult strategies, and also no correlation in the separate language contexts. Finally, the extent to which the child continued to code-mix after an adult strategy is shown in the last column of Table 2 and in Figure 4. The numbers are small (partly because the child did not always take a turn after an adult response to her code-mixing), but we can see that it looks as though adult repetition is the only strategy which may discourage the child from continuing to code-mix, and yet it is not considered by Lanza to be the most monolingual strategy. On the other hand, minimal grasp, which is supposed to be the most monolingual, seems to lead paradoxically to the continuation of code-mixing in more cases than not. 3.2.6 Discussion A possible explanation for the lack of a clear overall correlation between the parental strategy scores and the child responses may be that some of the parental strategies were wrongly coded because of difficulties applying Lanza’s scheme of classification. For example, Lanza considered “minimal grasp” to be the most monolingual strategy in that it most constrained the child to use the language of the interlocutor. She defines it as “Adult requests clarification” (Lanza 1997:262), and says that “the parent relies primarily on the child to resay the repairable utterance, by using, for example, ‘I don’t understand’, ‘Say that again’, or wh-interrogatives” (Lanza 1997:263). At the other end of the continuum is the strategy “Adult code-switches”, which Lanza considers to be maximally negotiat-

Factors accounting for code-mixing in an early developing bilingual

ing a bilingual context, and thus encouraging the child to produce utterances in both languages. An adult response to a child mix is coded as a code-switch if the adult follows the child in changing to the other language. This occurs in the example below from our data, where the child was being recorded in a Spanish context in conversation with her father. The father was showing the child (aged 1;9) a picture of a table in a book and asking her to name it. (See 870329sf.cha in deuchar.zip at http://childes.psy.cmu.edu/win/biling.) *FAT: %eng: *CHI: *FAT: *CHI: %eng: *FAT: %eng:

eso qué es? what’s that? mmm table. table? mesa. table gracias thank you

As the example (in CHAT format, see MacWhinney 2000) shows, the child first gave the (inappropriate) name for the object in English, to which the father replied table? We coded this as involving a code-switch since the father had followed the child in changing language, but because of its question intonation one could argue that it functions more like minimal grasp, the strategy at the other end of the continuum, and indeed it has the result expected of minimal grasp, that is the child stops code-mixing and produces the appropriate word in the other language. Lanza (1997: 267) herself provides examples of the adult repeating the child’s mixed utterance, but with question intonation, and suggests that this will count as a minimal grasp if the adult follows up in the same turn with a clear statement that language choice is at fault, whereas if the adult continues in the inappropriate language in the same turn without comment, this will count as a code-switch. In the example above, however, the father does not continue his turn. Another difficulty arising with minimal grasp is the question as to how the child knows that their inappropriate language choice has given rise to the parental response, rather than some other aspect of their utterance. This question arises in the following example from the data, where the child (aged 1;9, 870402eg) is interacting with the mother in the English context, and being shown a picture of a baby. The child’s use of a Spanish word baño (‘bath’) leads to the mother responding with the question What?

171

172 Margaret Deuchar and Rachel Muntz

*CHI: %eng: *MOT: *CHI: *MOT:

baño. bath what? baby. baby yes.

In this example it is not clear whether the mother is responding to the inappropriate choice of language or the semantically incorrect lexical item. If the latter, the mother’s response could be interpreted as what Lanza would call a “Move on” strategy, which she considers to be the most bilingual after the code-switch. However, we coded it as a minimal grasp, the most monolingual strategy. A similar problem arises with an adult strategy taking the form of Mm? or Hm? where one also can question whether the child will target the problem as being language choice. In our data we found some indication that the child understood Mm? or Hm? as a request for repetition rather than reformulation in the other language, as she would tend to repeat her code-mixed utterance after these responses. On the other hand, she seemed slightly more likely to reformulate after a more explicit version of the minimal grasp strategy such as Yo no entiendo (‘I don’t understand’) or What? Lanza (1997: 271) acknowledges that the coding of some strategies will be problematic. She gives as an example the parent’s use of “Hm? or What? followed by a Repetition within the same turn”, and raises the question as to whether this should be coded as a minimal grasp or repetition. In her analysis she codes such responses as repetition, on the grounds that the child was not given time to respond to the Hm? or What? Repetition as a strategy is intermediate between the most monolingual and the most bilingual. Aside from the problem of coding, there is the problem of whether the quantification of the strategies gives an accurate view of their impact. Lanza views the strategies as “contextualization cues” (cf. Gumperz 1982) and draws attention to Auer’s (1995: 123) insight that “The same cue may receive a different interpretation on different occasions” (Lanza, 2001: 216). Lanza illustrates this in relation to the minimal grasp in her data from Siri, where she suggests that a minimal grasp from Siri’s mother is more likely to lead to a change of language than a minimal grasp from Siri’s father. Lanza suggests that this is because Siri has been socialized into different interactional styles with her two parents. However, this begs the question of how the process of socialization works, since Lanza’s earlier work has been interpreted as suggesting that the greater frequency of one strategy in comparison to another would play a role in determining the specific interaction style.

Factors accounting for code-mixing in an early developing bilingual

As indicated in Section 3.1, Lanza advocates a more qualitative approach to the analysis of child and adult bilingual interaction. However, it is not clear how overall differences between adult styles can be established without at least some quantification. We shall now consider two additional types of quantitative analysis in order to explore the possibility that they may lead to more insights into the factors which determine code-mixing in developing bilinguals.

4. Modelling Hypothesis 4.1 What is the modelling hypothesis? In our investigation of the Parental Discourse Hypothesis we focused, as described, specifically on responses by the adult interlocutors to instances of the child’s code-mixing. However, these adult responses represented only a small proportion of the total language input to the child, and it may be that the child uses the adult input in general as a model to target. This is known as the “modelling hypothesis”, and is justified by Comeau, Genesee and Paquette (forthcoming): “it is well recognized that, in the long run, children acquire the sociolinguistic and pragmatic constraints and patterns that characterize adult language usage; this is likely to be as true of code-mixing as it is of other features of language use”. Thus one might expect the child to adjust the proportion of code-mixing in their speech so that it is equivalent to that in the adult input. As Genesee et al. point out, however, the evidence for the modelling hypothesis is conflicting. Goodz (1989), in a study of four developing bilinguals in Canada, found that there were significant positive correlations between the child’s and the mother’s code-mixing over various sessions for three of the children, but with the father’s for only one of the children. On the other hand, Genesee et al. (1995), in a study of five two-year old Canadian developing bilinguals, found no evidence of a relation between parental and child mixing. However, Nicoladis and Genesee (1997), in a longitudinal study using some of the same children in a group over four sessions from when they were aged 2;0 up to 3;6, found that parental and child codemixing were correlated in the final two recording sessions but not in the first two. One of the most successful attempts to date to demonstrate the relation between adult and child code-mixing appears to be Genesee et al.’s (1995) quasi-experimental study of four developing bilinguals who increased their rates of code-mixing in the second session recorded with a previously unknown bilingual who deliberately

173

174 Margaret Deuchar and Rachel Muntz

used a higher rate of code-mixing in the second than the first session. However, it is important to emphasize that this result is achieved with interlocutors who were not previously familiar to the children, and so cannot be strictly compared with the results in the present study. It raises the paradoxical possibility that strangers may be better models than members of the child’s family, or at least that the factors involved with familiar interlocutors are more complex, possibily including aspects of the history of previous interactions. 4.2 Testing the Modelling Hypothesis in the case study 4.2.1 Data analysis In testing the Parental Discourse Hypothesis as described in Section 3.1 above, the child’s rate of mixing in the fourteen transcripts listed in Table 2 was calculated. In order to test the Modelling Hypothesis, the same transcripts were used to calculate the adult interlocutors’ mixing rates, in order to determine whether there was a relation between the adult and child mixing rates as would be suggested by the Modelling Hypothesis. The mixing rates of the adult interlocutors were calculated in a similar way to those of the child, as described in Section 3.2.4. Where a transcript contained data from more than one interlocutor, mixing rates were calculated for the adults as a whole in that session, as well as individually. (In two transcripts, 870728er and 870823ej there was one or more child interlocutors, but this only affected the overall mixing rate in 870823ej, to be discussed below.) As in the analysis of code-mixing by the child, ambiguous utterances were excluded, as were utterances that were direct imitations of a previous utterance by the child. Once the adult mixing rates had been calculated for each transcript, correlations with the child rates were calculated. 4.2.2 Results The adult mixing rates in each transcript are shown in Table 4, alongside the child’s mixing rates. In Table 4, information is also provided about the raw numbers, and about the individual mixing rates for the adults. The adult mixing rates are also shown in Figure 5, which can be compared with the child’s mixing rates as shown in Figure 1. A glance at Figure 5 shows that there is generally more code-mixing by the adults in the Spanish than the English context, except when the child was aged 2;2 (cf. Section 3.2.5). This is probably because the Spanish-speaking adults also speak English, whereas the English-speaking adults do not all speak Spanish. The overall relation between the adult and child code-

Factors accounting for code-mixing in an early developing bilingual

Table 4.Adult and child code-mixing rates CHAT file- Language No. of No. of Rate of No. of No of Rate of name context child ut- child child mix- adult mixed adult terances mixed ing (%) utterances adult mixing utterances utterances (%)

1;7.5 1;7.8 1;9.5 1;9.9 1;11.4 1;11.7 2;0.26 2;1.4 2;1.30 2;2.5 2;5.2 2;5.5 2;6.2 2;6.21

870129eg 870201sf 870329sf 870402eg 870528eg 870531sf 870720sk 870728er 870823ej 870829sp 871126eg 871129sf 871226sf 880114eg

Rate of codmixing (%)

Age of child

10 9 8 7 6 5 4 3 2 1 0

English Spanish Spanish English English Spanish Spanish English English Spanish English Spanish Spanish English

53 84 127 207 131 175 133 119 68 217 214 253 146 84

23 11 32 63 10 19 11 0 16 4 1 15 11 0

43.4 13.1 25.2 30.43 7.63 10.86 8.27 0 23.53 1.84 0.47 5.93 7.53 0

618 588 301 447 271 266 192 146 166 289 170 266 229 196

6 22 8 6 2 9 17 0 3 5 0 2 0 0

0.97 3.74 2.66 1.34 0.74 3.38 8.85 0 1.81 1.73 0 0.75 0 0

English Spanish

1;7

1;9

1;11

2;1

2;2

2;5

2;6

Age of child

Figure 5.Adult codemixing rates.

mixing rates is shown in Figure 6. In general, we can see that while the child’s rate of code-mixing declines, the adults’ stays more or less constant. The relation between the adult and child code-mixing rates can be viewed separately for the English and Spanish contexts in Figures 7 and 8 respectively. Correlations between the child’s and the adults’ rate of code-mixing were calculated overall, and separately for the English and Spanish context recordings, and the results can be seen in Table 5. Table 5 shows no significant overall correlation

175

176 Margaret Deuchar and Rachel Muntz

50 45 Codemixing rate (%)

40 35 Child Adult

30 25 20 15 10 5 2;6.21

2;6.2

2;5.5

2;5.2

2;2.5

2;1.30

2;1.4

2;0.26

1;11.7

1;11.4

1;9.9

1;9.5

1;7.8

1;7.5

0

Age of child

Codemixing rate (%)

Figure 6.Adult and child codemixing rates in all contexts. 50 45 40 35 30 25 20 15 10 5 0

Child Adult

1;7.5

1;9.9

1;11.4

2;1.4

2;1.30

2;5.2

2;6.21

Age of child

Figure 7.Child and adult codemixing rates in English context.

between the child’s and the adults’ mixing rate. However, it shows that if recordings in the English context are considered alone, there is (just) a significant correlation between the adult and child mixing. This result will be discussed in the next section. In the Spanish context, however, no correlation was found between adult and child mixing.

Factors accounting for code-mixing in an early developing bilingual 177

Codemixing rate (%)

30 25 20

Child Adult

15 10 5 0 1;7.8

1;9.5

1;11.7

2;0.26

2;2.5

2;5.5

2;6.2

Age of child

Figure 8.Child and adult codemixing rates in Spanish context. Table 5.Correlation between child’s and adults’ code-mixing rates Child’s mixing rate

Adults’ mixing rate (both contexts) Adults’ mixing rate (English context) Adults’ mixing rate (Spanish context)

n

Pearson’s r

p (2-tailed significance)

14

0.084

0.774

7

0.758*

0.049

7

0.110

0.815

*Correlation is significant at the 0.05 level (2-tailed)

4.2.3 Discussion Overall, we do not see a clear statistical relation between adult and child mixing, suggesting that we do not find overall support for the Modelling Hypothesis. It is possible that the (just) significant correlation that we find in the English context is due to the anomalous nature of recording 870823ej, made at age 2;1.30, and discussed in Section 3.2.5. As Figure 1 shows, the child’s rate of mixing rises considerably, probably because of the unusual circumstances of the recording. The unusual nature of this context may also account for the adults’ rise in mixing rate also. So the context itself may have caused the mixing rates of both the adults and children to rise, rather than the adults modelling the appropriate amount of mixing for the child.

178 Margaret Deuchar and Rachel Muntz

5. Dominance 5.1 Measuring dominance Deuchar and Quay (2000: 9–10) discussed the various ways in which the term “dominance” has been used in the literature, showing that it is sometimes used to refer to aspects of language production as presumed to represent relative proficiency in the two languages, but sometimes to aspects of the input. Here we shall focus on the question of relative proficiency and ask two questions in relation to our study:1) was the child more proficient in one language than the other? and 2) if so, how did this affect her code-mixing? Genesee, Nicoladis and Paradis (1995: 615) view dominance in terms of relative proficiency and predict “a general tendency for bilingual children to mix elements from their dominant language when using their non-dominant language, rather than vice versa, because many of the linguistic structures for communication are lacking in the non-dominant language”. On the basis of three recording sessions within a three-week period, they calculated four indices of relative dominance in their bilingual subjects’ two languages: MLU, “Upper bound”, “MMU” and “Word types”. “Upper bound” was “the number of morphemes in the longest utterance in the record of each child’s language” (Genesee, Nicoladis and Paradis 1995: 620); “MMU” = “Multi-morphemic utterances” was the percentage of all utterances longer than one morpheme (from both languages) which was produced in each language, and “Word types” was based on the percentage in each language of the total number of word types produced. The children’s scores in each language on the four indices then led to them being classified as either dominant in one of their languages, or “balanced” in both. Nicoladis and Genesee (1996) took a slightly different approach in that they used only MMU and Word types to measure dominance, and since each of these scores is a percentage, they then averaged the two percentage scores in each language to arrive at an overall dominance percentage score. The dominance percentage scores would add up to 100, so that if a child had a dominance score of 60% in one language, for example, he or she would have a score of 40% in the other. Since they analysed recordings made for each child at seven different points in time, the child would have different dominance scores at different points in time. Yip and Matthews (2000) attempted to measure language dominance in a Cantonese-English bilingual child by measuring MLU in words in each language,

Factors accounting for code-mixing in an early developing bilingual 179

and assuming that the dominant language “will have a higher MLU value than the less dominant one” (Yip and Matthews 2000). However, they recognized that MLU may not be a reliable way of comparing the child’s stage of development in languages of different morphological types, a point made also by Nicoladis and Genesee (1996), who decided to therefore exclude it from their measures. 5.2 Findings on dominance and code-mixing Genesee et al. (1995) found in their investigation of five roughly two-year old developing French-English bilinguals in Canada (see also Section 4.1) that, according to the measures described in Section 5.1 above, one child was balanced, one dominant in French, and two or possibly three dominant in English. In order to examine the possible relation between dominance and the child’s code-mixing, they compared the extent of mixing in recordings with the French and English-speaking parent. The dominant children were found to produce more inter-utterance mixing speaking the non-dominant than the dominant language. The balanced children, however, produced similar amounts of mixing to both parents. Nicoladis and Genesee (1997) also investigated the relation between dominance and code-mixing in seven developing French-English bilinguals at the ages of 2;6 and 3;6, again pursuing the hypothesis that children would code-mix more in their non-dominant than their dominant language. At ages 2;6 and 3;0 this was true for four out of the five children who were dominant in either English or French. (The two balanced children code-mixed about the same in the two languages.) At age 3;6 it was true for all five children. Nicoladis and Genesee also found an interaction between dominance and parental code-mixing as exhibited in three distinct patterns: (1) both balanced children and their parents decreased their code-mixing over time; (2) children who were dominant in a language did not code-mix much in that language, nor did their parents, and neither changed much over time; (3) children who were dominant in one language increased their code-mixing in the non-dominant language over time, as did their parents. We shall return to these findings in Section 5.3.3 when considering which category our case falls into if any. 5.3 Investigating dominance in the case study Assuming the term “dominance” to refer to relative proficiency in the two languages, the next task was to decide on a measure of dominance in the pairs

180 Margaret Deuchar and Rachel Muntz

of transcripts available. A quantitative measure was preferable, since it would allow us to investigate, in a way analogous to our focus on child and adult codemixing, whether there was a correlation between the child’s rate of mixing across sessions and her language dominance. If Genesee et al.’s (1995: 615) prediction, mentioned in Section 5.1 is correct, we would expect the child to mix more in the context of the language in which she is not dominant, and less in the context of the language in which she is dominant. This might be reflected in a negative correlation between her dominance scores and the rate of mixing in each context. 5.3.1 Data analysis Following Genesee et al. (1995) we decided to calculate the child’s MLU in each language, the Upper Bound in each language, the percentage of MMUs in each language, and the percentage of Word Types in each language. However, rather than using these four measures to determine relative dominance in each language (i.e. whether English or Spanish was dominant), we decided to adopt Nicoladis and Genesee’s (1996) method of arriving at a percentage dominance score on the basis of the MMUs and Word Types in each language, and to use MLU and Upper Bound as additional indicators of dominance. Another departure from Genesee et al.’s (1995) approach was that whereas they based their calculations on pooled data from three recording sessions for each child, one with the father, one with the mother, and one with both parents, we made separate calculations for each recording session, and then averaged the results for each pair of sessions at a given age. Averaging the percentage results for MMU and Word Types in each language then gave us a dominance score in each language at each age. The advantage of this approach is that it avoids the distortion that could arise in the MMU scores if the recording in one context happens to be longer than that in another (this sometimes happened despite the investigators’ attempts to make all recordings last half an hour) or if the adult in one context encourages the child to speak more than the adult in the other context (cf. also Lanza 2000: 230). At the same time, this approach takes account of the likely role of the context in leading to a predominance of utterances from one language rather than the other, in that the percentage of utterances in each language is averaged across contexts at each age. This will be illustrated in the description of the calculation of MMUs in Section 5.3.1.3 below. 5.3.1.1 Calculation of MLU. For each transcript, the utterances in English were considered separately from the utterances in Spanish. Non-language-specific

Factors accounting for code-mixing in an early developing bilingual

Table 6.Mean length of utterance (MLU) and Upper Bound (UB) Age of child

CHAT filename

Language MLU context English*

MLU Spanish*

UB UB Dominant English Spanish language

1;7.5 1;7.8 1;7 1;9.5 1;9.9 1;9 1;11.4 1;11.7 1;11 2;0.26 2;1.4 2;1 2;1.30 2;2.5 2;2 2;5.2 2;5.5 2;5 2;6.2 2;6.21 2;6

870129eg 870201sf both 870329sf 870402eg both 870528eg 870531sf both 870720sk 870728er both 870823ej 870829sp both 871126eg 871129sf both 871226sf 880114eg both

English Spanish

1.00 (25/25) 1.00 (76/76) 1.00 1.18 (118/100) 1.21 (80/66) 1.19(198/166) 1.33(12/9) 1.30(130/100) 1.30(142/109) 1.34 (134/100) – 1.34 (134/100) 2.41 (41/17) 2.33 (233/100) 2.34(274/117) 1.00 (1/1) 2.66 (266/100) 2.64(267/101) 2.11 (211/100) – 2.11(211/100)

1 1 1 2 3 3 4 4** 4 3 7 7 4 3** 4 7 4** 7 3 6 6

Spanish English English Spanish Spanish English English Spanish English Spanish Spanish English

1.00 (39/39) 1.00 (23/23) 1.00 1.07 (32/30) 1.32 (132/100) 1.26 (164/130) 1.29 (129/100) 2.17 (39/18) 1.42 (168/118) 1.50 (15/10) 2.56 (256/100) 2.46 (271/110) 1.88 (120/64) 1.50 (6/4) 1.85(126/68) 3.00 (300/100) 1.63 (26/16) 2.81(326/116) 1.50 (18/12) 2.32 (183/79) 2.21(201/91)

1 1 1 4 3** 4 2 4** 3 4 – 4 4 8 8 1 11 11 10 – 10

Neither

?

English

English

Spanish

?

?

*Calculated on the basis of 100 utterances or less if less than 100 occur. **Upper Bound is based on a mixed utterance.

utterances were excluded, but mixed utterances were counted in both the English and the Spanish set of utterances. Utterances containing unintelligible material were excluded. A sample of 100 utterances in each language was used where there were more than 100 utterances in that language; otherwise all utterances were used. If there were more than 150 utterances, utterances 51–150 were used, and if there were between 51 and 150 utterances, the last 100 were used. The rules for counting morphemes were as outlined by Brown (1973: 54) for English and by Siguan, Colomina and Vila (1990: 56) for Spanish. In counting bound morphemes, attention was paid to the productivity criteria outlined by Pizzuto and Caselli (1994: 144) so that an inflection would only be counted if it was used in the same or previous recording sessions with at least two distinct verbs, and if the same verb also appeared in the same or previous recording sessions with a different inflection. It must be emphasized, as indicated above in Section 5.1, that cross-linguistic comparison of MLU may not be very meaningful since English and Spanish differ morphologically: in

181

182 Margaret Deuchar and Rachel Muntz

particular, Spanish has richer morphology than English. This is why we did not include the MLU measurements in our dominance calculations. However, they are listed for the purposes of information in Table 6. As the table shows, we have given MLU figures in each language for each recording, but have also included the overall MLU in each language across pairs of recordings. 5.3.1.2 Calculation of Upper Bound (UB). This was based, as in Genesee et al. (1995), on the number of morphemes in the longest utterance in each language. As in the calculation of MLU, a separate figure was established for each language in each recording session, and then the longest utterance overall in each language (across each pair of recordings) is noted. This is shown in Table 6. Since the length of mixed utterances was counted as part of each language as outlined in Section 5.3.1.1 above, the Upper Bound number was occasionally based on a mixed utterance. This is indicated by a double asterisk in Table 6. In the final column of Table 6 we have indicated where possible whether the child is more dominant in English or Spanish at each age, based on MLU and UB indicators alone. This will provide a basis for comparison with the quantitative assessment of dominance based on MMU and word types, as outlined above, and shown in Table 7. 5.3.1.3 Calculation of MMUs. This was a percentage figure calculated for each age at which a pair of transcripts was available, and was based on all the multimorphemic utterances in each language in each transcript. To arrive at the percentage scores in each language (which would add up to 100 at a given age), the total number of multi-morphemic utterances in each member of a pair of recordings was counted for each language. This meant essentially subtracting the number of one-morpheme utterances from the total number of utterances in each language group. The number of Spanish multi-morphemic utterances would be added to the number of English multi-morphemic utterances and the relative percentage of Spanish versus English multi-morphemic utterances calculated. For example, in file 870528eg (age 1;11) there are 30 English multimorphemic utterances and 3 Spanish multi-morphemic utterances, making a total of 33 multi-morphemic utterances. This 3:30 ratio of Spanish:English can be expressed as a percentage: 9.09% Spanish versus 90.91% English. As this is an English context file, this result is not unexpected. The same procedure was then followed to obtain the MMU English and MMU Spanish figures for the paired Spanish file (870531sf): 26.09% English and 73.91% Spanish. Finally, the two English MMUs were averaged to produce a total English MMU for the pair of

870129eg 870201sf both 870329sf 870402eg both 870528eg 870531sf both 870720sk 870728er both 870823ej 870829sp both 871126eg 871129sf both 871226sf 880114eg both

1;7.5 1;7.8 1;7 1;9.5 1;9.9 1;9 1;11.4 1;11.7 1;11 2;0.26 2;1.4 2;1 2;1.30 2;2.5 2;2 2;5.2 2;5.5 2;5 2;6.2 2;6.21 2;6

Spanish English

English Spanish

English Spanish

Spanish English

English Spanish

Spanish English

English Spanish

CHAT filename Language context

Age of child 0 0 0 14.29% (3/21) 75.00% (39/52) 44.65% 90.90% (30/33) 26.08% (12/46) 58.49 % 9.30% (4/43) 100% (95/95) 54.65% 69.77% (30/43) 0.88% (1/113) 35.33% 100% (147/147) 3.70% (6/162) 51.85% 5.62 (5/89) 100% (44/44) 52.81%

MMU English 0 0 0 85.71% (18/21) 25.00% (13/52) 55.36% 9.10% (3/33) 73.91% (34/46) 41.51% 90.70% (39/43) 0% (0/95) 45.35% 30.23% (13/43) 99.12% (112/113) 64.68% 0% (0/147) 96.30% (156/162) 48.15% 94.38 (84/89) 0% (0/44) 47.19%

MMU Spanish 72.22% (13/18) 20.59% (7/34) 46.41% 25.93% (21/81) 75.00% (48/64) 50.47% 91.67% (77/84) 17.44% (15/86) 54.56% 15.66% (13/83) 100% (83/83) 57.83% 73.61% (53/72) 5.38% (7/130) 39.50% 99.53% (211/212) 9.09% (15/165) 54.31% 7.09% (9/127) 100% (123/123) 53.55%

Word types English

Table 7.Multi-morphemic utterances (MMU), word types and dominance score

27.78% (5/18) 79.41% (27/34) 53.59% 74.07% (60/81) 25.00% (12/64) 49.54% 8.33% (7/84) 82.56% (71/86) 45.44% 84.34% (70/83) 0% 42.17% 26.39% (19/72) 94.62% (123/130) 60.50% 0.47% (1/212) 90.9% (150/165) 45.69% 92.91% (118/127) 0% 46.45%

Word types Spanish

53:47

53:47

37:63

56:44

57:43

48:52

46:54

English

English

Spanish

English

English

Spanish

Spanish

Dominance Dominant score language

Factors accounting for code-mixing in an early developing bilingual 183

184 Margaret Deuchar and Rachel Muntz

files and the same is done with the Spanish MMUs. In the case of this pair of files at age 1;11, the average English MMU is (90.91%+26.09%)/2 = 58.49% and the average Spanish MMU is 41.51%. 5.3.1.4 Calculation of Word Types. The CLAN programme FREQ from the CHILDES system (see MacWhinney 2000) was run on each of the transcript files in order to obtain a list of word types and tokens. The word types were then identified as either English, Spanish or non-language-specific. The nonlanguage-specific items were deducted from the word type total, and the percentage of English versus Spanish words was then calculated. The percentages of English words found in both files at each age were averaged to obtain an overall Word Type score for English at that age, and the same was done for Spanish. The results are shown in Table 7. 5.3.2 Results As indicated above, Table 7 shows the MMU and WT scores for each language and in each language context, as well as overall scores for each of the seven age points. The final column of Table 7 combines these two scores as outlined in 5.3.1 above to show the relative dominance of English versus Spanish at each age, in terms of a ratio. The child’s overall dominance pattern is shown in Figure 9. We can see from Figure 9 that the child shows a very slight Spanish dominance at ages 1;7 and 1;9, and that this then changes to a slight English dominance at ages 1;11, 2;1, 2;5 and 2;6. The exception is at age 2;2, where the child no doubt appears to be Spanish dominant because of the unusual context of recording 870823ej, discussed above in Section 3.1.4. The dominance ratios

Table 8.Correlation between child’s dominance and mixing rate Child’s mixing rate

Child’s dominance* (both contexts) Child’s dominance (English context) Child’s dominance (Spanish context)

n

Pearson’s r

p (2-tailed significance)

14

−0.402

0.154

7

−0.666

0.102

7

−0.043

0.927

*The correlations represented in this table are based on the English dominance scores; the results based on the Spanish scores were identical, but with positive rather than negative Pearson’s r.

Percentage

Factors accounting for code-mixing in an early developing bilingual

100 90 80 70 60 50 40 30 20 10 0

Spanish English

1;7

1;9

1;11

2;1

2;2

2;5

2;6

Age of child

Figure 9.The child’s dominance pattern.

in Table 7 are compatible with the indications of dominance in Table 6, based on MLU and UB. However, it should be emphasized that, given the small differences in the child’s English and Spanish dominance throughout the period of the recordings, it would probably be more appropriate to describe her as “balanced” than as dominant in one language rather than the other. Nevertheless, as indicated in Section 5.3, correlations were calculated between the child’s mixing rate and her dominance scores at each age. The results are shown in Table 8. As can be seen in the Table, no correlation is found between the child’s dominance and her mixing rate, whether all recordings in both contexts are taken into account, or whether the recordings in the English and Spanish contexts considered separately. 5.3.3 Discussion In Section 5.1 we raised the question as to whether the child was more proficient in one language than another, and if so, whether it affected her code-mixing. However, as indicated in Section 5.3.2, there was very little difference between her English and Spanish proficiency over time, and what difference there was did not correlate with variation in code-mixing. We suggested above that it would be more appropriate to describe her as “balanced” rather than dominant in one language rather than another. We reported in Section 5.2 that Nicoladis and Genesee (1997) found that balanced children and their parents decreased their code-mixing over time, and Figure 6 certainly suggests that we found something similar in our case. In the case of this child, then, dominance was not a factor in her code-mixing.

185

186 Margaret Deuchar and Rachel Muntz

6. How do we account for the lack of correlations? We have conducted three separate quantitative analyses of the relation between the child’s rate of code-mixing at seven points in time and the factors of adult strategy in response to child mixing, adult mixing in general, and child dominance. We did not find an overall correlation between the child’s mixing and any of these three factors. Two of the factors examined concern the input to the child, and one concerns her proficiency in terms of dominance. As outlined in the discussion of dominance in Section 5.3.2 above, the child’s languages seem to be fairly well balanced throughout the period of the recordings, yielding relatively constant dominance scores. This means that the lack of a correlation between dominance and mixing is unsurprising. More surprising, perhaps, is the lack of correlation between the aspects of the input examined and the child’s mixing. Reasons for this lack were discussed in Sections 3.2.6 and 4.2.3 respectively, but an additional reason might be that we have not sufficiently taken into account developmental factors affecting the child herself. Vihman (1985: 313) draws attention to the assertion by Kagan (1981) that self-awareness develops towards the end of the child’s second year and goes hand in hand with increasing awareness of the standards of adults as well as motivation to conform to them. Linguistic evidence of this new awareness includes the use of evaluative words and self-descriptive utterances. In the case of bilinguals, Vihman suggests that this awareness would also be reflected in the establishment of appropriate language choice. To explore the possibility that the development of self-awareness might be a factor in the changing patterns of mixing in our case-study, we conducted a preliminary analysis of some of the relevant linguistic evidence. Kagan (1981: 48, 123–124) suggests that children approaching two years old tend to use evaluative language to comment on instances of the violation of norms. An example of this in our data is the use of the expression oh dear: this is produced relatively frequently in the second half of the child’s second year, usually in response to an unexpected event, such as an item being dropped, or something being out of place. We calculated the tokens of oh dear (counted as one lexical item) as a percentage of the tokens of all lexical items in each pair of recordings at each age, and the results were as in Figure 10. This shows that the percentage of oh dear lexical items was around 5% at age 1;7, 4% at age 1;11, and then dropped to zero or virtually zero at the ages from 1;11 onwards. This relatively

Factors accounting for code-mixing in an early developing bilingual 187

6

Percentage

5 4 3 2 1 0 1;7

1;9

1;11

2;1

2;2

2;5

2;6

Age of child

Figure 10.Tokens of oh dear as a percentage of the total.

frequent production of oh dear around the ages of 19–21 months could be explained by the child’s new awareness of norms and particular sensitivity to violation of these norms when unexpected events occur. Self-descriptive utterances are defined by Kagan (1981: 67) as “utterances that occurred when the child was engaged in an action and spoke a word or sentence referring to that action” and “utterances containing the words I, my, mine, or the child’s name, together with a predicate”. In our data, the child’s first term of self-reference was her own name, which appeared at 1;3. It occurred in early multi-word utterances noted in the diary at age 1;7, but infrequently in the recordings until age 1;11. At this age a quarter of the child’s multi-word utterances included her own name, and this increased to a peak of 35% at age two, as shown in Figure 11. It seems likely that this peak may coincide with a peaking of the new self awareness. So what light does this throw on the development of language choice in the child? As shown in Figure 2, there is an overall decline in child mixing which could be explained in terms of the child’s gradual increase in self-awareness and gradual adjustment to adult norms of appropriateness in language choice. In addition, Figure 2 shows a change between ages 1;9 and 1;11 from a predominance of mixing in the English context to almost exclusive mixing in the Spanish context. As suggested in Section 2, the greater mixing in the English context at ages 1;7–1;9 might be explained by the location of the recordings being associated with the Spanish context. From age 1;11 onwards, however, the

188 Margaret Deuchar and Rachel Muntz

40 35

Percentage

30 25 20 15 10 5 0 1;7

1;9

1;11

2;1

2;2

2;5

2;6

Age of child

Figure 11.Percentage of child utterances including own name.

child mixes more in the Spanish than the English context, which perhaps reflects her growing awareness that her Spanish-speaking interlocutors are all bilingual and can understand her English, whereas many of her Englishspeaking interlocutors are monolingual. We may note that it is also from 1;11 onwards that we found linguistic evidence of self-awareness on the part of the child, in the preliminary analysis outlined above.

7. Conclusion Our quantitative analysis of various factors in relation to child mixing has failed to show any clear correlations. In relation to dominance this can be explained in terms of the relative balance of the two languages, and in relation to the input we have argued that developmental factors may also be playing a role in the reduction of mixing over time. Such factors may obscure the effect of the input. In particular, we have briefly examined some examples of linguistic evidence for the development of self-awareness around the time that the child’s mixing patterns begin to reflect her awareness of the norms and competence of her interlocutors.



Factors accounting for code-mixing in an early developing bilingual 189

Note * Thanks are due to the UK Economic and Social Research Council for a grant (C00232393) for the collection of the data, and to the British Academy for supporting the analysis reported here. We should also like to acknowledge the participation of undergraduate participants in the Bangor Bilingualism Seminar held in 1999 and 2000, especially Charlotte Lovie, James Russell, Holly Pinkney and Hannah Wallace. Finally, we are grateful for comments on an earlier draft to Fred Genesee, Elizabeth Lanza, Marilyn Vihman and two anonymous reviewers. Any remaining shortcomings are of course our own.

References Auer, P. 1995. “The pragmatics of code-switching: A sequential approach”. In One Speaker, Two Languages: Cross-disciplinary Perspectives on Code-switching, L. Milroy and P. Muysken (eds), 115–35. Cambridge: Cambridge University Press. Brown, R. 1973. A First Language: The early stages. Cambridge, MA: Harvard University Press. Comeau, L., Genesee, F. and Paquette, L. (forthcoming). The Modeling Hypothesis and child bilingual code mixing. International Journal of Bilingualism. Deuchar, M. and Quay, S. 1999. “Language choice in the earliest utterances: A case study with methodological implications”. Journal of Child Language 26: 461–475. Deuchar, M. and Quay, S. 2000. Bilingual acquisition: Theoretical implications of a Case Study. Oxford: Oxford University Press. Genesee, F., Nicoladis, E., and Paradis, J. 1995. “Language differentiation in early bilingual development”. Journal of Child Language 22: 611–31. Goodz, N. 1989. “Parental language mixing in bilingual families”. Infant Mental Health Journal 10 (1): 25–44. Gumperz, J. 1982. Discourse Strategies. Cambridge: Cambridge University Press. Juan-Garau, M. and Pérez-Vidal, C. 2001. “Mixing and pragmatic parental strategies in early bilingual acquisition”. Journal of Child Language 28: 59–86. Kagan, I. 1981. The Second Year. Cambridge, Mass.: Harvard University Press. Köppe, R. 1997. Sprachentrennung im frühen bilingualen Erstspracherwerb Französisch/ Deutsch. Tübingen: Gunter Narr Verlag. Lanza, E. 1992. “Can bilingual two-year-olds code-switch?” Journal of Child Language 19: 633–658. Lanza, E. 1997. Language Mixing in Infant Bilingualism: A Sociolinguistic Perspective. Oxford: Clarendon Press. Lanza, E. 2000. “Concluding remarks: language contact — a dilemma for the bilingual child or for the linguist?” In Cross-linguistic Structures in Simultaneous Bilingualism, S. Döpke (ed.), 227–246. Amsterdam: John Benjamins. Lanza, E. 2001. “Bilingual first language acquisition: A discourse perspective on language contact in parent-child interaction”. In Trends in Language Acquisition Research, J. Cenoz and F. Genesee (eds). Amsterdam: John Benjamins.



190 Margaret Deuchar and Rachel Muntz

MacWhinney, B. 2000. The CHILDES Database: Tools for Analyzing Talk. 3rd Edition. Vol. 2: The Database. Mahwah, NJ: Lawrence Erlbaum Associates. Mishina, S. 1999. “The role of parental input and discourse strategies in early language mixing of a bilingual child”. Multilingua 18 (4): 343–367. Nicoladis, E. and Genesee, F. 1996. “A longitudinal study of pragmatic differentiation in young bilingual children”. Language Learning 46 (3): 439–464. Nicoladis, E. and Genesee, F. 1997. “The role of parental input and language dominance in bilingual children’s code-mixing”. In Proceedings of the 21st Annual Boston University Conference on Language Development, Volume 2, E. Hughes, M. Hughes and A.Greenhill (eds), 439–464. Nicoladis, E. and Genesee, F. 1998. “Parental discourse and codemixing”. The International. Journal of Bilingualism 2 (1):85–99. Pizzuto, E. and Caselli, M. C. 1994. “The acquisition of Italian verb morphology in a crosslinguistic perspective.” In Other children, other languages: Issues in the theory of language acquisition, Y. Levy (ed.), 137–187. Hillsdale, N. J.: Erlbaum. Siguan, M., Colomina, R. and Vila, I. 1990. Metodología para el Estudio del Lenguaje Infantil. Vic: Abril Editorial. Vihman, M. M. (1985). Language differentiation by the bilingual infant. Journal of Child Language 12: 297–324. Yip, V. and Matthews, S. J. 2000. “Syntactic transfer in a Cantonese-English bilingual child”. Bilingualism: Language and Cognition 3(3): 193–208.



Syllable final consonants in Spanish and German monolingual and bilingual acquisition* Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo Research Center on Multilingualism, University of Hamburg

1.

Introduction

CV syllables are believed to be the unmarked canonical syllables, from a synchronic as well as developmental perspective. According to the well-known Jakobsonian typology,1 all languages possess syllables beginning with a consonant, but not all languages possess syllables ending in a consonant, and still fewer languages exhibit syllables ending in more than one consonant. This typology of syllable patterns is also mirrored in acquisition: Children first produce CV (consonant vowel) syllables before they produce syllables with final consonants and consonant sequences.2 In onset-rhyme theory, syllable-final consonants belong to the part of the syllable known as the coda.3 Theories of phonological acquisition have dealt with the later emergence of codas in various ways. Principles and Parameters theory accounts for the difficulty of codas by requiring that the Branching rhyme parameter (Rhymes can branch into a nucleus and coda) be initially not set or set to NO (Fikkert 1994), and only later be set to YES, i.e. to branching. Optimality Theory accounts for the initial stage of CV syllables by the high ranking of the markedness constraint NoCoda (Prince and Smolensky 1993). At later developmental stages, this constraint will be demoted, depending upon the complexity of the target language, allowing syllable types with codas to emerge. This paper presents findings on the acquisition of codas by German and Spanish monolingual and bilingual children. Both German and Spanish contain

192 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

syllables with codas, but whereas German allows many different consonants and consonant clusters at the end of syllables, Spanish has strong restrictions on what segments can appear in this position. This implies that the acquisition task for the child acquiring German should be more difficult than for the child acquiring Spanish: The German child has more to learn, because markedness constraints will have to move a longer way down the hierarchy of constraints, in order to arrive at the grammar of the target language (but see Section 2.4 for an alternative prediction). In the first part of the paper, we present results on the acquisition of codas by monolingual German and Spanish children. Our findings show, in fact, that German-speaking children acquire codas earlier than Spanish-speaking children, a result, we believe, that relates both to the greater frequency of codas in German than in Spanish and to the importance of codas in the phonology of German (i.e., binarity and quantity sensitivity). In the second part of the paper, we turn our attention to the acquisition of codas by bilingual German and Spanish children. Here we are interested in the relationship between the phonological systems of the two languages. Do they develop independently or do they influence each other? Our results show that there is indeed interaction between the phonological systems of the two languages, in that GermanSpanish bilingual children produce codas earlier in Spanish than monolingual Spanish-speaking children do. The bootstrapping effect for codas in Spanish parallels findings in other areas of bilingual acquisition such as morphosyntax (Gawlitzek-Maiwald and Tracy 1996). The paper is organized as follows: In Section 2, we contrast the syllable structure of German and Spanish and present relevant literature on the acquisition of codas by monolingual and bilingual children. In Section 3, we present the methodology; in Section 4, we present the monolingual and bilingual results on coda acquisition. The different patterns of coda acquisition shown by the German and Spanish monolinguals are compared to the bilingual results, which indicate a clear interaction between the two systems. Section 5 discusses the implications of these results for monolingual as well as bilingual research, and Section 6 summarizes the conclusions.

2. Acquisition of syllable structure: Codas In order to grasp the task of acquiring syllable final consonants both in German and in Spanish, we will first consider the syllable structure differences in the two

Acquisition of codas in Spanish and German 193

languages. We will then turn to the predictions that can be made as regards monolingual and bilingual acquisition, taking into account the results of previous research in this field. 2.1 Syllable structure differences in German and Spanish Our analysis is based on current theoretical accounts of the syllable that assume the following internal organization: The syllable is divided into onset and rhyme. The onset dominates prevocalic elements; the rhyme, all other elements. The rhyme is further divided into the nucleus (vowel) and coda (consonant(s) following the vowel). This study will restrict itself to the rhyme part of the syllable, more specifically to codas. Spanish has much less complex syllable structure than German. In wordfinal position, Spanish rhymes consist of a single vowel, a diphthong, or a vowel plus consonant. Only a restricted set of consonants may appear as codas in Spanish, predominantly coronal segments: /n/, /r/, /l/, /s/, /ð/, /θ/ (Harris 1983). Medial rhymes allow other Points of Articulation for nasals by means of assimilation to the following onset, as well as for obstruents (e.g., /p/ and /k/ are allowed, although they may be converted to /θ/ in some dialects). Medially there appear a few clusters, /ns/ and /rs/, the former often being reduced to /s/. In contrast, the German rhyme allows many more possibilities. It consists of minimally two positions, as suggested by forms with a long vowel or diphthong, and forms with a vowel and a consonant. However, the consonants which appear as codas in German may be labial, coronal, or dorsal, not just coronal as in Spanish. In addition, German allows rhymes of more than two positions: a long vowel followed by a consonant, a short vowel followed by two consonants, or even a long vowel followed by two consonants. In German, the difference between long and short vowels is relevant to syllable structure: Long vowels take up two positions in the nucleus; short vowels, one position (Hall 1992, Wiese 1996). Thus, with respect to many aspects of rhyme structure, German is more complex than Spanish. Figure 1 shows typical syllables in Spanish (1a), representing the words yo ‘I’, ley ‘law’, sol ‘sun’, and German (1b), representing the word Hahn ‘cock’, Mund ‘mouth’, Mond ‘moon’. Findings on adult corpora indicate that Spanish has 26.5% syllables with codas whereas German has 67% (Meinhold and Stock 1980). Delattre (1965: 42), basing his count on 2000 syllables taken from narrative and dramatic texts in German and Spanish, had arrived at similar results: 63% closed syllables in German and 28% in Spanish. Roark and Demuth (2000) have reported

194 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

s

s

O

R

s

O

R

N N

O

R

N

o

Ûi

e

l

s

N

C

o

l

Figure 1a.Syllable structures in Spanish. s

s

O

R N

h a

O C

˜

s

n

R N

m

~

O C

n

R N

t

m o

C ˜

n

t

Figure 1b.Syllable structures in German.

comparable trends using child-directed speech in English and Spanish. They found the mean percentage of coda consonants in English to be 59% and in Spanish 25%. Thus it is apparent from both adult and child-directed data that there are salient differences between the frequency of codas in Spanish and Germanic languages such as German and English. We now consider previous literature findings on coda and rhyme development in monolingual and bilingual children and introduce some of the assumptions of Optimality Theory as regards coda acquisition. 2.2 Acquisition of codas 2.2.1 Monolingual acquisition of codas It is widely proposed that the first stage in syllable structure development is consonant-vowel (CV) and that later stages are characterized by the development of codas, complex onsets and complex codas, and complex nuclei (see Levelt, Schiller and Levelt 1999/2000 for a recent proposal on typologies of syllable acquisition). Although an initial stage without codas is claimed to be a universal stage for all children, recent studies with English- and Germanspeaking children report the presence of codas from the onset of word production. In English, for example, Bernhardt and Stemberger (1998: 375) report that

Acquisition of codas in Spanish and German 195

one child, Morgan (age 0;11), produced codas in her first word up ([‘aph]) and on no occasion deleted a word-final coda. Bernhardt and Stoel-Gammon (1996) report that 35% of a group of normally developing children produced two of their first ten words with coda consonants. Similarly in German, Grijzenhout and Joppen’s (1999) subject, Naomi (age 1;2), produced forms with codas (e.g., an [an] “on” and ab [ap] “up”) in her first words. Findings on coda production in Spanish are less forthcoming but several reports in the literature indicate codas are produced later in Spanish than in English or German, that is, around the age of 1;10 or later (Garlant 2001, Lléo to appear, Roark and Demuth 2000). A direct comparison of syllable structure development in monolingual German and Spanish children was performed by Lleó et al. (1996), employing data from the PAIDUS study.4 They found no significant differences in the production of onsets but significant differences in the production of codas between the two groups. Specifically, they found that the German children produced significantly more codas than the Spanish children at two points of speech production, which they termed W1 and W2, following de BoyssonBardies and Vihman (1991). The former corresponds to 15 word types in one recording session, and the latter to 25. The superiority of codas in German at W1 was only marginally significant, whereas at W2 it already was highly significant. Lleó et al.’s (1996) findings indicate that relative frequency of codas in the target language influences children’s production of codas at a very early developmental stage. 2.2.2 Bilingual acquisition of codas Paradis (1996) reanalyzed data from Leopold’s longitudinal study (1949/71) of his bilingual (English-German) daughter Hildegard. Although the mean percentages of target monosyllabic words with codas are relatively similar in both languages (77% in English and 71% in German), Hildegard produced significantly more monosyllabic words with codas and more types of coda consonants in German than in English from the time period, 1;6 to 2;0. This suggests that even when two target systems are similar with respect to a linguistic feature, bilingual children may still distinguish them in some way. Paradis interprets this finding as evidence for two differentiated phonological systems. In a similar vein Ingram (1981/82) studied the phonological structures of an Italian-English bilingual child, L, at 2;0 years of age. L produced 81% of closed syllables in English and only 36% in Italian. Notice that these percentages parallel the relatively high frequency of closed syllables in target English as

196 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

opposed to target Italian. Based on these and related findings, such as the preference for monosyllables in English and ample use of reduplication in Italian, Ingram concluded that L had two separate phonologies. A comparable conclusion is reached by Johnson and Lancaster (1998) in their study of a Norwegian-English bilingual child, Andreas, between 1;2 and 1;11. Andreas produced more codas and a greater variety of coda consonants in English than in Norwegian. These and further phonological differences between the two lexica allow the authors (1998: 294) to conclude that Andreas “did not treat English and Norwegian as a unified language system”. On the other hand, they are cautious to conclude that two distinct phonological systems are at work. An important aspect of their findings is that Andreas’ development does not exactly match the development of monolingual children in either language. Garlant (2001) also reported differences between the coda production patterns of monolingual and bilingual children. She studied the phonological patterns of a group of three monolingual Spanish (mean age = 2;10) and three bilingual Spanish-English children (mean age = 2;6), using language samples in which specific stimulus words were elicited. Bilingual children demonstrated greater use of word-final phonemes in Spanish (particularly the phoneme /s/) than the monolingual children. They deleted codas 9% of the time in Spanish compared to the monolingual children who deleted codas 30% of the time. Certain phoneme contrasts could not be tested, however, because the monolingual children tended to avoid producing certain words with word-final consonants, whereas the bilingual children did not display these avoidance patterns. In sum, studies of bilingual children focusing on coda production have shown both differentiation and interaction between the two language systems. 2.2.3 Theoretical background This paper assumes Optimality Theory as a valid framework for studying phonological development, as has been demonstrated by the insightful work on prosodic acquisition, conducted by Levelt, Schiller and Levelt (1999/2000) or by Pater (1997), to just mention two examples. Whereas classical Generative Phonology describes phonological phenomena as the result of applying certain processes or rules to inputs in order to arrive at outputs in a derivational manner, Optimality Theory evaluates a set of output candidates that are relatable to an input by defining the optimal or more harmonic candidate. The grammar does not generate outputs, it consists of a hierarchy of constraints; the constraints are universal, but the hierarchy is language-specific, i.e. it is defined in each language depending on which outputs are allowed and which ones are banned.

Acquisition of codas in Spanish and German 197

Constraints are mainly of two types: markedness and faithfulness constraints. The former define unmarked structures, like syllables with onsets and without codas, whereas the latter minimize the discrepancies between input and output, i.e. they preserve identity between input and output. Markedness constraints relevant to this study are constraints regulating syllable structure, as for instance Onset, by which syllables are supposed to have onsets, and especially NoCoda, by which codas are not allowed. We have seen in Section 2.1 that there is a greater variety of codas in German — they can have all Points and Manners of Articulation (PA and MA) (except [+voiced] in the case of obstruents) — than in Spanish.5 Most theoretical treatments of codas imply that what is at stake is not a single coda condition, but several ones (Kaye 1989, Goldsmith 1990). Goldsmith expresses this by postulating that the coda position is a secondary licenser, which depending on the language, licenses different PAs, like [labial], [dorsal], etc., and different MAs, like [continuant]. In OT terms, the NoCoda constraint is short-hand for a set of constraints, like NoCoda[lab], NoCoda[-dors], NoCoda[cont], etc. We will be assuming an OT perspective on acquisition. According to this theory, the fact that at the beginning children omit codas and tend to produce predominantly CV syllables implies that their grammar is guided by structural constraints (Gnanadesikan 1995, Smolensky 1996). That is, children’s grammars at the initial state have markedness constraints like NoCoda outranking faithfulness constraints. 2.3 The effect of a bilingual context on coda acquisition The simultaneous acquisition of two dissimilar languages like German and Spanish, with their very different status for the NoCoda constraints, should shed light on bilingual acquisition and the issue of language interaction in the acquisition process. It is well-known that children manifest a lot of variation in the process of acquiring the same target language. Observing one single child acquire two different languages can give us a great deal of information on the relationship between the two systems being acquired, since individual variation can then be ignored (de Houwer 1990, Paradis and Genesee 1996). In the context of bilingual acquisition, i.e. simultaneous acquisition of two first languages, it has been suggested that either the two languages are autonomously acquired or they interact. In the latter case, three types of interaction have been proposed: transfer, acceleration or retardation (Paradis and Genesee 1996). In the case of transfer, one of the languages the child is acquiring is being influenced by some

198 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

grammatical phenomenon belonging to the other language. Acceleration and retardation refer to the development of some characteristic belonging to both target languages, which is acquired faster or slower in one of the languages in comparison with monolingual acquisition. 2.4 Overview of predictions on coda acquisition Given the general view that codas pose a certain degree of difficulty in acquisition, the difficulties posed by German codas should be greater than those posed by Spanish codas. This prediction is supported by the greater variety of codas in German than in Spanish (see Section 2.1 above). If we view grammar as a hierarchy of constraints, it is obvious that a language like German, that allows most of its consonants to appear in the coda, requires that markedness constraints like NoCoda be demoted underneath faithfulness constraints. In Spanish, the NoCoda constraints have a much shorter way to move; that is, the NoCoda constraints are ranked higher in this language, given the many restrictions against codas that we saw in Section 2.1. The longer way that the NoCoda constraints must be demoted in German reflects the higher complexity of the acquisition task in this language. According to this view, German children should need more time to acquire codas than Spanish children. However, in the literature (see especially Lindner and Johnston 1992) it has been argued that inflectional morphology tends to be learned faster by SLI German children than by SLI English children, due to its higher functional load in German. In a similar vein, Bates et al. (1987) show that both German- and Italianspeaking aphasics make fewer errors in article production than English-speaking aphasics, due to the higher morphophonological load of articles in the former languages. Because the rich inflectional morphology of German is often realized by means of codas, they might contribute a more salient trigger for the learner.6 Neverthelesss, as far as the grammatical import of codas in the target language, Spanish is comparable to German, in the sense that many final codas express inflectional morphology in Spanish, as well (see Lleó submitted). Therefore, the relatively similar import of final codas in the two languages, German and Spanish, does not lead to any clear predictions as regards acquisition. Two reasons could be adduced in favor of codas being acquired earlier in German than in Spanish (see Lleó to appear). First, the higher frequency of codas in German offer more evidence for the demotion of the NoCoda constraint (see Boersma and Hayes 2001 and Broselow to appear, who present a model of acquisition within OT, according to which constraint ranking and

Acquisition of codas in Spanish and German 199

reranking is sensitive to the frequency of the input categories). Second, the dominance of markedness constraints in the initial state affects not only syllable structure, but all levels of structure. Constraints on prosodic word structure lead to early truncation patterns in English, German and other languages. The resulting word shape is generally a Minimal Word, which is the least marked prosodic structure (Pater 1997, Salidis and Johnson 1997). The Minimal Word contains a foot consisting either of two syllables or of a single heavy syllable, comprised of two positions in the rhyme. In the case of monosyllables consisting of a short vowel followed by a consonant, the coda must be maintained, otherwise the resulting syllable would be light and the Word Minimality constraint violated. In monosyllables, which are very frequent in early German but not in Spanish (see Figures 3a–b below), the prosodic need to keep the coda overrides the effects of NoCoda. This should lead German children to produce more codas than Spanish children. Regarding the bilingual situation, since codas are present in both languages, the emergence of transfer is implausible. On the other hand, depending on the rate of monolingual coda acquisition in each language, interaction in terms of acceleration or retardation is plausible. Supposing that codas are acquired at a faster or at a slower rate in one of the languages, this could lead to acceleration or to retardation in the other language. Both would be a sign of interaction of the languages being acquired, with a positive effect in one case — this has been termed bilingual bootstrapping in Gawlitzek-Maiwald and Tracy (1996) — and a negative effect in the other.

3. Data and methods The data analyzed in this paper stem from several projects on phonological acquisition of German and Spanish, conducted in Hamburg with the support of the DFG: the monolingual project BIDS-PAIDUS, the bilingual project PEDSES and the bilingual project B3/E3 of the Research Center on Multilingualism. In the monolingual projects several monolingual children acquiring German in Hamburg (Germany) and Spanish in Madrid were followed longitudinally from the onset of word production until approximately 3;0 years. PEDSES as well as the present bilingual project B3/E3 analyze the simultaneous acquisition of German and Spanish by several children born in Hamburg, whose mothers are native speakers of Spanish. The children are audio- and video-recorded in their homes in unstructured play situations.

200 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

Three monolingual children have been randomly selected for German, Bernd, Marion and Thomas, and three for Spanish, José, María and Miguel. The three children of the bilingual project PEDSES, Irene, Robert and Stefan, have been included, and two children have been randomly selected out of B3/E3, Nils and Simon. The six monolingual children have only been exposed to either German or Spanish, whereas the five bilinguals were born in Germany, but were exposed to both languages from birth. Although the bilingual children did usually not receive much Spanish input outside their families, the primary caregiver during the first two years of life was the Spanish-speaking mother. The parents followed the “une personne une langue” rule, each talking in his/her native language to the child. The mothers would switch, however, to German in the presence of German speaking friends or family. The fathers, in general, had a very limited command of the Spanish language, except in the case of Simon’s father. Most children began attending a German day care center at 3;0, after the time considered in this article, except for Nils, who already attended one beginning at age 1;5. The data considered begin with the first produced codas until about 2;3 years of age or later, depending on the availability of the recordings, some of which are still being transcribed. In general, only spontaneous word productions have been taken into consideration. The recordings followed the naturalistic design, in which an interviewer and one of the parents played and talked to the child, while a second research assistant made the sound and video recording. In the case of the bilingual children, the languages have been recorded in separate sessions. If one of the parents was present, it had to be native speaker of the language in which the recording was taking place. Two different research teams made the recordings. We thus followed the one person — one language design, as well. The recordings have been glossed and phonetically transcribed by native speakers of the target language. Intertranscriber as well as intratranscriber reliability as far as presence vs. absence of coda always reached more than 90%. The child utterances have been entered into an electronic database (within 4th Dimension for the Macintosh). The monolingual analysis is based on 3,459 utterances in Spanish and 3,140 in German, and the bilingual analysis is based on 1,172 utterances in Spanish and 2,942 in German.

Acquisition of codas in Spanish and German 201

4. Results The results on coda production are shown on Figures 2–5 and in Tables 1–3. 4.1 Monolingual

Percentages

Figure 2 presents the group percentages of target coda production on a bimonthly basis — each group includes the mean percentages for three children. Tables 1a and 1b display the results for each individual child and separates codas into several positions within the word, showing also the MLU at each time point. The positions refer to word-final and word-medial stressed syllables. Word-medial unstressed syllables with codas appear seldom and therefore have not been considered, whereas word-final unstressed syllables with codas have been included, because they are very representative in both languages: endings like plural of nouns, especially in Spanish, or several verb endings in both languages are expressed with a final consonant on the second syllable of a trochee. Besides this position, in Spanish we have also separated a so-called proclitic position, which corresponds to the coda of articles like el and un. Percentage scores refer to the number of words in which codas were produced as a ratio of the total number of target words with codas. These figures merely show whether coda position is prosodically present and do not take into account the accuracy of coda production. Glides were included in the category of codas. Thus, the production of the Spanish word pez /pεθ/ ‘fish’ as [pεÛI] would still be counted as an example of coda production although the child produced the coda [ÛI] rather than the target coda /θ/ (see Lleó submitted for arguments supporting this decision). 100 90 80 70 60 50 40 30 20 10 0

German Spanish

1;1-1;4

1;5-1;6

1;7-1;8

1;9-1;10 1;11-2;0 2;1-2;2

Figure 2.Percentages of codas produced by the monolinguals.

2;3-2;4

202 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

Table 1a.Coda production by monolingual children: Spanish Age

MLU Final Stressed

José 1;2 1,00 1;3 1,08 1;5 1,00 1;6 1,00 1;7 1,02 1;9 1,07 1;10 1,20 1;11 2,15 2;0 2,15 2;2 2,69 2;3 2,12 María 1;2 1,00 1;3 1,08 1;4 1,05 1;6 1,03 1;7 1,19 1;10 1,21 2;0 1,13 2;1 1,53 2;2 1,75 2;3 1,84 2;4 2,26 Miguel 1;1 1,00 1;2 1,00 1;3 1,17 1;5 1,08 1;6 1,10 1;7 1,25 1;8 1,36 1;10 1,81 2;0 1,76 2;2 2,20 2;3 2,32

0/1 0/1 1/22 7/29 0/10 0/21 4/31 6/30 5/22

0% 0% 5% 24% 0% 0% 13% 20% 23%

1/5 0/6 0/16 0/37 0/37 0/19 3/19 1/13 10/36 7/32

20% 0% 0% 0% 0% 0% 16% 8% 28% 22%

1/1 0/6 9/9 33/41 24/27 5/8 11/16 5/11

Final Unstressed

Medial Stressed

0/1

0%

0/1 0/6 0/5 0/10 1/14 0/11 3/6

0% 0% 0% 0% 7% 0% 50%

2/5 3/8 1/34 12/63 4/77 12/83 35/74 13/46

40% 37% 3% 19% 5% 14% 47% 28%

2/3 8/39 31/60 24/43 19/34

67% 21% 52% 56% 56%

0% 0% 0% 0% 0% 0% 0% 0%

1/2 4/8 0/1 0/9 1/5 9/23 3/13 6/23 2/21

50% 50% 0% 0% 20% 39% 23% 26% 10%

0/2 2/8 2/5 2/4 1/3 4/10 0/4 3/8 2/15

0% 25% 40% 50% 33% 40% 0% 38% 13%

1/5 1/10 6/33 2/49 2/57 2/31 16/53 4/36 19/78 11/81

1/2

50%

3/6 3/3 7/11 2/8 4/8 19/33 14/17

50% 100% 64% 25% 50% 58% 82%

0% 25% 52% 24% 56% 71% 91%

4/7 57% 2/3 67% 4/16 25% 4/11 36% 19/27 70% 71/107 66% 54/93 58% 27/44 61% 88/128 69% 35/50 70%

4/7 100% 0% 4/10 1/2 100% 5/7 80% 9/13 89% 23/37 62% 9/12 69% 26/34 45% 6/11

0% 0%

Total

0/1 0% 0/1 0% 0/1 0% 2/6 33% 4/31 13% 8/69 12% 14/81 17% 12/147 8% 48/188 26% 65/158 41% 40/108 37%

0/1 0/6 0/7 0/4 0/1 0/6 0/11 0/13

0/1 0/1

Proclitic

57% 40% 50% 71% 69% 62% 75% 76% 55%

0/3 2/8 22/42 5/21 9/16 32/45 10/11

20% 10% 18% 4% 4% 6% 30% 11% 24% 14%

Two aspects of the results should be emphasized with respect to the numbers and positions of the produced codas. Coda production is much higher in German than in Spanish at all time points. This is clearly shown by the

Acquisition of codas in Spanish and German 203

Table 1b.Coda production by monolingual children: German Age Bernd 1;0 1;1 1;5 1;6 1;8 1;10 1;11 2;0 2;1 2;2 2;3 2;4 Marion 1;1 1;2 1;3 1;4 1;5 1;6 1;8 1;9 1;10 1;11 2;0 2;2 Thomas 1;5 1;6 1;7 1; 8 1;9 1;10 1;11 2;0 2;1

MLU

Final Stressed

Final Unstressed

1,00 1,00 1,00 1,00 1,13 1,18 1,49 1,19 1,13 1,16 1,16 2,09

0/3 4/5 53/57 41/53 48/51 59/59 65/65 57/58 94/102

0% 80% 93% 77% 94% 100% 100% 98% 92%

1,00 1,00 1,00 1,00 1,00 1,02 1,14 1,08 1,10 1,52 1,34 2,06

8/8 12/15 12/13 12/20 12/13 44/46 24/27 44/48 46/49 33/34 107/122

100% 80% 92% 60% 92% 96% 89% 92% 94% 97% 88%

1,00 1,00 1,04 1,03 1,01 1,20 1,50 1,81 2,28

23/29 15/23 62/62 23/23 26/26 69/69 71/72 40/43

79% 65% 100% 100% 100% 100% 99% 93%

Medial Stressed

Total

0/1

0%

0/1

0%

0/11 3/21 2/22 0/7 7/19 8/30 29/39 23/25 54/65

0% 14% 9% 0% 37% 27% 74% 92% 83%

0/14 16/41 73/103 60/80 81/98 96/120 154/164 117/123 238/260

0% 39% 71% 75% 83% 80% 94% 95% 92%

0% 0% 38% 8% 13% 9% 67% 92% 86% 72%

8/8 12/27 12/18 26/86 14/32 51/78 26/42 85/117 123/136 98/111 158/193

100% 44% 67% 30% 44% 65% 62% 73% 90% 88% 82%

0% 45% 36% 33% 82% 84% 83% 81%

23/36 25/45 76/99 33/53 63/70 152/172 147/159 156/183

64% 56% 77% 63% 93% 88% 92% 85%

9/15 18/24 19/20 26/28 29/31 60/60 37/40 90/93

60% 75% 95% 93% 93% 100% 93% 97%

0/6

0%

0/29 1/6 5/16 1/4 23/42 43/50 29/35 28/39

0% 17% 31% 25% 55% 86% 83% 72%

0/6 0/5 14/37 1/13 2/16 1/11 18/27 34/37 36/42 23/32

0/3 0/16 5/12 6/18 27/25 36/47 38/41 50/59

0% 0% 42% 33% 93% 77% 93% 85%

0/4 10/22 9/25 4/12 14/17 47/56 38/46 66/81

comparison of the group results of Figure 2. Whereas the German monolinguals produce as a whole very high percentages, always above 50% (except at 1;5–1;6), reaching almost 90% of coda production by the end of the second

204 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

year, the Spanish monolinguals never reach 50% of coda production as a group, not even at 2;4. If we look at the individual children in Table 1, we can see much agreement between the German children: they all reach very high values (over 70%) at 1;10 (Thomas already at 1;8). Only at the early time points, are there percentages lower than 50% in the case of Marion (until 1;6) and Bernd (until 1;8). In fact, all German children exhibit very high percentages of coda production from the beginning, with the exception of Bernd, who begins to produce over 50% at 1;10. The Spanish monolinguals show much more variation: whereas José and María hardly reach 40% even after 2;0 — José’s highest value is 41% at 2;2 and María’s, 30% at 2;1 — Miguel has higher percentages, which already reach 70% at 1;7. If we do not look at chronological age but take into account developmental criteria, like MLU, German children begin to produce codas with an MLU of 1, whereas two of the Spanish children, José and María, reach a substantial MLU of almost 2 before they begin with coda production.7 One-tailed t-tests on percentages of coda production by the two monolingual groups have been executed for ages 1;7–2;0 and 2;1–2;4. They show significantly higher percentages of coda production in German than in Spanish at both ages: p = 0.041 at 1;7–2;0 and p = 0.045 at 2;1–2;4. Likewise, a Mann-Whitney U-Test indicates a significant higher coda production by monolingual German in contrast to monolingual Spanish for ages 1;7–2;0 and 2;1–2;4 and for MLU values both under and over 2,00 (U = 0, (3, 3), p = 0.050). Besides the quantitative differences between the two groups during this development period, there is an important difference in relation to the position in which codas appear in the two languages. As Tables 1a and 1b show, in German there is a clear tendency for codas to appear in final position, whereas in Spanish there is a certain preference for medial (and proclitic) codas, especially in the case of José and María, whereas Miguel produces high percentages of codas independently of their position within the word. Mann-Whitney U-tests show that this preference for final over medial codas is significant in German (U = 0, (3, 3), p = 0.050), but not in Spanish. The main reason for the lack of clear-cut results in Spanish as far as position is concerned lies in the high degree of variability among the individual children: if we add the proclitic position to the medial one, which can be then considered non-final, José and María have a clear preference for this non-final position, whereas Miguel does not. If we only look at stressed syllables, it could be argued that the reason for the German tendency in favor of final codas lies in the fact that in German many of the early words are monosyllabic. In Spanish, on the other hand, the preferred word pattern is that of a trochee, monosyllables playing a very limited role.

Acquisition of codas in Spanish and German 205

Percentages

100 90 80 70 60 50 40 30 20 10 0

Monosyllables Disyllables Multisyllables

1;1

1;2

1;3

1;4

1;5

1;6

1;7

1;8

1;9 1;10 2;0

Figure 3a.Percentages of monosyllables, disyllables and multisyllables in the Spanish children’s lexica.

There is thus the tendency to produce the coda of the stressed syllable in each language, the medial coda in Spanish and the final coda in German. This is especially clear in German, whereas in Spanish, as argued above, this preference is only clear in the case of José and María, and it is even more obvious if we add the proclitic position to the medial one.8 Figures 3a and 3b show the percentages of monosyllables and disyllables typical of the two groups. The percentages refer to the cumulative vocabulary of each child, calculated in terms of word types, not tokens. In Spanish disyllables clearly outnumber all other words. In German there is a preference for monosyllables during the first time periods, and a trade-off with disyllables after 1;5, although monosyllables are highly represented all along. Monosyllables in Spanish hardly reach 10% during all time points. Another important difference relates to the segments appearing as the first codas in the two languages: in Spanish the first codas are glides, whereas in German the first produced codas include nasals, liquids and obstruents. More varied consonantal types are produced in Spanish, as well, but they make their appearence later. Some examples including all time points are given in the Appendix. 4.2 Bilingual Regarding the bilingual data, we present findings on five children, Irene, Robert and Stefan, from the project PEDSES, and Nils and Simon, from B3/E3. The group results, in Figure 4, show that children begin to produce relatively high percentages of target codas in both languages no later than at 1;9–1;10. When

Percentages

206 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

100 90 80 70 60 50 40 30 20 10 0

Monosyllables Disyllables Multisyllables

1;1

1;2

1;3

1;4

1;5

1;6

1;7

1;8

1;9

1;10

Figure 3b.Percentages of monosyllables, disyllables and multisyllables in the German children’s lexica.

compared to the monolinguals, coda production in German is very similar, reaching more than 80% at 1;11–2;0. In general, the percentages of syllables with codas in the German productions of the bilingual children individually resemble those of the monolingual German speaking children, but two of the children, Simon and Stefan show slightly lower values: at 1;6 Irene reaches 65% of coda production, at 1;7 Nils reaches 76% and Robert 94%, whereas at 1;10 Simon and Stefan produce 43% and 39% codas, respectively. Nevertheless, Simon reaches 80% in the following month and after that, his values grow steadily. Only Stefan, out of all the bilingual children, shows a delay of several months: he reaches 86% at 2;3. The group percentages of target codas produced in Spanish by the bilinguals are greater than those of the Spanish monolingual children, reaching 50% at 1;9–1;10 and 70% at 2;1–2;2. Whereas two of the Spanish monolingual children, José and María, never reached 50% of coda production during the period studied — only Miguel reached substantial percentages — all bilingual children steadily produced more than 50% of target codas beginning no later than at 1;11: Nils at 1;8, Irene and Stefan at 1;10, and Robert and Simon at 1;11. That is, even Stefan reaches higher percentages of coda production in Spanish, in spite of the slight delay we detected in German. One-tailed t-tests show significantly higher percentages of codas in bilingual Spanish, compared to monolingual Spanish, at age 2;1–2;4 (p = .039), but not at age 1;7–2;0 (p = .111), and no significant differences between bilingual and monolingual German: at 1;7–2;0, p = .292, and at 2;1–2;4, p = .486. Because we are interested in the relationship between the two grammars within the bilinguals, the group results give us a general view, but the relevant

Acquisition of codas in Spanish and German 207

Table 2a.Coda production by bilingual children: Spanish Age

MLU Final Stressed

Irene 1;1 1,00 1;3 1,00 1;4 1,00 1;6 1,13 1;7 1,16 1;10 1,17 2;1 1,18 2;4 1,65 Nils 1;3 1,00 1;4 1,00 1;5 1,05 1;6 1,00 1;7 1,02 1;8 1,06 1;9 1,06 Robert 1;1 1,00 1;2 1,00 1;3 1,00 1;4 1,00 1;6 1,00 1;7 1,00 1;11 1,11 2;0 1,14 2;1 1,23 2;2 1,00 2;3 1,33 2;4 1,79

Final Unstressed

Medial Stressed

Proclitic

Total

0/1 0/3

0% 0%

2/5 0/11 11/11 1/1 34/42

40% 0/2 0% 1/1 100% 0/3 100% 100% 1/1 100% 0/1 81% 17/19 89% 2/14

2/2 6/6 1/1 2/7 6/6 13/13

100% 100% 1/1 100% 100% 0/1 29% 0/2 100% 1/1 100% 4/4 100% 13/14 93% 11/13

1/1

100%

1/1

100%

4/7 0/1

57% 0%

24/31 11/15 6/6 4/5 5/6 48/52

77% 73% 100% 80% 83% 92%

4/7 0/1 1/3 38/76 54/71 17/21 6/7 7/16 74/84

57% 0% 33% 50% 76% 81% 86% 44% 88%

8/8 13/14 36/36 10/10 1/1 1/1 14/14

100% 93% 100% 100% 100% 100% 100%

0/1 1/31 7/20 1/5 1/1 1/9 12/18

0/1 0/3

0% 0%

2/8 1/16 11/11 2/4 58/81

25% 6% 100% 50% 72%

100% 100% 33% 22% 100% 88%

0% 0/1 0% 0/1

0% 0%

0% 0/1 14% 5/6

0% 83%

1/1 0% 0/1 0% 100% 1/1 85% 0/2

2/2 100% 8/8 0% 1/3 2/9 100% 12/12 0% 37/42

0% 3% 35% 20% 100% 11% 67%

perspective is offered by the individual data. Focusing on the percentages of codas by each individual child (Table 2), we have seen that in German the values of the bilinguals are not substantially different from those of the monolinguals: only Simon and Stefan showing slightly lower values and a relative delay. In Spanish, the picture is different: Leaving aside values with very low numbers of tokens, all children begin to produce substantial numbers and high percentages of codas before they reach the age of 2;0, Nils already at 1;8. Only

208 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

Table 2a.(continued) Age Simon 1;4 1;5 1;6 1;7 1;8 1;9 1;10 1;11 2;0 2;2 2;3 2;4 Stefan 1;2 1;5 1;6 1;8 1;9 1;10 2;0 2;1 2;2 2;3

MLU Final Stressed

1,00 1,00 1,00 1,04 1,19 1,33 1,00 1,21 1,28 1,44 1,80 2,13

2/19 1/3 0/6 1/37 1/5 0/16 5/14 8/16 7/8 12/14 36/64 37/44

11% 33% 0% 3% 20% 0% 36% 50% 88% 86% 56% 84%

1,36 1,29 1,00 1,50 1,00 1,56 1,54 1,83 2,00 1,97

2/5

40%

2/4 0/1 0/2 2/3 3/5 3/3 2/4 5/6

50% 0% 0% 67% 60% 100% 50% 83%

Final Unstressed

0/1

Medial Stressed

0/1

0%

1/1 0/12 0/4 13/29 7/12 7/17 16/38 17/23

100% 0% 0% 45% 58% 41% 42% 74%

Proclitic

Total

1/1 1/1 0/1 0/0 11/17 10/15

2/19 11% 1/3 33% 0/7 0% 1/38 3% 2/6 33% 0/28 0% 6/19 32% 26/50 52% 16/23 70% 38/52 73% 88/146 60% 92/113 81%

0%

4/4 2/2 19/21 25/27 28/31

100% 100% 90% 93% 90%

2/3

67%

1/1

100%

0/1 1/1 0/1 0/2

0% 100% 0% 0%

3/6 1/3 8/10 3/4

50% 33% 80% 0/1 75% 1/4

100% 100% 0% 0% 65% 67%

2/5 2/3 2/4 1/2 0/2 2/3 6/12 5/7 0% 10/16 25% 9/16

40% 67% 50% 50% 0% 67% 50% 71% 63% 56%

Stefan advances more slowly and produces 71% of target codas (only 5 tokens) at 2;1. Only during the first time periods is coda production in the case of most bilinguals not higher than for monolinguals. The findings from the bilingual children suggest that, although their coda production in Spanish is more advanced than for the monolinguals, in German it is still more advanced than in Spanish. However, there is a certain correlation between the coda production in both languages of some of the bilinguals: Simon and Stefan, who have the lowest coda production in German before 1;10 and 1;8, respectively, also produce very few codas in Spanish until that point: Simon’s codas in Spanish substantially increase at 1;10 and Stefan’s at 2;0. On the other hand, those children that reach relatively high values in German, Irene at 1;6 and Nils at 1;4, are the first to produce more codas in Spanish, too: Irene at 1;10 and Nils at 1;8. Only Robert, who is very advanced in coda production

Percentages

Acquisition of codas in Spanish and German 209

100 90 80 70 60 50 40 30 20 10 0

German Spanish

1;1-1;4

1;5-1;6

1;7-1;8

1;9-1;10 1;11-2;0

2;1-2;2

2;3-2;4

Figure 4.Percentages of codas produced by the bilinguals.

in German (at 1;6 he produces 7 out of 10 target tokens) needs a few more months in Spanish: he finally produces a very high number of codas (although amounting only to 50%) at 1;11. There seems thus to be a correlation in coda production between the two languages of the bilingual children, which points to an interrelation between the two grammars. In general, the Spanish of the bilinguals as compared with the Spanish of the monolinguals reaches a higher percentage of target coda production, which makes their Spanish percentages look closer to the German percentages of both, monolinguals and bilinguals. As far as word position, in German, bilinguals produce codas more often in final stressed syllables, like the monolinguals. But there is a difference in the position in which codas are produced in Spanish. As we have seen above, in Spanish, monolinguals as a group showed a weak tendency to produce more codas medially; individually, only José and María showed a clear tendency, whereas Miguel did not have any preference for either position. In Spanish, bilinguals produce codas more often in final stressed syllables, like German children do. This is especially the case for the codas produced during the earliest time periods. In Spanish the codas produced until 1;7 by Nils, until 1;10 by Simon and until 2;0 and 2;1 by Stefan and Irene respectively, occur more often in final stressed syllables, although given the relatively low numbers of tokens, the values do not reach statistical significance. Moreover, if we look at the segments produced, bilinguals do not begin exclusively producing glides in Spanish, as most Spanish children do, but begin with nasals, liquids and obstruents. This is illustrated in Table 3, where the superscripts indicate the number of times a particular sound was uttered in the coda.

210 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

Table 2b.Coda production by bilingual children: German Age Irene 1;1 1;2 1;3 1;4 1;6 1;7 1;10 2;1 2;4 Nils 1;3 1;4 1;5 1;6 1;7 1;8 1;9 Robert 1;1 1;2 1;3 1;5 1;6 1;7 1;8 1;11 2;0 2;1 2;3 2;4

MLU

Final Stressed

Final Unstressed

1,00 1,00 1,00 1,00 1,96 1,53 1,00 1,74 1,85

1/1 0/1 1/1 21/30 16/29 1/3 30/42 118/126

100% 0% 100% 70% 55% 33% 71% 94%

1,00 1,01 1,05 1,12 1,03 1,13 1,30

24/27 9/13 3/3 50/53 64/71 88/100

89% 69% 100% 94% 90% 88%

0/7 0% 1/1 100% 19/38 50%

1,00 1,00 1,00 1,00 1,02 1,00 1,09 1,11 1,20 1,48 1,67 2,15

5/8 10/11 10/10 13/13 18/18 19/19 86/87 91/92

63% 91% 100% 100% 100% 100% 99% 99%

2/2 6/6 1/1 10/10 24/27 14/14 27/28 154/156

Medial Stressed

Total

1/1 0/1 1/1 33/51 20/43 1/3 32/44 198/220

100% 0% 100% 65% 47% 33% 73% 90% 0% 71% 56% 100% 76% 87% 78%

70% 94% 100% 100% 85% 100% 95% 99%

7/15 47% 2/9 22%

5/6 2/5

83% 40%

1/1 100% 37/46 80%

1/1 43/48

100% 90%

0/3 0/7 0/3

0% 0% 0%

11/20 0/3 22/28

55% 0% 79%

0/3 24/34 9/16 3/3 61/80 65/75 129/166

38% 100% 87% 100%

7/10 16/17 11/11 23/23 45/53 36/36 146/153 324/327

100% 100% 100% 100% 89% 100% 96% 99%

3/8 3/3 33/38 79/79

Summarizing the data on coda acquisition, bilingual children resemble German monolingual children in several aspects of coda production. They produce similar percentages of simple codas, although with a mild delay in a few cases, produce them in the same word positions and produce the same types of segments. On the other hand, bilingual children do not resemble the Spanish monolingual children. They produce a higher percentage of target

Acquisition of codas in Spanish and German

Table 2b.(continued) Age Simon 1;3 1;4 1;5 1;6 1;7 1;8 1;9 1;10 1;11 2;0 2;1 2;2 Stefan 1;2 1;3 1;4 1;5 1;6 1;7 1;8 1;9 1;10 2;0 2;1 2;2 2;3

MLU

1,00 1,07 1,02 1,02 1,03 1,06 1,00 1,06 1,14 1,40 1,41 1,55 1,00 1,33 1,29 1,46 1,00 1,00 1,58 1,47 1,43 1,40 1,80 1,25 1,85

Final Stressed

2/2 2/3

100% 67%

11/13 103/107 91/102 14/15 28/30

85% 96% 89% 93% 93%

5/8 2/13 2/8 0/4 1/3 8/10 12/13 6/12 16/24 26/40 7/8 18/20

63% 15% 25% 0% 33% 80% 92% 50% 67% 65% 88% 90%

Final Unstressed

Medial Stressed

0/1 7/15 29/30 8/8 13/13

3/32 0/17 1/4 3/7 0/2 2/16 4/20 28/37 15/16 19/25

9% 0% 25% 43% 0% 13% 20% 76% 94% 76%

3/32 0/17 3/6 5/10 0/2 13/30 114/142 148/169 37/39 60/68

9% 0% 50% 50% 0% 43% 80% 88% 95% 88%

2/2 1/2 0/1 0/1 1/1 0/1 3/6 2/5 2/7 3/4 12/14 1/3 0/1

100% 50% 0% 0% 100% 0% 50% 40% 29% 75% 86% 33% 0%

2/2 6/10 3/16 6/16 1/7 1/4 14/23 17/24 9/23 24/34 40/63 10/15 19/22

100% 60% 19% 38% 14% 25% 61% 71% 39% 71% 63% 67% 86%

0% 47% 97% 100% 100%

1/2 4/7 0/2

50% 57% 0%

3/7 3/6 1/4 5/6 2/9 2/4 1/1

43% 50% 25% 83% 22% 50% 100%

Total

codas than Spanish monolinguals, the first produced codas are in final position and they are segmentally more varied than those of the monolinguals.

5. Discussion According to the tenets of Optimality Theory, the early appearance of codas in German poses a basic problem. As it was mentioned in Section 2.2, markedness constraints are supposed to outrank faithfulness constraints in the initial stage (Smolensky 1996), which should ban the production of codas, because of the NoCoda constraint. Nevertheless, as pointed out in Section 2.4, other markedness

211

212 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

Table 3a.Segments produced by Nils (bilingual) in German in coda position until 1;7 Age

Final stressed

Medial

1;4 1;5 1;6 1;7

l19 oÛ φ n s Û62 l6 b m Û6 n2 ] l20 θ s tθ k n ~Û f2 Û66 ~Û 5 Û?7 oÛ fÛ ÛI

p l6 ˆ 2 \ φ

Table 3b.Segments produced by Nils (bilingual) in Spanish in coda position until 1;9 Age

Final str.

1;4 1;5 1;6 1;7 1;8 1;9

n2 n3 θ2 Û6 l l˜ ÛI l n4 ] n5 s3 θ2 l Û62

Final unstr.

Medial

θ

s s8 θ

Proclitic l

n l k2 k 3 n3 x ç

l

constraints, not only faithfulness, could be responsible for why NoCoda is not complied with in early German. The early vocabularies of German children are characterized by a high percentage of monosyllables, as we have seen (Figure 3b), and these are the words that are initially produced with codas by the children. If words with a short vowel in the nucleus, e.g. Milch ‘milk’, Mund ‘mouth’, Schiff ‘boat’, were produced without a coda, simply as CV according to the NoCoda constraint, they would obviously violate binarity at the foot level. That is, the constraint Ft-Bin prescribes that feet are binary by having two syllables or, in the case they just have one syllable, they must have two positions in the rhyme, either a long vowel or a short vowel followed by a coda. Thus, in the case of monosyllables with a short vowel, the only way binarity can be fulfilled is by having a coda. Furthermore, in the literature on German stress it has been proposed that only superheavy syllables attract stress, i.e. syllables containing rhymes with three positions, either CVVC or CVCC (Féry 1998), thus, syllables always ending in a consonant. The presence of a consonant in such syllables may be interpreted by young children as a necessary condition for their being stressed, which commits them into producing the consonant independently of the length of the vowel.9 In general German children prefer CVC at the early stages, by which V can be either tense or lax, with a certain preference for it being lax (Kehoe and Lleó to appear). In Spanish 80% of the vocabulary is accented on the penultima syllable

Acquisition of codas in Spanish and German

(Sosa 1999: 59), that is, in this language trochees are by far the preferred metrical foot. This has a clear correspondence in that disyllabic words are the most frequent form of children’s utterances, as we have seen in relation to Figure 3a. Disyllables fulfill binarity at the syllabic level, which means that codas are not needed in this language for binarity reasons, or in order to assign stress, and are thus in general left unpronounced by the children at first. In the case of José and María, the first codas are generally produced on the stressed syllable, that is, in the non-final syllable. María even adds some codas in this position, which are not present in the target language words. José as well as María seem to be guided by a constraint requiring that stressed syllables be heavy, regardless of there being one more syllable in the same (disyllabic) foot. This is reminiscent of the constraint StressToWeight (STW), by which stressed syllables must have two rhyme positions. This constraint refers to phonetic (not phonological) weight, which is a consequence of a syllable being stressed (see Benua 1995) and begins to be implemented later on, whereas in some analyses of German phonology, codas are a prerequisite for a syllable to be stressed and they are implemented by German children from the start. As mentioned in Section 2.4, frequency of codas in German is also a reason for the early production of codas in this language. In Optimality Theory this frequency argument has a decisive importance, because the overwhelming numbers of codas in this language lead the child to demote the NoCoda constraint very early (Boersma and Hayes 2001). We thus interpret the very different courses of coda acquisition in German and Spanish as due to a combination of prosodic requirements (binarity constraints and foot make-up in a vocabulary with high percentages of monosyllables) and frequency effects.10 The findings on syllable structure by bilingual children indicate that the phonological systems of these bilingual children do not develop totally independently but influence each other. The results indicate a positive influence of German on Spanish, leading to an acceleration effect with respect to this one aspect of syllable structure production, codas. Compared to the monolingual Spanish children, bilingual children develop codas in Spanish at a faster rate. The German system appears to be unaffected. Since codas are more complex in German than in Spanish, the effects of simultaneous acquisition of these two languages were not necessarily predictable. A negative influence of Spanish onto German might have also resulted, leading to a delay of coda production in German, compared to monolinguals. This is not what our data indicate; rather they indicate a positive effect of bilingualism, in the sense that Spanish codas develop at a faster pace than in monolinguals.

213

214 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

A further important question refers to the reason for such an acceleration effect. We hypothesize that the input plays an important role. The high frequency of codas in the German input certainly provides overwhelming evidence for the demotion of the NoCoda constraint. A more subtle reason, a structural one, though, has been adduced above for the coda production of monolinguals, in the sense that codas are often needed in German in order to satisfy the requirements of binarity. Bilingual children produce codas in their German utterances, like the monolingual children, and this accelerates the production of codas in Spanish. On the other hand, it is interesting that precisely monosyllables, e.g. dos ‘two’, más ‘more’, pez ‘fish’, tren ‘train’, etc., are the words that have their first codas pronounced in Spanish. The compliance with binarity is not as absolute in Spanish as it is in German. That is, although binarity at the level of the foot manifests itself by means of trochees being the preferred metrical structure, monosyllables do not comply with it, because the language allows monosyllables with the form CV, like sí ‘yes’, no ‘no’, té ‘tea’, do ‘musical note’. At least for some children, like the ones reported on in this paper, the minimal word constraint, operative in German, seems to be passed from German into Spanish. Going back to the hypotheses on the relationship between the two languages of the bilingual presented in Section 1.3, bilinguals in Spanish manifest a positive effect, in the sense that they arrive sooner than the monolinguals at the target language. In the literature on bilingual acquisition, such positive effects are attributed to bilingual bootstrapping (see Gawlitzek-Maiwald and Tracy 1996 and Johnson and Lancaster 1998 for a similar proposal in the phonology). Once the child is led to demoting NoCoda in German, he/she will be led to demoting it in Spanish, too. Thus, it must be the influence of the German grammar on the Spanish one, which leads to such an acceleration. This is possible, though, if we accept that the two growing grammars of the bilingual interact.

6. Conclusion Our data on coda acquisition have shown that monolingual German children produce higher percentages of codas than monolingual Spanish children. This difference has been interpreted as due to the respective different effects of the prosodic structure of the two languages. Data on bilingual phonological acquisition in this same realm have shown that around the age of 1;4 to 2;0 years, the two developing systems interact. Bilingual children in German produce the same amount of codas as monolingual children. In Spanish, they



Acquisition of codas in Spanish and German

produce a greater amount of codas; a pattern that can be characterized as acceleration. Two concluding remarks can be drawn from our analyses. In the first place, the input has a decisive weight in the area of syllable structure. The overwhelming number of German codas plus the Spanish ones and the structural import of codas in German lead to early production of codas in Spanish. In the second place, it can be claimed that in some areas, bilingualism has a beneficial influence. This is suggested by the rapid development of codas in the bilingual children compared to the Spanish monolingual children.

Notes * The research reported in this paper was conducted in the Research Center on “Multilingualism” in Hamburg, supported by the Deutsche Forschungsgemeinschaft and the University of Hamburg, to which we express our gratitude. 1. See Jakobson (1941/1968) for a seminal formulation of this typological property in terms of universals and Prince and Smolensky (1993: 85–96) for a characterization of the typology in terms of Optimality Theory. The universals underlying this claim have recently been called into question on the basis of the Australian language Arrernte, in which the best analysis seems to favor the syllable type VC as basic instead of CV. See Breen and Pensalfini (1999). 2. The literature supporting this claim is abundant. One of the first to notice the basic role of onsets as opposed to codas in the acquisition of English is Branigan (1976). See Levelt, Schiller and Levelt (1999/2000) for a recent analysis. 3. Government Phonology and some recent acquisition work stemming from this theory analyze final consonants not as codas but as onsets of empty syllables or onset-nucleus syllables (Goad and Brannen 2000). This is not however the framework adopted in this paper. 4. Monolingual cross-linguistic project on the phonological acquisition of German and Spanish (Parameter Fixing in German and Spanish), supported by a grant of the Deutsche Forschungsgemeinschaft (DFG) to the first author. 5. German codas can also consist of consonant clusters, but we will concentrate on the acquisition of the coda position, and will disregard the effects of having more than one consonant in that position. 6. As suggested by an anonymous reviewer. 7. Notice, though, that Miguel begins to produce a substantial number of codas with a very low MLU (1,25), as the German children do. Since this measurement does not seem to be a superior predictor of coda production than age, in the remainder of the paper we will concentrate on age as the criterion for matching the coda data. 8. In Lleó (submitted) it is argued that the coda of many proclitic protodeterminers receives a secondary stress, which would explain the early production of such codas by some Spanish children.

215

216 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

9. In the literature on syllabic acquisition there is a debate on when vowel length is acquired. Although Fikkert (1994) proposes a rather late acquisition in Dutch, Kehoe and StoelGammon (2001) and Kehoe and Lleó (to appear) have shown that the tense lax distinction is acquired much earlier in English and in German, respectively. We will not enter into this debate here. 10. A reviewer has suggested attributing the early production of codas in German to Menzerath’s law, in the sense that “there is a negative correlation between the length of words as measured in syllables, and the length of syllables as measured in phonemes” (see Fenk and Fenk-Oczlon 1993: 11). This correlation seems to be due to an economy principle in the processing of (linguistic) information, related to the limitation of our immediate memory span for auditive pattern recognition (Fenk and Fenk-Oczlon 1993: 21ff). We do not know what memory span is available at the one-word stage. If this were the reason for not omitting codas in German, they should not be omitted independently of the vowel being tense (or long) or lax (or short). Data on coda production in English as well as in German (Kehoe and Stoel-Gammon 2001, Kehoe and Lleó to appear) show that the consonant is more readily omitted after tense than after lax vowels (at first, especially in English, tenseness substitutes length). This being the case, we think that Menzerath’s law is too coarse a sieve to distinguish between coda retention and coda omission in German and thus cannot account for the child data on codas.

Appendix. Examples of codas produced by monolingual children 1. Monolingual Spanish José

1;7,27 1;9,2

[βub.βhf] [b>Ûu] [p7j] [p7Ûi .da] [te˜Ûi ] [thfÛi .tha˜] [‘æÛI.thf˜h] [ta˜n.th6] [fæn.th6] [‘7ç.t7] [t7n] [ha.bi.a.‘u.na.βes] [ban.de]

pulpo flor pez lengua tren asusta esto diente elefante este ten había una vez grande

‘octopus’ ‘flower’ ‘fish’ ‘tongue’ ‘train’ ‘(it) frightens’ ‘this’ ‘tooth’ ‘elephant’ ‘this’ ‘take’ ‘once upon a time’ ‘big’

1;3,6 1;4,21

[t‘7˜s] [aÛI.t6h] [‘aÛi .ki˜.ja]

tres cuento rastrillo

‘three’ ‘tale’ ‘rake’

1;6,3

[aj.dæ˜] [ndaj.j7] [aj.d7.di] [‘æl.xu.wf]

diente diente calcetín el cubo

‘tooth’ ‘tooth’ ‘sock ‘the bucket’

1;10,3 1;11,23 2;0,3 2;2,12 2;3,13 María

1;10,17

Acquisition of codas in Spanish and German 217

2;0,12

[kfn] [a.v~Ûi .j!] [7Ûi .do] [a.vu˜l] [‘w7.l!.m"l] [s7ÛI] [βe].g!]

jabón a guardar esto azul huele mal seis venga

‘bacon’ ’(let’s) put it away’ ’this’ ‘blue’ ‘(it) stinks’ ‘six’ ’come on’

[phIç˜] [v7].ga] [‘!.pIç] [u.b7ÛI.je] [pa.p7l] [el.bes] [ba.di.do˛] [t7s] [6s.b!.>7.tes] [i.βus.k!n.gwfn.dow]

lápiz venga lápiz un peine papel el pez pajaritos tres espaguetis y buscan cuentos

‘pencil’ ‘come on’ ‘pencil’ ‘a comb’ ‘paper’ ‘the fish’ ‘birds’ ‘three’ ‘spaghetti’ ‘and they look for tales’

[‘a˜m] [˙a.βa.b7n] [b7t] [hin.t7ln] [hant]

Arm arbeiten Bett hinstellen Hand

‘arm’ ‘to work’ ‘bed’ ‘to put (down)’ ‘hand’

Marion 1;1,7 1;2,23 1;3,27 1;4,23 1;5,24 1;6,7 1;8,3 1;9,1 1;10,5 1;11,25 2;0,19 2;2,2 2;3,19

[baj] [baˆ] [baj] [haÛIç] [hap] [mam] [d~th] [wIf] [‘aÛ~f.maχ] [dIft] [pas.d!.‘aÛIn] [m~nth] [mãÛI.n6.hant]

Ball Ball Ball heiß ab Mann Loch Schiff aufmachen Stift passt da rein Mund meine Hand

‘ball’ ‘ball’ ‘ball’ ‘hot’ ‘off’ ‘man’ ‘hole’ ’ship’ ‘to open’ ‘pencil’ ‘(it) fits’ ‘mouth’ ‘my hand’

Thomas 1;6,0 1;7,1 1;8,16 1;9,0 1;10,4 1;11,2 2;0,6

[nas] [aÛuf] [v7k] [‘ap] [‘en.t6] [‘aph.ho˜ln] [hIn.wt7.l6n]

nass auf weg ab Ente abholen hinstellen

‘wet’ ’on’ ‘gone’ ‘off’ ‘duck’ ‘pick up’ ‘to put (down)’

2;1,09 2;2,11 2;3,11 2;4,22 Miguel 1;2,1 1;3,0 1;5,1 1;6,3 1;7,26 1;8,23 1;10,18 2;0,20 2;2,1 2;3,7

2. Monolingual German Bernd

1;8,6 2;0,23 2;1,12 2;2,17 2;3,22

218 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

2;1,3 2;2,11 2;3,15

[h~nth] [‘a]sth] [hox]

Hund Angst hoch

‘dog’ ‘fear’ ‘high’

References Bates, E., Friederici, A. and Wulfeck, B. 1987. “Grammatical morphology in aphasia: Evidence from three languages”. Cortex 23: 545–574. Benua, L. 1995. “Identity effects in morphological truncation.” In Papers in Optimality Theory, J. Beckman, L. Walsh Dickey and S. Urbanczyk (eds), 77–136. Bernhardt, B. H. and Stemberger, J.P. 1998. Handbook of Phonological Development. From the Perspective of Constraint-based Nonlinear Phonology. San Diego: Academic Press. Bernhardt, B. H. and Stoel-Gammon, C. 1996. “Underspecification and markedness in normal and disordered phonological development”. In Children’s Language 9, C. E. Johnson and J. H. V. Gilbert (eds), 33–54. Malwah, NJ: Lawrence Erlbaum. Boersma, P. and Hayes, B. 2001. “Empirical tests of the Gradual Learning Algorithm”. Linguistic Inquiry 32: 1–44. Branigan, G. 1976. “Syllable structure and the acquisition of consonants: The great conspiracy in word formation”. Journal of Psycholinguistic Research 5: 117–133. Breen, G. and Pensalfini, R. 1999. “Arrernte: A language with no syllable onsets”. Linguistic Inquiry 30: 1–25. Broselow, E. to appear. “Unmarked structures and unmarked rankings”. To appear in The International Journal of Bilingualism. De Boysson-Bardies, B. and Vihman, M. M. (1991). “Adaptation to language: Evidence from babbling and first words from four languages”. Language 67: 297–319. De Houwer, A. 1990. The Acquisition of two Languages from Birth: A Case Study. Cambridge: CUP. Delattre, P. 1965. Comparing the Phonetic Features of English, German, Spanish and French. Heidelberg: Julius Groos. Fenk, A. and Fenk-Oczlon, G. 1993. “Menzerath’s law and the constant flow of linguistic information”. In Contributions to Quantitative Linguistics, R. Köhler and B. B. Rieger (eds), 11–31. Dordrecht: Kluwer Academic. Féry, C. 1998. “German Word Stress in Optimality Theory”. Journal of Comparative Germanic Linguistics 2: 101–142. Fikkert, P. 1994. On the Acquisition of Prosodic Structure. The Hague: Holland Academic Graphics. Garlant, M. 2001. Spanish phonological patterns of young Spanish-English bilinguals. MA Thesis, University of Arizona. Gawlitzek-Maiwald, I. and Tracy, R. 1996. “Bilingual bootstrapping”. Linguistics 34: 901–926. Gnanadesikan, A. E. 1995. Markedness and Faithfulness Constraints in Child Phonology. University of Massachusetts at Amherst: Ms. ROA-67.

Acquisition of codas in Spanish and German 219

Goad, H. and Brannen, K. 2000. “Syllabification at the right edge of words: parallels between child and adult grammars”. In Issues in the Phonology-phonetics Interface. Amsterdam/ Philadelpia: John Benjamins. Goldsmith, J. A. 1990. Autosegmental and Metrical Phonology. Oxford: Blackwell. Grijzenhout, J. and Joppen, S. 1999. “First steps of German phonology: A case study”. ROA-304–0399. Hall, T. A. 1992. Syllable Structure and Syllable-Related Processes in German. Tübingen: Niemeyer. Harris, J. W. 1983. Syllable Structure and Stress in Spanish. A Nonlinear Analysis. Cambridge, Mass.: MIT Press. Ingram, D. 1981/82. “The emerging phonological system of an Italian-English bilingual child”. Journal of Italian Linguistics 1981/82: 95–113. Jakobson, R. 1941/1968. Kindersprache, Aphasie und allgemeine Lautgesetze. Uppsala: Alqmvist and Wiksell. Child Language, Aphasia and Phonological Universals. The Hague/ Paris: Mouton. Johnson, C. E. and Lancaster, P. 1998. “The development of more than one phonology: A case study of a Norwegian-English bilingual child”. International Journal of Bilingualism 2: 265–300. Kaye, J. 1989. Phonology: A Cognitive View. Hillsdale, NJ, Hove and London: L. E. A. Kehoe, M. and Lleó, C. to appear. “The acquisition of phonological vowel length: A longitudinal analysis of three German-speaking children”. Journal of Child Language. Kehoe, M. and Stoel-Gammon, C. 2001. “Development of syllable structure in Englishspeaking children with particular reference to rhymes”. Journal of Child Language 28: 393–432. Leopold, W. 1949/71. Speech Development of a Bilingual Child, Vol. I-II. Evanston: Northwestern University Press. Levelt, C. C., Schiller, N. O. and Levelt, W. J. 1999/2000. “The acquisition of syllable types”. Language Acquisition 8, 237–264. Lindner, K. and Johnston, J. R. 1992. “Grammatical morphology in language-impaired children acquiring English or German as their first language: A functional perspective”. Applied Psycholinguistics 13: 115–129. Lleó, C. to appear. “Child prosody and filler syllables: Looking into Spanish through the optimal window of acquisition”. To appear in Selected papers of the 4th International Conference on the Acquisition of Spanish and Portuguese, S. Montrul (ed). Somerville: Cascadilla Press. Lleó, C. submitted. “Prosodic licensing of codas in the acquisition of Spanish”. Lleó C., Prinz, M., El Mogharbel, Ch. and Maldonado, A. 1996. “Early phonological acquisition of German and Spanish: A reinterpretation of the continuity issue within the Principles and Parameters Model”. In Children’s Language 9, C. E. Johnson and J. H. V. Gilbert (eds), 11–31. Mahwah, NJ: Lawrence Erlbaum. Meinhold, G. and Stock, E. 1980. Phonologie der deutschen Gegenwartssprache. Leipzig: Bibliographisches Institut. Paradis, J. 1996. “Phonological differentiation in a bilingual child: Hildegard revisited”. In Proceedings of the 20th Annual BUCLD, Vol. 2, A. Stringfellow, D. Cahana-Amitay, E. Hughes, A. Zukowski (eds), 528–539. Somerville, Massachusetts: Cascadilla Press.



220 Conxita Lleó, Imme Kuchenbrandt, Margaret Kehoe and Cristina Trujillo

Paradis, J. and Genesee, F. 1996. “Syntactic acquisition in bilingual children: Autonomous or Interdependent?” Studies in Second Language Acquisition 18: 1–25. Pater, J. 1997. “Minimal violation and phonological development”. Language Acquisition 6: 201–253. Prince, A. S. and Smolensky, P. 1993. Optimality Theory: Constraint Interaction in Generative Grammar. Ms.: Rutgers University and University of Colorado. Roark, B. and Demuth, K. 2000. “Prosodic constraints and the learner’s environment: a corpus study”. In Proceedings of the 24th Annual BUCLD, Vol. 2, S. C. Howell, S. A. Fish, T. Keith-Lucas, 597–608. Somerville, Massachusetts: Cascadilla Press. Salidis, J. and Johnson, J. 1997. “The production of minimal words: A longitudinal case sudy of phonological development”. Language Acquisition 6: 1–36. Smolensky, P. 1996. “On the comprehension/production dilemma in child language”. Linguistic Inquiry 27: 720–731. Sosa, J. M. 1999. La entonación del español: su estructura fónica, variabilidad y dialectología. Madrid: Cátedra. Wiese, R. 1996. The Phonology of German. Oxford: Clarendon Press.



Interrogative elements as subordinators in Turkish Aspects of Turkish-German bilingual children’s language use* Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

This study has been conducted within the larger context of the project SKOBI and its research questions concerning the acquisition and development of linguistic connectivity in Turkish and German in bilingual children. The phenomenon analyzed is approached from the perspective of language typology, sociolinguistics, language psychology and acquisition theory. The paper is intended as a mainly qualitative study on one phenomenon in the language use of five Turkish-German bilingual children displaying innovative constructions of a kind not to be found in the data of the monolingual control group. It does, however, contain some additional quantitative information to give some idea of the relative frequency of the construction type in question. The paper starts with (1) the presentation of our basic pre-assumptions and research questions, (2) then takes a cross-linguistic look at the procedural structure of wh-elements in interrogative and non-interrogative usages and (3, 4) looks at their specific syntactic and discursive roles in monolingual Turkish in contrast to German. Part 5 presents the innovations under way in embedded constructions in the bilingual Turkish data; Part 6 is a quantification of the phenomenon, based on our data. Part 7 explains these constructions with regard to a possible expansion of their procedural compositionality. The conclusion (8) attempts to re-relate the findings to the more general discussion of language change on the basis of language contact.

1.

Preliminary remarks on language contact and bilingualism

Contact-induced cross-linguistic influence and resulting language change have been a widely discussed topic not only within a Turcological framework, but in theoretical linguistics and language acquisition studies as well. The need for

222 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

interdisciplinary work in the fields of bilingual first language acquisition and language change studies has been on the agenda for some time, see for instance Lightfoot (1991), Kaiser (1998, 2000), Brendemoen et al. (1999), Kroch (2001). Most recently, Meisel (2001), who is committed to a generative framework, stated the need to identify grammatical areas likely to cause problems in (bilingual) acquisition and to find convincing evidence of actual parameter resetting in acquisition data, possibly in the grammars of the respective weaker one of the two acquired languages in situations of unbalanced bilingualism.1 In this context, Meisel advances the notion of “incomplete acquisition” or “transmission failure”, thus focusing on parts of a grammar not acquired, rather than on new structures added to a grammar. This is quite in keeping with the idea that a given grammar can only realize one option of a given parameter. From this perspective, the emergence of a new structure is not really conceivable without the loss of another one instantiating its parametric opposite. — This, however, is not quite what we seem to be observing in our data, namely, a functional expansion of grammatical forms which coexist with their functionally more restricted, monolingual counterparts. With respect to language contact, the history of the Turkic languages can equally be characterized as one of numerous migrations and resulting contact situations involving Iranian, Slavic, Uralic, and Mongolic languages, as well as Chinese and Arabic (cf. Johanson 1992). Examples of Turkish languages or varieties influenced by language contact include Karaim and Gagauz in Eastern Europe, Chaladj and Kashgay in Iran and Chakasian in Southern Siberia.2 Due to migration from Turkey, linguistic groups of considerable size have been emerging in Western Europe for more than three decades. Millions of Turkish immigrants have lived in various countries, particularly in Germany, a situation which has led to interlingual discourse as a new form of bilingual interaction (see Rehbein 2001b).3 The new linguistic situation is influenced by the effects of both a partial separation from the old speech community and new contact situations with various European languages. The communities in Turkey and in Europe differ from each other structurally and therefore in linguistic respects,4 the differences vary according to the language domains, though. On the one hand, especially with regard to family structures, the spoken and homileic forms of discourse are more characteristic of Turkish than German society and occur more frequently. On the other hand, Turkish immigrants are confronted with more written, textual, constellations in their environments than in Turkey. The overall difference concerns all those settings in which institutional forms of discourse tend to emerge, and wherein the

Interrogative elements as subordinators in Turkish 223

productive use is required and, consequently, the acquisition of certain connective means is fostered. To sum up, bilingual Turkish children in Germany do not acquire Turkish in the same constellations as monolingual Turkish children growing up in Turkey do.5 It is in this sense that we speak of language contact phenomena in terms of discourse phenomena (Rehbein 2001a). The present study is situated within the larger context of the project SKOBI6 and its functional-pragmatic approach concerning the development of linguistic connectivity in the Turkish and German of bilingual children. — The research aim of SKOBI is to document and analyze the use of linguistic means of discourse connectivity from a longitudinal developmental perspective.7 Of special interest in this connection is the acquisition of linguistic procedures, which serve as a means to relate linguistic expressions to each other and to the general frame of knowledge (as well as lack of knowledge) of the involved speakers and hearers (see Part 2). In contrast to projects which study the monolingual development of discourse abilities (e.g. Berman and Slobin 1994), the SKOBI informants acquire their languages in a multilingual discourse context and in a situation of permanent language contact, in which they are confronted with formally differing, but functionally alternative linguistic structures. Internal connectivity of utterances in bilinguals is realized through “functional elements” — types of linguistic units that occur as single words in Indoeuropean, whereas in Turkish, they are often agglutinated as suffixes. While some of our data match typological distinctions like these quite well, others pose a number of problems. These concern the usage of interrogative elements (wh-elements in generative terminology)8 in non-interrogative constructions as used by bilinguals, in contrast to monolingual children. Interestingly, although “question words” are seen as function words by Muysken (2000: 164), and in manifestations of language contact in bilingualism as particularly sensitive, they are indeed single words in both languages.9 However, the terminology covers only some aspects of the bilingual Turkish-German development of function words in grammar and discourse. Our data comprises recorded and transcribed discourses from ten children between the ages of four and thirteen; five of them are Turkish-German bilinguals and the other five are monolingual in Turkish (52 discourses on the whole; 26 by monolinguals and 26 by bilinguals). For the present study, we took 797 utterances containing question words (517 used by monolinguals and 280 used by bilinguals) as a basis for our analysis, and extracted from them a subset of 169 utterances with non-interrogative usages of wh-elements (93 by monolinguals and 76 by bilinguals).

224 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

The paper proceeds in such a way as to gradually bring out the formal differences and common characteristics of the Turkish and German non-interrogative wh-constructions and their differential functions as they emerge in the children’s linguistic data (Parts 2 through 5). Part 6 is quantitative in that it establishes the relevance (in terms of frequency) of the complementizer-like usage of wh-elements in our data. Part 7 studies the phenomenon from the perspective of a procedural analysis in a functional-pragmatic sense and shows in what respect the language use — and possibly internal grammar — of the bilingual children studied seems to differ from that of the monolingual control group.

2. Wh-elements from a cross-linguistic perspective Linguistic units can be reconstructed from their role in speech actions. For this purpose, Functional Pragmatics provides among others the following categories: speaker (S) and hearer (H), the knowledge domains of S (: ΠS) and H (: ΠH) with elements of knowledge and lack of knowledge (: π, ¬π) together with the knowledge verbalized in the propositional act (: content) and its separate elements (: p, q, r etc.); in addition, the illocutionary and utterance acts (the latter including syntax, morphology, phonology, prosody) and in particular the theory of ‘linguistic fields’ (operative, expeditive, deictic, symbol and expressive fields) with the linguistic procedures that go along with it.10 After this theoretical framework, linguistic units frequently classified as “functional elements” are supplied with different procedures according to their linguistic field: for instance, “demonstratives” focusing H on external linguistic objects, including speaker and hearer, belong to the deictic field; “prepositions” that represent ontological relationships and “auxiliaries” that represent abstract processes to the symbolic field. The operative field, wherein language processes language, is the domain of morpho-syntactic phenomena and, because of the different functional sectors, it is structured varyingly: third-person “(subject-/ object-)pronouns” establish phoric reference to nominals; the “word order” operates on the topology of phrasal positions (constituents); “particles” containing operative procedures take single constituents into their scope and relate it to a discourse context; “definite articles” and “conjunctions” (complementizers) arise out of the change from deictic field to the operative field (paraoperative field transposition); therefore, a great deal of linguistic units responsible for carrying out utterance-internal connectivity can be assigned to the operative or to the deictic field. Altogether, the functional-pragmatic grammatical analysis

Interrogative elements as subordinators in Turkish 225

requires access to those linguistic units that form the basis of linguistic procedures. According to this approach, it might be said that, in the course of bilingual language acquisition, it is the linguistic procedural architecture of wh-elements and their constructions which develops and, if need be, changes. For this purpose, we shall make a theoretical analysis: first, of the individual structures of monolingual Turkish and German, and then of the structures of Contact Turkish, to show how the inner architecture of the expressions and their associated functions relate to specific principles of construction in these languages. Until we conduct a comparative analysis, which includes relevant linguistic procedures, it is impossible to reconstruct the innovative aspects of Turkish wh-constructions. At the same time, the following problems arise: namely, in Turkish as well as in German, “question words” can be classified as belonging to the same sector of operative fields, but the change of linguistic procedures during a transition from a “question word” to a “complementizer” applies only to German; in Turkish, for language-typological reasons, such a procedural change in a “question word” is not possible. How then can the productive non-interrogative wh-changes which occur in the Turkish of bilingual children be explained procedurally? 2.1 The cross-linguistic basis of wh-words If one takes a cross-linguistic look at the common basic characteristics of wh-constructions in both languages, the most striking feature is that wh-elements occur in both interrogative and non-interrogative constructions. Wh-words are therefore words with wh-features — but not necessarily words which, per se, always contribute to the realization of a question.11 Let us take a short look at the procedural composition of wh-words. (F1) Cross-linguistic basic characteristics of wh-words i. A first linguistic procedure corresponds to the wh-part, whose morphology marks a lack of knowledge (: knowledge deficit) felt by S; it is executed with ‘wh’, which, functionally speaking, opens an access to a specific propositional element; it is operative and not per se directed towards the hearer. ii. A second linguistic procedure, which is contained in the wh-word, verbally outlines the domain of S’s knowledge deficit. This procedure is also not per se directed to the hearer, but relates operationally to the linguistically classified representation of unknowing.12 In many languages this procedure is realized by case morphemes and adpositions.

226 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

iii. Wh-words are equipped with a directional potential (‘drift’) towards that knowledge domain which will fill the missing knowledge. 2.2 The cross-linguistic configuration of the interrogative usage of wh-words An interrogative wh-construction is realized in the speech situation with a propositional and illocutionary act. The cross-linguistic common features doubtless have something to do with the universal structures and functions of the basic human speech pattern of asking/answering questions. The pattern involves activating the “basic cross-linguistic configuration of the interrogative” between S and H by switching to the interrogative mode (cf. Rehbein 1999b). In order for wh-words to contribute to an interrogative illocution by accessing the hearer’s knowledge, the wh-construction needs to be directed towards H. (F2) The basic cross-linguistic configuration of the interrogative iv. Utterances made under certain conditions allow the wh-words to realize their potential drift with regard to the knowledge of H; these “certain conditions” are the different linguistic realizations of illocution which are specific to a language. When activated by the illocution of the question, the potential drift directs the wh-element towards the knowledge of the hearer (ΠH-domain), so that wh-words become genuine question words. v. The illocutionary act of asking a question leads H to conduct a mental search of his/her own knowledge in order to fill the space that S has outlined as his/her knowledge deficit. The concept of “illocution” refers to the action of gaining interactive linguistic access to the knowledge of H.13 vi. Discourse condition: wh-elements in interrogative constructions initiate a speech action sequence, implying a change of speaker, in which the request for knowledge is processed in a question-answer pattern. The result of the request is then verbalized as the propositional element q in the answer. 2.3 The cross-linguistic basis of the non-interrogative usage of wh-words In their non-interrogative usage, wh-words14 occur in utterances or parts of utterances without a hearer-directed interrogative illocution. This is the case, among other possibilities,15 when the wh-construction is syntactically embedded or otherwise semantically integrated within a superordinate proposition. Depending on the typological characteristics of the language in question, the

Interrogative elements as subordinators in Turkish 227

wh-element may or may not be syntactically involved in the process of subordination (see Parts 3 and 4 for more detail). The general discourse function of the embedded wh-construction may be described in the following way: the process of mentioning a knowledge deficit (which is placed somewhere in the interactional space without any specific drift), indicated by the wh-words and in which the hearer may or may not participate, takes the place of a sequential pattern of question and answer. But its task is to mark an empty propositional space within the framework of a superordinate construction instead of explicitly naming it. Most commonly, a verbum sentiendi is used to verbalize a higher level of propositional knowledge and thus produces a “self-answer” which is usually voiced in epistemic form (as a conjecture, assumption, knowledge deficit etc.) about a (missing) element of knowledge. For the non-interrogative uses of wh-constructions, we generally find the following structures: (F3) The cross-linguistic basic configuration of non-interrogative wh-usage iii¢ The wh-words unfold their potential drift, not in relation to the knowledge of H (: ΠH), but to the knowledge of S (: ΠS), within whose framework the single π-element has deficits. iv¢ The wh-construction with a propositional content p is embedded in another construction (= matrix construction) with propositional elements in such a way that the illocution is executed (only) by means of the construction with a propositional content q¢, into which the wh-construction is embedded (cf. Radford 1997, Section 7; see the comparison of minimalist with the pragmatic analysis base in Rehbein 1999b: 112, footnote 39). Thus, the propositional content q¢ is the verbal representation of the knowledge domain of S (verbalizes ΠS). v¢ H participates in the search for knowledge in the Π-domain of S by means of inference and co-construction (: normal hearer activity). vi¢ Discourse condition: within the context of the speaker’s concatenation of speech actions, a propositional content p, part of which is unknown to S, is embedded into another propositional content q¢ with an illocution. When p is embedded in q¢, the potential drift of the wh-words is brought to bear on S’s own knowledge domain ΠS with the wh-words accessing a verbal representation of ΠS in the shape of q¢. This is an operative procedure in language processing. In both uses, the wh-words have a potential drift in common, but differ according to how the wh-words participate in the illocutionary act of the utterance.

228 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

3. Interrogative and non-interrogative uses of wh-constructions in German With regard to German wh-words16 and to what has been said in Part 2, the procedural, compositional analysis by Rehbein (1999b) needs fine-tuning with respect to their functions as general and specific markers of a lack of knowledge, their operative and bi-procedural qualities and the way in which their hearerdirectedness depends on additional linguistic means. Some examples: w-er (who) w-essen (whose) w-es-halb (why) w-elch (which)

w-as (what) w-as für ein (what kind of) w-a-rum (why) w-ann (when)

w-ie (how) w-ie-viel (how much)

w-o (where) w-o-hin (where to) w-o-mit (with what) w-o-durch (by what means) w-o-her (where from)

w-ie-so (why) etc.

In German, there are various aspects of the realization of the illocution: The finite verb is in V2 position, the wh-word is often at the start of the utterance in topic (theme) position (e.g. in pre-V2; other forms of word order are possible). In some cases, the realization of the illocution may become evident from the “institutional constellation”, as, for example, when a doctor queries a patient (Rehbein 1993), or by directly addressing H, which S does with non-verbal means (facial expressions and gestures denoting query). The interrogative use of wh-constructions in spoken German can be observed in the transcript (1): Five-year old Binnaz and her interlocutor Nesli are talking about Binnaz’s kindergarten and her friends there. In score area 11, Nesli asks Binnaz: Was magst du am liebsten im Kindergarten? ‘What do you like best about kindergarten?’ In score area 13, Binnaz asks Nesli: Äh äh was liest du? ‘Eh eh what are you reading?’17 (1) Binnaz EFE7dt Kass 0661 Dok 1 PF 11–16 040500/SKO/EFE 7/Binnaz/Familie/Esen/Sony GX-WM670/661 220600/Esen/1:40/Atik/1:30/Tas¸demir/1:10/211100/Sony TC KB 920S Nes: Nesli, interviewer; Bin: Binnaz, girl, 5, goes to the kindergarten; Ben: Bengi, Binnaz’s sister, 9, 3rd form of primary school; Fer: Ferda, interviewer

Interrogative elements as subordinators in Turkish 229

11

Nes

Was magst du am liebsten im Kindergarten? ((3s))

TS-Nes What do you like best about kindergarten?

((Mikrofongeräusch)) 12

Nes

Hu’bs!

Malen. • • Malen. Nur malen? Nichts

TS-Nes INT

13

Drawing. • •

Drawing.

Just drawing?

Nothing else?

Bin

• Malen.

Ja‘.

TS-Bin

• Drawing.

Yeah.

Nes

anderes?

Und hast

TS-Nes

And do you

Bin

[I'⁄hm]. ((3s)) Äh äh was [liegst] du?

TS-Bin

INT (no)

Eh

[for: nein

14

Nes

eh

what are you reading? [for: liest

du da Freunde?

TS-Nes have friends there?

15

Bin

[Hmhµ]. • Soll ich dir sagen, wie die

TS-Bin

INT (yes)

Nes

Ja, sag mal, wie die heißen.

TS-Nes

Yeah, why don't you tell me what their names are.

Bin

heißen?

TS-Bin

16

• Do you want me to tell you what their names are?

Bin TS-Bin

Katharina, Aylin, Katharina, Aylin, Jaklin

Jaklin, ((atmet ein)) Mandy • ist auch meine Freundin. ((breathes))

Mandy • is my friend, too.

(F4) Interrogative use of German wh-words i. The procedures of the wh-element represent a knowledge deficit and a specification of its domain, respectively; the procedural composition of the wh-element is as described in 2.2. ii. The potential drift of the wh-element to the knowledge domain of H gets activated here because of the matrix-clause characteristics of the construction it is part of, namely its illocutionary force. iii. The realization of the illocution is represented in German by the finiteness of the predicate, a V2 position of the finite verb, a clause-initial scope position of the wh-element, a specific intonation of questioning a.o. iv. and v. as stated: H’s search for knowledge and his/her reply, containing what was found, form part of a sequence of speech actions.

230 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

In German finite subordinate clauses, the wh-word still represents a knowledge deficit, but, due to the absence of an illocution in the subordinate construction, is not directed to H. Instead, the referential potential of the wh-element becomes directed towards the superordinate construction, within which the subordinate construction is assigned a phrasal role. To return to (1), in score areas 13f, Nesli asks Binnaz, whether she has any friends in the kindergarten; to which Binnaz responds with an utterance containing an embedded wh-construction: Soll ich dir sagen, wie die heißen? ‘Do you want me to tell you what their names are?’, which is then reiterated by Nesli in score area 15: Ja, sag mal, wie die heißen! ‘Yeah, tell me what their names are.’ For comparison, the interrogative use is taken up again as (2), the non-interrogative uses as (3) and (4) respectively: (2) Was magst du am liebsten im Kindergarten? ‘What do you like best about kindergarten?’ (3) Soll ich dir sagen, [wie die heißen]? ‘Do you want me to tell you what their names are?’ (4) Ja, sag mal, [wie die heißen]! ‘Yeah, tell me what their names are.’

Several characteristics of the non-interrogative use are important: the nonrealization of an illocution, the connectivity (subordination in this instance), and the syntactic means by which these two are represented. Typologically, the lack of an illocution is formally represented by the finite verb’s being placed in a construction-final position. (It may be worthwhile noting in this connection that both main and embedded clauses exhibit finite verbs.) The linkage to the superordinate construction is achieved by means of a complementizer, which, in German as in most Indo-European languages, is a morphologically unbound element in a construction-initial position. Crucially, in German (but not in all other Indo-European languages), this complementizer happens to be the very wh-word, so that it might also make sense to speak of a kind of synthesis between the wh-feature and the subordinating element into one wh-complementizer. The propositional content of the embedded clause is separated by a strict syntactic boundary from the content of the main clause; the boundary itself is “bridged” by means of the wh-complementizer. For the wh-element then, this amounts to two changes with respect to the interrogative use: first, the illocutionary force of the construction is deactivated; second, a connective function realized by an operative procedure is now activated.18

Interrogative elements as subordinators in Turkish

Redder (2001) calls this role of the complementizer an “illocution stop”. It is interesting to note that this illocution-stop function exists in German but not in Turkish, which has no subordinate clause syntagmas with finite verbs. It would therefore not be suitable to characterize the non-interrogative use of wh-words in Turkish as illocution-stopping. When placed in the complementizer position, the wh-elements activate an additional operative procedure which causes them to change its linguistic field quality in the sense of a “field transposition” and assume the role of the connector to the superordinate propositional content. What Binnaz achieves with her non-interrogative wh-constructions in terms of discourse pragmatics, can be described as follows: in score area 14–15, she utters a question regarding the names in a non-interrogative form (wie die heißen ‘what they’re called’) using a wh-construction to hint at knowledge of names which makes her interlocutor (Nesli) curious by allowing her to participate, as it were, in the search for knowledge. In the matrix construction, S (Binnaz) takes her stance regarding what H (Nesli) is supposed to do with the knowledge, namely that H (Nesli) should ask for the next speech action, the naming of names, that S (Binnaz) has already mentally planned, and thus reveal the knowledge only hinted at in the wh-construction. As expected, H (Nesli) answers: Ja, sag mal, wie die heißen! ‘Yes, tell me what they’re called!’ (score area 15). In this manner, S (Binnaz) has integrated the knowledge hinted at in the wh-construction into the illocution of the query. In terms of rhetoric, S (Binnaz) makes it more attractive for H to ask for the revelation S has planned to make anyway because she lets H (Nesli) interact to share the knowledge deficit and its remedy by incorporating the wh-construction. Here we must also point out a condition that has a bearing on bilingual language processing (see 4.2.3), namely both speakers (score area 14 and score area 15) place the wh-construction after the superordinate matrix construction. This observation may be important to note since it has repercussions with respect to the communicative planning and organization of the discourse and distinguishes German from Turkish. Generalizing the discussion above one might summarize the communicative functions of the wh-constructions as follows: (a) to outline, without actually naming it, a nominal that S cannot or will not verbalize and thus to involve H; (b) to formulate an “attitude” in the superordinate matrix construction with regard to the knowledge hinted at in the wh-proposition; (c) to place the content of the wh-construction in theme position (‘pre-field’) or rheme position (Brinkmann 1962: 654–661).19 When playing a rhematic role, subordinated

231

232 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

wh-constructions, being clauses with clausal length, weight and complexity, tend to be placed in the post-verbal position (‘post-field’), unlike nominal rhematic elements and unlike their Turkish counterparts (comparable to the phenomenon of heavy NP shifts in English, but see Part 4.2.3 for more information on this point). (F5) Non-interrogative use of wh-words in German i. The matrix clause is finite, verbalizes the knowledge domain of S and contains an illocution. ii. The syntagma of the subordinate clause is opened with the wh-element. iii. The wh-element represents the specifically marked knowledge deficit of S, albeit without actualizing its potential hearer-directedness. iv. The wh-word stops the illocution; the utterance is no longer directed to H and/or makes exothetic reference (cf. Hohenstein and Kameyama 2002) to the speaker as holder of the domain of the knowledge deficit. v. The wh-element serves as a connector between the main and the embedded construction, linking the latter to the superordinate proposition. vi. The verb of the embedded clause is finite and placed in a final position. vii. The subordinate construction can occur in a post-verbal position (‘postfield’) with respect to the main clause even when it is rhematic.

4. Wh-constructions in monolingual Turkish20 In Turkish, we are concerned with, amongst others, the following expressions: ne (what) kim (who)21 hangi (which) nece (which language) kaç (how many) nere (which place), nerede (where), nereye (where to), nereden (where from), nereli (where from (descent)) niçin (why), neden (why), nasıl (how) etc. ne zaman (when) ne kadar (to what degree) etc. As a general fact, Turkish is a ‘wh-in-situ’ language with the wh-element usually remaining in a pre-predicative focus position (cf. Erguvanlı 1984 for this term and generally for word order in Turkish, Rehbein 1995 for a contrastive approach and Watanabe 2001 for one recent generative analysis in a more widely typological perspective). The generative term ‘in situ’ accounts for the

Interrogative elements as subordinators in Turkish 233

observation that the element in question is not moved to any clause-initial topic or complementizer position with scope over the whole construction, but rather remains in its basic position. This situation applies to interrogative and noninterrogative, embedded and non-embedded, constructions alike: (5) Bunu niye yaptı? dei-acc why do-pst ‘Why did she/he do this?’ (6) Bunu niye yaptıg˘ını bilmek isterdim. dei-acc why do-par-pss3sg-acc know-vn want-prs-pcop-1sg ‘I would like to know why she/he did this.’

The realization or non-realization of the interrogative illocution then seems to occur independently of the syntactic position of the wh-element. 4.1 Interrogative use of wh-words in monolingual Turkish (F6) Interrogative use of Turkish wh-words i. The procedure performed with the initial ‘n’ or ‘k’ component signalizes S’s knowledge deficit, i.e. the speaker’s lack of a π-element; this is an operative procedure. ii. The part of the word following ‘n’ or ‘k’ marks the specific domain of S’s knowledge deficit, in other words the knowledge domain in which the missing π-element should be sought for.22 iii. The potential drift of the wh-element to a knowledge domain (of H or S) becomes activated when the question word is used. The wh-element is most often rhematic and tends to be placed in situ in the pre-predicative focus position, independently of whether an interrogative illocution is realized or not. Here too, the general statements of Section 2 apply. (F7) wh-words in monolingual Turkish questions iv. In interrogative wh-constructions, Turkish actualizes the inherent potential drift of the wh-words, which is directed to ΠH with a finite construction and specific sequences of pitch (or, in self-sufficient utterances, as free word questions using simple question words), thereby performing the illocutionary act of asking a question.23

234 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

4.2 Non-interrogative use in monolingual Turkish The non-interrogative usage of Turkish wh-words differs from their interrogative use as formulated in step (iv). Importantly and in contrast to German, Turkish does not dispose of morphologically unbound, construction-initial complementizers. Basically, two options exist: syntactic subordination by morphological means on the non-finite verb stem and semantic integration of finite constructions into a superordinate proposition, which can occur under certain conditions. 4.2.1 Non-interrogative uses in non-finite constructions In Turkish, the non-interrogative use of wh-words forms part of a non-finite construction, such as an embedded participle construction acting as a complement, with the whole wh-construction integrated into the superordinate sentence scheme by means of a case-marking relationship (other possible means of integration are converbial constructions or postpositional phrases). Embedded clauses display a morphology (and partially syntax) untypical of main clauses.24 There are no finite forms in embedded clauses, nor are there morphologically unbound, clause-initial complementizers.25 The potential drift of the wh-words to a knowledge domain is “passed along” via the non-finite construction to the superordinate finite one. The total construction, based on the overall plan, realizes one illocutionary act, which is passed to the subordinate structures via a number of different relationships of syntactic government. The following example is to illustrate the various respects in which noninterrogative uses of wh-constructions in monolingual children’s Turkish differ from their German counterparts (presented above). The relevant utterance can be studied in score areas 41–42: (7) Can Bes Kass 0403 Dok 3 PF 32–42 310894/ENDFAS/Bes/Türe/Familie/Türker/Sony WM-F2041/0403a 091096/Dur/1:58/Yamaha AX-440 Can: Can, boy, 9; Sin: Sinem, interviewer; Mut: Can’s mother; Bet: Betül, neighbour; Ban: Banu, neighbour 32

Can

batar. • Hatta, e kumsalda da bi tane de ye‚il bi cam

TL-Can

even

IJ beach-LOC

TS-Can

Even,

I also found a green piece of glass at the beach ,

also one piece

also green

one glass

Interrogative elements as subordinators in Turkish 235

33

34

35

Can

buldum, bi tane de bundan, böyle bi ye‚il bi ta‚ buldum.

TL-Can find-PST-1SG

one piece

TS-Can

a piece of that,

Can

too

DEI-ABL

so

TS-Can

Such a stone I found,

one stone find-PTE-PCOP-1SG

büyüklükdeydi.

DEI

it was big.

It was that

Onu kaybettiime çok üzüldüm. ((4s)) DEI-ACC lose-PAR-PSS1SG-DAT

I wasvery sad to lose it.

Sin

H⁄

TL-Sin

IJ

well

shininess-PSS3SG

TS-Can Its, you know, shininess.

be sad-PST-1SG

bulmu‚tum.

green

so

Green.

I had found such a stone.

I-EQU

TS-Can

I think so, too.

Sin

Íu çok enteresan. • Midye gibi.

TL-Sin

DEI very

TS-Sin

That is very interesting.

Can

interesting

ta‚ benim çok ho‚uma

shell

one stone

Bence de. • Evet. Bu

TL-Can find-PTE-PCOP-1SG

TL-Can stone DEI1SG-GENvery

too

yes

DEI

Yes.

I find

like(POP)

It is like a shell.

gidiyo.

delight-PSS3SG-DATgo-PRS-3SG

TS-Can this stone very nice.

((7s)) Onlara • bakiyim.

Ban 39

very

Böyle yani parlakl⁄⁄. ((1s)) Ye‚il. ((4s)) Böyle bi ta‚

TL-Can so

38

DEI big-PCOP

green,

TS-Can size.

Can

one stone find-PST-1SG

green

TL-Can size-LOC-PCOP

36 Can

37

one green

((1s)) Íöyle bi ta‚ bulmu‚tum, ye‚il, o büyüktü. Íu

TL-Can

Can

so

I found such a green stone.

TL-Ban

DEI-PL-DAT

TS-Ban

Let's look at those.

look-OPT1SG

Sin

Al

TL-Sin

take-NEG-

TS-Sin

I'm not having

(Meyvenizi) suyu?

Mut TL-Mut

fruit-PSS2SG-ACC

TS-Mut

Juice?

Bet Ban

Ben bi de görebilir miyim?

TL-Ban DEI1SG one too see-MOD-PRS Q-1SG TS-Ban May I have a look, too?

water-PSS3SG

(

).

236 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

40

Kalp ‚eklinde,

Can TL-Can

heart

TS-Can

Look, it's in the form of a

Sin

mayay⁄m. Te‚ekkür ederim.

TL-Sin

OPT1SG

thank

TS-Sin

any.

Thanks.

make-PRS-1SG

Banu?

Bet Ban

(

Banu

).

TL-Ban

41

Can

form-PSS3SG-LOC

H⁄‘h⁄’. IJ

bak.

((2s)) Ben kalp ‚eklinde

TL-Can look

DEI1SG heart

TS-Can heart.

I don't know how it came to be in

Siyah (

Bet

form-PSS3SG-LOC

).

black

42

Ban

H⁄. Evet.

H⁄.

TL-Ban

IJ

IJ

Can

Yes

nas⁄l olduunu

TL-Can how

bilmiyorum.

become-PAR-PSS3SG-ACCknow-NEG-PRS-1SG

TS-Can the shape of a heart.

Bet

Tesadüf (oluyo

TS-Bet

chance

Ban

).

be-PRS

Chance happens.

‚

In Can’s utterance in score areas 41–42, the subordination is marked morphologically on the infinite verb form, in this case a nominalized form, with an ordered series of at least three suffixes, e.g.: lexical verb form + nominalizing suffix + possessive agreement marking + case marking. (8) illustrates how the non-finite construction (kalp ¸seklinde nasıl oldug˘unu ‘heart shape-pss3sg-loc how become-par-pss3sg-acc’) is governed by finite bilmiyorum ‘I don’t know’ via the accusative case marking on the nominalized form ol-dug˘-u-nu (‘become-par-pss3sg-acc’): (8) Ben [kalp ¸seklinde nasıl oldug˘unu] dei1sg [heart form-pss3sg-loc how become-par-pss3sg-acc bilmiyorum. know-neg-prs-1sg ‘I don’t know how it came to be in the shape of a heart.’

Interrogative elements as subordinators in Turkish 237

The wh-element nasıl ‘how’ remains in situ in a pre-predicate focus position. Although it does mark the existence of a knowledge deficit as well as its specific domain, it does not participate in realizing an interrogative illocution with direct access to H, since the non-finiteness of the construction and its syntactic ‘integration’ (cf. Ehlich 1999a) preclude the realization of any independent illocutionary force. The fact that the question ceases to be directed towards H, is then directly related to the wh-construction being embedded. As to the wh-element in itself though, it has no part in the syntactic process of embedding, nor does it contribute to the realization or non-realization of the illocutionary force. The role of the Turkish wh-element is then one of marking a knowledge deficit, independently of whether anything beyond the mere mention of this deficit is to happen or not. In terms of discourse pragmatics, what happens is that instead of directly imposing the mentioned knowledge deficit on behalf of the hearer, and prompting her to fill in the gap in the knowledge, the speaker (Can) simply marks a knowledge deficit and relates it to the negative answer part bilmiyorum ‘I don’t know’ in the matrix construction (thus leaving it to the hearer to fill the knowledge gap when listening). The knowledge deficit is placed by the speaker in the interactional space and suggested as an object of consideration, reflection or comment; the hearer (Betül) is invited to participate in the search for knowledge, but not directly requested to verbalize an answer by means of an interrogative illocution.26 The hearer’s task here is best described as one of inference or co-construction, where the co-actor (Betül) performs her H-role by speculating or commenting on the specific empty space: (9) (score area 42:) Bet: Tesadüf (oluyor). chance (happens

This utterance is a good illustration of the normal role of H when wh-words are used in non-interrogative ways: oluyor ‘happens’ picks up the oldug˘unu ‘become-par-pss3sg-acc’ of the previous utterance, but in stating Tesadüf (oluyor) ‘chance happens’, Betül, in the role of H, provides no answer to fill the gap in S’s knowledge but co-constructs the utterance made by S with its empty space in the domain of knowledge and thus participates in S’s process of reflection. Taking into consideration the complete fragment of the transcript (7) above, we have shown that embedded wh-constructions do indeed possess specific interactive functions, but it seems important to note that their illocutionary and connective role is less significant than in German.

238 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

(F7) Non-interrogative embedded wh-construction in monolingual Turkish i. (predominantly) non-finite verb form in subordinate phrase (participle, verbal noun) ii. with possessive (subject agreement) suffix iii. plus case marking suffix iv. wh-element in situ, i.e. in pre-predicate focus position within the embedded clause v. mostly position of the embedded clause in front of the matrix clause. 4.2.2 Non-interrogative uses with finite constructions There are, however, contexts in which finite forms can be used (for a more detailed version of the following classification, see Karakoç and Herkenrath 2002). These usages fall into two formal categories. One possibility is to simply connect the two part-utterances without any morpho-syntactic connectivity marker other than mere syntactic adjacency. This usage is restricted to a certain class of verba sentiendi and dicendi (cf. Kennelly 1993; see Kornfilt 1997: 1ff for the use of the verb demek ‘to say’ as a matrix verb to finite embedded constructions conveying direct speech). The other possibility involves the quotative particle diye ‘in/by saying’ as a connecting device. Turkish verba dicendi and sentiendi generally fall into two groups according to whether they integrate finite constructions with or without the quotative particle diye. In this context, we are interested in the integration of finite wh-constructions without diye only. In these cases, a finite construction with a wh-word is directly followed or, occasionally, preceded (depending on the theme-rheme-structure of the utterance as a whole, cf. Part 6) by another finite verbum sentiendi or dicendi. The interrogative construction is thus part of a longer concatenation of finite constructions, which may also involve the use of connective particles like ki or the quotative particle diye. Such finite constructions are morpho-syntactically paratactic, but semantically this is a matter of propositional hypotaxis. Importantly, the concatenation of finite constructions never involves any kind of complementizers whatsoever, for nor can the mentioned particles be analyzed as such, nor do the wh-elements display any connectivizing functions. This usage can be studied by looking at an excerpt of the transcription (10).

Interrogative elements as subordinators in Turkish 239

(10) Seda EFE 8 tk Kass 0602 Dok 4 PF 60–67 080200/SKOBI/EFE 8tk/Sed/Familie/Güler/Sony WM-F2041/602ab 270400/Tas¸demir/1:40/Atik/1:20/Tas¸demir/1:40/Esen/1:30/30082000/ Sony-TC-FX44 Sed: Seda, girl, 7, 1st form of primary school; Ken: Kenan, interviewer 60

Sed

Yengeçi ald⁄lar ((güler))

TL-Sed

cray fish-ACC

TS-Sed

They took the crayfish ((laughs)) into their hands.

’

Ken 61

62

Ald⁄lar.

TL-Ken

take-PST-PL

TS-Ken

They took it.

Sed

ellerine.

Ondan sonra n' apt⁄lar

TL-Sed hand-PSS3PL-DAT

DEI-ABL

TS-Sed

What they did after that, I don't know.

aft er-POP what do-PST-PL

Ken

Ellerine ald⁄lar. • Hiç korkmuyorlar m⁄?

TL-Ken

had-PSS3PL-DATtake-PL

at all

TS-Ken

They took it into their hands.

Were they not afraid at all?

Sed

be afraid-NEG-PRS-PL

bilmiyorum. H⁄’h⁄‘ Korkmad⁄lar hiç.

TL-Sed know-NEG-PRS-1AG

63

t ake-PST-PL

IJ

be afraid-NEG-PST-PL

Q

E:

at all

IJ

TS-Sed

They were not afraid at all.

Ken

Sen k/ sen korkuyo musun? Yen/

TL-Ken

DEI2SG

TS-Ken

Are you afraid?

Sed

Eh, I might have

be afraid-PRS

Q-2SG

Can you take a

deniycektim ama ((güler))

TL-Sed t ry -FUT-PCOP-1SG

[Yok,

but

no

TS-Sed tried, but ((laughs))

Ken

No no no

yegeç/ yengeci alabilir misin? Dokunabilir misin

TL-Ken

Crayfish-ACC

take-MOD

Q-2SG

touch-MOD-PRS

TS-Ken crayfish into your hand?

Q-2SG

crayfish-DAT

Can you touch a crayfish? [laughing

64

Sed

yok. Yok, yok]. Korkar⁄m ben.

TL-Sed no

no

TS-Sed no.

Ken TL-Ken

yengece?

no

E:: böyle ölü

be afraid-PRS-1SG DEI1SG

IJ

so

I´m afraid.

Eh

even when it's like

He’

Öyle mi?

IJ

So

Q

dead

240 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

65

Sed

bile olsa korkar⁄m. Görmekten bile korkar⁄m böyle.

TL-Sed even

be-CDCOPbe afraid-PRS-1SGsee-VN-ABL

TS-Sed dead, I´m afraid.

66

67

alt houghbe afraid-PRS-1SGso

I´m even afraid of looking at it.

Sed

Onda n s onr a • ye nge ç buldula r . Sonr a n'a pt⁄la r

TL-Sed

t hen

TS-Sed

Then • they found a crayfish.

cray fish

find-PST-3PL

Ken

Eve t.

Sed

bilmiyorum. Betül y⁄lan av⁄na ç⁄kt⁄.

t hen

what do-PST-3PL

What they did with it afterwards,

TL-Sed know-NEG-PRS-1SG TS-Sed I don't know.

Betül went snake hunting.

Ken In talking about a group of children, who played with a crayfish, the monolingual girl Seda, who is 7 years old, says (score areas 66–67): (11) Sonra n’ aptılar bilmiyorum. later on what-wh do-pst-3pl know-neg-prs-1sg ‘What they did (with it) afterwards, I don’t know.’

The construction Sonra n’aptılar bilmiyorum. ‘What they did after this, I don’t know.’ is perfectly grammatical in this context. The verbum sentiendi is placed after the proposition it refers to. Importantly and interestingly, the wh-element n’ (< ne) ‘what’ is in its original prepredicate focus position. The wh-construction in score areas 66–67 n’aptılar ‘what did they do/what they did’ is finite so that, syntactically, the realization of an illocutionary act is not automatically precluded. With the second finite verb (bilmiyorum ‘I don’t know’) tagged on (after a short caesura) in an asyndetic, paratactic manner, H is not given any interactional space to follow the wh-construction with his answer, and thus the potential drift directed to the hearer’s knowledge is neutralized before it can trigger Kenan’s answer sequence. The speaker, Seda, provides the answer herself, by linking the two finite syntagmas in the discourse in rapid succession. She thereby creates a rhetorical question-answer pattern which is in effect a “battery” of speech actions (cf. Ehlich and Rehbein 1977 on this term). One could conclude that finiteness alone is not sufficient to fully realize the illocution in Turkish either. Perhaps it is necessary for languages in general, to use an additional feature to completely address the hearer, e.g. with a pause providing an opportunity to answer, or by the speaker refraining from giving his own answer. This would mean that syntactic subordination, by means of

Interrogative elements as subordinators in Turkish 241

non-finiteness, is not the only possibility for neutralizing the potential directedness of a wh-element to the hearer; other procedures may have similar effects. 4.2.3 Order of words and clauses At this point, a closer look at a general feature of Turkish in Turkey might be useful, since it will be important when it is contrasted with the Turkish spoken in Germany. The data shows (as in (11)) that in Turkish, where a paratactic string of finite wh-elements occurs before finite matrix verbs, as described above, or when verbs occur in a serial construction, the forms are not simply set one after the other, but enter into a specific determination relationship: the following finite element affects the previous finite verb in a non-reciprocal fashion, creating the effect that it is subordinating the previous finite verb (“quasi-superordination”), or that the second predicate is of a higher order than the previous one (cf. Figure 1). “Pre-predicate position” Predicate 1 Wnite element 1 “quasi-subordinated”

A

B determines A

before

Predicate 2 (“higher order”) Wnite element 2 “quasi-superordinated” B

Figure 1.Counter-running relationship between sequence and determination within concatenations of finite but non-interrogative wh-constructions as a discourse preference in Turkish spoken in Turkey.

The typical counter-running rules of sequence and determination which Mundy (1955) and Johanson (1975, 1991a) described for the intra-sentential word order of Turkish spoken in Turkey can obviously be applied, with limitations, to extra-sentential word order in discourse, i.e. here to the relationship between determination and trans-sentential sequence of finite verbs. The preceding finite verb assumes the role of a pre-predication which specifies the predicate (on this term see Erguvanlı 1984, Rehbein 1995). This pre-predication of a finite element seems to be responsible for creating the impression that the preceding finite wh-construction is subordinate to the following construction which, then, adopts the function of a matrix, as described in the above example (11). Moreover, in Turkish, the inverse order is also possible: both finite and infinite subordinated or quasi-subordinated constructions can occur after the finite verb of a matrix construction when, in terms of the topic-focus structure

242 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

of the construction as a whole, they play a non-rhematic role in the sense of a devrik cümle construction (cf. Erdal 1999: 80ff for a more detailed analysis of the Turkish post-field as an anti-topic/anti-focus position). German is different from monolingual Turkish here in that the subordinate construction, when rhematic, tends to follow the finite verb of the matrix construction (unlike less complex rhematic elements like for instance NPs) yielding an ‘A before B’ sequence. The previous predicate, then, seems to be of a higher order than the following.27 When thematic,28 it tends to precede the matrix verb.

Predicate 1 (“higher order”) Wnite element 1 “quasi-superordinated” A

A determines B

before

Predicate 2 Wnite element 2 “quasi-subordinated”

B

Figure 2.Parallel relationship between sequence and determination of superordinated matrix construction A and non-interrogative wh-construction B as a discourse preference in German.

Anticipating the argumentation to come (6.3) one should note here that the acquisition of parallelism in sequence and determination, which German prefers in discourse matrix constructions, is obviously an essential pre-condition for the changes in functions noted in contact Turkish use of non-interrogative wh-constructions.29 (F8) Non-interrogative quasi-embedded finite constructions with wh-elements in monolingual Turkish i. finite wh-construction: no subordinating morphology on the verb, specifically no case marking that might relate this to a matrix verb; ii. the construction containing the finite verbum dicendi or sentiendi, e.g. bilmiyorum ‘I don’t know’, is not a matrix clause in the sense of a syntactic subordination relationship; iii. no hearer-directedness in the mentioning of the knowledge deficit (hearer is given no opportunity to provide an answer); iv. sequence and determination run counterwise in discourse: A before B, B determines A (A stands in pre-predicative position before B).

Interrogative elements as subordinators in Turkish 243

5. Embedded wh-constructions in bilingual children’s language use The following section presents examples of our informants’ use of wh-elements in the guise of complementizer-like particles. As will be seen, the children use almost any wh-element to connect finite embedded structures to a matrix construction. The repertoire of linguistic means to establish this kind of connectivity seems to partly draw on options marginally available in Turkish; for the remaining part, the constructions seem to have been inspired by German schemata. (12) Selim EFE7tk Kass 0501 Dok 2 PF 1–6 270495/ENDFAS/EFE7/Sel/Familie/Karton/Sony WM-F2041/0501a 260695/Karton/1:58/Karaca/1:22/Yamaha AX-440 Sel: Selim, boy, 12;2;Yıl: Yılnur, interviewer 1

[ He he ].

Sel Yıl

Okul kita b⁄d⁄r.

TL-Yıl

school

TS-Yıl

It's a textbook.

O okül/ okul [kita b⁄n] iç inde

book-PSS3SG-COP

DEI school-

book-GEN

[affirmative

2

• Nas⁄l/ • böyle

TL-Sel

always write-PRS

TS-Sel

It's always written there.• How /•

Yıl

3

[for: kitab⁄n⁄n

((1,5s)) Hep yaz⁄yo.

Sel

how

so

they always

neler var?

TL-Yıl

LOCwhat-PL

Sel

he p gonu‚uyo. Ve ‚e y de [ya p⁄yoh] • ka se ti [görüyoh],

there are

TL-Sel

always speak-PRS

and

TS-Sel

sort of like speak.

And we always do stuff • we see the tape,

thing

also do-PRS-1PL

tape

[for: yap⁄yoruz

4

inside-PSS3SG-

What's in this textbook?

Sel

onda n [s ona ] höre n ya p⁄yoh.

TL-Sel

DEI-ABL

TS-Sel

and then we listen to it.

POP

hear

do-PRS-1PL

• • Ne yi höre n

Yıl

TL-Yıl

what-ACC hear

TS-Yıl

What do you hear? [for: sonra

see-PRS-1PL

[for: görüyoruz

244 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

5

Sel

Ee, nas⁄l [gonu‚uyorlar]. (

TL-Sel

IJ

TS-Sel

)…

speak-PRS-3PL

How they speak.

Yıl

yap⁄yosunuz? Neyi duyuyosunuz?

O

TL-Yıl

do-PRS-2PL

what-ACC hear-PRS-2PL

DEI

What do you hear?

Those people.

TS-Yıl

6

how

Sel

He

TL-Sel

IJ

TS-Sel

Yeah.

Ee, Kevin ((1,5s))

Yıl

insanlar.

Kimler var mesela?

TL-Yıl

people-PL

who-PL

TS-Yıl

there arefor instance

Who is there for instance?

(13) Bilingual contact Turkish: Yıl: Neyi hören yapıyorsunuz? what-wh-acc hear do-prs-2pl ‘What do you listen to?’ Sel: Ee, [nasıl gonus¸uyorlar]. ij [how-wh speak-prs-3pl ‘How they speak.’ (14) (13) should read in standard Turkish: Sel: Ee, [nasıl konus¸tuklarını]. ij [how-wh speak-par-pss3pl-acc

In this example, the interviewer and Selim are co-constructing a questionanswer sequence. Selim is telling the interviewer about his English textbook and listening to tapes. The interviewer wants to know what the children listen to on the tapes. Selim answers nasıl gonus¸uyorlar ‘how they speak’. In terms of argument structure, the answer, nasıl konus¸uyorlar ‘how they speak’, is the direct object of the verb “hören yapıyorsunuz” (listen do-prs-2pl) and thus fits into the syntactic scheme designed by the question. Our argumentation is that the wh-construction is indeed subordinated, depending on the preceding utterance as on a matrix construction, semantically as well as syntactically. Selim, however, instead of marking this syntactic dependency by means of a case-marked nominalized form, nasıl konus¸tuklarını (how-wh speakpar-pss3pl-acc), uses a finite form here, gonus¸uyorlar (speak-prs-3pl), so that the construction seems to have some characteristics in common with (F8) illustrated by example (10). What distinguishes this example from the monolingual use, however, is that the matrix verb, a verbum sentiendi, does not belong

Interrogative elements as subordinators in Turkish 245

to the group of verba sentiendi with which such constructions occur in monolingual Turkish (cf. Karakoç and Herkenrath 2002). In terms of information structure, the utterance nasıl konus¸uyorlar, as an answer to the question preceding it and with respect to the preceding general talk about the tape listening activities, has a rhematic function. In the sense that both parts form a propositional whole, then, nasıl konus¸uyorlar is post-posed and carries a new informational value. It functions as a rheme and as such is being put into the postfield slot of the matrix construction. (This postposed rhematic position is also manifested in complex sentences as it will be illustrated in examples (16) and (20) below.) The syntactic position of the wh-element is ambiguous: due to the shortness of the construction it could be analyzed both as in situ and as constructioninitial.30 We analyze the wh-element nasıl (how-wh) as a combined wh-element plus complementizer. The procedures of which the wh-element is composed are slightly altered in that they have come to operate on the proposition of the subordinate structure and to connect it to the verb of the (preceding) matrix construction. There are other occurrences of this kind of phenomenon in our data, in which the children use other wh-elements, such as ne ‘what’, kim ‘who’, hangi ‘which’, as connecting elements to a “higher” matrix construction. (15) Sema EFE 7tk Kass 0500a Dok 2 PF 77–85 270495/ENDFAS/EFE7/Sel/Familie/Karton/Sony WM-F2041/501a 080695/Karton/1:66/Karaca/1:32/Tas¸demir/1:50/Babur/1:10/280801/ Yamaha AX-440 Yıl: Yılnur, Turkish speaking female interviewer; Sem: Sema, girl, 11;1 77

78

Sem Y⁄l

Öyle • iyi diil.

Hep küçük ‚eylerlen. H⁄⁄. Beenmiyorsun. Sizin

TL-Yıl

your

TS-Yıl

Who is your sports

Sporda Herr Neumann.

Sem TL-Sem

sports-LOC

TS-Sem

Sports... Mr Neumann.

Y⁄l

öretmen kim oluyo • sporda?

TL-Yıl

teacher

TS-Yıl

teacher?

who

to be-PRS

sports-LOC

Mr

Neumann

O DEI

Is

246 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

79

80

E' e'.

Sem TL-Sem

IJ

TS-Sem

No.

Y⁄l

sizin s⁄n⁄f öretmeniniz mi?

TL-Yıl

your

TS-Yıl

that your form teacher?

form

teacher-PSS2PL

Kim o sizin s⁄n⁄f

Q

who

DEIyour

Frau Münder und • Frau Brockmann.

Sem TL-Sem

Ms

TS-Sem

Ms Münder and Ms Brockmann.

Y⁄l

öretmen?

TL-Yıl

teacher

Münder

and

• MS

Brockmann.

Ûki tane two

TS-Yıl

81

82

83

84

piece

So there are

Sem

H⁄⁄⁄‘.

TL-Sem

IJ

Frau Münder de Sport yap⁄yo ama Frau

TS-Sem

Yeah.

Ms Münder also teaches sports,

Hµ

Münder

also sports

do

but

but Mr

Y⁄l

var yani?

TL-Yıl

EXIST PTC

TS-Yıl

two of them?

Sem

Herr Neumann… Frau Münder hep bize bak⁄yo kim

IJ

TL-Sem

Mr

TS-Sem

Neumann...

Sem

öyle bi ‚ey yap⁄yo. Hep böyle ba⁄r⁄nca yaz⁄yo. • • O/

Neumann...

Frau

Münder

always DEI1PL-DAT look-PRS

TL-Sem

such

TS-Sem

thing.

Sem

Herr Neumann da Spor bizlen yap⁄yo.

a

who

Ms Münder alwayswatches us (to see) who is doing such a

thing

do-PRS

always like this

shout-CV

write-PRS

DEI/

She always writes when shouting.

TL-Sem

Mr

TS-Sem

Neumann also teaches us sports.

Neumann

also sports

He/ Mr

DEI1PL-INS do-PRS

Evet • • O

Y⁄l

85

form

Who is it then, your form teacher ?

TL-Yıl

yes

DEI

TS-Yıl

Yes.

Who teaches

Sem

• • Politik? • • C⁄k • • • Herr

TL-Sem

••

TS-Sem

• • Politics?

Y⁄l

politik dersini kim veriyo?

TL-Yıl

politics

TS-Yıl

those politics classes?

class

who

give-PRS

politics

IJ

Mr

Mr

Interrogative elements as subordinators in Turkish 247

In (15), Sema is talking about her teachers and, in particular, about one teacher’s reactions to things happening in the classroom: Frau Münder hep bize bakıyo kim öyle bir ¸sey yapıyo. ‘Ms Münder always looks at us to see who is doing such a thing’ (score area 82–83). The subordinate construction kim öyle bi ¸sey yapıyo ‘who is doing such a thing’ has a finite predication and is postponed to the matrix construction Frau Münder hep bize bakıyo ‘Ms Münder always looks at us’, being subordinated to the matrix construction by means of kim ‘who’:31 (16) Frau Münder hep bize bakıyor [kim öyle bir ¸sey Frau Münder always dei1pl-dat look-prs [who-wh such a thing yapıyor]. do-prs ‘Ms Münder always watches us to find out who is doing such a thing.’

While in a bilingual Turkish construction like (16) the interrogative word kim ‘who’ is being used as a complementizer, in a monolingual standard construction the embedded clause would have been linked to the matrix verb by means of the quotative particle diye (in/by saying) as in (17) or, generally, by means of morphological subordination suffixes on the infinite verb, similar to the construction in (14). Important to note, however, is the non-involvement of the wh-element in the process of subordination. (17) Frau Münder [kim böyle bir ¸sey yapıyor diye] hep bize Frau Münder [who-wh such a thing do-prs say-c always dei1pl-dat bakıyor. look-prs

As the following example will illustrate, however, the emergence of innovative complementizers does not preclude the use of the more traditional, morphologically marked, means of subordination: (18) Enis EFE7tk Kass 0690/0691 Dok 2 PF 25–30 23012001/SKOBI/Efe7tk/Eni/Familie/Güler/Sony WM-F2041/690a 22012001/Esen/1:40/Tas¸demir/1:30/Tas¸demir/1:40/Babur/1:10/060601/ SonyWE835S Eni: Enis, boy, 8, 2nd form of primary school; Ken: Kenan, interviewer

248 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

25

((1s)) [Ihm’] ((1s)) O zaman da • • • paralar

Eni TL-Eni

DEI time

TS-Eni

And then• • •

Ken

too

there is money

Hmhm

Hm

TS-Ken

26 Eni TL-Eni

money-PL

IJ

var, gerçek para [diil] de oyuncak paralar. • Onlarla there is real

TS-Eni

money

not

but

toy

money-PL

DEI-PL-INS

not real money, but play money

I put those •

[for: deil

27

Eni

koyuyo(n) • ⁄hm ((1,8s)) koyarken, o zaman say⁄yom

TL-Eni

put-PRS-1SG

TS-Eni

28

put-CV

when putting them, then I count

Eni

kaç para var, [ordaya] yaz⁄yom.

TL-Eni

how

TS-Eni

how much money there is I write it down there.

money

there is

there-DAT

thattime

((1,8s)) Bi de one also

And they

Hmhm

TS-Ken

IJ

TS-Ken

IJ

TL-Eni

• • • bi ‚eyler boyuyolar, kaç tane olduunu one thing-PL

paint-PRS-3PL

TS-Eni

also • • • paint something,

Eni

say⁄yom.

TL-Eni

count-PRS-1SG

how many piece

bazen

be-PAR-PSS3SG-ACCsometimes

I sometimes count, how many there are.

• • Böyle. So.

TS-Eni

Ken

count-PRS-1SG

write-PRS-1SG

Ken

29 Eni

30

IJ

ehm

••

Evet.

This way.

• Hm‘

TS-Ken TS-Ken

Yes.

Being an obligatorily transitive verb, saymak ‘to count’, quite unlike bilmek ‘to know’ in example (10), yields an ungrammatical construction in Standard Turkish when used without a direct object. One way to create a connective relationship between saymak ‘to count’ and the question kaç para var ‘how much money there is’ would be to subordinate the interrogative part to the construction containing saymak ‘to count’. This, in turn, could only be done by means of an infinite form with the appropriate subordinating morphological cluster:

Interrogative elements as subordinators in Turkish 249

(19) [kaç tane oldug˘unu] bazen sayıyorum. [how many-wh piece be-par-pss3sg-acc sometimes count-prs-1sg ‘I sometimes count how many pieces there are’

The use of the wh-element kaç ‘how much’ to mark the subordinate character of the construction again resembles the typical procedure used in German. The linear ordering of lexical elements in this construction seems to reflect a sort of weakening of the transitivity of saymak ‘to count’. (20) o zaman sayıyorum [kaç para var] dei time count-prs1sg [how much-wh money existing ‘then I count how much money there is’

Here the word kaç ‘how much’ no longer has any illocutionary force as a question; instead, it introduces an object phrase as a complementary clause to sayıyorum ‘I count’; the matrix clause (o zaman sayıyom) is finite, the “relative clause” in postposition linked by kaç ‘how much’ is finite and represents an independent syntagma in the subordinate clause. However, one utterance later (score area 28–30), Enis is using a complement construction which is formed by means of a participle item plus possessive suffix and accusative case marking: Kaç tane oldug˘unu bazen sayıyorum ‘I sometimes count how many there are’. This is said in the original Turkish way. So, at least some of the children seem to have both possibilities at their disposal and, rather than completely giving up the monolingual forms, they add to them the innovative bilingual ones and continue to use both forms. 6. Quantitative aspects32 Let us take a look at the frequency of the innovative constructions in our data. The data consists of recorded and transcribed discourses from five TurkishGerman bilingual children between four and thirteen years of age, plus a monolingual control group of the same size (52 discourses on the whole; 27 by monolinguals and 26 by bilinguals).33 The presentation of the data focuses on the overall usage of wh-constructions, as such (in absolute figures), abstracting away from the age-related acquisitional process.34 Table 1 presents for each child the ages at the time of recording and the number of discourses as well as the overall number of wh-constructions and the number of non-interrogative wh-constructions. The bilingual children seem to be using more non-interrogative constructions (27% in contrast to 18% of the monolinguals).

250 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

Table 1.Total of wh-constructions (interrogative and non-interrogative) and noninterrogative wh-constructions for individual children Monolingual

Bilingual

Azim Gözde Seda Melike Tahir total

Burçin Selim Sema BinnazKubat total

age at first 6;11 recording age at last 8;10 recording checked dis- 9 courses total of wh- 115 construct. non-interr. 23 wh-construct.

4;8

6;10

6

4;3

–

9;11

8;11

7;10

5;6

5

–

8;3

6;10

6

4;10

–

12;9

12;8

11;7

6;4

5;9

–

8

1

3

6

27

4

6

11

3

2

26

212

45

32

113

26

132

57

42

23

42

15

11

2

517 [= N] 93 [= 18%]

16

37

13

6

4

280 [= N] 76 [= 27%]

Table 2 contrasts the occurrences of all non-interrogative wh-constructions with respect to the morpho-syntactic means of connectivity by which they are integrated into the superordinate matrix construction: Table 2.Linguistic means for non-interrogative constructions for individual children Monolingual

Bilingual

Azim Gözde Seda MelikeTahir total

BurçinSelim Sema BinnazKubat total

total of wh- 115 construct. non-interr. 23 wh-construct. finite non-int.17 wh-construct. finite non-int.1 wh-construct. with diye non-finite 5 wh-construction

212

45

32

113

517

26

132

57

42

23

280

42

15

11

2

16

37

13

6

4

28

11

9

2

14

35

13

6

4

–

2

1

–

93 [= N] 67 [72%] 4 [= 4,3%]

–

–

–

–

–

76 [= N] 72 [= 94.7%] –

14

2

1

–

22 [=23,7%]

2

2

–

–

–

4 [=5.3%]

It turns out to be the case that both monolingual and bilingual children use more finite non-interrogative wh-constructions with no morpho-syntactic connectivity marker in particular (including constructions with the quotative particle diye) than non-finite ones. Besides, the bilingual children tend to prefer

Interrogative elements as subordinators in Turkish

finite non-interrogative (subordinate) wh-constructions to constructions with the quotative particle diye and make little use of non-finite constructions with subordinating morphology, in contrast to monolinguals (~5% vs. ~24%). Moreover, bilinguals tend to use these constructions with a range of matrix verbs that seems to exceed that of the monolinguals quite considerably. As to the verba dicendi and sentiendi used as matrix verbs governing non-finite wh-constructions without the involvement of diye, we mainly found bilmek ‘to know’ and demek ‘to say’ in the data of the monolinguals. Bilingual children additionally use verbs like unutmak ‘to forget, saymak ‘to count’, yazmak ‘to write’, yapmak ‘to do’, bakmak ‘to look’, lesen yapmak ‘to read’, söylemek ‘to say’, ög˘renmek ‘to learn’, messen yapmak ‘to measure’, hören yapmak ‘to listen’, duymak ‘to hear’, experiment yapmak ‘to carry out an experiment’, göstermek ‘to show’, top çekmek ‘to shoot a ball’, koymak ‘to put’, plan yapmak ‘to make a plan’, olmak ‘to happen’. Most of these are verbs which monolinguals do not use in this way. In order to be able to show the differences between the monolingual and the bilingual usages, however, further differentiations need to be made: – – – –

the specification of the verba dicendi and sentiendi used in the matrix construction (as given above);35 the position of the wh-construction in relation to the matrix construction (pre-posed [+pre] or not pre-posed [−pre]);36 the position of the wh-element in situ [+sit] or not in situ [−sit];37 rhematic [+rhe] vs. non-rhematic characteristics [−rhe] of the wh-constructions post-posed to the matrix construction [−pre].

Table 3 shows the distribution of rhematic and non-rhematic pre- and postposed wh-constructions and of wh-elements in situ [+sit] and in constructioninitial position [−sit], in non-interrogative constructions. Obviously, the bilingual children do produce more instances of non-in-situ wh-elements [−sit] (28.95%, i.e. 22 out of a total of 76 non-interrogative wh-constructions) than the monolingual children (5.38% or 5 out of 93). Regarding the placement of the wh-construction in relation to the matrix construction, bilinguals seem to use a lot more wh-constructions following their matrix verbs (= [−pre]) (81.58% on the whole, i.e. 62 out of 76) than monolinguals do (16.12% or 15 out of 93). But, in particular and as far as our corpus has been checked, bilinguals only tend to place rhematic wh-constructions in a post-field position (= [−pre][+rhe]), a phenomenon which does not at all appear in the data of the monolinguals (61.84% vs. 0%). All in all, bilinguals seem to display a greater tendency to move wh-elements

251

total of wh-construct. non-interr. wh-construct. [+ sit] [− sit] [? sit] [+ pre] [+ & −pre] [−pre] [−pre] [+rhe]

212

42

19 2 21 32 – 10 –

115

23

13 – 10 16 4 3 –

9 2 4 13 1 1 –

15

45

7 1 3 10 – 1 –

11

32

Melike

1 – 1 2 – – –

2

113

Tahir

49 [= 52%] 5 [= 5.38%] 39 [= 42%] 73 [= 78.5%] 5 [= 5.4%] 15 [= 16.1%] –

93 [= N]

517

total

4 8 4 1 – 15 11

16

26

Burçin

Seda

Azim

Gözde

Bilingual

Monolingual

8 8 21 11 – 26 20

37

132

Selim

Table 3.Non-interrogative constructions; positions of wh-elements and of wh-constructions

3 4 6 2 – 11 9

13

57

Sema

2 – 4 – – 6 4

6

42

Binnaz

– 2 2 – – 4 3

4

23

Kubat

17 [= 22.37%] 22 [= 28.95%] 37 [= 48.68%] 14 [= 18.42%] – 62 [= 81.58%] 47 [= 61.84%]

76 [= N]

280

total

252 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

Interrogative elements as subordinators in Turkish 253

to a construction-initial position and to postpone wh-constructions to the matrix construction, even when these play a rhematic role.

7. Explanation: A micro-change in the procedural composition of Turkish wh-elements 7.1 Functional expansion of Turkish wh-words in bilingual child discourse For the following discussion let us recap on the three conditions for the use of wh-words which became apparent in the above examples of contact Turkish. (F9) Embedded constructions with wh-elements in bilingual contact Turkish i. finiteness of the embedded clause, i.e. no subordinating morphology, such as nominalizing, possessive, and case marking on the verb; ii. wh-element in complementizer position of the embedded clause; iii. post-field position (“Nachfeld”) of the embedded clause after the matrix clause Æ sequence and determination in the discourse are parallel: A before B, A determines B (B in postposition after A); wh-construction in postposition after the matrix construction; iv. rhematic status of the postposed wh-construction. Condition (i) should be kept in mind because the absence of non-finite subordinating morphology is what creates the communicative need to mark the subordination in some other form, e.g. with a “functionally changed” wh-word. Conditions (iii) and (iv) are concerned with the processing of propositional contents in discourse. Condition (ii) states that the Turkish spoken in Germany displays a more flexible usage of the in-situ position of Turkish wh-words. In our data, they occur in the clause-initial position, which could be argued to be comparable to that of complementizers in German, acquiring, as it seems, the ability to be used there. — With the rhematic embedded wh-construction in the post-field position, a parallel relationship between sequence and determination of the superordinate matrix construction A and the following non-interrogative wh-construction B is created as a discourse preference — an important precondition for the functional changes operating to make the contact Turkish wh-words analogous to their German counterparts. When condition (ii) is given we must look at the inner architecture of the Turkish wh-words in language contact contexts with regard to how they

254 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

function as complementizers. Therefore, the condition for use (ii) in the contact language German should be specified as follows: iia. In German — and in the spoken German of the bilingual speakers as well — wh-constructions are embedded when the wh-word subordinates the subordinate clause of a structural phrase position to the superordinate matrix construction, and when the method it uses is to “slot” the subordinate clause into the position that the superordinate construction has reserved for the phrase, thereby “opening up” the clause for processing by the proposition in the matrix construction. The subordinate clause can, through the wh-word (: complement opening), then figure within the representation of the speaker’s propositional knowledge in the superordinate clause. In this manner, the part-syntagmas of a complex utterance in German are processed successively. iib. One further important procedure is at work, which we shall analyze by referring to the concept of the “illocution stopper” suggested by Redder (2001).38 In her paper, Redder identifies an essential role for the complementizer in German in its ability to “stop” the illocution in the subordinate clause. The explanation for the change of functions described in conditions (i)–(iv) is as follows: an adaptation to the processing methods of German discourse occurs within the complex wh-constructions of bilingual speakers. The internal procedural quality of Turkish wh-words changes in such a way that they can link the subordinate clause to the phrasal slot of the superordinate construction, thereby connecting it to the propositional content of the main clause. The Turkish wh-elements then acquire an additional operative procedure that becomes activated in non-interrogative usage only and that enables them to occupy a new syntactic position within a matrix construction and to link a finite syntagma to another finite syntagma,39 working as a single connector word. However, what is crucial for bilingual usage is that the directedness of the wh-element is now turned towards the matrix construction. The result of the acquisition of both partial procedures is a wh-word with its own, independent operative potential, namely the ability to integrate the subordinate wh-construction into the proposition and the illocution of the superordinate construction: this process of expansion to include other procedures is called ‘adoption of procedures’. If, then, Turkish wh-words appear during a bilingual discourse, in a complementizer-like position at the beginning of a subordinate clause, with a finite verb, the operative procedure of this syntactic position is performed with

Interrogative elements as subordinators in Turkish 255

them, whereby they make the syntactic link between the subordinate clause and the superordinate proposition; over and above that, having made the connection to the proposition, the Turkish wh-word is responsible for integrating the clause into the superordinate illocution and, probably, vice versa. Contact Turkish wh-words, in complementizer-like positions, thus adopt an independent operative procedure, which, for reasons of syntax, they do not have in the Turkish spoken in Turkey. The operative linkage of constructions and the integration into the superordinate illocution this involves is, in essence, performed by one single word set in a specific syntactic position and has ceased to be managed solely by means of a combination of non-finite suffixes; this being so, adoption of such procedures is evidence of the analytic, i.e. Indo-European, type of language gaining at the expense of the agglutinative. Let us recap the changes of functions seen for non-interrogative wh-constructions in contact Turkish as shown by the data collected from our bilingual speakers: (F10) Embedded constructions with wh-elements in bilingual contact Turkish i. Cancellation of the infiniteness of the embedded clause, i.e. no subordinating morphology, such as nominalizing, possessive, and case marking on the verb; ii. final position of the finite element of the embedded clause; iii. post-field position (“Nachfeld”) of the embedded wh-construction after the matrix clause: (speech processing), rhematic information of the post-posed wh-construction; iv. wh-element in complementizer position of the embedded clause; v. the Turkish wh-word becomes able to link the embedded clause to an empty propositional space in the superordinate syntagma Æ the linking procedure is an operative one: expansion of function; vi. the illocution is delegated to the superordinate construction; ((v) and (vi) comprise the ‘adoption of procedures’); vii. the connectivity between the embedded clause and matrix clause is transferred via relational syntactic procedures to individual words (wh-words); viii. the discourse purposes of wh-constructions used in the non-interrogative apply. 7.2 Aspects of the clausal order of wh-constructions in bilingual discourse The above analysis should now be placed within its theoretical context, the theory of language processing in the form of language-specific verbal planning (see 4.2.3

256 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

above). In the case of query processing, the knowledge deficit, which is marked by the question words, would have to be remedied by the hearer with knowledge from outside the utterance, whereas in the case of processing an integrated, complex construction, the knowledge deficit marked by the question words would have to be remedied by the speaker, in other words by a process of reflection on the syntactically organized knowledge structures.40 The linguistic processing of elements specifically marked as unknown, which need to be integrated into the total propositional content of an ongoing complex utterance act, is an act of self-reflection and could be considered as a reference to own knowledge when speaking. This concept would apply equally to German and Turkish. (F11) Principles for processing the bilingual discourse i. the syntax is influenced by the discourse situation, which imposes short sentences and short subordinate clauses; ii. the desire to subordinate appears during speech through the postponement of propositional contents which are mentally planned as dependent elements; iii. involving H when S consults with own knowledge activates a process of coconstruction and inference as H’s contribution to the speech acts of S; iv. leaving symbol field names open during conversation proves to be a rhetorical device to encourage the mental interaction and participation of H. Based on these language processing principles, it would seem that a bilingual discourse is emerging that lays the foundations for polyglot skills, one of the languages being that variety of Turkish which is used by bilingual speakers and which shares certain discourse principles with German. These seem to manifest themselves morpho-syntactically in the finiteness of the embedded clause predicate, the position of the wh-element as well as the position of the whole embedded clause in relation to the superordinate construction of which it is a part. The last two characteristics are not always as easy to identify as one might like, because word order in monolingual Turkish is variable and not simply definable in a strictly syntactic sense. It largely depends on and interacts with discourse-related notions such as topic and focus marking, definiteness vs. indefiniteness and the like (cf. e.g. Erguvanlı 1984). A wh-element may thus indeed be moved to a clause-initial position if it happens to represent the topic of the utterance. Moreover, whenever the embedded construction consists of no additional elements, the wh-element may appear in the preverbal focus position and still be the first element in the clause. As for the position of the embedded wh-construction as a whole, it may be placed after the finite predicate of the matrix construction in monolingual Turkish, if associated with a backgrounding

Interrogative elements as subordinators in Turkish 257

(i.e. non-rhematic) type of information (cf. Erguvanlı 1984: 50ff). The use of post-posed wh-constructions for presenting rhematic information, however, seems to occur in bilingual discourse only.

8. Towards a new variety? On the background of the data presented and of what has been said above, the innovative constructions used by the children strongly suggest that changes as to the grammatical representation of the elements in question are under way. These crucially concern their procedural compositionality. Pragmatically speaking, the procedural linguistic potential of the interrogative elements receives an additional, operative, component, and, by this, undergoes a change within the operative field in that the elements integrate the new operative procedure or, to put it in traditional terms, incorporate a further word class structure.41 However, in terms of functional grammar, the change of the elements’ procedural compositionality cannot be explained as the process of grammaticalization because it takes place within the grammatical domain. Rather, it should be seen that the grammar of the bilingual children becomes “larger” than that of the monolinguals in that the elements in question come to contain a more complex procedural structure with more functions. The question now emerges as to whether these findings can, together with those treated in our other studies, be interpreted as the auspices of a new variety of Turkish developing in Germany or, on a larger scale, in Western Europe (cf. Rehbein 1999a, 2001b). Varieties42 are characterized by a certain stability of use, conventionalization, and wide-spread use within a population of speakers. Johanson (1991b, 1992, 2001) considers this possibility rather questionable as a long-term perspective because the linguistic isolation from the original community, which has always been typical of such situations, is lacking here, mostly due to the presence of Turkish media, successful language maintenance efforts, ongoing contact with Turkey and new immigrants who, having acquired Turkish in Turkey (also in institutional settings), come to Germany as adults. In our opinion, however, it still makes sense to speak of some kind of isolation, meaning that many of those types of discourse and communicative encounters which are central for the acquisition and active development of certain types of connectivity seem to be different in the environment of our bilingual informants. These form a widely spread group of people in Germany and Western Europe communicating

258 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

with their peers in at least two languages in an interlingual discourse and in constellations which cannot totally be controlled by social groups using Standard Turkish and/or one of the European standard languages. What does need further consideration, from a psycholinguistic point of view, is the specific bilingualism of the children in question. Being bilinguals, they are likely to activate a “bilingual mode of speaking” in the sense of Grosjean (2001, referring to Weinreich 1953, Hasselmo 1970, Clyne 1972 a.o.), when addressing an equally bilingual listener, so that, even when speaking Turkish, their German remains latently activated in their mental background and influences language processing. This not only involves syntactic phrase structures, but all the levels at which connectivity occurs. Herkenrath, Özdil and Rehbein (2000) show the emergence of a bilingual mode in which connective means from both languages are combined. In speaking Turkish, then, they may sometimes mentally rely on German structures. As connectivity in discourse strongly relies on the copresence of the speaker and hearer and their shared knowledge, in this case on the interactants’ knowing about each other’s bilingualism, the emergence of new varieties is a consequence of the spoken language use of bilinguals, having to do with speaker-hearer interaction and with the planning and processing of linguistic actions. The term “language contact situation” receives an additional dimension in this connection: besides being understood as the mental presence of two linguistic systems in bilingual individuals, it is also to be conceived as an indicator of the mental presence of two language systems in discourse interaction between speakers and listeners. Since Turkish as used in Western European immigrant communities is predominantly a language of spoken discourse, its contact variety is influenced by a spoken discourse variety of German. If understood as deviations or differentiations from a standard use, varieties are discourse phenomena and as such largely a function of spoken language. In this respect, the use of the wh-complementizer may probably not be traced back to one of the contactinduced mechanisms of language change as collected by Thomason: “… codeswitching, code alternation, passive familiarity, ‘negotiation’, second-language acquisition strategies, bilingual first-language acquisition, and change by deliberate decision.” (see Thomason 2001: 129). Rather it is to be explained as the adoption of a linguistic function effecting propositional and illocutionary linkage which is influenced by the use and order of subordinated rhemes in the postfield position — a practice adopted from German discourse, which serves as a functional expansion of the Turkish discourse. In this respect, such a linguistic practice cannot be taken as a straightforward transfer of a surface or

Interrogative elements as subordinators in Turkish 259

deep structure or simply of a feature of one language German to another language Turkish. It is more plausible to interpret this adoption of linguistic functions as the creation of an innovative word structure within bilingual Turkish discourse being triggered by something like a German substrate, and, in this sense, as an enrichment of the Turkish language. Whatever the longterm perspective may turn out to be, however, the case of Turkish in Western Europe still offers a fairly good possibility to study a contact situation with interaction and processes of change in statu nascendi. The situation of a Turkish variety about to emerge in Germany may best be characterised in terms of a tension between Contact Turkish — touched by the above mentioned contact-induced mechanisms of language change — and Standard Turkish. The dynamics of this tension evolving in interlingual discourse may be taken as a nutrient of preventing the bilinguals’ shift towards monolingualism and at the same time providing conditions for the emergence of linguistic varieties.

Abbreviations 3sg abl acc cau cdcop cop cv cvcop dat dei dim equ fut gen icop ij ins loc mod neg opt par

3rd person singular Ablative Accusative Causative Conditional copula Copula Converb Converbial copula Dative Deixis Diminutive Equative Future Genitive Indirective copula Interjection Instrumental Locative Modal Negation Optative Participle

pas Passive pcop Past tense copula pl Plural pop Postposition prs Present tense pss Possessive pst Past tense ptc Particle pte Postterminality q Interrogativity marker rec Reciprocal ref Reflexive vn Verbal Noun S: H: p: Π S: ΠH: π: ¬π:

Speaker Hearer Propositional content knowledge domain of S knowledge domain of H element of knowledge lack of knowledge



260 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

Notes * This study was funded within the framework of the SFB 538 Mehrsprachigkeit (Collaborative Research Center 538 Multilingualism) by the Deutsche Forschungsgemeinschaft (DFG). Our thanks include the participants of the International Colloquium on Interaction of Language Systems in Translation, Bilingual Language Acquisition and Language Change at the University of Hamburg (18–20 October, 2001), further Lars Johanson, Carol Pfaff, and Dan I. Slobin during their respective guest stays at the SFB as well as Angelika Redder. We would like to particularly thank Kemal Güler for his critical contributions to the general discussion, time-consuming help with the quantification and his sharp eye for all kinds of slips and mistakes. Additional thanks go to Ezel Babur, Ünal Bilir, Nesrin Esen, Filiz Tas¸demir and Esra Yaman for their valuable help with the collection and transcription of the data as well as to Veronica Trespalacios for checking our English. Needless to say, without the children’s and their families’ friendly cooperation and patient participation in our longitudinal recording sessions, there would have been no data in the first place. Last but not least, we are grateful to the anonymous reviewers for their helpful comments. 1. See Schlyter (1993) and Pfaff (1994) for investigations into the issue of what kind of areas these might be. 2. Publications concerning Turkic languages having undergone syntactic or structural change include e.g. Csató (1994) on Karaim, Menz (1999) on Gagauz, Brendemoen (1999) on Black Sea Turkish, and Erdal (2002) on Krymchak. 3. Cf. House and Blum-Kulka (1986) for the term ‘interlingual’ in connection with interlingual communication. The term ‘interlingual discourse’ focuses on the multilingual interaction between bilinguals. 4. See Johanson (1991b, 2001) who mentions these two factors as prerequisites for processes of change. For publications dealing with Turkish as an immigrant or minority language in Western Europe see the works of Boeschoten (1990), Backus 1996, Backus & Heijden 1998, Brendemoen et al. (1999), Türker (2000), Akıncı and Jisa (2000), Pfaff (e.g. 1993), Rehbein (1985, 2001b) and Johanson (1991b, 1992, 2001) as well as the references therein. 5. Whereas language acquisition is classically understood as the development of syntax in a stricter sense and considered to be more or less completed by the age children enter school, those aspects of language we are concerned with in our project, namely discourse ability, is rather acquired later. This can imply that the period of acquisition under study here may differ in various respects from the initial period of language acquisition. 6. The project SKOBI (“Sprachliche Konnektivität bei Türkisch-Deutsch bilingual aufwachsenden Kindern” — Linguistic Connectivity in Turkish-German Bilingual Children), with Jochen Rehbein as its principal investigator, is one of the projects of the Collaborative Research Center 538 Multilingualism (Sonderforschungsbereich Mehrsprachigkeit) at the University of Hamburg. The objective of the project is to document the use of connective means in the children’s speech as it emerges throughout preschool and primary school years (for a definition of connectivity cf. Rehbein 1999a).

Interrogative elements as subordinators in Turkish 261

7. See for instance Aarsen (1996) and Schroeder (1999) for the analysis of various grammatical areas from the perspective of their contribution to discourse continuity in Turkish; none of these works, however, deals with wh-constructions. 8. We use the term ‘wh-element’, which has come to be used generally in linguistic literature (based on the English initial phoneme). Possible alternatives such as ‘question word’ or ‘interrogative element’ have the drawback that they tend to evoke a connotation of interrogativity, which in fact may be deactivated as often as not. Besides, the term ‘wh-element’ is quite clear in that it cannot be taken to include other markers of interrogativity, such as particles and bound morphemes signalizing interrogativity in propositional questions. 9. Muysken (2000) adds ‘question words’ (referring to Jakobson 1971) to the subgroup ‘shifters’ and, correctly so, not to the functional categories proper. 10. For a definition of these categories within the context of functional pragmatic discourse analysis, global reference is made to the overviews in Ehlich (1991, 1994, 1999b) and Rehbein (2001a), with reference to Turkish in Rehbein (2002) and the literature therein. 11. In many languages, questions with question words are differentiated from ‘propositional questions’ (traditionally also known as ‘sentence questions’); in Turkish, propositional questions are marked, above all, by the particle mI; in German, they are mostly realized through intonation, utterance-initial position of the finite verb etc. ‘Propositional questions’ are not discussed in this paper. — The theory of questions this paper makes use of is influenced by the following studies: Ehlich and Rehbein (1986), Ehlich (1990), Rehbein (1993, 1994, 1999b). 12. For the insight that wh-words are ‘outlining words’ (‘Umrißwörter’) see Brinkmann (1962). The specific knowledge deficit with regard to a specific element (: ¬π) in the domain ‘Π’ is naturally to be classed in the category knowledge. The specific knowledge domain can be outlined by several procedures, such as in Tk. ne-re-den (from where) or in Germ. w-ie-viel (how many) (see next two sections below). 13. This concept of illocution in the sense of linguistic S-to-H-action is specific to the functional-pragmatic framework and seems to be absent in many other approaches, notably UG-based aproaches, which also attempt to integrate discourse-related phenomena, such as topic, focus, and illocution, into a phrase-structure-based model of analysis, but lack the action-theoretic concepts of speaker, hearer, and different kinds of discourse knowledge. 14. According to Brugmann (1911: 348) a number of Indo-European languages developed interrogatives into relatives, e.g. Armenian, Albanian, Latin, Baltic-Slavic, Greek and Germanic (the last three more recently). This historical process could scarcely be explained by the term ‘reanalysis’. Behaghel (1923: 361) postulates for Indo-European that the two uses existed side by side from relatively early on: “Die fragende und indefinite Bedeutung [von wer-waz — die Verf.] stehen schon indogermanisch nebeneinander; welche älter ist, läßt sich nicht mit Sicherheit sagen. (The interrogative and indefinite meaning [of wer-waz — the authors] already co-existed in Indogermanic; one cannot say for certain which is older.)” The thesis of this paper goes one step further, however, structurally, the wh-words are assumed to have the same procedural core, differences appearing only when they are used in different types of wh-constructions, which in turn are bound into varying discourse structures. Eissenhauer classifies the German wh-words wo, wer and was (where, who, what) by their

262 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

communicative function by demonstrating that S uses these words “to refer to a knowledge domain where she/he believes the hearer is storing those knowledge elements which belong to the proposition of the immediately following subordinate clause [= the wh-construction — the authors]” (Eissenhauer 1999: 286). To be more precise, however, it must be said that by using wh-words in non-interrogative manner, S allows H to share in the knowledge that S contemplates, supposes, outlines etc. when s/he (S) verbalizes the questions s/he is asking herself/himself (and not H, as the above wording suggests); H is thus actively involved in S’s knowledge process and the co-construction of H is, so to speak, “ramped up” with rhetorical methods. 15. We classified rhetorical and exothetic questions as non-interrogative usages, cf. Karakoç and Herkenrath (2002). 16. In the Indo-European languages the wh-element is derived from a complex, aspirated initial |*ko/u|-sound of a proto-form and its assimilation with other words in a different case (cf. e.g. Kluge 1989); from |*ko/u| the Romance group selected the qu§ sound [“qui”], the Germanic group the h-/w- sound, whilst a number of Satem languages, e.g. Persian, took the k-/c- sound. Certainly, there are other lines of realization, e.g. Armenian (/o/), Attic Greek (/tis/) (see Brugmann 1911: 348ff for Indo-European in detail). Other types of language likewise reveal the importance of the initial phoneme: in Arabic, for example, the wh-elements start with ka-/ma- sounds, Japanese has ik-/do-/na-, Estonian m-/k-. 17. Due to the contextual presentation of the data in transcript form, the examples generally contain a number of different phenomena, some of which may only become relevant as the paper proceeds. In example (1), non-interrogative usages of wh-constructions do occur, alongside the immediately relevant interrogative ones. They will be needed later on. 18. Redder (1990) equates the syntactic position of the wh-element with the position where German normally places words which link the subordinate to the main clause. 19. According to Lehman (1984) and Ozil (1993) wh-words, respectively, relative clauses introduced by wh-elements, are themselves nominals. The nominal takes up a constituent position (nominal/prepositional) in an interrogative clause; in a relative clause without a reference noun, there appears in German an entire clause, which, in the higher matrix, takes on the position of a nominal, respectively, a prepositional constituent. We agree that the nominalization (and thus also its referred state of affairs) is “not determined” and has no specific reference (Ozil 1993: 77). The position of the constituent in the higher matrix clause is, however, not identical with the wh-word that initiates such a constituent position. The question word verbalizes much more a search scheme with a class-like reference process that morphologically, in its essence, is adjoined with the case and adpositional components merged with the wh-morpheme; however, it is not to be taken as a nominalization (in contrast to Lehmann’s and Ozil’s interpretation). Rather, the operative linguistic procedure inherent to the question word has an affinity (= ‘drift’) to nominals through its case and adpositional components. 20. Grammars of Turkish simply describe the characteristics of wh-words and the means of subordination from a morphological or, sometimes, morpho-syntactic point of view (e.g. Banguog˘lu 1959: 367ff, 562ff, Lewis 1967: 72f, 158ff, Underhill 1976: 69ff, 325, Kornfilt 1997: 9–40, 45–53). In doing so, they generally abstract away from their functions with

Interrogative elements as subordinators in Turkish 263

respect to discourse knowledge and speaker-hearer interaction. If and when they explicitly contrast the interrogative usages against the non-interrogative ones, they do so in a morphosyntactic perspective. 21. Hangi (which) stems from /k/ or /q/ (qangi, qanyu) and at an earlier stage qaçan ‘when’, too (L. Johanson, personal communication). 22. Cf. the general characterizations in Section 2. 23. This is not the place to discuss the details of phonological givens. 24. See for instance Kornfilt (1997) for an overview on the morpho-syntax of subordination in Turkish or Haig (1998) for an analysis of Turkish relative clauses, combining syntactic and functional (in the sense of Simon Dik) approaches. An analysis relating the insights of modern syntactic theory to functional-pragmatic research questions (i.e. looking at what these subordinating syntactic means do in terms of discourse connectivity and with respect to the model of speaker and hearer knowledge) has not been found in the literature. 25. Particles such as ki, çünkü etc., borrowed from Persian (meaning ‘that’ and ‘because’, respectively, in Persian), are not used as complementizers in Turkish; cf. Johanson (1975) for this argumentation; see Güler and Rehbein (2003) who argue that ki is a particle with the meaning of a colon by means of which the utterance may modify its illocutionary force. 26. H is certainly addressed in a more immediate manner than if S were to direct the question to himself in the sense of an exothesis (cf. Hohenstein and Kameyama 2002). 27. In Rehbein (1992: 558), and with reference to complex constructions processed in German discourse, a word order where the last rheme within a series of embedded ones is of the highest prominence and gets the highest salience is called ‘cascade construction’. 28. Cf. the interesting study by Kim (1995) who recognizes three discourse functions of noninterrogative wh-constructions in topicalized pre-field and pre-pre-field positions: gist marking of the prior talk, disagreement/repair and affecting display. 29. Kahrel (1985) in quoting Dik (1983) discusses the typological distinction of pre-field and post-field languages which may help to clarify the difference between Turkish and German concerning the word order of finite predicates in spoken discourse. According to Kahrel, there exists a general pattern: Prefield – Centre – Postfield with a verbal predicate, a head noun or an adjective as possible centers, and with relators, like adpositions, case markers and subordinating devices, as in indirect questions placed in the area in front of or after the center. The position of relator elements depends on the type of the language, e.g., Abkhaz, Basque, Quechua, Tamil, Turkish, Urdu, Japanese, Navaho a.o. are pre-field languages (SOV word order), whereas Finnish, Yoruba, Bahasa, Egyptian, Dutch, Vietnamese, Italian (SVO word order), German a.o. are post-field languages. The remark which is of interest here concerns the situation “when a pre-field language places its indirect questions in the post-field.” (Kahrel 1985: 171). Kahrel says that this happens to be the case in Basque but there, a ‘quote particle’ (ea) is preposed to the “embedded question” — a situation which reminds us of Turkish constructions postponed to a matrix predicate with the particle ki in front; where, in fact, ki just indicates that the following syntagma is syntactically not subordinated to the previous predicate.

264 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

30. Cf. Müller (1998) for an approach relating the concept of syntactic ambiguity to the interaction of grammatical systems in bilingual first language acquisition. According to Müller, grammatical properties of one language may be relied on as a momentary “relief strategy” to disambiguate structurally ambiguous surface strings in the other language. 31. An anonymous reviewer suggested that the two partial utterances in this example could be interpreted as two independent utterances and that an analysis of intonation patterns or of the prosodic structure would be needed to shed more light on the matter. We think, however, that independently of the phonological facts, the two constructions depend on each other in terms of their illocutionary status and that the part-utterance kim öyle bir ¸sey yapıyor ‘who is doing such a thing’ cannot be interpreted as a hearer-directed question with an interrogative illocution of its own. Instead, the propositional contents of the two parts are linked together in such a way that the illocutionary force of the second part-utterance is cancelled and the construction, as a whole, is syntactically subordinate to the preceding construction. — To add further evidence to our interpretation, a look at (12) seems to be helpful: in (12), there is a change of speaker, something that one cannot but consider as a clear disruption between the two part-utterances. In spite of this very distinct caesura, the second part-utterance nasıl konus¸uyorlar ‘how they speak’ certainly does not display any kind of interrogative or otherwise independent illocution. 32. We would like to seize the opportunity to express our special thanks to Kemal Güler for undertaking the crucial and time-consuming task of counting wh-constructions of various categories in the major part of the corpus used here. Without his help, the quantification would have been less complete and the corpus would have had to be considerably smaller. 33. The number and length of the discourses by the individual children vary for several reasons. — The data quantified for this paper constitute some 20% of our entire corpus, which is still growing. In particular, with regard to the SKOBI children (Seda, Melike, Tahir, Binnaz, and Kubat), we are at the beginning of a longitudinal study. See Karakoç and Herkenrath (2002) for quantitative information based on a larger corpus. 34. See Karakoç and Herkenrath (2002) for a more detailed analysis of the developmental aspects of the phenomenon and of the systematic use of the verba dicendi and sentiendi. 35. The verba dicendi and sentiendi fall into two groups according to whether, when employed as matrix verbs with finite wh-constructions, they are used with or without the quotation particle diye. 36. Constructions in which the wh-construction is inserted between two occurrences of the same verbum dicendi are marked with [+ & −pre]. 37. There are quite a few occurrences, in which, due to the shortness of the construction, it is not possible to decide whether the wh-element is placed in situ or has been moved into a construction-initial position. These occurrences, we labeled as [?sit]. 38. For the quality of conjunctions and other complementizer words to link a subordinate clause to a superordinate one by means of word-internal pseudo-deictic, i.e. para-operative, procedures see Redder (1990). 39. Cf. the analysis of the non-interrogative use of wh-words in Section 2.3 above.

Interrogative elements as subordinators in Turkish 265

40. Non-interrogative uses of wh-words express a need to extend the contents of propositional elements and to delineate them in the discourse. Sankoff and Brown (1980) believe these ordering relationships lead to the creation of relative clause constructions in many of the world’s languages. 41. The explanation given focusses a process which may also be described as a change of the subcategories of a “functional element”, i. e. as a change of the “question word” from the subcategory ‘shifter’ to that of a “complementizer” (cf. Muysken 2000: 164f). Nevertheless, the phenomenon analyzed above is not covered by any of the types of code-mixing (insertion, alternation, congruent lexicalization) mentioned there. 42. There is no generally accepted clear-cut definition of the term “variety”. If one chooses to rely on the definition suggested by Labov in (1972), the term “variety” serves to cover various aspects of the heterogeneity of a language (Adamzik 1998). These may comprise categories as diverse as dialect, jargon, sociolect, as well as register, functional style, etc. This implies two things: first, varieties are conceived as differentiations within one language; second, they appear as coherent, discrete systems in their own right, as “languages within a language”. — A variety may be spoken by a certain group of people and its use may be restricted to certain situations (Gumperz 1994, Milroy and Milroy 1997; cf. Erdal 2002 on register-specific means for relativization in Krymchak). The use of a variety can be systematic in that its characteristics enter into play in several discourse occurrences and involve several informants. If one takes dialect as a prototype of variety, neither its restriction to specific speech situations nor the fact that not all speakers always produce the phenomena in question is especially surprising, because all of these restrictions apply to dialectal language use as well.

References Aarsen, J. 1996. Relating events in two languages. Acquisition of cohesive devices by TurkishDutch bilingual children at school age. Tilburg: University Press. Adamzik, K. 1998. “Fachsprachen als Varietäten”. In Fachsprachen. Ein internationales Handbuch zur Fachsprachenforschung und Terminologiewissenschaft Vol. I, L. Hoffmann, H. Kalverkämper and E. H. Wiegand (eds), 181–189. Berlin: De Gruyter. Akıncı, M. A. and Jisa, H. 2000. “Development of Turkish clause linkage in the narrative texts of Turkish-French bilingual children in France”. In Studies on Turkish and Turkic Languages, A. Göksel and C. Kerslake (eds), 317–324. Wiesbaden: Harrassowitz. Backus, A. 1996. Two in One. Bilingual Speech of Turkish Immigrants in The Netherlands. Tilburg: UP. Backus, A. & van der Heijden, H. 1998. “Life and birth of a bilingual: The mixed code of bilingual children and adults in the Turkish community in the Netherlands.” In: Johanson, 527–551. Banguog˘lu, T. 1959. Türkçenin Grameri. Ankara: Türk Dil Kurumu Yayınları. Behaghel, O. 1923. Deutsche Syntax. Eine geschichtliche Darstellung. Heidelberg: Winter.

266 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

Berman, R. and Slobin, D. (eds). 1994. Relating Events in Narrative. A cross-linguistic developmental study. Hillsdale, NJ: Erlbaum. Boeschoten, H. E. 1990. Acquisition of Turkish by Immigrant children. A multiple case study of Turkish children in the Netherlands aged 4 to 6. Wiesbaden: Harrassowitz. Brendemoen, B. 1999. “Greek and Turkish language encounters in Anatolia”. In Brendemoen et al. 1999: 353–378. Brendemoen, B., Lanza, E. and Ryen, E. (eds). 1999. Language encounters across time and space. Studies in language contact. Oslo: Novus Press. Brinkmann, H. 1962. Die deutsche Sprache. Gestalt und Leistung. Düsseldorf: Schwann. Brünner, G. and Graefen, G. (eds). 1994. Texte und Diskurse. Methoden und Forschungsergebnisse der Funktionalen Pragmatik. Opladen: Westdeutscher Verlag. Brugmann, K. 1911. Grundriss der vergleichenden Grammatik der indogermanischen Sprachen. II. Bd. Vergleichende Laut-, Stammbildungs- und Flexionslehre nebst Lehre vom Gebrauch der Wortformen der indogermanischen Sprachen. Strassburg: Trübner. Bührig, K. and Matras, Y. (eds). 1999. Sprachtheorie und sprachliches Handeln. Festschrift für Jochen Rehbein zum 60. Geburtstag. Tübingen: Stauffenburg. Clyne, M. 1972. Perspectives on language contact. Melbourne: Hawthorne Press. Csató, É. Á. 1994. “On word order differences between Turkish and Karaim”. Dilbilim Aras¸tırmaları 1994: 54–61. Dik, S. C. 1983. “Two constraints on relators and what they can do for us”. In Advances in Functional Grammar. S. C. Dik (ed.), 267–298. Dordrecht: Reidel. Ehlich, K. 1990. “Zur Struktur der psychoanalytischen ‘Deutung’”. In Medizinische und therapeutische Kommunikation. Diskursanalytische Untersuchungen. K. Ehlich, A. Koerfer, A. Redder and R. Weingarten (eds), 210–227. Opladen: Westdeutscher Verlag. Ehlich, K. 1991. “Funktional-pragmatische Kommunikationsanalyse”. In Verbale Interaktion. Studien zur Empirie und Methodologie der Pragmatik, D. Flader (ed.), 127–143. Stuttgart: Metzler. Ehlich, K. 1994. “Funktionale Etymologie”. In Brünner and Graefen 1994: 68–82. Ehlich, K. 1999a. “Der Satz. Beiträge zu einer pragmatischen Rekonstruktion”. In Redder and Rehbein 1999: 51–68. Ehlich, K. 1999b. “Funktionale Pragmatik — Terme, Themen und Methoden”. Deutschunterricht in Japan 4: 4–24. Ehlich, K. and Rehbein, J. 1977. “Batterien sprachlicher Handlungen”. Journal of Pragmatics 1: 393–406. Ehlich, K. and Rehbein, J. 1986. Muster und Institution. Untersuchungen zur schulischen Kommunikation. Tübingen: Narr. Eissenhauer, S. 1999. “Wer, wo, was und Relativsätze ohne Bezugsnomen”. In Sprachtheorie und sprachliches Handeln, K. Bührig and Y. Matras (eds), 281–289. Tübingen: Stauffenburg. Erdal, M. 1999. “Das Nachfeld im Türkischen und im Deutschen”. In Johanson and Rehbein 1999: 53–94. Erdal, M. 2002. “Relativisation in Krymchak”. In Scholarly Depth and Accuracy. A Festschrift to Lars Johanson. Lars Johanson Armag˘anı, N. Demir and F. Turan (eds), 117–136. Ankara: Grafiker. Erguvanlı, E. E. 1984. The function of word order in Turkish grammar. Berkley: UC Press.

Interrogative elements as subordinators in Turkish 267

Grosjean, F. 2001. “The bilingual’s language modes”. In One Mind, Two Languages. Bilingual Language Processing, J. L. Nicol (ed.), 1–22. Oxford: Blackwell. Güler, K. and Rehbein, J. 2002. ki — Form und Funktion eines Ausdrucks und seine Verwendung durch mono- und bilinguale türkische Kinder. Working paper of the project SKOBI. Research Center on Multilingualism at the University of Hamburg. Ms. (University of Hamburg, SBF 538). Gumperz, J. J. 1994. “Sprachliche Variabilität in interaktionsanalytischer Perspektive”. In Kommunikation in der Stadt. Teil 1, W. Kallmeyer (ed.), 611–639. Berlin: De Gruyter. Haig, G. 1998. Relative Constructions in Turkish. Wiesbaden: Harrassowitz. Hasselmo, N. 1970. “Code-Switching and Modes of Speaking”. In Texas Studies in Bilingualism, G. G. Gilbert (ed.), 179–210. Berlin: de Gruyter. Herkenrath, A. and Karakoç, B. 2002. Zum Erwerb von Verfahren der Subordination bei türkisch-deutsch bilingualen Kindern — Transkripte und quantitative Aspekte. Universität Hamburg, SFB 538: Working Papers in Multilingualism 37, Series B. Herkenrath, A., Özdil, E. and Rehbein, J. 2001. Aspects of Turkish-German Code Switching. Talk given at the Colloquium on Code-Switching, Accomodation and Multilingualism, Hamburg, 15th — 17th December 2000. Ms. (University of Hamburg, SBF 538). Hoffmann, L. (ed.). 1992. Deutsche Syntax. Ansichten und Aussichten. Berlin etc.: de Gruyter. Hohenstein, Ch. J. and Kameyama, Sh. 2002. “Exothesen”. In Bausteine diskursanalytischen Wissens, J. Rehbein and S. Kameyama (eds), (in prep.) House, J. and Blum-Kulka, S. (eds). 1986. Interlingual and Intercultural Communication Discourse and Cognition in Translation and Second Language Acquisition. Tübingen: Narr. Jakobson, R. 1971. “Shifters, verbal categories, and the Russian verb”. In Selected Writings II, 130–47. The Hague: Mouton. Johanson, L. 1975. “Some Remarks on Turcic ‘Hypotaxis’”. Ural-Altaische Jahrbücher 47: 104–118. Johanson, L. 1991a. “Bestimmtheit und Mitteilungsperspektive im türkischen Satz”. In Linguistische Beiträge zur Gesamtturkologie, L. Johanson (ed.), 225–242. Budapest: Akadémiai Kiadó. Johanson, L. 1991b. “Zur Sprachentwicklung der ‘Turcia Germanica’”. In Türkische Sprachen und Literaturen, I. Baldauf et al. (eds), 199–212. Veröff. d. Societas Ural-Altaica 29. Johanson, L. 1992. Strukturelle Faktoren in türkischen Sprachkontakten. Stuttgart: Franz Steiner. Johanson, L. (ed.) 1998. The Mainz Meeting. Proceedings of the Seventh International Conference on Turkish Linguistics. Wiesbaden: Harrassowitz. Johanson, L. 2001. Code interaction in linguistic diaspora groups. Paper given on October 12, 2001, at Professor Michael Clyne’s Seminar on Contact Linguistics at the Linguistic Department of Melbourne University. Ms. (University of Mainz). Johanson, L. and Rehbein, J. (eds). 1999. Türkisch und Deutsch im Vergleich. Wiesbaden: Harrassowitz. Kahrel, P. 1985. “Indirect questions and relators”. In Predicates and Terms in Functional Grammar, A. M. Bolkestein, C. de Groot and J. L. Mackenzie (eds), 165–181. Dordecht etc.: Foris.

268 Annette Herkenrath, Birsel Karakoç and Jochen Rehbein

Kaiser, G. 1998. “Sprachwandel durch Parameterwechsel? Kritische Betrachtungen generativer Sprachwandeltheorien am Beispiel der Entwicklung der Verbstellung im Französischen”. In Reanalyse und Grammatikalisierung in den romanischen Sprachen, I. Neumann-Holzschuh and J. Lang (eds), 53–73. Tübingen: Niemeyer. Kaiser, G. 2000. Dialect contact and language change. A case study on word-order change in French. Universität Hamburg, SFB 538: Working Papers in Multilingualism 3, Series B. Karakoç, B. and Herkenrath, A. 2002. Aspekte mono- und bilingualer Verwendung von wh-Konstruktionen im Türkischen. Talk given on July 18, 2002, at the Research Center on Multilingualism at the University of Hamburg. Ms. (University of Hamburg, SBF 538). Kennelly, S. D. 1993. Theta government in Turkish: Effects on IP dominated by DP. Ms. (Rutgers University). Kim, K.-H. 1995. “WH-clefts and left-dislocation in English conversation”. In Word order in discourse, P. Downing and M. Noonan (eds), 247–296. Amsterdam: Benjamins. Kluge, F. 19892. Etymologisches Wörterbuch der deutschen Sprache. Berlin: De Gruyter. Kornfilt, J. 1997. Turkish. London and New York: Routledge. Kroch, A. 2001. “Syntactic Change”. In The Handbook of Contemporary Syntactic Theory, M. Baltin and Ch. Collins (eds), 699–729. Oxford: Blackwell. Labov, W. 1972. “Rules for ritual insults”. In Studies in Social Interaction, D. Sudnow (ed.), 120–169. New York: Free Press. Lehmann, C. 1984. Der Relativsatz. Typologie seiner Strukturen, Theorie seiner Funktion, Kompendium seiner Grammatik. Tübingen: Narr. Lewis, G. L. 1967. Turkish Grammar. Oxford: UP. Lightfoot, D. 1991. How to set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press. Meisel, J. M. 2001. From bilingual language acquisition to theories of diachronic change. Universität Hamburg, SFB 538: Working Papers in Multilingualism 30, Series B. Menz, A. 1999. Gagausische Syntax. Eine Studie zum kontaktinduzierten Sprachwandel. Wiesbaden: Harrassowitz. Milroy, J. and Milroy, L. 1997. “Varieties and Variation”. In The Handbook of Sociolinguistics, F. Coulmas (ed.), 47–64. Oxford: Blackwell. Müller, N. 1998. “Transfer in bilingual first language acquisition”. Bilingualism: Language and Cognition 1(3): 151–193. Muysken, P. 2000. Bilingual Speech. A typology of code-mixing. Cambridge: UP. Mundy, Ch. S. 1955. “Turkish syntax as a system of qualification”. In BSOAS 17, 279–283. Ozil, S¸. 1993. “Die Begriffsbildung durch Relativsätze ohne Bezugsnomen im Deutschen und verbalnominale Konstruktionen im Türkischen”. Diyalog 1: 69–90. Ankara. Pfaff, C. 1993. “Turkish language development in Germany”. In Immigrant Languages in Europe, G. Extra and L. Verhoeven (eds), 119–157. Clevedon: Multilingual Matters. Pfaff, C. 1994. Early bilingual development of Turkish children in Berlin. Ms. (FU Berlin). Radford, A. 1997. Syntactic theory and the structure of English. A minimalist approach. Cambridge: CUP. Redder, A. 1990. Grammatiktheorie und sprachliches Handeln: ‘denn’ und ‘da’. Tübingen: Niemeyer.



Interrogative elements as subordinators in Turkish 269

Redder, A. 1992. “Funktional-grammatischer Aufbau des Verb-Systems im Deutschen”. In Hoffmann 1992: 128–154. Redder, A. 2001. Illokutionsstopper. Talk given at the 6. international conference on funktional pragmatic, Köln, 29.11. — 1.12.2001. Ms. (University of Munich). Redder, A. and Rehbein J. (eds). 1999. Grammatik und mentale Prozesse. Tübingen: Stauffenburg. Rehbein, J. 1985. “Zur Zweisprachigkeit türkischer Schüler. Ein Bericht über Untersuchungen der sprachlichen Handlungsfähigkeit türkischer Schüler im ehemaligen Krefelder Grundschulmodell”. Materialien Deutsch als Fremdsprache 25: 265–279. Rehbein, J. 1992. “Zur Wortstellung im komplexen deutschen Satz”. In Hoffmann 1992: 523–574. Rehbein, J. 1993. “Ärztliches Fragen”. In Arzt-Patienten-Kommunikation, P. Löning and J. Rehbein (eds), 311–364. Berlin: de Gruyter. Rehbein, J. 1994. “Zum Klassifizieren ärztlichen Fragens”. In Medizinische Kommunikation. Diskurspraxis, Diskursethik, Diskursanalyse, A. Redder and I. Wiese (eds), 147–170. Opladen: Westdeutscher Verlag. Rehbein, J. 1995. “Grammatik kontrastiv — am Beispiel von Problemen mit der Stellung finiter Elemente”. Jahrbuch Deutsch als Fremdsprache 21: 265–292. Rehbein, J. 1999a. “Konnektivität im Kontrast. Zu Struktur und Funktion türkischer Konverbien und deutscher Konjunktionen, mit Blick auf ihre Verwendung durch monolinguale und bilinguale Kinder”. In Johanson and Rehbein 1999: 189–243. Rehbein, J. 1999b. “Zum Modus von Äußerungen”. In Redder and Rehbein 1999: 91–142. Rehbein, J. 2001a. “Konzepte der Diskursanalyse”. In Text- und Gesprächslinguistik. 2. Halbband. HSK, K. Brinker, G. Antos, W. Heinemann and S. F. Sager (eds), 927–945. Berlin: De Gruyter. Rehbein, J. 2001b. “Turkish in European Societies”. In Lingua e Stile XXXVI-2: 317–334. Rehbein, J. 2002. “Pragmatische Aspekte des Kontrastierens von Sprachen — Türkisch und Deutsch im Vergleich”. In Eröffnungsreden und Tagungsbeiträge des VII. Türkischen Germanistikkongresses. Hacettepe Üniversitesi Ankara: UP (in press). Sankoff, G. and Brown, P. 1980. “The origins of syntax in discourse: A case study of Tok Pisin relatives”. In The Social Life of Language, G. Sankoff (ed.), 211–255. Pennsylvania: UP Press. Schlyter, S. 1993. “The weaker language in bilingual Swedish-French children”. In Progression and regression in language. Sociocultural, neuropsychological and linguistic perspectives, K. Hyltenstam and A. Viberg (eds), 289–308. Cambridge: CUP. Schroeder, C. 1999. The Turkish nominal phrase in spoken discourse. Wiesbaden: Harrassowitz. Thomason, S. G. 2001. Language Contact. Edinburgh: UP. Türker, E. 2000. Turkish-Norwegian Codeswitching. Evidence from Imtermediate and Second Generation Turkish Immigrants in Norway. Oslo: Unipub forlag. Underhill, R. 1976. Turkish Grammar. Cambridge, MA: MIT Press. Watanabe, A. 2001. “Wh-in-situ Languages”. In The Handbook of Contemporary Syntactic Theory, M. Baltin and C. Collins (eds), 203–225. Oxford etc.: Blackwell. Weinreich, U. 1953 dt. 1977. Sprachen in Kontakt. Ergebnisse und Probleme der Zweisprachigkeitsforschung. Mit einem Nachwort von A. de Vincenz. München: Beck.



Child L2 acquisition An insider account* Mary Aizawa Kato UNICAMP

1.

Introduction

This paper is a metalinguistic reflection on matters of competence, proficiency, acquisition and loss, using myself as the subject of an early acquisition of Brazilian Portuguese (BP. as L2, comparing it with my competence in Japanese (J), my L1, and marginally with my later acquisition of English as L3.1 I will focus not only on the development of grammatical competence but also on communicative competence. I will be using competence for the former and proficiency for the latter. In the presentation, I will assume that the mind is an organ composed of specialized modules, in the sense of Fodor (1983), in addition to a multifunctional central system which has access to the outputs of the specialized modules. Competence will be understood as the outcome knowledge of a specialized module of the mind (Grammar, or I-language according to Chomsky 1986) and its use in sentence parsing independent of context. As for proficiency, it is assumed to be the use of Grammar in comprehension and production, understood as the outputs of a multifunctional system which makes use not only of the output of Grammar but also of other modules, such as the semantic and pragmatic ones, as well as the articulatory and perceptual ones. The result of this interaction is what is understood as the use of language in context. In Section 2, I will give a short account of my linguistic history. In Section 3, I will talk about my L1, both from the competence and the proficiency perspectives. In Section 4, I will present a theoretical discussion of what was involved in my L2 acquisition, taking into consideration that Japanese and BP have different settings in the major parameters. I will make conjectures on the role of my L1 in my L2 competence, focusing on the residues of my L1 in my L2. In

272 Mary Aizawa Kato

Section 5, I will come back to the central issue of the organization of the mind and see if it is possible to use consciousness as a clue to see how the specialized module of grammar and the central system enter into the processing of the languages that I know, including my L3, English. In the sixth section, I will venture some hypotheses concerning different types of L2 learning. In the conclusion I will summarize the main points of the paper.

2. My linguistic history I am a speaker of three languages: a. Japanese as L1 b. Brazilian Portuguese (BP) as my first L2 and c. English as my second L2, referred to here as L3 I was a monolingual until I was six years old and, though I was born in Brazil, I lived in a district of Sao Paulo, where my neighbors were Japanese and my street friends were also Japanese. There was no TV at the time, so the only input I had was Japanese. Before going to school, I learned some basics of Japanese writing, which enabled me to read children’s stories. I also had already begun to learn that bound morphology was written in hiragana, a syllabic system, while content words had to be coded in ideograms.2 However, my literacy in J. was interrupted when I entered school, especially with the breaking of World War II. As was the case of many immigrant children all over the world, I started to acquire my second language (BP) when I started school at six years of age. Competence and proficiency only in L1 were not considered a handicap and I was put to learn reading and writing at the same time as other children. I can’t remember how I acquired BP, but I remember that by the end of the year, I was as talkative as any other child besides being the first to get a storybook as a prize for being the first to achieve the ability to read and write. My guess is that my acquisition of L2 took the first year of schooling, namely one year, the same amount of time that children normally take to acquire the grammar of his/her mother tongue. We can say, in turn, that it was instantaneous — and, therefore, that I had the same access to UG as a one-year-old child. It is also possible that my cognitive development was a positive factor in speeding up my acquisition, especially in the learning of written skills. If there is a protolanguage period in the development of language, in the sense of Bickerton (1990), I seem to have skipped it, or gone through it in a very fast manner.3

Child L2 acquisition 273

As for English, I started learning it when I was twelve, at regular school and also in a language school, that is, learning took place in the context of ordered input, and negative data. My proficiency today is dependent on the domain of conversation and on the interlocutor. I am quite fluent in English when the subject is linguistics and my understanding is near-native when the interlocutor has higher education. On the other hand, I have trouble with dialectal differences, with uneducated speakers and with talking to children and animals. My proficiency in Japanese, on the other hand, can be said to be complementary in domain to that of English.

3. Aspects of my Japanese and that of other members of the Japanese community Of the languages that I speak, I am most proficient in my L2, BP. My proficiency in L1, Japanese, is more limited. I can say that it has fossilized since the disappearance of the first generation of Japanese, as the descendants rarely use Japanese with co-etarian and their descendants, contrary to German descendants in Brazil, who have kept German for three generations. What has also contributed to the deterioration of my Japanese is the fact that I consider myself illiterate in this language. This limited the enrichment of the substantive vocabulary and did not favor the maintenance of what I had already learned. Literacy can thus be a positive factor in the development of proficiency, though it has nothing to do with competence, as we will see later. Until I started interacting with members of my husband’s family, my Japanese resembled child Japanese.4 When speaking with my parents I had always used my own name to refer to myself, even as an adult. Second person and third person were referred to by the person’s name. Japanese has no real pronouns; what it has is a complicated system of address forms. There are eight forms only to refer to the first person, a fact that is comprehensible if they are understood as different address forms.5 Women and men use different forms. Forms differ according to formality and intimacy. Learning pronouns is lexical learning and socially conditioned. To give an example, in a family situation, a man has to refer to himself with the first person boku. A woman, on the other hand, has to use watashi. In order to make a boy use the proper first pronoun, parents and elder brothers and sisters often call a boy by the pronoun boku.6 My parents did not go into the trouble of “teaching” me how to refer to myself and I continued to use my name, something not unusual in Japanese families.

274 Mary Aizawa Kato

I started using address forms when I got married and I had to use proper address forms with each interlocutor. It was like second language learning. I also learned the third person pronouns kare (he/him) and kanojo (she/her), as these are rarely used in spoken language. In most situations it is the person’s name that is used, often accompanied by an honorific morpheme (Kato-san, Emy-chan). On that occasion I also had to learn pairs of verbal forms that indicate whether the subject is the speaker or the hearer. The form for the speaker is deferential and the form for the hearer has a politeness component. Thus if I ask someone whether he wants to eat x, I say (1a), but if I want to answer that I want to eat x, I have to use (1b): Kore agari-massu-ka? this (you) eat-pres question particle ‘Will you have this?’ b. Hai, itadaki-massu yes (I) eat ‘Yes, I will.’

(1) a.

In my child Japanese, the interaction with my mother involved only the unmarked symmetrical verb taberu (to eat). I had no other type of relation to condition the use of asymmetrical pairs. (2) a.

Taberu? eat? ‘Will you have this?’ b. Taberu. eat ‘Yes, I will.’

Besides names, the use of null categories was pervasive both in subject and object position as can be seen in (2). In my first exposure to the verbal asymmetrical forms I saw myself in a terribly embarrassing situation. As my mother-in-law asked me (3a) I answered with (3b) causing everyone to laugh. Kore agari-massu-ka? this (you) eat-pres question particle ‘Will you have this?’ b. *Hai, agarimassu. yes eat ‘Yes, I will.’

(3) a.

Child L2 acquisition 275

Thus, negative response from my in-laws was quite effective in learning what I should not do. The asymmetrical pairs again involved social and lexical learning, which I only mastered partially, even after I started to interact in my husband’s family, where a lot of asymmetrical and differentiated relations took place. But my proficiency in Japanese grew fast at that time and I was able to maintain social conversations with various interlocutors for lengthy periods of time. However, after the disappearance of the first generation of Japanese, including my parents and my in-laws, my use of Japanese decreased significantly and today I am conscious that my Japanese is undergoing a process of fossilization as it happens with most members of the community. It is almost impossible to maintain long conversations without lexical borrowings from Portuguese and code-switching between Japanese and BP.7 The functional skeleton, however, is still there and what I do is borrow words from the substantive classes. As it is difficult to do so with verbs, as they need to be inflected for the Japanese morphology, what I do and what I see other Japanese descendants doing is use the Japanese light verb suru combined with a noun, a verb stem or a nominalized word in Portuguese: (4) a.

divorcio-suru divorce do b. divorcio-shita divorce did

esconde-suru hide do esconde-shita hide did

mudança-suru moving do mudança shita moving did

Though in many respects my Japanese has fossilized, I discovered quite recently, sitting in a Japanese syntax course at USC, that I can give well-formedness judgements in quite complicated tests that involve long distance extractions, and cross-over and weak cross-over phenomena. It is true that proficient speakers of foreign languages can often provide sound judgements about grammaticality of sentences in their L2, however distant the constructions are from their L1. But what surprised me was the fact that despite my problems with communication in Japanese, the judgement tasks were not a problem at all. Language loss is often seen as the reverse process of language acquisition, but what is observed in child language is the omission of functional categories (Radford 1990, Rizzi 1994, Meisel 1994) while what the Japanese community does, both the first and second generations, is to preserve the functional skeleton and borrow the lexical items of the dominant language. As an example, when my 85-year old aunt was still alive she called my other aunt in Japan and said:

276 Mary Aizawa Kato

(5) Segunda-feira-ni telegrama-o manda-shimashita monday-ni telegrama-o send-shimashita ‘I sent the telegram on Monday.’

And she was not even conscious that she was mixing languages. A more serious phenomenon is the mixing of Portuguese pronouns, a closed word-class, in the speech of the less educated Japanese and Japanese descendants speaking their own language. Io-ga mita I Port-ganom sawJap. ‘I saw it.’ b. Ocê-ga warui-no da you Port-ganom guilty-no da ‘You are the guilty one.’

(6) a.

By doing this, they reduce a complicated system of address forms to a simple biunique system. Concluding this part, we can say that learning my L1 did not stop at the critical age, but was completed with new social interactions. However, it was mostly lexical learning, which we know has no age limitation, and can be attributed to proficiency and not competence. After becoming quite proficient, my L1 started to decline, as was the case of most people in the community, even the ones born in Japan. However, I presented some linguistic and behavioral clues that allowed me to conclude that my competence remained unharmed by fossilization: a) no functional category has been lost and b) the intuition as to what is well-formed or ill-formed has also been preserved.

4. Conjectures about my acquisition of Portuguese as L2 4.1 Parameter values in my L1 and my L2 In the Principles and Parameters Model (Chomsky 1981, 1986), the acquisition of L1 is understood as a selective process. Universal Grammar (UG) is postulated as the state So, common to all children and homogeneous within the human species. Far from being an empty state, UG is defined as a set of invariant properties of natural languages — the Principles, and a set of variable properties — the Parameters. Given this set of Parameters, the acquisition is completed when all of them have the value set as (+) or (−), according to the child’s

Child L2 acquisition 277

linguistic experience. In traditional GB terms, my learning of Japanese as L1, among other characteristics, consisted of: a. b. c. d. e.

the setting of the X-bar schema as a head-final parameter the positive setting to morphological case the positive setting regarding the nominative accusative case system the positive setting of the null subject parameter the negative setting of the V-to-I movement etc.

I will now speculate on how my acquisition of Portuguese took place. The question that arises is whether I had direct access to UG, or whether I started from an indirect access, through the principles and parameters set in the acquisition of Japanese.8 This is not an easy question to answer, since early L2 learners do not exhibit the incompleteness features of late L2 learners.9 If, as I assume, I had direct access to UG, acquisition was merely a selection process involving, for some parameters, the opposite value of the Japanese ones. The corresponding list would be the following: a.¢ b.¢ c.¢ d.¢ e.¢

the setting of the X-bar schemata as head-initial the negative value to morphological case the positive setting regarding the nominative accusative case system the positive setting of the null subject parameter the positive setting of the V-to-I movement etc.

Today, however, parameters are better understood as associated to morphological rather than syntactic properties. Thus, in learning Japanese, I learned that it has an inventory of case-markers, postpositional in nature, containing among them the nominative and accusative cases. The learning of these postpositions allowed me to learn that Japanese is a language with morphological case, with head-final word order, and that belongs to the nominative-accusative system, namely three of the traditional parameters.10 (7)

Taroo-ga Kimiko-ni CD-o agueta. Taroo-nom Kimiko-dat CD-accus gave ‘Taroo gave a CD to Kimiko.’

(8)

case morphemes = {-ganom, -oaccus, -nidat, …}

As for my acquisition of Brazilian Portuguese, the absence of postpositions and the presence of prepositions lead me to set the directionality parameter as head initial.

278 Mary Aizawa Kato

Em casa; para Pedro; com Maria at home for Peter with Maria b. prepositions = {em, para, com ….}

(9) a.

Case is not overtly exhibited in morphology in BP, except for pronouns, which must act as clues to set the nominative/accusative parameter. The acquisition of pronouns is then crucial to learning the basics of syntactic form. Eu telefonei.  b. Você me telefonou  c. Você telefonou para mim. I called   you meclitic called   you called to me b. personal pronouns = {eunom, meaccus/dat clitic, mimdat, …}

(10) a.

As nouns do not carry case morphology, abstract case features appear in the lexicon and are selected to be added to nouns before numeration (cf. Chomsky 1995). (11) a.

Maria(+nom) telefonou para Pedro(+dat) Maria telephoned to Peter ‘Maria called Peter.’ b. Cases = [[+nom], [+dat], [+accus]]

Once these items from the lexicon are acquired, we expect error-free forms. For instance, once prepositions are acquired, they always appear in the correct position as pre-positions. In effect I have never heard a Japanese immigrant or descendant producing forms such as in (12), where prepositions are analyzed as postpositions:11 (12) a. *mamãe pra mummy for ‘to/for Mummy’ b. *casa em house in ‘at home’

However, OV forms are possible to be heard in the speech of young children acquiring Portuguese (cf Gonçalves 1997), and I assume it was possible in my early Portuguese. (13) a.

(eu) [bolo comê] (I [cake eat ‘I eat cake.’

Child L2 acquisition 279

b. (eu) eh … [bentô comeu] (I ah  [snack ate ‘I, ah, ate a snack.’ c. (eu) [“gameboy” tem] na minha casa.12 (I [“gameboy has in the my house ‘I have a gameboy at home.’

But OV forms can also be heard in adult immigrant speech without an overt subject, as are most of the forms produced by the child: (14) a.

pizza comeu pizza ate ‘I/you/he/she/we/they ate pizza.’ b. conta pagou bill paid ‘I/you/he/she/we/they paid the bill.’

The languages involved in my acquisition were Japanese, which has a consistent OV order, and BP, which has VO as the unmarked order. Can we say that forms such as (13) and (14) result from a transfer operation from Japanese word order? Though this is a possibility, we may consider another hypothesis: that the OV forms are due to topicalized forms in the target language. In effect, BP is a topic prominent language and structures where the object is dislocated or topicalized are very common in the vernacular. Moreover, topicalization is very common in acquisition, perhaps because the topic can be merged in-situ, or even be generated as an independent discourse chunk However, BP has become a partially non-null subject language and the omission of the referential pronominal subject in such contexts is impossible today (cf. Duarte 1993).13 This is due to the fact that the agreement inflection in BP lost its pronominal nature.14 (15) a.

A pizza, ele comeu. the pizza he ate ‘The pizza, he ate it.’ a¢. *A pizza, pro ate. b. A conta, ele pagou. the bill he paid ‘The bill, he paid it.’ b¢. *A conta, pro pagou.

280 Mary Aizawa Kato

Japanese, on the other hand would never have a pronoun in such position:15 (16) a.

Pizza-wa, __tabeta. pizza-top __ate ‘The pizza, I/he/she ate.’ b. Okanjo-wa, __harata bill-top __paid ‘The bill, I/he/she paid.’

The difference between the two languages disappears when the subject is lexical: (17) a.

A pizza, o Pedro comeu ontem. the pizza the Pedro ate yesterday ‘The pizza, Pedro ate yesterday.’ b. Pizza-wa Pita-ga kinoo tabeta. the pizza Peter yesterday ate ‘The pizza, Peter ate yesterday.’

Going back to Gonçalves’ examples in (13), we see that he has represented them as canonical SOV sentences with a null subject, and not as [Topic [pro [OV]]] sentences. There is one example in his data that suggests that the structure I am assuming may be the correct one: (18) ele [eu bati] he [I beat ‘Him, I beat.’

If the forms in (13) and (14) are topicalized sentences, what the speaker has transferred is not the OV order from L1, but the unrestricted null subject feature of Japanese to BP, where the phenomenon is much more constrained, rather than the OV order of Japanese. The few cases in which Gonçalves’ children have an overt subject may be analyzed as a recursive topic construction [Topicsub [Topicobj [pro [V pro]]]], which is possible in both Japanese and BP. (19) a.

Maria, o carro, o pneu furou. Maria the car the tyre went-flat ‘Maria, her car had a flat tyre.’ b. Maria-wa, kuruma-wa, taia-ga panku-shita. Maria-top car-top tyre-subj went-flat ‘Maria, her car had a flat tyre.’

We have mentioned the case of the null subject in BP, which has a very limited

Child L2 acquisition 281

distribution when compared to a canonical null subject language like Italian, Spanish or Japanese. A similar distributional contrast is studied by SilvaCorvalán (1993) in Spanish-English bilinguals. She studies the use of the zero complementizer in L1 by speakers of Spanish who have acquired English as L2. Quoting Weinreich’s “permeability of linguistic cells”, she argues that “the permeability of grammars to foreign influence does not depend on its structural weaknesses, but rather on the existence of superficially parallel structures in the languages in contact…” (p. 20). She shows that the zero complementizer has a very peripheral use in Spanish, only in complement clauses of certain verbs, and a less restricted occurrence in English, where it can also occur in relative clauses. Influenced by L2, the bilingual speaker starts using the zero complementizer more productively in L1 Spanish than monolinguals. See also Müller and Hulk (2001), who compare a similar restricted to unrestricted licensing of null objects in the languages of simultaneous bilingual children. If my hypothesis is right, my early Portuguese may have contained more null subjects than is today licensed in this variety of Portuguese. Considering Japanese to be the “default” case, I had to reset my grammar, as the nonuniform morphology of BP required a more restricted use of the null subject. In other words, I had to learn that, regarding the null subject, BP is a sub-set of Japanese. Sentences like (15), abundant in the input, must have worked as positive evidence for the BP type of null subject. The difference between the L2 speakers of Portuguese like my parents, who learned it after the critical age, and the descendants like me seems to be that while I reset my grammar to restrict the domain where null subjects can occur, my parents’ Portuguese stayed fossilized in the “default” value. Summarizing these considerations, we can suggest that, regarding the head directionality parameter, selection was the result of direct learning of the functional lexicon, no transference having been involved in this domain. However, in the case of the null subject, as Japanese and Portuguese are in a subset relation, transfer may have occurred. The claim is, therefore, that in my acquisition of BP both direct and indirect access took place. 4.2 Residues of my L1 in my L2 But can we claim that my Portuguese is exactly like the Portuguese of L1 learners? We can say that in terms of proficiency it can be superior to that of many speakers. However, there are two domains where my capacity is not equivalent to that of a Brazilian born speaking Portuguese:

282 Mary Aizawa Kato

1. in the gender agreement domain, where it is not entirely automatic16 2. in prosody, identifiable as a little foreign-sounding by Brazilians and as “nissei-like” by other descendants17 I make mistakes in agreement in gender when writing, errors that I consciously correct when editing the text.18 So I am certain that I make similar errors when speaking. I edit my husband’s speech all the time correcting his gender agreement errors, though I myself make them when not monitoring my speech. As for prosodic evidence, if I am on the phone speaking to strangers, I know when, on the other side, there is a Japanese descendant, and I have also been identified as such by the other side. This seems to be due to certain aspects of my prosody and my tone of voice. We have to consider that prosody can be correlated to syntactic/semantic functions or to more expressive and interpersonal functions like anger, surprise, sex distinction, politeness, etc. Since I acquired a native-like syntax and semantics, the prosodic features in my BP that resembles that of a female Japanese speaker must be those that correlate with the latter. Like tone of voice, these prosodic features, once acquired, seem to be inevitably transferred to L2 As for gender agreement, why wasn’t it completely acquired when in all other aspects I seem to have full competence in BP? Following Kato (1994), I want to relate gender in BP to phonological harmony. There I proposed that in at least some Romance languages, gender can be acquired as a phonological relation since the article and the noun endings are vowel harmonic: (20) a. a casa b. o cavalo

‘the house’ ‘the horse’

Though much more opaque, pronouns and agreement morphemes also exhibit phonetic similarity: vowel harmony or alliterative relation:19 (21) a. Io quiero b. Tu hiciste c. nosotros vamos

‘I want’ ‘you did’ ‘we go’

Kato shows that the child plays very early with these rhymes: (22) a. ô bolo b. a casa

‘the cake’ ‘the house’

Raquel 1;9,8 (apud Kato 1994: 146)

(23) a. ô vô b. (voc)ê lê

‘I go’ ‘you read’

Raquel 1;9,8 (apud Kato 1994: 147)

Child L2 acquisition 283

The capacity to perceive these phonological regularities in agreement, like everything else in phonology (cf. among others, Werker 1995), seems to appear very early in the child and maybe the critical age for its acquisition is earlier than most syntactic phenomena. Agreement mistakes are common in L2 learners that have passed the critical age. In near native learners like me, errors are rare when the elements involved are superficially adjacent. However, distance or turn-shifting seems to break the link. Literacy helped to improve my performance, but a more complex parsing seems to bring back earlier stages of performance.

5. Peripheral and central modules of the mind For Chomsky linguistic knowledge is unconscious knowledge. We can say that one aspect that distinguishes competence and proficiency is the level of consciousness and automation that one has of the linguistic processes. Consciousness about one’s own performance is greater in L2 than in L1. Below we have this problem quoted in a research of Italian being acquired by a Japanese (Caruso 1990). (24) quando io sono/era s/ero? ero studen/studente when I __/__/__/__/___ student ‘when I was a student’

Thus, editing one’s errors is a frequent phenomenon in L2 speakers. Let us now take an example of the unconscious nature of L1. Speakers of Portuguese use the articles in a completely unconscious way. They only become partially conscious of their use, through schooling and metalinguistic knowledge. But to say that the definite article is used when something is definite does not help a Japanese learning Portuguese, whose language does not have the articles. For them it is incomprehensible when we tell them that we can say (25) Um cachorro me mordeu. a dog meclitic bit ‘A dog bit me.’

with an indefinite article, since there is nothing indefinite about the dog. Though Portuguese is one of my L2, my use of the article is completely unconscious. Just recently I learned metalinguistically that bare nouns in Portuguese can have a definite generic interpretation. But it is not in any

284 Mary Aizawa Kato

context that we can have bare plurals and have definite generic interpretation as in (26a). Sometimes the interpretation is existential/indefinite as in (26b): Eu adoro cenouras / as cenouras I love carrots / the carrots ‘I love carrots.’ b. Eu achei livros de sintaxe/alguns livros de sintaxe. I found books of syntax /some books of syntax ‘I have found books/some books on syntax.’

(26) a.

And this has no crosslinguisic validity. It is part of an abstract parameter that involves other properties. Thus, languages that allow bare plurals with a definite generic meaning are the ones that allow null objects (Raposo 1998). The other Romance languages license only bare plurals with existential reading. Speakers “know” these relations, though unconsciously. Second language learners, on the other hand, want to know why and when to use everything. In my learning of English I had to learn each rule for the use of the article. However, the knowledge of rules does not lead to error-free utterances. Thus, Brazilians quite proficient in English are often heard saying things such as (27) *I play piano and guitar. ‘I play the piano and the guitar.’

transferring the use of bare singulars from the Portuguese vernacular. Citing another domain, though I have learned that English has obligatory subjects and objects I easily catch myself saying things like (28), where the neuter pronoun it (expletive in a) is involved. Here, the transfer can have operated from Japanese to English since Japanese also licenses null subjects and objects. But Japanese descendants, who do not speak Japanese, like my daughters, exhibit a similar behavior. (28) a. *I think __ is easy. b. *I bought it without trying __.

As my knowledge of Portuguese is mostly unconscious, I can say that I am almost a simultaneous bilingual when Portuguese is concerned, confirming well-known studies like Johnson and Newport´s (1989), which shows that L2 acquired between 3 to 7 has the same characteristics of L1. The same cannot be said about my L3, English, which I started learning much later. The difference does not have to do with the simultaneous learning of reading and writing since

Child L2 acquisition 285

my acquisition of Portuguese involved also my literacy process. The difference could be due to age, but I believe it had also to do with the fact that the input was ordered, not natural, and with a lot of negative evidence. Besides the level of consciousness, another aspect that distinguishes L1 from L2 is the “capacity” to control one’s own perception. Bilinguals often give testimony about their failure in L2 production when under stress, when under the effect of alcohol or drugs or even when they are tired. These effects are much less felt with L1. As for perception, I have heard fewer comments. People say, for example, that, when listening to the radio or the TV in a foreign language, they increase the volume, which I do when watching a film in English, but not in Portuguese. I have also had another interesting experience in a bilingual setting. When I was in Hawaii waiting for a dolphin show to start, I was trying to concentrate on the book I was reading and trying to ignore the pre-show announcements and jokes of the announcer. I managed to switch off when the announcements were in English, but could not do so when they were in Japanese. First language is like the capacity to see. You cannot fail to see if your eyes are open. Similarly, you cannot fail to hear human talk if the language spoken is L1. My experience tells me that my reaction to Portuguese speech is the same as that with Japanese: it interferes with whatever I am doing. These facts take us back to the distinction between peripheral specialized modules and central systems. According to Fodor (1983), the processes in the former have, among others, the following characteristics: – – – –

they are specific of a domain of knowledge (e.g. vision, I-language) they are automatic (the eyes cannot fail to see if they are open; the unconscious nature of the computational operations of the mind) they are encapsulated (one’s vision knows when an object moves and when the eyes move) they are fast (the reaction is immediate when such processes are tested)

A multifunctional central system, on the other hand, has the following characteristics (Garfield 1989): – – –

it has access to the outputs of the specialized modules (Logical Form and Phonetic Form) it has limited or no access to intermediate stages (e.g. intermediate stages in the syntactic derivation) it is slow (the reaction can be slow when operations in such central system are tested)

286 Mary Aizawa Kato

We can presume, from the facts described above, that L1 has a component of the specialized type. L2 may have it if acquired before the critical age and under certain conditions. In the case of my English it seems that it lacks some of the properties of the specialized module like automaticity and speed. If perception and production in a second language acquired after the critical age exhibit different properties from L1, what can explain these differences? In the following section, I will first try to account for the L1 residues in my first L2, and then I will venture an explanation for the differences between my first and second L2.

6. Some hypotheses concerning the acquisition of L2 The explanation of my residual Japanese prosody and my errors in gender agreement will be based on a theory of different times for setting parameters and different critical ages, for each component of the grammar. People have been asking themselves whether the invariable Principles emerge all at once or whether each of the Principles is activated at a certain stage of maturation. The same is asked about Parameters in general. A lot of research has been made on the early acquisition of prosody and phonology (cf. Pinker 1994, Herschensohn 2000, and references therein). On the other hand, proposals have been advanced regarding prosodic change triggering morpho-syntactic change (cf., among others, Adams 1987), which raise the question as to the possibility of prosody and phonology constraining morphosyntactic selection. My hypothesis here is that prosodic parameters start to be set before everything else, phonology coming next and morphosyntactic parameters being fixed last. I also assume that there may be partial overlapping of the three domains. Critical ages for the setting of the three types of parameters would be different. Let us call CAPR the critical age to set prosodic parameters; CAPH the critical age to set phonological parameters and CAMS the one to set morphosyntactic parameters. The competence of L2 learners will have to do with how early they have started. If s/he has started before CAPR s/he can be syntactically competent but have failures in prosody and phonology: if s/he started learning before CAPH s/he may be syntactically and phonologically competent, but not prosodically competent. This hypothesis also predicts that there will be no L2 learner that is prosodically and phonologically competent but not syntactically competent. Now we can venture an answer to my lack of automaticity in gender

Child L2 acquisition 287

agreement and my residual Japanese prosody. My initiation in BP may have skipped an early stage of acquisition of prosodic features that do not correlate with syntactic form. As for gender agreement in BP, if it is both phonological and syntactic in nature, I may have started my learning after a certain stage of phonological acquisition where such agreement facts are processed, namely when we have an overlapping period of phonological and syntactic acquisition. What can we say, now, about my acquisition of English? First consider that the syntax of English is much closer to that of BP than the syntax of Japanese. It is more plausible that what functioned as the initial state for my English was BP and not Japanese, my L1.20 A Brazilian learner can transfer merge operations quite successfully from BP to English, but not from Japanese to English.21 The problems appear when movement operations are at play.22 Let us illustrate this with V-to-I movement, which is set positively in BP and negatively in English (examples in 29), and Aux-to-Comp movement, marked positively in English, but not in BP (examples in 30).23 (29) a. John always kisses Susan. a¢. [IPJohn I [always [VPkisses Susan]]] b. João beija sempre a Susana. João kisses always the Susana b¢. [IPJoão beijaV [sempre [VP tv a Susana]]] (30) a. Where can John be? a¢. [CP wherewh can [IP John tI [VP be twh]]] b. Onde o João pode estar? where the Joao can be b¢. [CP Ondewh [IP o João pode [VP estar twh]]]

However, BP has a small domain where structures similar to the English ones in (29) and (30) are found. Thus, the frequency adverbs can appear preverbally in BP if the adverb is negative or if the adverb is focalized (cf Moino 1996): (31) a.

João NUNCA beija a Susana João NEVER kisses the Susana ‘John never kisses Susan.’ b. João SEMPRE beija a Susana. João ALWAYS kisses the Susana ‘It is ALWAYS that John kisses Susan.’

As for Aux-to-Comp movement, BP may optionally exhibit this order with the copulas ser and estar in contemporary BP, but not with modals:24

288 Mary Aizawa Kato

(32) a.

Quem é você? who are you b. Onde estão os meninos? where are the boys?

Unlike English, where Aux-to-Comp movement is obligatory, BP allows optional movement of Aux-to-Comp: (33) a.

Quem você é? who you are? ‘Who are you?’ b. Onde os meninos estão? where the boys are? ‘Where are the boys?’

Aux-to-Comp movement may be considered a simplified property that divides languages into those that have C filled with an auxiliary element in wh-questions and those that do not.25 Except for the copulas, BP is a language that has no Aux-to-Comp movement. English, on the other hand, allows any auxiliary: be, have, do and the modals. A theory of parameters does not, in principle, account for language internal variation. But Chomsky (1981, 1986) asserts that the setting of parameters results in core grammars, an idealization. He admits that actual languages allow a marked periphery, where competing constructions may appear. This is the case of Aux-to-Comp movement in BP. While in its core grammar there is no Aux-to Comp, its periphery allows sentences with Aux-to-Comp movement. For Roeper (1999), the existence of such variation makes every speaker a bilingual, what he calls universal bilingalism.26 If L1 is the basis for the acquisition of L2, and L1 retains more than one choice of UG, then we can assume that my BP could have been used to build the grammar of English. We can say, then, that my learning of English, was, in many aspects, a cognitive decision of making something that was a property of the periphery in BP into a core feature of English. Perhaps the lack of automaticity, the errors and the incomplete nature of my second L2 have to do with such decisions.

7. Final considerations My early acquisition of BP as L2 was shown to have resulted in something akin to L1, except for some prosodic features, and some errors in gender agreement.



Child L2 acquisition 289

My assumption that there are different critical ages for the acquisition of prosody, phonology and syntax explains why there are different types of L2 learners. The earlier the L2 acquisition of a learner, the closer his/her competence will resemble that of an L1 learner. I have shown, on the other hand, that, though I had BP as L2, my proficiency in it is superior to that in Japanese. The problems with my Japanese today have more to do with proficiency than with competence, as it affects the use of the substantive class of words, a domain that has fossilized. However, my competence, namely the functional skeleton and the construal relations are intact, and my reaction when I hear Japanese, even when I don’t want to, reveals that competence is not degenerating. In other words, once acquired, competence cannot be lost, unless there are brain damages. Proficiency, on the other hand, can decrease with lack of use of the language. As for my English, which was acquired after the critial age and without natural input, I suggested that the base to its learning was my first L2, not only its core properties, but also what appears in its periphery, endorsing SilvaCorvalán’s proposal of the permeability of grammars, cited in Section 4.2.

Notes * I am grateful to Ruth Lopes, Jairo Nunes and the two anonymous reviewers for the valuable comments and suggestions, and Marcello M. Rosa for his editing contributions. 1. Though I started learning English in my early adolescence, its features are similar to what is said about adult L2 acquisition. 2. How much knowledge of functional words I acquired through the hiragana system is a topic I dealt with in Kato (1985). 3. Bickerton defines protolanguage as similar to pidgins, a purely associative system. 4. All the members of his family were born and educated in Japan. 5. The emperor has one only for himself. 6. This happened with my husband when he was a kid, as he was the first male after three sisters. And his sisters call him boku even today. 7. When speaking with Japanese from Japan, the source of borrowing and code-switching is English. 8. I am using the terms “direct” and “indirect” access, following Meisel (2000), who does not believe that the terms “full access” and “no access” are appropriate to discuss L2. 9. There is an extensive literature concerning the critical age for the acquisition of L2. I refer the reader to Herschensohn’s (2000) thesis for an up-to-date overview. Our assumption here

290 Mary Aizawa Kato

is that below the critical age (before puberty), the acquisition of L2 can be fairly close to that of L1. See further discussion in Section 6. 10. The reviewer finds it hardly convincing that the head-directionality parameter can be set by the pure evidence of postpositions or prepositions, mentioning that German has morphological case and prepositions. In my opinion, the existence of prepositions seems to be a sufficient trigger for the child to assign the head-initial value to the directionality parameter, at least for functional heads. Morphological case is independent as it can appear with a language with prepositions (German) or postpositions (Japanese). Japanese case markers are real postpositions and not bound case morphemes as in German since they may appear governing a whole phrase or sentence, which is not possible in German. 11. The only exception is found in Gonçalves (1997), who shows that his Japanese children (6 to 8 years old) learning Portuguese produce possessive phrase in BP in prenominal position like in Japanese, in variation with the proper BP postnominal position. Thus, the same child produces, in the same dialogue, the possessive phrase in prenominal (i) and in postnominal (ii) positions. (i) Raru é nome de gente, né… Raru is name of people isn’t it (ii)

Mas, de quem é o cachorro? but of whom is the dog? “But, whose is the dog”?

do Raru do papai of Raru of daddy “Raru’s daddy’s” What is important, however, is that the preposition is prenominal like in BP and not postpositional like in Japanese. The child seems to have raised the whole possessive phrase, but maintaining the head initial nature of the preposition. (iii)

12. Notice that the adjunct appears in the strict BP order, namely with a head-initial construction. 13. According to Duarte, subject pronouns in BP are categorically null when they are nonreferential (expletives, quantifier-bound subjects), but are overt when referential. Japanese, on the other hand, can have an empty category for both referential and non-referential subjects. This makes the null subject cases of BP a subset of the Japanese ones. 14. The remaining null subjects cannot be analyzed as licensed by morphology as in other Romance null subject languages. For a detailed discussion, see Kato (1999). 15. What seems to explain the possibility or impossibility of the null subject is a uniform verbal paradigm. Languages that have a uniform paradigm, whether because it has no agreement in any person, or because it has a distinctive agreement in every person are considered uniform (cf. Jaeggli & Safir 1989). BP has a defective paradigm, which explains its partial pro-drop nature. 16. Lack of agreement is a phenomenon also found in natives, but lack of agreement in gender is never found among natives, even among illiterates. 17. “Nissei” refers to second generation of Japanese.

Child L2 acquisition 291

18. Vowel harmony between Noun and Adjective is often broken when the adjective is not adjacent to the Noun. Reference to a previous masculine Noun is often made through a feminine pronoun in subsequent clauses. 19. The first person eu (pronounced [ô]) appears in vowel harmony with all four conjugations; the indirect second person você/cê, rhymes with all, but the first conjugation agreement: 1st ô falo ô leio ô parto ô ponho

cê fala cê lê cê parte cê põe

20. But an analysis in the same vein can be made if Portuguese had been my L1, in which case I would be endorsing the thesis that UG is indirectly accessible to L2 learning, through L1, when learning takes place after the critical age. For interesting recent work along these lines, see Herschensohn (2000). 21. The acquisition of English as L2 for someone who has only Japanese as L1 must have a very different account since the learner cannot resort to transfer operations. 22. Moreover, as “merge” is less costly than “move” (Chomsky 1995), its transfer may also be less costly than transfer of “move” operations. 23. We are giving a simplified account of the contrast. Of course movement has to be triggered by some strong feature in C, which I assume to be a strong Tense-feature. 24. Later, with schooling, Brazilians learn that Aux-to Comp movement is possible in literary Classic writing with any type of auxiliary (Lopes Rossi 1993): Que remédio pode el-rei dar a essas cousas? what medicines can the king give to these things? 25. What Rizzi (1996) considers “dynamic agreement”, as the head of Comp has no element that checks the wh-features. 26. In this particular case, Aux-to-Comp movement is restricted to a sub-domain of the lexicon, a possibility that Roeper also considered.

References Adams, M. 1987. Old French, Null Subjects and Verb Second Phenomena, UCLA: PhD Dissertation. Bickerton, D. 1990. Language and Species. Chicago and London: The University of Chicago Press. Caruso, M. 1990. “Italiano L2: prime observazioni sullïnterlingua di un soggetto giapponese”. Quaderni 6: 11–20. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986. Knowledge of language. New York: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass: The MIT Press.

292 Mary Aizawa Kato

Duarte, M. E. 1993. “Do pronome nulo ao pronome pleno”. In Português Brasileiro: uma Viagem Diacrônica, I. Roberts and M. A.Kato (eds), 107–128. Campinas: UNICAMP. Fodor, J. A. 1983. The Modularity of Mind. Cambridge, Mass: The MIT Press. Garfield, J. L. 1989. Modularity in Knowledge Representation and Natural-Language Understanding. Cambridge, Mass: The MIT Press. Gonçalves, S. L. 1997. A Aquisição do Português como Segunda Língua: o caso das crianças Yuba. Unicamp: MA Thesis. Herschensohn, J. 2000. The Second Time Around: Minimalism and L2 acquisition. Amsterdam/Philadelphia: John Benjamins. Jaeggli, O. and Safir, K. 1989. The Null Subject Parameter. Dordrecht: Kluwer. Johnson J. and Newport, E. 1989. “Critical period effects in second language acquisition: The influence of maturational state on the acquisition of English as a second language”. Cognitive Psychology 21: 60–99. Kato, M. A. 1994. “A theory of null objects and the development of a Brazilian child grammar” In How Tolerant is Universal Grammar?, R.Tracy and E. Lattey (eds), 125–154, Tübingen: Niemeyer. Kato, M. A. 1995 [1985]. O Aprendizado da Leitura. 4th ed. São Paulo: Editora Martins Fontes. Kato, M. A. 1999. “Strong pronouns and weak pronominals in the null subject parameter”. Probus 11(1): 1–37. Lopes-Rossi, R. M. A. 1993. “Estudo diacrônico sobre as interrogativas do português do Brasil”. In Português Brasileiro: uma Viagem Diacrônica, I. Roberts and M. A.Kato (eds), 307–342. Campinas: UNICAMP. 307–342. Meisel, J. M. 1994. “Getting FAT: Finiteness, agreement and tense in early grammars”. In Bilingual First Language Acquisition, J. M. Meisel (ed.), 89–129, Amsterdam/Philadelphia: John Benjamins. Meisel, J. M. 2000. “Revisiting Universal Grammar”. In Arbeiten Zur Mehrsprachigkeit 1: 25–31. Hamburg, Sonderforschungsbereich 538/Universität Hamburg. Moino, R. E. L. 1996. “Preenchimento das fronteiras V…V”. In Gramática do Português Falado: Vol IV., M. Basílio (ed.), 219–244, Campinas: UNICAMP. Müller, N. and Hulk, A. 2001. “Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages”. Bilingualism: Language and Cognition 4 (1): 1–21. Pinker, S. 1994. The Language Instinct: How the Mind Creates Language. New York: William Morrow. Radford, A.1990. Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Raposo, E. 1998. “Definite/zero alternations in Portuguese” In Romance Lingusitics: Theoretical Perspectives, A. Schwegler, B Tranel and M.Uribe-Etxebarria (eds), 197–212. Amsterdam/Philadelphia: John Benjamins. Rizzi, L. 1994. “Root infinitives as truncated structures in early grammars”. 18th Annual Boston Conference on Language Development. Rizzi, L. 1996. “Residual V-second and the wh-criterion” In Parameters and Functional Heads: Essays in Comparative Syntax, A. Belletti and L.Rizzi (eds), 63–90, New York/ Oxford: Oxford University Press.



Child L2 acquisition 293

Roeper, T. 1999. “Universal Bilingualism”. Bilingualism: Language and Cognition 2 (3): 169–186. Silva-Corvalán, C. 1993. “On the permeability of grammars” In Linguistic Perspectives on Romance Languages, W. J. Ashby, M. Mithun, G. Perissinotto and E. Raposo (eds), 19–44. Amsterdam/Philadelphia: John Benjamins. Werker, J. 1995. “Exploring developmental changes in cross-language speech perception”. In An Invitation to Cognitive Science, L. R. Gleitman and M. Liberman (eds), 87–120, Cambridge, Mass: MIT Press.



The Verb–Object parameter in simultaneous and successive acquisition of bilingualism* Anja Möhring and Jürgen M. Meisel University of Hamburg

1.

Introduction

In this paper, we present some results from a larger study contrasting simultaneous and successive acquisition of bilingualism. Our aim is to determine how the age of onset of language acquisition influences syntactic development. In this discussion, we focus on parameterized principles of Universal Grammar (UG), since we assume that they refer to properties of grammars which are subject to maturational change; see Smith and Tsimpli (1995) and Meisel (2000). In the present contribution, we focus on the acquisition of the OV/VO parameter, as it has been proposed by Neeleman and Weerman (1999). According to these authors, a number of word order properties in which OV and VO languages differ from one another, result from distinct settings of this parameter which is defined by Neeleman and Weerman (1999) in terms of different grammatical domains for head-government. More specifically, they argue that different settings of the OV/VO parameter lead to structural differences between English (VO), and Dutch (OV) with respect to basic word order, particle constructions, and Verb–Adverb–Object constructions, etc. In other words, these constructions form a cluster of surface phenomena related to this parameter. To the extent that the acquisition of such phenomena is the effect of setting a parameter to a specific value, they should therefore be expected to emerge simultaneously in the speech of learners. In this study we investigate the effects of setting the OV/VO parameter in the acquisition of German (OV) and French (VO) by bilingual (2L1) children acquiring both languages simultaneously from birth, as compared to the acquisition of German as a second language (L2) by adult learners whose native

296 Anja Möhring and Jürgen M. Meisel

languages are Italian, Spanish, and Portuguese, all VO languages. Based on the assumption that the human language acquisition device is designed for multilingualism, we view simultaneous acquisition of two or more languages as an instance of multiple “first” language acquisition, i.e. the course of development will not differ substantially from that in the respective monolingual L1 development, and it should lead to native competence in each of the languages. This capacity, however, declines in the course of later childhood, probably beginning around the age of four years, cf. Hyltenstam and Abrahamsson (to appear). Consequently, adult second language learners should exhibit different acquisitional patterns.1 With respect to the setting of the OV/VO parameter, the prediction is that bilinguals set the parameter to the respective value in each of the languages, even if they differ in their parameter settings. Adult L2 learners, on the other hand, are predicted not be able to set the OV/VO parameter to a target value differing from the source value; consequently, we expect to find that the cluster of phenomena related to the parameter are learned separately in the course of adult second language acquisition. The specific hypothesis we want to test can thus be formulated as in (1): (1) 2L1 children set the OV/VO parameter in each of their languages to the respective values. The cluster of structural phenomena related to this parameter will therefore appear in each of the languages within a short period of time, i.e. during the same acquisitional phase. L2 learners do not have access anymore to parameter values not instantiated in their L1. They therefore have to learn these phenomena independently of one another and distributed over various acquisitional phases.

This paper is organized as follows. First, we briefly introduce the OV/VO parameter, as proposed by Neeleman and Weerman (1999). We then discuss some implications of this approach for the acquisition of constructions related to this parameter in French and German. In the next step, we present the results of an empirical analysis of the acquisition of these phenomena by two French-German bilingual children and of two adult Portuguese and Spanish learners of German. To conclude, we offer an interpretation of these findings, focusing on some crucial differences between simultaneous and subsequent acquisition of syntax.

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 297

2. The OV/VO parameter The fact that languages differ with respect to the order in which verbs and their complements are arranged has long been recognized as an essential property of grammars. In typological studies, the surface order of these elements is used as a basic criterion of classification, distinguishing between OV (Object–Verb) and VO (Verb–Object) languages, and it is related to a number of other word order patterns, e.g. the position of attributive adjectives, the possessor-possessed sequence, etc., see Greenberg (1963). In the tradition of generative syntax, the OV–VO distinction is generally understood as an abstract property of syntax, defined at an underlying level from which the various surface patterns need to be derived.2 In view of the fact that grammars of natural languages differ systematically in this respect, this is interpreted as resulting from a parametric difference, frequently referred to as the head-directionality parameter which determines whether the head of an XP precedes or follows its complement, see, for example, Chomsky (1981), Kayne (1983), Koopman (1984), Stowell (1981), or Travis (1984). From the perspective of language development studies, a particularly interesting approach is the one suggested by Neeleman and Weerman (1999) which takes into account recent developments in syntactic theory, i.e. the Minimalist Program (cf. Chomsky, 1995). As opposed to many competing theories, it preserves an essential property of syntactic parameters suggested by Chomsky (1981): the clustering of apparently unrelated surface phenomena which are related to a specific setting of the parameter. This allows for empirical scrutiny of claims regarding parameter setting in language acquisition since the phenomena dependent on a particular setting of a parameter should become available to the child simultaneously, i.e. as soon as the parameter is set to one of its possible values. On the basis of the Minimalist Program (Chomsky 1995), Neeleman and Weerman (1999) develop a theory of Case and word order.3 In line with Chomsky (1995), they argue that linear order is not stipulated in syntax. Neeleman and Weerman propose that no fixed argument positions exist within the clause; instead arguments are freely generated, e.g., either to the right or to the left of the verb. Conditions imposed at PF, having to do with Case Theory and the licensing of empty categories, determine which one of the possible argument positions is chosen. Furthermore, they suggest that the thematic rules mapping syntax onto an interpretation take Case as their point of reference in syntax. Case Theory is therefore relevant at both LF and PF. Case, in this framework, refers to the functional marker on an argument rather than to a

298 Anja Möhring and Jürgen M. Meisel

fixed structural position. Neeleman and Weerman assume that arguments which bear overt morphological case project a case shell and that the head of this shell is filled with morphological material. In languages without overt case morphology, the case shell is empty and must, like all empty categories, be appropriately governed (Empty Category Principle; Rizzi 1990). Arguments with empty case shells must therefore be in a position at PF where they are properly governed. The Head Directionality parameter and constraints on the domain of proper government determine in what position the arguments surface at PF with respect to the verb. If the case shell is filled, the DP need not be governed and, consequently, the surface position of the argument may be much freer. As for the OV/VO parameter, Neeleman and Weerman (1999) conclude that it is to be defined in terms of two grammatical domains: a) the domain of Θ-role assignment (exclusively defined in terms of hierarchical relations), and b) the domain of head government. OV/VO variation is understood as the consequence of the direction of head government in the language being learnt, this holding at PF. Head government depends on two variables: domain and direction. The authors propose that a particular setting of the directionality parameter automatically entails a specific setting of the domain parameter.4 The general definition of head government reads as in (2), quoted from Neeleman and Weerman (1999: 24). (2) Head Government (general): A head α governs β iff a. α precedes/follows β, and b. α and β are contained in the same phrasal domain.

Neeleman and Weerman (1999: 24) argue that both syntactic and phonological information is accessible at PF. Consequently, the phrasal domain “can either be defined syntactically or phonologically.” As the task which must be met at the PF interface is to prepare sentences for the requirements of pronunciation, Neeleman and Weerman assume that there is a strong inclination to shift to phonological information as soon as possible. The domain of head government for VO languages is therefore defined in prosodic terms, i.e. the verb and the object are in the same phonological phrase (Neeleman and Weerman 1999:25f); see (3). In OV languages, on the other hand, the object and the verb are not contained in the same phonological domain. In order to establish a head government relation between these elements, syntactic information must be accessed. The head and its governees must be in the same syntactic phrase, that is, in the same m-command domain; see the definitions given in (4).

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 299

(3) Head Government (VO) A head α governs β iff a. α precedes β, and b. α and β are contained in the same Φ.5 (4) Head Government (OV) A head α governs β iff a. α follows β, and b. α and β m-command each other.

Neeleman and Weerman (1999) demonstrate how different settings of the OV/VO parameter lead to structural differences between English (VO), and Dutch (OV), e.g. basic word order, particle constructions, and Verb–Adverb– Object constructions. These structures may thus be understood as forming a cluster of surface phenomena related to this parameter. Similar clusters should therefore be found in other VO and OV languages, respectively. Head government is defined phonologically in Romance languages (VO) and syntactically in German (OV). Note that Neeleman and Weerman argue that internal arguments in all these languages (English, Dutch, German and the Romance languages) bear an empty case shell and must therefore be properly governed.6 The resulting structural differences between German and French are summarized in Table 1. It shows that these languages differ, as predicted, with respect to surface verb complement order in main and embedded clauses, adverb placement, and particle constructions. As for the latter, however, the difference is obscured by the fact that French does not allow for such constructions at all. The comparison between Dutch and English, on the one hand, and German and French, on the other hand, is further complicated by some uncertainties involving Verb–Adverb–Object constructions. Neeleman and Weerman (1999: 39) refer to structures “in which adverbials may freely be inserted between the object and the verb” as scrambled constructions. However, the term “scrambling” is ordinarily used to refer to constructions which result from object raising. Moreover, the question of whether scrambling should be restricted to cases of rearrangement of two or more objects, or whether it may also refer to instances of object raising past adverbs, raises some controversy, see e.g. Haider (2000). In order to avoid terminological confusion, we refrain from adopting Neeleman and Weerman’s definition of scrambling. This does not affect the fact that Adverb–Verb–Object sequences are structurally linked to the OV/VO parameter.

300 Anja Möhring and Jürgen M. Meisel

Table 1.Some similarities and differences between German and French with respect to phenomena linked to the OV/VO parametera

Verb-complement order in main clauses

German

French

vO Ich sehe den Mann ‘I see the man‘

vO Je vois l’homme ‘I see the man’

vOV vVO Ich habe den Mann gesehen J’ai vu l’homme I have the man seen ‘I have seen the man’ ‘I have seen the man ‘ OVv vVO Verb-comple…, qu’il a vu l’homme ment order in …, dass er den Mann gesehen hat ‘that he has seen the man’ embedded claus- …, that he the man seen has’ ‘…, that he has seen the man’ es vV DP Adv Adverb place- Adv DP Vv …, dass er gestern den Mann gesehen hat Il a vu l’homme hier ment …, that he yesterday the man seen has’ ‘He has seen the man yesterday’ ‘…, that he has seen the man yesterday’ DP Adv Vv (*)vV Adv DP …., dass er den Mann gestern gesehen hat *Il a vu hier l’homme …, that he the man yesterday seen has ‘He has seen yesterday the man’ ‘…, that he has seen the man yesterday’ Il m’a raconté aussitôt la nouvelle He me has told immediately the news ‘He immediately told me the news’ PP Prt Vv – Particle con…, dass sie vom Flugzeug abgesprungen structions ist …, that she from the plain off jumped is ‘…, that she has jumped off the plain’ *Prt PP Vv – * …, dass sie ab vom Flugzeug gesprungen ist …, that she off from the plain jumped is ‘…, that she has jumped off the plain’ DP Prt Vv – …, dass sie das Telefon abkaufen wollen …, that they the telephone PT-buy want ‘…, that they want to buy the telephone’ *Prt DP Vv – * …, dass sie ab das Telefon kaufen wollen …, that they PT the telephone buy want ‘…, that they want to buy the telephone’ a

Small “v” stands for finite verb forms, capital “V” for non-finite forms.

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 301

According to the analysis by Neeleman and Weerman (1999), the grammaticality of Dutch OAdvV sequences follows from the fact that the domain of head government is defined in syntactic terms. In English, on the other hand, adverb placement between the verb7 and the object is ruled out by the phonologically defined domain of head government, since Object–Adverb–Verb sequences cause closure of the phonological phrase; the verb and the object would no longer be in the same phonological phrase and, consequently, head government is blocked. This is illustrated by (5) and (6). (5) a. Mary always drinks milk. b. *Mary drinks always milk. (6) a. *

V¢ Ø-DP8

V¢ V

AdvP

b. {VAdvP}{Ø-DP}

As opposed to English, French does exhibit adverb insertion.9 (7) Marie boit toujours du lait Mary drinks always  milk ‘Mary always drinks milk’

This difference between English and French is not, however, due to properties of the OV/VO parameter but to the fact that French, as opposed to English, allows for verb raising; see Pollock (1989). In other words, French VAdvO patterns are the result of verb raising into the functional domain and, consequently, are not related to the OV/VO parameter as is the case in OV languages. Our contention thus is that the position of adverbs between objects and verbs results from different syntactic processes in French and German. We expect to corroborate this claim by our analysis of the acquisition data, below.

3. Implications of the OV/VO parameter for the acquisition of French and German 3.1 The OV/VO parameter in bilingual first language acquisition The goal of this section is to explore the empirical consequences following from the assumption that children, in developing their first languages, set the OV/VO

302 Anja Möhring and Jürgen M. Meisel

parameter to one of the two values offered by Universal Grammar (UG). Neeleman and Weerman (1997) discuss this issue with respect to Dutch and English. They emphasize that not all the phenomena related to the parameter may reasonably be expected to emerge simultaneously since some of the necessary structural prerequisites have not yet been acquired at the point of development when the parameter is set, e.g. embedded clauses. Indirect empirical evidence is, nevertheless, available because, once the parameter is set, errors in using structures linked to the parameter should not occur. As for children acquiring French and German simultaneously, a possible acquisitional problem consists in the fact that they have to set the OV/VO parameter to opposing values in the two languages. In accordance with our basic assumption that the human language faculty enables children to develop more than one linguistic competence, we nevertheless propose the following hypothesis in (8) which will be tested in the empirical part of this paper. (8) Although German and French exhibit conflicting settings of the parameter (French: VO; German: OV), children acquiring French and German simultaneously will set the OV/VO parameter to the appropriate value in each of the languages. The structural properties dependent on a specific parameter value will therefore emerge during the same acquisitional phase.

More specifically, our prediction is that, just as in monolingual acquisition, one should not find target deviant sequences with respect to the cluster of phenomena listed in Table 1 in the speech of the bilingual children, once the parameter is set to the respective value in German and French. If, however, constructions linked to the parameter do not appear at all in the speech of these children, during a given period, the explanation for this absence must rely on facts independent of the OV/VO parameter. 3.2 The OV/VO parameter in adult second language acquisition As mentioned briefly in the Introduction above, we start from the assumption that the easily observable differences between first and second language acquisition are, to a large extent, due to the fact that second language learners do not have access any more to the language making capacity (Slobin, 1985) which is available to the child. More specifically, the claim is that due to neurolinguistic maturation some properties of this capacity become inaccessible beyond the limits of the critical period for language acquisition. In grammatical terms, we assume that these properties are the ones which are commonly

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 303

described as parameterized principles of UG, see Smith and Tsimpli (1995). Consequently, non-parameterized principles continue to guide the acquisition of second languages, whereas parameterized principles not instantiated in the L1 of an L2 learner, as well as parameter values not activated in the course of L1 development, are not accessible any more to the L2 learner; see Meisel (2000). An obvious consequence is that L2 learners can tap their L1 grammatical knowledge in acquiring the L2, but they cannot “reset” parameters, or, phrased more correctly, they cannot set parameters to different values, as opposed to the ones on which they have been fixed in the course of L1 development. In order to avoid misunderstandings, we should hasten to add that this does not mean, of course, that L2 acquisition is impossible. Rather, it entails that different learning mechanisms are operative in L1 and in L2 acquisition, resulting in different acquisitional patterns and, in all likelihood, in different kinds of knowledge about the target language. With respect to parameterized principles, the prediction is that the various phenomena related to a parameter need to be learned separately and inductively. Moreover, in order to make up for the lack of information offered by UG, L2 learners will resort to general, i.e. not language-specific, learning mechanisms; see Meisel (1991, 1997a,
b, 2001). In the present context, it is obviously not possible to present a more elaborate discussion of the theoretical foundations of this approach. This brief outline of the position advocated here must suffice, but we can refer to a number of other authors who have developed similar ideas in more detail, e.g. Bley-Vroman (1989), Clahsen and Musyken (1986), Hawkins and Chan (1997), among many others. With respect to the acquisition of the OV/VO parameter, the hypothesis which can be deduced from this theoretical position is the following in (9): (9) Second language learners acquiring a language in which the OV/VO parameter is set to a different value than in their L1 cannot change the setting of this parameter to the value of the target grammar. In order to be able to use the L2 in a native-like fashion, they need to learn the constructions related to the L2 parameter setting independently of one another.

Earlier research on the acquisition of word order by second language learners provides some preliminary evidence supporting this hypothesis. It indicates that some of the phenomena listed in Table 1 are used almost without errors, from relatively early on, whereas other represent major difficulties for L2 learners; see, for example, Meisel, Clahsen and Pienemann (1981), Clahsen and Muysken (1986), Müller (1998), Neeleman and Weerman (1997), Towell and Hawkins (1994). We will return to this problem in Section 5, below.

304 Anja Möhring and Jürgen M. Meisel

4. The OV/VO parameter in 2L1 acquisition (French/German) 4.1 The children In what follows, we present an analysis of the language use of two FrenchGerman bilingual children, Pascal and Annika.10 The recordings of these children are part of the corpus gathered by the research group DuFDE (Deutsch und Französisch — Doppelter Erstspracherwerb — German and French — Simultaneous First Language Acquisition).11 Both children grew up in a bilingual setting where each parent uses his or her respective native tongue when communicating with the children, thus following the well known strategy of ‘une personne, une langue’, first mentioned in the literature on bilingual acquisition by Ronjat (1913). The children were videotaped every second week. More detailed information about the children and the empirical methods employed are given by Köppe (1994). Table 2 displays their ages and MLU ranges during the period analyzed in the present paper. Table 2.Summary of recordings Child

Age

MLU

Pascal

French: 1;05–3;05 German: 1;05–3;06

French: 1–5.44 German: 1–5.79

Annika

1;09–2;09

French: 1–3.55 German: 1–3.5

4.2 The development of German and French In what follows, we examine the emergence and development of the structural cluster related to the OV/VO parameter in order to determine whether the empirical predictions of hypothesis (8) can be corroborated. 4.2.1 Verb–Object order in main clauses Although German underlying word order is OV, surface main clause word order is vO for constructions with simple finite verbs and vOV for complex verbal constructions involving a modal or an auxiliary combined with a nonfinite verbal element.12

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 305

14 12 10

vO

8

VO

6

vVO

4 2 9 ,2

4

06 3;

0

,1 04

3;

1

,0 01

3;

6

,1

,1

11 2;

6 ,1

2;

09

9 ,1

06 2;

1

05 2;

5

,2 04

2;

8

,2 02

2;

9

,2 01

,0

2;

3 ,1

00 2;

11 1;

09

,1

1

0 1;

Number of utterances with VO structure

In French, surface as well as underlying order is normally VO, independently of whether the verb is finite or not (see Table 1 for both languages). Once the OV/VO parameter has been set to the specific value, we expect to find OV structures with non-finite verbs in German, but vO structures with finite verb-forms. In French, only VO sequences should appear. The developmental pattern is illustrated in Figures 1 and 2, based on the analysis of Pascal’s data, and in Figures 3 and 4, based on the analysis of Annika’s data.

Age

Number of utterances with VO structure 1; 09 ,1 1 1; 11 ,1 3 2; 00 ,0 9 2; 01 ,2 8 2; 02 ,2 5 2; 04 ,2 1 2; 05 ,1 9 2; 06 ,1 6 2; 09 ,1 6 2; 11 ,1 1 3; 01 ,0 3; 0 04 ,1 3; 4 06 ,2 9

Figure 1.Pascal: German VO structures in main clauses.

14 12 10

Ov

8

OV

6

vOV

4 2 0

Age

Figure 2.Pascal: German OV structures in main clauses.

14 12 10 8

vO

6

VO

4 2 1 09

;3

9 2;

2;

08

;1

1 ;0

1 2;

07

;0

8 2;

06

;1

6 2;

04

;1

2

03 2;

02

;2

0 2;

2;

01

,1

7 ,1

8

00

,1

2;

1;

1;

10

,2

1

0 09

Number of utterances with VO structure

306 Anja Möhring and Jürgen M. Meisel

Age

14 12 10

Ov

8

OV

6

vOV

4

OvV

2 1 2;

09

;3

9 08

;1

1 2;

;0 07

;0

1 2;

06

04

;1

8 2;

6 ;1 03 2;

2;

2 2;

02

;2

0 ,1 01

,1

7 2;

00

10

,1

8 2;

1;

1;

,2

1

0 09

Number of utterances with OV structure

Figure 3.Annika: German VO structures in main clauses.

Age

Figure 4.Annika: German OV structures in main clauses.

Our analysis reveals for both children a marked preference for OV order in German main clauses during an earlier phase, mostly involving non-finite verb forms. vO patterns begin to be used more frequently during a later phase, once the children use finite verb forms productively. Complex verbal constructions generally need more time to emerge; this is why vOV patterns are attested only during later phases in the data. Our claim is that the emergence of VO structures is the result of (finite)

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 307

verb raising and depends, therefore, on the acquisition of finiteness. Following Meisel (1994), we define productive usage of finite verbs in terms of the emergence of at least two different agreement markers with different verb types. As for Pascal, we find first evidence for productive use of finite verbs at age 1;10,28 (MLU ~ 1.70). Annika starts producing different verb types with two agreement markers from age 2;03,30 (MLU ~ 2.05) onwards. Although the data from both children confirm this observation, we also find exceptions in the data of both children, e.g. object placement in front of finite verbs or after non-finite verbs. Focusing on the fixing of the OV/VO parameter, we concentrate here on the children’s German utterances with VO placement. We counted 12 (15%) out of a total of 82 utterances with (v)VO order in Pascal’s data and 7 (17%) out of 41 utterances with VO order in Annika’s data. Examples are given in (10). (10) a.

öh guauen ein schiff eh build-inf a ship ‘to build a ship’ b. ja, will auch essen dieses yes want also eat-inf this ‘yes I also want to eat this’ c. [faloän] das das [forgotten-pt that that ‘forgotten that’

Pascal 1;11,28

Pascal 2;08,17

Annika 2;03,30

Annika starts producing complex verbal constructions only as of 2;09 onwards. Although we did not analyze her speech systematically after the age of 2;09, we did examine the following three recordings in order to get an impression of her usage of complex verbal constructions. All in all, we counted eight utterances with complex verbal constructions, only one of them exhibiting vVO order; see example (11). (11) nein muss a hauen den bär no must-inf a hit the bear ‘no must hit the bear’

Annika 3;00,15

At this point, we cannot offer a satisfactory explanation for what causes the appearance of these pattern. A systematic comparison with the acquisition patterns in the French of other bilingual and of monolingual children could perhaps shed some light on this problem, but this is beyond the scope of the present paper. At any rate, what matters for our purposes is that such phenomena

40 30 VO

20

OV

10

4 ,1

0 3;

04

,0

3 3;

01

,1

7

10

,1

2;

2;

06 2;

2;

08

,0

2

5 05

,0

,0

7

2

04 2;

2;

2;

01

02

,1

,1

4

9 ,0 00

2;

09

,1

1

0 1;

Number of utterances with VO and OV structure

308 Anja Möhring and Jürgen M. Meisel

Age

40 30 VO

20

OV

10

1 ;3 09

;1

9 2;

08

07

;0

1 2;

1 ;0 06

2;

2;

8

2;

04

;1

6

2;

03

;1

2 ;2 02

,1

0 2;

01

,1

7 2;

00

,1

8 2;

10 1;

09

,2

1

0 1;

Number of utterances with VO and OV structure

Figure 5.Pascal: French VO and OV structures in main clauses.

Age

Figure 6.Annika: French VO and OV structures in main clauses.

are extremely rare during the early acquisitional phase, i.e. during the period when, according to our hypothesis, the children set the OV/VO parameter. Note also that OV represents the clearly dominant pattern throughout the entire period of investigation. We take these facts as constituting evidence in support of the claim that both children have acquired the German Verb–Object order in main clauses.

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 309

In French too, the development of Verb–Object structures in both children corroborates our predictions. In other words, as with French monolingual children, one finds VO order in constructions with lexical objects almost exclusively, although Annika’s data exhibit slightly more variability than Pascal’s. Figures 5 and 6 display the corresponding developmental patterns in Pascal’s and Annika’s acquisition of French. These figures demonstrate that verb object placement is generally stable in both children during the earliest period of syntactic development in French. In other words, in their first multi-word utterances, they use only VO sequences up to the ages of 2;04,07 (Pascal) and 2;04,18 (Annika). Admittedly, however, the evidence for this early phase is rather scanty in Annika’s case since she only uses one Verb–Object sequence during this period. All we can say, therefore, is that the available evidence does not stand in conflict with our hypothesis. Figures 5 and 6 also demonstrate that VO order remains the dominant word order pattern during the whole period of observation. During the later phase, we find several OV strings in the data of both children. Ferdinand (1997) also observed utterances of this type in the data of monolingual French children which she interprets as the result of fronting of constituents. She argues that these structures occur only in a later developmental phase following phase I during which only subject clitics are placed before finite verbs. In phase II, constructions are observed in which other constituents too are fronted, e.g. objects. Ferdinand concludes that these fronted objects bear contrastive focus and that the children raise them into a sentence-initial Focus position. Hulk and van der Linden (1996) study Verb–Object placement in the data of a French-Dutch bilingual child, Anouk. Just like Ferdinand (1997), they find OV structures during the developmental phase following the beginning of Anouk’s productive use of subject clitics. But whereas object fronting is rare in monolingual French children, this phenomenon seems to be more frequent in bilingual children. Hulk and van der Linden (p. 98) explain the higher frequency of these structures with “the (indirect) influence of Dutch, a head final language where [XP Vi] orders are very frequent”. More data from bilingual as well as monolingual children need to be examined in order to decide whether this explanation can indeed be confirmed. The following examples illustrate the kind of OV structures which appear in the data of the children studied here. We would like to point out that 12 out of a total of 16 French OV utterances of both children contain preverbal ça, an element frequently used with contrastive focus. This seems to support the explanation proposed by Ferdinand (1997).

310 Anja Möhring and Jürgen M. Meisel

(12) a.

un un tracteur veux a a tractor want ‘(I) want to have a tractor’ b. (ça) enleve(r)? (that take away ‘(can you) take it off?’ c. ça faut mettre that must put ‘must put it (here/there)’ d. ah ça connais ah that know ‘ah (I) know that’ e. non ça seul chercher là no that alone search-inf there ‘(I’m) searching that one alone there’ f. non (mais) à maman porter no (but to mom bring-inf ‘no but (I) bring it to mom’ g. là dans [edad] (= regarde) là dans there in look there in ‘in there look in there like this and comme ça et puis ça faire like this and then that do-inf then you have to do that’

Pascal 2;04,07

Pascal 2;05,05

Pascal 2;08,17

Pascal 3;02,23

Annika 2;04,18

Annika 2;06,18

Annika 2;09,31

Note further that VO structures clearly represent the dominant pattern, OV structures appearing much less frequently (Pascal: 2;04,07–3;04,14: 8 OV (4%); 192 VO (96%); Annika: 2;01,24–2;09,31: 8 OV (29%); 20 VO (71%)). Kaiser (1994) has shown that Pascal starts using subject clitics productively at the age of 2;03, one month before he uses the first OV structure. First productive usage of subject clitics in Annika’s data is observed at the age of 2;00,17, long before the first occurrence of OV structures. In other words, we find in Pascal’s and Annika’s data OV structures only in a later developmental phase, as is the case for monolingual children. (13) a.

il dort ‘he sleeps’ b. il est là ‘he is there’

Annika 2;00,17 Annika 2;00,17

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism

Since we do not find qualitative differences between monolingual and bilingual children, we interpret our findings as evidence supporting the claim that the children set the Verb–Object parameter to the respective value in each of the languages as soon as they begin to use multi-word utterances productively. 4.2.2 Verb–Object order in embedded clauses In German embedded clauses, verbal elements normally appear in clause-final position, thus resulting in OV order. In fact, both the finite and the non-finite verb are placed clause-finally, but in the reversed order, as compared to main clauses, if both are present, i.e. OVv. Only in embedded clauses introduced by the conjunction denn ‘for’ is vO(V) order required, and in colloquial speech of some regional varieties, weil ‘because’ and obwohl ‘although’ allow for both, vO(V) and O(V)v order. Our prediction is that once the OV/VO parameter is set to the OV value in German, only OV patterns should be attested in embedded clauses with conjunctions licensing OV order. This is precisely what we find in the data of both children. The following examples illustrate Pascal’s use of Verb–Object sequences in embedded clauses. Examples with non-finite verbs are in bold letters. (14) a.

weil ich — ich will viele sammeln because I  I want a lot collect ‘because I want to collect a lot’ b. weil ich keine feder habe because I no feather have ‘because I have no feather’ c. weil — {der/die} hat keine flügel because  {he/she} has no wings ‘because (s)he has no wings’ d. weil der keine federn haben because he no feathers have-inf ‘because he to have no feathers’ e. doch, das geht wenn man so krack macht but it works when you so crack make ‘but it works when you make crack, this way’ f. ich möchte
du machst das I want [that] you do this ‘I want you to do this’

Pascal 2;10,27

Pascal 2;11,11

Pascal 2;11,24

Pascal 2;11,24

Pascal 3;02,23

Pascal 3;04,14

311

312

Anja Möhring and Jürgen M. Meisel

g. weil da- da kann man sich aua machen because  there can you yourself ouch do ‘because you can hurt yourself there’ h. und wenn man alles auszieht and when you all put off ‘and when you put all off’ i. we (?= wenn) ma plumps macht when you bump make ‘when you fall’ j. wenn man alle legt when you all put ‘when you put all’

Pascal 3;04,14

Pascal 3;04,14

Pascal 3;05,16

Pascal 3;05,16

We find a predominance of OV order in Pascal’s German subordinate clauses with both finite and non-finite verbs. With respect to constructions with nonfinite verbs, we observe only (v)OV order. On the basis of these findings, we want to claim that Pascal has already set the parameter to its appropriate value when, at age 2;10,27, he starts using embedded clauses with verbs and objects in German. The fact that Pascal places the object in front of the non-finite verbs in embedded clauses (14a., d., g.) confirms this claim. At first glance, Annika’s data appear to exhibit more variability than Pascal’s concerning word order in subordinate clauses. This is illustrated in (15). Utterances with non-finite verbs are again given in bold letters. weil ich noch nicht keller gemacht hab because I still not cellar done have ‘because I still have not done (it) in the cellar’ b. weil ich hab doch geburtstag because I have adv birthday ‘because it’s my birthday’ c. wei [is] hab doch gar [nist] gebur- burstag because [I have adv at all [not birthday ‘because it’s not my birthday’ d. weil das habe ich von mama because this got I from mom ‘because I got it from mom’ e. weil das will ich nicht because that want I not ‘because I don’t want to have it’

(15) a.

Annika 2;08,19

Annika 3;01,26

Annika 3;02,10

Annika 3;02,10

Annika 3;02,24

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism

f.

wenn ich da die schieb so so so! when I there it push this way this way this way ‘when I push it this way’ g. wenn ich die gelbe füllung (sprenge) when I the yellow filling (blow up ‘when I blow up the yellow filling’ h. weil hab kein, kein bett gehabt because have no no bed had-pt ‘because I didn’t have a bed’ i. sondern also einen großen tisch gehabt but thus a big table had-pt ‘but I had a big table’ j. weil, weil ich will nur ein, ein because because I want only a a luftballon haben balloon have-inf ‘because I only want to have a balloon’ k. weil ich will das because I want that ‘because I would like to have it’

Annika 3;02,24

Annika 3;02,24

Annika 3;02,24

Annika 3;02,24

Annika 3;02,24

Annika 3;02,24

Although Annika uses vO structure in embedded clauses more frequently than Pascal, the examples in (15) show that most of these sentences are introduced by weil. In cases where Ov order is required (15f and g), Annika produces the target-like word order pattern. Moreover, Annika uses only (v)OV order in embedded clauses with complex verbal structures (15a, h, i, j). This indicates that Annika has already set the OV/VO parameter to the target-like value. In French too, we expect the children to use only VO order in embedded clauses, once the correct parameter value has been fixed. This is exactly what we find in our data. As soon as Pascal and Annika start producing embedded clauses, they use VO order almost exclusively. In (16) we list several examples found in the data of the children. Examples with infinitives or participles are in bold letters. (16) a.

parce que il a mordu de petits garçons, because he has killed (de) little boys le lion the lion ‘because the tiger has killed (the) little boys’

Pascal 2;09,01

313

314 Anja Möhring and Jürgen M. Meisel

b. oh c’est quoi qu’il a fait ça oh that’s what that he has done that ‘oh what’s that that he has done’ c. parce que mouille mon chemisier because wet my blouse ‘because I wet my blouse’ d. hein parce que toi toi as ça cassé eh because you you have it broken ‘eh because you broke it’ e. [pastö] elle a la caméra là [because she has the camera there ‘because she has that camera’ f. (parce que) euh (mon) papa a fait ça (because euh (my father has done that ‘because my father has done that’

Pascal 2;09,16

Annika 2;08,19

Annika 2;10,27

Annika 3;02,10

Annika 3;06,16

We found only one example in Annika’s data of a vOV string that might be due to the influence of German. We conclude from these findings that the children’s use of subordinates confirms the results obtained from the analysis of word order patterns in main clauses. The OV/VO parameter is set to its appropriate value from very early on. The empirical confirmation of the predicted clustering effect, however, has to be interpreted in light of the fact that embedded clauses appear only several months after the setting of the parameter. Our findings nevertheless support the claim that the parameter has been set to its appropriate value since the expected word order patterns emerge as soon as embedded clauses are attested in the data. 4.2.3 Verb–Adverb–Object constructions Another word order pattern which Neeleman and Weerman (1999) show to be related to the OV/VO parameter, consists of Verb–Adverb–Object constructions. German, as an OV language, allows insertion of adverbs between the verb and the object. Consequently, two word order patterns are possible, AdvOV and OAdvV. Neeleman and Weerman (1997) and Schaeffer (2000) found that, in the acquisition of Dutch as L1, both word order patterns appear as soon as adverbials, verbs and definite object-DPs are combined. We expect to find the same in Pascal’s and Annika’s German data. Figures 7 and 8 give an overview of the Object–Adverb–Non-finite verb combinations found in Pascal’s and Annika’s German utterances.

6

4 OAV AOV 2

6 ;1

4

05

;1

3 3;

3;

04

;2

7

02

;1

3;

00

;1

3;

11 2;

2;

2;

10

08

;1

1

3

7 ;1

7 ;0 04

2;

2;

02

01

;2

;2

5

8

0 2;

Number of utterances with OAV or AOV sequence

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism

Age

6

4 OAV AOV 2

1

2;

09

,3

3

9

,0

2;

09

,1

9

08 2;

07

,2

1 2;

2;

07

,0

8

1

,1

2;

06

,0

8

06 2;

05

,1

8 2;

2;

04

,1

0

6

,3

2;

03

,1

6 ,0

03 2;

03 2;

02

,2

2

0

2;

Number of utterances with OAV or AOV sequences

Figure 7.Pascal: German Verb–Adverb–Object Structures.

Age

Figure 8.Annika: German Verb–Adverb–Object Structures.

These figures indeed confirm that, as soon as verbs, adverbs and object-DPs are combined, OAdvV and AdvOV sequences emerge simultaneously in the data of both children. The first OAdvV strings (Pascal: 2;01,28, Annika: 2;02,22) appear even before the first AdvOV sequences (Pascal: 2;02;25; Annika: 2;03,06). Since OAdvV order is assumed to be linked to the VO/OV parameter, this result indicates that the parameter is set to the correct value in German at 2;01 (Pascal) and at 2;02 (Annika).

315

8 6 VOA

4

VAO

2

01 ; 2; 00 01 ;2 2; 8 02 ;1 2; 2 05 ; 2; 05 05 ;1 2; 9 06 ; 2; 02 06 ;1 2; 6 08 ; 2; 17 09 ;1 2; 6 10 ;1 2; 3 11 ; 3; 11 01 ;0 3; 0 02 ; 3; 23 04 ;1 4

0 2;

Number of utterances with VOA or VAO sequences

316 Anja Möhring and Jürgen M. Meisel

Age

8 6 VOA

4

VAO

2

6 ,1

,1

6

06 3;

1 3;

05

,2

4 ;2

03 3;

;2

6

02 3;

9 ;2

01 3;

;1

5

00 3;

7

00

;2

3;

;3

1

10 2;

9 2;

09

;1

1 ;0

08 2;

07 2;

06

;0

1

0 2;

Number of utterances with VOA or VAO sequences

Figure 9.Pascal: French Verb–Adverb–Object Structures.

Age

Figure 10.Annika: French Verb–Adverb–Object Structures.

In French, the empirical assessment of our claim is complicated by the fact briefly discussed in Section 2 above, that although French as a VO language does not allow for adverb insertion between the verb and its object, surface patterns with adverbs in this position are in fact possible as a result of verb raising. From this it follows that these patterns, VAdvO and VOAdv, should indeed be used by children acquiring French, but they should not emerge

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism

simultaneously, as is the case in German and Dutch. Instead, VAdvO structures are predicted to appear as a consequence of the acquisition of finiteness, a prerequisite for verb raising. This is indeed the case, as is illustrated by Pascal’s and Annika’s use of these sequences, displayed in Figures 9 and 10. Pascal’s acquisitional pattern indicates that, unlike in his German, French Verb–Object–Adverb sequences occur quite early in his utterances (2;01;00; MLU: 2.14), whereas the first Verb–Adverb–Object patterns are attested more than four months later (2;05;19; MLU: 3.63). Assuming that the productive use of subject clitics indicates the availability of the functional category in which [+finite] is checked, it has been shown in 4.2.1, above, that Pascal acquires finiteness in French at the age of approximately 2;3 (MLU ~ 2.42). It is important to note that before this age and hence before the acquisition of finiteness, we do not find VAdvO sequences in Pascal’s French. Annika does not use Verb–Adverb–Object sequences at all until the age of 2;10,13 (MLU: 3.55). But when she starts combining these elements, the first string exhibits VAdvO order. As was shown in 4.2.1, Annika starts using subject clitics productively at the age of 2;00,17 (MLU ~ 1.48). Evidence for finiteness is hence available long before her first VAdvO sequences. Both findings constitute strong support for our claim that VAdvO order is the result of verb raising and not part of the OV/VO parameter. Consequently, it depends on the OV/VO parameter neither in its grammatical analysis nor with respect to the developmental chronology. These results therefore provide further evidence, albeit indirectly, that the children have set the parameter to the respective values in German and French. 4.2.4 Particle constructions The last phenomenon to be considered concerns the placement of verbal particles. Neeleman and Weerman argue that in OV languages like Dutch, particles are adjoined to the verb. Their position in the clause is immediately to the left of the verb ([V Prt(P) V]). Particles, even if they project, will not block head-government in OV languages.13 In OV languages like German and Dutch, we expect to find only structures with particles left-adjoined to non-finite verbforms; see Table 1. This is indeed what Bennis et al. (1995) found in their investigation of the acquisition of Dutch as a first language. The analysis of the use of German by both French-German children studied here lends further support to this approach. As soon as they begin to produce particle constructions, Annika (2;02) and Pascal (1;11) employ only O-Prt-V constructions. The only error we could find during the entire period under

317

318

Anja Möhring and Jürgen M. Meisel

investigation occurred at a late point of development in the recordings with Pascal (3;04). (17) and (18) offer examples of these cases in the German recordings with both children. In these examples, the particles are in normal print, and participles or infinitives appear in bold letters. (17) a.

tor (= tür) aumachen (= aufmachen) door up open-inf ‘open the door’ b. das abnehmen! it off take-inf ‘take it off’ c. auge(n) zumachen eye(s) prt-close-inf ‘close eyes’ d. (muss) es da reinstecken (must it there in put-inf ‘must put it in there’ e. ich hab’ die aufgemacht I have them prt-open-pt ‘I have opened them’ f. da kann man doch die säge nehmen raus there can you adv the saw take-inf out ‘over there you can take out the saw’

(18) a.

das auch, ankleben it too on stick-inf ‘stick it on, too’ b. und und das [awfaln] (= aufräumen) and and it [clear-inf up ‘and clear it up’ c. das abmachen it off take ‘take it off’ d. zumachen ohren prt-close ears ‘close your ears’ e. ohren zumachen ears prt-close ‘close your ears’

Pascal 1;11,28

Pascal 2;01,28

Pascal 2;10,13

Pascal 3;00,17

Pascal 3;04,14

Pascal 3;04,14

Annika 2;02,22

Annika 2;03,30

Annika 2;06,18

Annika 2;09,31

Annika 2;09,31

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 319

The fact that both children produce O-Prt-V constructions from early on indicates that they have set the OV/VO parameter to the respective value, Pascal already at 1;11, Annika at 2;02. 4.3 Summary of the analysis of bilingual language acquisition The analysis of the data of two bilingual French-German children revealed that both children set the OV/VO parameter at a very early point in development to the respective value in each of the languages. Although Annika’s use of both languages, French and German exhibits more variability than that of Pascal, our conclusion is strongly supported by the developmental pattern of the structurally linked phenomena found in both children. Although we observed some unusual sequences in the data of both languages, contrary to what we expected, we assume that these patterns do not present evidence for an incorrect setting of the OV/VO parameter. This claim is justified by the following: (A) These ‘deviant’ strings occur less frequently than the target-like ones. (B) Sequences of this type are also found in the data of monolingual children (cf. Ferdinand 1997). The difference between monolingual and bilingual children is only a quantitative and not a qualitative one. (C) Incorrect setting of the OV/VO parameter implies that all clustering phenomena should not be used in the target-conform fashion. But although we found some nonstandard strings representing some of the phenomena in question, no developmental phase exists during which all clustered phenomena are simultaneously used in a not-target like way. The analysis of our production data suggests that the OV/VO parameter is already fixed to the respective value at approximately age 1;09,11 in both languages for Pascal (German: MLU ~ 1.07; French: MLU ~ 1.06) when first OV sequences in German and VO sequences in French with non-finite verbs occur. In Annika’s data, evidence for the target-like parameter setting is found at age 2;01,24 (MLU ~ 1.31) in German and at age 2;02,22 (MLU ~ 1.33) in French. One may indeed surmise that the OV/VO parameter is really fixed even earlier but that the children are not yet able to use the longer and possibly more complex sequences connected to the parameter. This claim is confirmed by findings from comprehension studies indicating that this parameter is set at an even younger age; see e.g. Guasti et al. (2001). At any rate, we believe we have presented convincing evidence in favor of the hypothesis given in (8). Most importantly, our analysis revealed that the cluster of phenomena related to the OV/VO parameter (Verb–Object order in main and embedded clauses, adverb placement and particle constructions) are all used by the children as soon as the

320 Anja Möhring and Jürgen M. Meisel

required contexts appear in the data, and that they are generally used in the fashion predicted by the syntactic analysis of the parameter.

5. The OV/VO parameter in the acquisition of German by adult second language learners Let us now turn to the analysis of second language acquisition in order to examine whether the hypothesis presented in (9) above can also be confirmed. 5.1 The learners The following discussion is based on the data of two L2 learners of German, Zita and Ana. The recordings are part of the ZISA corpus.14 Zita’s first language is Portuguese. She was first recorded during her 14th week in Germany, and the recordings continued until her 132nd week in the country. Ana’s native language is Spanish. Her recordings last from her 16th through her 65th week in Germany. Subsequently, she was recorded several times in Spain, where she spent several weeks, and again twice after she had returned to Germany. For more information about the learners and the empirical methods employed, see Meisel, Clahsen and Pienemann (1981) and Meisel (1991). 5.2 The acquisition of German word order phenomena Both Portuguese and Spanish are VO languages, just like other Romance languages, including French, studied in the preceding section. Portuguese and Spanish should therefore exhibit the same properties as French with regard to the OV/VO parameter. There exists a large body of studies dedicated to the acquisition of word order phenomena by L2 learners, many of them investigating OV languages like German and Dutch and VO languages like English and the Romance languages. Several of these have argued that the parameter which is here referred to as the OV/VO parameter cannot be “reset” in adult second language acquisition, i.e. it cannot be fixed to a value different from that of the L1 of the learners; see, for example, Meisel, Clahsen and Pienemann (1981), Clahsen and Muysken (1986), and Neeleman and Weerman (1997). The results of the following analysis support these findings and suggest that L2 learners need to learn the cluster of

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism

phenomena listed in Table 1 separately, as is predicted by the hypotheses presented in (1) and (9). 5.2.1 Verb–Object order in main clauses Earlier work of the research team ZISA, e.g. Meisel, Clahsen and Pienemann (1981), Clahsen, Meisel and Pienemann (1983), has shown that Romancespeaking learners of German initially produce only VO sequences. (19) Ich studieren in Porto I to study in Porto ‘I study in Porto’ (Example from Clahsen and Muysken, 1986)

Number of utterances with VO-structure

At this stage, the learners show no evidence of verb-final placement or V2 placement of finite verbs. During a second phase, they begin to place verbs at the end of the main clauses, and subsequently, they begin to place the finite verb in second position in main clauses. Finally, finite verbs begin to appear in final position in embedded clauses. Figures 11–14 illustrate Zita’s and Ana’s use of Verb–Object sequences.

25 20 15

vO VO

10 5 0 14

18

27

32

36

44

50

58

72

89 100 107 113

Length of stay in weeks

Figure 11.Zita: German VO structures in main clauses.

321

Number of utterances with OV structure

322 Anja Möhring and Jürgen M. Meisel

25 20 15

Ov OV

10 5 0 14

18

27

32

36

44

50

58

72

89 100 107 113

Length of stay in week

Number of utterances with VO-structure

Figure 12.Zita: German OV structures in main clauses. 25 20 15

vO VO

10 5 0 16 19 22 24 26 34 36 39 50 53 59 61 65 70 72 74 78 Length of stay in weeks

Figure 13.Ana: German VO structures in main clauses.

Number of utterances with OV-structure

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 323

25 20 15

Ov OV

10 5 0 16 19 22 24 26 34 36 39 50 53 59 61 65 70 72 74 78 Length of stay in weeks

Figure 14.Ana: German OV structures in main clauses.

These figures show that both Zita and Ana produce only V-O sequences during an initial phase of acquisition, regardless of whether the verb-form is finite or non-finite. In fact, this phase extends over a period of more than one year, and only during the 67th week (Zita) and 50th week (Ana) of residence, respectively, do they begin to use OV sequences, mostly with infinitives (cf. Clahsen and Muysken, 1986; Meisel, Clahsen and Pienemann, 1981). Whereas with the bilingual children one finds only a short period during which they use both finite and nonfinite strings (in which infinitives are combined rarely with subjects; cf. Ferdinand 1996 for French) Zita and Ana use infinitives during the entire period of investigation, and they frequently combine them with subjects. We found two examples with vVO structure in Ana’s early utterances, given in (20). (20) a.

eh der junge eh wollen eh helfen  the boy eh want-inf eh help ‘the boy wants to help the grand-mother’ die die die großmutter the grand-mother b. großmutter großmutter wollen grand-mother grand-mother want-inf ‘grand-mother wants to take the bus’ nehmen die bus take the bus

Later on, she also starts producing vOV sequences.

Ana 19th week

Ana 19th week

324 Anja Möhring and Jürgen M. Meisel

(21) [b]erkstatt (= werkstatt) eh hat mein freund das auto [garage eh has my friend the car abgeschleppt prt-towed away ‘my friend towed the car away to a garage’

Ana 59th week

Zita starts using complex verbal sequences only during a later phase of development. Both, vOV and vVO sequences appear in her data. All in all, we counted 53 utterances with VO (75% of all utterances with non-finite verb forms) and only 18 utterances (25%) with OV order in Zita’s data. Similarly in Ana’s data where we found 44 VO sequences (86%) and only 7 with OV (14%). Remember that the distribution of word order patterns in the utterances of the children revealed a rather different picture since they predominantly place verbs in final position. In other words, our analysis shows that the L2 utterances do not exhibit an acquisition pattern for OV, vO, and vOV sequences resembling that of the German 2L1 data. We therefore conclude that the children fix the OV/VO parameter at an early stage of acquisition, whereas the L2 learners seem to learn the placement of verbs and objects only gradually. 5.2.2 Verb–Object order in embedded clauses Müller (1998) studied the acquisition of word order regularities in embedded clauses, analyzing the language use of another adult L2 learner of German from the ZISA corpus, Bruno whose L1 is Italian, also a VO language. She finds that, during an extended period of time, Bruno only produces SVO order in embedded clauses. The first OV sequences occur only during the 60th week of residence. Müller argues that Bruno learns the word order of German embedded clauses in an item-by-item fashion, depending on the complementizer introducing the embedded clause. The examples in 22 and 23 illustrate that Zita and Ana, too, prefer Verb– Object placement in German embedded clauses during an extended period of time. The examples with non-finite verb forms appear in bold letters. wenn frowein sprechen eh spanisch when miss speak-inf eh Spanish ‘when the young lady speaks Spanish’ b. aber wenn de arbeitsamt nich eh sagen ja wenn but when the job center not eh say-inf yes when die sage “nein” they say “no” ‘but when the job center says ‘no’’

(22) a.

Zita 58th week

Zita 97th week

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 325

c.

wenn die sage “nein” Zita 107th week ‘when they say “no”’ d. ich habe geld von bezahl die zimmer wenn Zita 111st week I have money from pay the flat when ich habe arbeit I have a job ‘If I had a job, I would have (some) money to pay the flat’ e. und jetzt darf ich nicht mehr kommen ne ja Zita 113rd week and now must I not still come-inf ne yes die die — die hier die die junge sag auch ich ich the the  the here the the boy say too I I kenne die mann aber ich sag know the man but I say ‘and now I’m still not allowed to come — the boy here, too, I say too’ f. wenn du eine mußten so machen portugiesisch Zita 132nd week when you a must so make-inf Portuguese ‘if you had to do this in Portuguese’

In fact, Zita probably produces the first OV pattern only during the very last recording, in the 132nd week of her stay in Germany (22f.). But in this example, it is difficult to decide which element functions as object (‘eine’ or ‘portugiesisch’), i.e. even here it is not obvious whether she has indeed placed the verb in final position of a subordinate clauses. With respect to VO sequences with nonfinite verb forms, we find only VO patterns. As Zita starts using OV order in main clauses in the 41st week of residence, this finding cannot be taken as evidence for parameter-fixing. (23) a.

wann ich liebe ein person ein mann und diese Ana 34th week when I love a person a man and this mann liebt mir man loves me-dat ‘when I love a man and this man loves me’ b. qui haben nicht geld Ana 50th week qui have not money ‘who doesn’t have money’ c. weil eh die leute — er er brauch braucht Ana 53th week because eh the people  he he needs needs ein minimo ein minimo die leute in die kurs ne a minimo a minimo the people in the class ne ‘because the people need a minimum amount in the class’

326 Anja Möhring and Jürgen M. Meisel

d. wenn geld ich habe Ana 59th week when money I have ‘if I had money’ e. weil ich in ein andere schule unterricht hat Ana 65th week because I in another school lessons have-3ps.sg ‘because I take lessons in another school’ f. aber wenn ich brauche ein text kommentieren Ana 65th week but when I need a text comment-inf ‘but when I must comment on a text’ g. ja wann ich mache eine verbrechen — Ana 70th week, then yes when I make a crime  in Spain after 2 versprechen ja ich eh ja weeks with promise yes I eh yes German visitor ‘when I promise something’ h. ja wann ich habe ein titel ne Ana 2 weeks in yes when I have a title ne Germany again ‘and when I have a degree’ i. und wann möchten eine andere auto — come se dice Ana 6 weeks in and when want another car Germany again ‘and when you want to have another car’

Ana uses the first OV string in the 59th week of residence, but she continues producing VO sequences until the end of the recordings. With respect to utterances containing non-finite verbs, we observed mainly VO sequences (examples 23 b and i) and only one vOV string in the 65th week of stay. Since she already uses OV patterns in main clauses in the 50th week of residence, this finding cannot be explained as resulting from the setting of the OV/VO parameter. We conclude that this finding, in combination with what has been observed about her use of main clause word order, provides unambiguous evidence against the assumption of parameter (re)setting in this case of adult L2 acquisition. There can be no doubt that word order regularities are learned separately for main and subordinate clauses. 5.2.3 Verb–Adverb–Object constructions In (24) and (25), utterances are listed containing non-finite Verb–Adverb– Object combinations found in Zita’s and Ana’s use of German, indicating that during most of the period under investigation both learners place adverbs variably in different positions.

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 327

(24) a.

nee isch vielleicht mal gucken eins Zita 55th week no I perhaps some day see-inf one ‘no, perhaps I will see one some day’ b. machen die stempel da ne Zita 97th week make-inf the stamp there no ‘stamp it there’ c. ganz bestimm nisch spreschen nur deutsch Zita 100th week certainly not speak-inf only German nur portugies only Portuguese ‘we certainly do not speak only German but Portuguese’ d. immer schokolade mitbringen so Zita 107th week always chocolate along bring-inf so ‘she always brings some chocalate’ e. mein schwester machen ein bisschen an meine Zita 111st week my sister slam-inf a little prt my schwager brother-in-law ‘my sister teases my brother-in-law a little’ f. muss ganz gut portugiese schreiben Zita 132nd week must quite good Portuguese write-inf ‘must write fairly good Portuguese’

(25) a.

hören hören eh gestern die Ana 39th week hear-inf hear-inf eh yesterday the programm eh hollandaise für spanisch channel eh hollandaise for spanish ‘yesterday (we) heard the Dutch channel in Spanish’ b. und eh auch machen ein fête international in Ana 39th week and eh also make-inf a fête international in hollande Holland ‘and we also had an international party in the Netherlands’

Zita and Ana only insert adverbs between non-finite verb forms and objects if the object follows the verb. In other words, OAdvV sequences are not attested in this corpus, only VAdvO and AdvOV sequences. We interpret this fact as evidence suggesting that Zita and Ana have not learned to produce OAdvV sequences during the period under investigation and that the word order patterns listed do consequently not result from a setting of the OV/VO parameter to the OV value.

328 Anja Möhring and Jürgen M. Meisel

5.2.4 Particle constructions Finally, with respect to particles, we have shown that they are consistently adjoined to the left of infinitives in the utterances of the bilingual children. (26) and (27) offer exhaustive lists of the corresponding utterances in the recordings with Zita and Ana; particles appear in normal and non-finite verbs bold letters. die frau machen die tür — tür — tür auf the woman open-inf the door prt ‘the woman opens the door’ b. ausfragen den jungen prt-question the boy ‘question the boy’ c. immer schokolade mitbringen so — always chocolate along bring-inf so  ‘always bring along chocolate’ d. nee ich eh die jack anziehen no I eh the jacket on put ‘I put the jacket on’ e. mein schwester machen ein bisschen an meine my sister slam-inf a little prt my schwager brother-in-law ‘my sister teases my brother-in-law a little’

(26) a.

(27) [b]erkstatt (= werkstatt) eh hat mein freund das auto garage eh has my friend the car ‘my friend towed the car away to a garage’ abgeschleppt prt-towed away

Zita 30th week

Zita 103rd week

Zita 107th week

Zita 107th week

Zita 111st week

Ana 59th week

In two of the five examples in (26), the particle is not attached to the left of the infinitival verb. Although this data base is too small to justify broad generalization, it may be possible to argue that these examples represent instances where a nonfinite verb form is raised. In this case, one could assume that Zita knows that particles are left-adjoined to infinitives in German. Ana produces only one utterances in which she combines a particle construction with an object. The particle is placed to the left of the infinitival verb. This suggests that Ana too has already learned the target-like placement of particles in German.

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 329

5.3 Summary of the analysis of second language acquisition This brief analysis of Zita’s and Ana’s acquisition of the word order phenomena dependent on the OV/VO parameter has shown very clearly that they use some of them in a more target-like way and that they deviate considerably from the target norm when using others. Importantly, one observes rather inconsistent use of each of these patterns. For example, when Zita and Ana start using OV and vOV sequences in main clauses, they continue to use VO and vVO order in embedded clauses. We conclude that Zita and Ana, unlike the bilingual children, learn each of the phenomena associated with the OV/VO parameter separately, probably beginning with particle constructions and main clause word order (although this continues to be used inconsistently during the entire period of observation); target-like word order in embedded clauses and adverb placement, on the other hand, do not seem to be acquired at all, during the period investigated. As opposed to the bilingual data, the L2 corpus does not contain OV, vO, vOV patterns. We interpret these facts as evidence supporting the hypotheses given in (9); The L2 learners do not set the OV/VO parameter to the value required by the grammar of German; instead, they learn some of the phenomena linked to the parameter individually.

6. Conclusions The goal of this paper has been to test the validity of three hypotheses, (1), (8), and (9), according to which the developmental pattern formed by the emergence of grammatical phenomena related to a syntactic parameter is significantly different in simultaneous as opposed to successive acquisition of bilingualism. Whereas in monolingual as well as in bilingual first language development the cluster of structures dependent on the parameter appear in a parallel fashion within one developmental phase, these constructions are not developmentally connected in L2 acquisition. We believe to have shown that these claims are indeed corroborated by our analysis of the word order patterns related to the OV/VO parameter, as postulated by grammatical theory. In 2L1 acquisition, they appear in the children’s speech as soon as the parameter has been set to a specific value. From very early on, basic word order in main clauses, Adverb– Object–Verb sequences, and particle constructions are used mainly as required by the target grammars. Embedded clauses appear only later, but word order in this sentence type is then immediately used in a target-like fashion.



330 Anja Möhring and Jürgen M. Meisel

In the acquisition of German as an L2, on the other hand, serious problems have been detected with respect to the use of these constructions. Whereas particle constructions seem to be acquired relatively early during the entire period investigated, target word order in main clauses is used inconsistently, sometimes in a target-like fashion, at other occasions in a target-deviant way. Other sequences, e.g. Verb–Object order in embedded clauses, are not acquired at all, during this period. These findings demonstrate that bilingual children, just like their monolingual counterparts, acquire the phenomena related to the VO/OV parameter simultaneously, thus indicating that they have set the parameter to the correct value. Adult L2 learners, on the other hand, do not exhibit the predicted clustering effect in their use of the corresponding sequences. We therefore conclude that the acquisition of the target grammar by L2 learners does not involve setting parameters to the required L2 values. Instead, these learners have to learn inductively the respective sequences which therefore emerge at different points of time during the course of L2 acquisition. Consequently, we claim that the observed differences between various acquisition types are not due to the fact that some of these learners acquire a second language, since bilingual children behave like monolingual L1 children. We hypothesize, instead, that such differences are to be explained as a result of age of onset of acquisition, i.e. due to maturational changes, adult L2 learners no longer have direct access to UG and are therefore not able to set parameters on values not instantiated in their L1.

Notes * This study has been carried out as part of the research project “Simultaneous and successive acquisition of bilingualism”, directed by Jürgen M. Meisel, funded by the Deutsche Forschungsgemeinschaft (German Science Foundation) within the Collaborative Research Center on Multilingualism, established at the University of Hamburg. We gratefully acknowledge this support. Furthermore, we would like to thank the following friends and colleagues for useful comments and suggestions on earlier drafts of this article: Susanne E. Carroll, Gisella Ferraresi, Maria Goldbach, Marc-Olivier Hinzelin, Regina Köppe, Esther Rinke, Tessa Say, and Fred Weerman. 1. Children younger than four years, on the other hand, should exhibit developmental patterns which are similar to those of bilingual children, see e.g. Möhring (2001) for a discussion of this question. 2. We are thus following Neeleman and Weerman (1999), as opposed to Kayne (1994) who proposes a universal spec-head-complement order.

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism

3. But note that Neeleman and Weerman (1999) do not depart completely from standard minimalist assumptions. They differ from Chomsky with regard to several points, e.g., in their assumption that A-movement is not possible. 4. This entails that the four logical possibilities are reduced to two possibilities. 5. Selkirk (1986) defines the formation of a phonological phrase Φ as in (i): (i) Close Φ when encountering ]XP. 6. Morphological case in German is discussed in Neeleman and Weerman (1999: 91–94). 7. In contrast to lexical verbs which remain in their base position, have, be and do are raised into the functional domain, see e.g. Emonds (1976). 8. Ø-DP stands for a DP preceded by an empty case shell. 9. Adverb insertion is not possible with all adverbs in French. See, for example, Cinque (1999) for an analysis of the distribution of French adverbs. 10. See Hinzelin and Möhring (2001) for a first analysis of Pascal’s acquisition of the OV/VO parameter. 11. This research, funded by the DFG (Deutsche Forschungsgemeinschaft) from 1986 through 1992, was carried out at the University of Hamburg under the direction of Jürgen M. Meisel. 12. Small letters indicate finite elements, capital letters in italics stand for non-finite elements and the remaining instances of ‘V’ refer to verbal elements in general, i.e. when finiteness is not relevant for the discussion. Verb forms which could not be analyzed unambiguously as either finite verb forms or as participles or infinitives were not counted. 13. In English, on the other hand, projecting particles must be separated from the verb. As PrPs between objects and verbs would block head government, VP shell forming must apply. This is illustrated in example (ii) and (iii): (ii) John looks the information up. (iii)

V¢ V¢

VP DP

V tV

PrtP

(example from Neeleman and Weerman 1999: 34) 14. Zweitspracherwerb italienischer (portugiesischer) und spanischer Arbeiter (1977–1982). The project, under the direction of J. M. Meisel, was funded by the Ministerium für Wissenschaft und Forschung des Landes Nordrhein-Westfalen (1977–1978) and by the Stiftung Volkswagenwerk (1978–1982).

331

332 Anja Möhring and Jürgen M. Meisel

References Bennis, H., den Dikken, M., Jordens, P. and Weissenborn, J. 1995. “Picking up particles”. In Proceedings of the 19th Annual Boston Conference on Language Development, D. MacLaughlin and S. McEwen (eds), pp. 70–81. Somerville, MA: Cascadilla Press. Bley-Vroman, R. 1989. “What is the logical problem of foreign language learning?”. In Linguistic Perspectives on Second Language Acquisition, S. M. Gass and J. Schachter (eds), Cambridge: CUP. Chomsky, N. 1980. Rules and Representations, 41–68. New York: Columbia University Press. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1995. The Minimalist Program. Cambridge: MIT Press. Cinque, G. 1999. Adverbs and Functional Heads. A Cross-Linguistic Perspective. New York, Oxford: Oxford University Press. Clahsen, H., Meisel, J. M. and Pienemann, M. 1983. Deutsch als Zweitsprache. Der Spracherwerb ausländischer Arbeiter. Tübingen: Narr. Clahsen, H. and 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: 93–119. Emonds, J. 1976. A Transformational Approach to English Syntax. New York: Academic Press. Ferdinand, A. 1996. The Development of Functional Categories. The Acquisition of the Subject in French. The Hague: Holland Academic Graphics. Ferdinand, A. 1997. “The development of phrase structure in child French”. Linguistics in the Netherlands: 85–96. Greenberg, J. H. 1963. “Some universals of language with specific reference to the order of meaningful elements”. In Universals of Language, J. H. Greenberg (ed.), 73–113. Cambridge, Mass.: MIT Press. Guasti, M. T., Nespor, M., Christophe, A. and Ooyen, B. van 2001. “Pre-lexical setting of the head complement parameter through prosody”. In Approaches to Bootstrapping: Phonological, Syntactic and Neurophysiological Aspects of Early Language Acquisition, Vol. I, B. Höhle and J. Weissenborn (eds), 229–248. Amsterdam/Philadelphia: John Benjamins. Haider, H. 2000. “Scrambling — what’s the state of the art?”. In The Acquisition of Scrambling and Cliticization, S. M. Powers and C. Hamann (eds), 19–40. Dordrecht: Kluwer. Hawkins, R. and Chan, C. Y. 1997. “The partial availability of Universal Grammar in second language acquisition: The ‘failed functional features hypothesis’”. Second Language Research 13: 187–226. Hinzelin, M.-O. and Möhring A. 2001. “The acquisition of subjects and verb–object order in bilingual children”. In Proceedings of the Colloquium on Structure, Acquisition, and Change of Grammars: Phonological and Syntactic Aspects. Volume II [University of Hamburg, 27th — 29th October 2000. Working Papers in Multilingualism Series B], K. F. Cantone and M.-O. Hinzelin (eds), 1–25. Hamburg: Sonderforschungsbereich 538/Universität Hamburg. Hulk, A. and van der Linden, E. 1996. “Language mixing in a French-Dutch bilingual child”. Toegepaste Taalwetenschap in Artikelen 55 (selected proceedings of EUROSLA 6): 89–101.

The Verb–Object parameter in simultaneous and successive acquisition of bilingualism 333

Hyltenstam, K. and Abrahamsson, N. (forthcoming). “Maturational constraints in second language acquisition”. In Handbook of Second Language Acquisition, C. Doughty and M. Long (eds). Oxford: Blackwell. Kaiser, G. 1994. “More about Infl-ection and agreement: The acquisition of clitic pronouns in French”. In Bilingual First Language Acquisition. French and German Grammatical Development, J. M. Meisel (ed.), 131–159. Amsterdam/Philadelphia: John Benjamins. Kayne, R. 1983. “Connectedness”. Linguistic Inquiry 14: 223–249. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge: MIT Press. Koopman, H. 1984. The Syntax of Verbs: From Verb Movement Rules in the Kru Languages to Universal Grammar. Dordrecht: Foris. Köppe, R. 1994. “The DUFDE Project”. In Bilingual First Language Acquisition. French and German Grammatical Development, J. M. Meisel (ed.), 15–27. Amsterdam/Philadelphia: John Benjamins. Meisel, J. M. 1991. “Principles of Universal Grammar and strategies of language use: On some similarities and differences between first and second language acquisition”. In Point-Counterpoint: Universal Grammar in the Second Language, L. Eubank (ed.), 231–276. Amsterdam/Philadelphia: 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/Philadelphia: John Benjamins. Meisel, J. M. 1997a “The acquisition of the syntax of negation in French and German: Contrasting first and second language development”. Second Language Research 13(3): 227–263. Meisel, J. M. 1997b. “The L2 Basic Variety: A discussion of Klein and Perdue”. Second Language Research 13(4): 374–385. Meisel, J. M. 2000. “Revisiting Universal Grammar”. D. E. L. T. A. (Revista de Documentaçâo de Estudos em Lingüística Teórica e Aplicada) 16 [Especial: Homenagem a Mary Kato]: 129–140. Meisel, J. M. 2001. Successive acquisition of languages: on how it differs from the siumltaneous acquisition of bilingualism — and why. Talk given at the Romanistentag in Munich, October 2001. Meisel, J. M. Clahsen, H. and Pienemann, M. 1981. “On determining developmental stages in natural second language acquisition”. Studies in Second Language Acquisition 3 (2): 109–135. Möhring, A. 2001. “The acquisition of French by German children of pre-school age: An empirical investigation of gender assignment and gender agreement”. In Eurosla Yearbook, S. Foster-Cohen and A. Nizegorodcew (eds). 171–193. Amsterdam/ Philadelphia: John Benjamins. Müller, N. 1998. “Die Abfolge OV/VO und Nebensätze im Zweit- und Erstspracherwerb”. In Eine zweite Sprache lernen. Referate anläßlich des Kolloqiums zum DaZ-Erwerb in Potsdam 1996, Wegener, H. (ed.), 92–122. Tübingen: Narr. Neeleman, A. and Weerman, F. 1997. “L1 and L2 word order acquisition”. Language Acquisition 6 (2): 125–170. Neeleman, A. and Weerman, F. 1999. Flexible Syntax. A Theory of Case and Arguments. Dordrecht: Kluwer.



334 Anja Möhring and Jürgen M. Meisel

Pollock, J. — Y. 1989. “Verb movement, universal grammar, and the structure of IP”. Linguistic Inquiry 20 (3): 365–424. Rizzi, L. 1990. Relativized Minimality. Cambridge: MIT Press. Ronjat, J. 1913. Le développement du langage observé chez un enfant bilingue. Paris: Champion. Schaeffer, J. 2000. The Acquisition of Direct Object Scrambling and Clitic Placement. Amsterdam/Philadelphia: John Benjamins. Selkirk, E. 1986. “On derived domains in sentence phonology”. Phonology Yearbook 3: 371–405. Slobin, D. I. 1985. “Crosslinguistic evidence for the Language-Making Capacity”. In The Crosslinguistic Study of Language Acquisition. Vol 2: Theoretical Issues, D. I. Slobin (ed.), 1157–1256. Hillsdale, NJ: Lawrence Erlbaum. Smith, N. and Tsimpli, I.-M. 1995. The Mind of a Savant: Language Learning and Modularity. Oxford: Blackwell. Stowell, T. 1981. Origins of Phrase Structure. PhD Dissertation, MIT. Towell, R. and Hawkins, R. 1994. Approaches to Second Language Acquisition. Clevedon: Multilingual Matters LTD. Travis, L. 1984. Parameters and Effects of Word Order Variation. PhD Dissertation, MIT.



Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP?* Tom Roeper University of Massachusetts

1.

Introduction

Linguistic Theory always contained a promise: the explanation of language acquisition. This promise has two natural dimensions: one profound dimension is the providing of insight into how children choose one grammar from among the set of possible grammars. Another dimension is equally important: the fundamental operations of grammar should be transparent and therefore find unique support in acquisition.1 A typical case is the concept of transformation: the perennial appearance of copying (Can I can come) supports the view that a transformation should be formulated as: copy then delete. This insight is maintained under the formulation of movement as Agree, Mark, Delete recently advocated by Chomsky (2000). In general, where adult grammars provide a conglomerate output, acquisition predicts that microscopic operations could appear as independent steps. In what follows we will develop an acquisition-oriented perspective on the question of the Universal Base Hypothesis and on the constraints governing Feature-checking. It then entails: A. an additional metric of economy, and B. an argument in behalf of a single Tns/Agr node. Finally we argue that Default choices in early child grammars support a theory of Multiple Grammars, where each module involves a different economy metric.2

336 Tom Roeper

1.1 Universal base and feature-checking acquisition Two questions lie at the heart of much recent theory: (1) a.

How much of the base structure is Universal and how much variation is there? b. How does Feature-checking work and how does it accommodate the fact that whole phrases and not just Features undergo movement? (Pied-piping).

At one extreme is the view that there is a fixed universal base: all nodes are the same in all grammars. At the other extreme is the view that a large variety3 of nodes are possible. Children might then have radically underspecified nodes in the process of determining the Feature content of nodes in their grammar.4 1.2 Separate nodes or a single node Two claims about the fundamental nature of tree-structure have recently been articulated. Chomsky (2000) has argued that there is only one TP node and Agreement features are located within it. Chomsky motivates this claim on abstract grounds: all nodes should represent interpretable features. If Agr is an uninterpretable feature, then it should not be an independent node. Guasti and Rizzi (2002) have argued for the Split-IP perspective which claims that Agr and Tense are separate nodes. Let us take a close look at the mechanisms involved. Under Chomsky (2000) the lexical item Probes for the features which will satisfy it in the Tense Phrase (see also Bobalijk 1995). The -s affix on a word like play carries both the Agr and Tense features. It moves from the VP up to the TP which has empty Feature positions that must be assigned a value and then delete:5

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 337

(2)

TP T [+Tns,+Agr] Goal

VP V V verb

aYx

play

-S[+Tns,+Agr] Probe

Guasti and Rizzi (2002) propose an alternative in which each feature is captured by a separate node and the derivation checks them off independently. (3)

AgrP A [+Agr] Goal

TP T [+Tns,] Goal

VP V

verb

(DP) aYx

-s[+Tns,+Agr] Probe

As we shall show, there is empirical evidence on behalf of both of these perspectives. It is a sign that theories are appropriate when subtle data is relevant to subtly different formulations. In order to make the argument, however, we must first refine the concept of Feature-checking itself.

338 Tom Roeper

2. Seek sublabel/path economy One of the surprising mysteries of grammar is the existence of Pied-Piping, the fact that the grammar machine may move more than is initially needed: (4) a. a picture of whom did he see b. who did he see a picture of

Why does the whole phrase a picture of whom move to the front when, as (4b) shows, it would be enough to move just who? We will suggest an answer below, following the analysis of Hollebrandse and Roeper (1996) but since both are grammatical, we must allow Feature-checking to operate on both of them. Nevertheless most speakers prefer (4b) and that preference deserves some attention. Note that, in principle, the same distinction, less contrastive intuitively, is found in cases like: (4) c. who did you talk to d. to whom did you talk.

How does Feature-checking find the crucial, but buried [+wh] in cases like (4a,
d)? To find the Feature, it is is necessary to pass down through the moved DP structure in cases like (4a) in order to find the [+wh] in the object of the PP (picture of whom). Checking must move through DP, D, NP, PP, NP, N to find whom. Chomsky (1995) labels this process Seek Sublabel, but he does not discuss the computational character of the operation. Seek Sublabel: “we have to have an elementary way to determine the features of alpha and K that enter into this checking relation no matter how deeply embedded these are in alpha and K” “F enters into a checking relation with a sublabel of K as a result of the operation (move)” (Chomsky 1995: 268)

Thus CP [+wh] must connect with whom [+wh] within the Pied-piped phrase:

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 339

(5) a.

CP [+wh] Spec

C

DP-x

did D

A

TP You

NP N picture

T trace

PP P

VP see

NP

DP Trace-x

N whom [+wh]

The alternate form requires Minimal Seek Sublabel because [+wh] is immediately dominated by SpecCP. In addition, the minimal tree allows the [+wh] feature to immediately C-command the verbal chain [C, TP, VP, t]: (4) b. who did you see a picture of fi Short Seek Sublabel, Immediate Feature Command

340 Tom Roeper

(6)

CP [+wh] Spec who-x

C did

TP you

trace

T VP

see

DP D a

NP N picture

PP P

NP

of

N trace-x

We argue that grammar should capture the notion of Seek Sublabel as a part of the economy system: (7) Economy prefers/requires minimal depth for Seek Sublabel

The consequence is that we can explain the intuition of most people that a picture of whom did you see feels more complex. Failure to satisfy a buried wh-feature can also produce a grammaticality difference.6 Pied-piping is excluded in the satisfaction of a [+wh] feature in this structure (8b): (8) a. I wonder who you have a picture of b. *I wonder a picture of whom you have (p.c. Kyle Johnson)

Clearly the [+wh] feature must be immediately satisfied, which entails rejection of any Feature depth beyond immediate domination. Now a new question arises: how can we capture the sharp grammaticality distinction in (8) while allowing a weaker “preference” in the contrast of (4a,
b)?

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 341

The notion of Seek Sublabel has become the subject of recent work on superiority in wh-movement, offering a variety of new formulations (Fitzpatrick 2002). Fitzpatrick outlines several approaches, none of which captures all of the data, mirroring the contrast between (4) and (8). The core contrast lies between Attract and Path measurement style constraints: Attract Constituent under C-command: Move closest constituent with relevant feature directly to Feature-checking domain. Path Quantification: Choose the shortest path in the tree to satisfy a feature. The technical statements from Fitzpatrick are: Attract Closest C-Command: G is the closest category in the sister of H iff there is no distinct category K such that K c-commands G and K bears a feature matching F.7 Path Quantification: G is the closest category in the sister of H iff there is no category K bearing a feature matching F such that the path P’ that would result from movement of K is smaller than the path P that would result from movement of G. It seems that (8a) is naturally captured by the spirit of Attract, while (4b) (and the superiority facts below) can be explained by invoking the spirit of Path. Moving a large constituent (a picture of whom) makes a short path, but fails to Attract the minimal category (whom). We will not develop technical definitions because our goal is primarily to broaden the relevant empirical domain. What difference is there between these constructions (4) and (8)? Note that one derives from lexical representation and how subcategorization works: wonder seeks visible immediate satisfaction of subcategorized information. Questions belong to a different domain: Illocutionary Force which is projected onto the CP, and always requires movement. This provides our first clue that locality may refract differently under the prism of different modules of grammar. For instance, a Path-theoretic account would explain why (9a) is superior to (9b): (9) a. who bought what b. *what did who buy

342 Tom Roeper

Movement of what out a VP creates a larger distance than moving the subject who in (9a). But now if we bury the wh-, (9c) then the path measurement becomes obscure: (9) c.

who did some pictures of whom disturb

movement from the object of disturb is not further than movement from deep within the subject phrase some pictures of whom. However if we take the Attract account, then (9b) has the same value as a phrase with no wh-word, since no buried [wh-] ever enters the computation: (9) d. who did some pictures of Bill disturb

Thus Attract seems to miss a weak sense of decreasing grammaticality and Path seems to be the one to enrich to where it can explain (9c). The fact that the judgments are obscure is one reason why the acquisition facts may provide a more powerful adjudicator among theories. We return to this question when we address Multiple Grammars and module-based notion of economy. First we build up relevant arguments from the verb-movement domain. 2.1 Verbal seek sublabel/path We can see these principles at work where negation occurs: (10)

TP Spec

Tx Tx

you

did

covert

NegP Neg

Vpx

not

V talk

Do-insertion allows the [+past] feature to be in a Feature-command position (sister) with respect to the lower VP.

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 343

(11) a.

T [+past] T [+past] do

VP +ed

However, normal verb-movement creates a small version of the Seek Sublabel/ Path problem. Movement of the verb+ed requires that the Verb move, but now the [+past]-ed affix is buried one node down inside of the V. After movement, the Tense node must look inside the V to find [+past]: (11) b.

T [+past] V V

T [+past]

talk

ed

Therefore, do-insertion offers a more economic form of Feature-Checking and Feature-command, and it fits our notion of Attract directly. One should ask again why the full verb moves at all. We now need an answer to that question.8 2.2 Acquisition evidence for minimal seek sublabel/path Now we can ask: if this theory is true, then is there acquisition evidence that spontaneously honors economy and chooses immediate domination (Attract) over depth (Path)? Hollebrandse and Roeper (1997) provide direct evidence in behalf of Immediate Domination. There are non-emphatic insertions of do in declarative sentences with no adult models. This occurs in English, Dutch, and German (see Schönenberger and Penner 1995): (12) a. b. c. d. e. f. g. h.

I do have juice in my cup I do taste them did wear Bea’s helmet I did paint yellow right here. I did put the brush in. I did paint it what did take this off do it be colored does it be on every day

344 Tom Roeper

i. j. k. l. m. n. o. p. q. r.

do it be colored You don’t be quiet Allison didn’t be mad This didn’t be colored did there be some does it be on every day does the fire be on every day do clowns be a boy or a girl did there be some A doggie did walk with Dorothy and the Doggie did hurt itself

(13) a.

ik doe ook verven ‘I do also paint’ b. ik does grapjes makken ‘I do grapes make’ c. hij doet taperecorder draaien ‘he does taperecorder turn’ d. wat doet ’ie bukken ‘what does he stoop’ CHILDES) wat doe jij zeggen e. ‘what do you say’ dat doe ik spelen ‘that do I play’ (from van Kampen 1996)

The preference for do-insertion over verb-movement suggests that the designation Last Resort is uninformative. One might more accurately say First Resort. See Schütze (2002) and Erb (2001) for a summary of a variety of dialects and contexts where this Default form predictably arises 2.3 Parallel wh-predictions Our initial wh- examples lead to parallel predictions. The child ought to spontaneously prefer Economic Seek Sublabel and Immediate Command. In fact, evidence from naturalistic data and experimentation over the past ten years reveals precisely this sort of phenomenon. Hoekstra, Koster, and Roeper (1992) found both naturalistic and experimental evidence that children will break the left-branch constraint in order to honor immediate Feature-command. Here are a few naturalistic examples:

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 345

(13) a.

how am I big for ‘how big am I’ b. How is it long for ‘how long is it’ c. keek eens hou ik groot ben ‘look how I big am’

In addition, Guasavara and Thornton (2001) found numerous examples of children extracting just a wh-word from a leftbranch in both naturalistic and experimental data: (15) who did you see’s book

The fact that these forms of economy appear to overrule whatever generally enforces the Leftbranch constraint and favors Pied-piping remains an interesting challenge from acquisition evidence for linguistic theory. It suggests that Pied-piping larger chunks invokes the principles that allow Path computation for wh-expressions, but the acquisition mechanism initially prefers Attract. (See Koster (2000) for arguments that Pied-piping is a domain for cross-linguistic parametric variation.)9 Therefore we find that our basic approach receives support from several quarters. 3. Agr and Negation: What forces deletion? A variety of theories have been put forward for the persistent fact that children may delete Agreement or Tense nodes in the course of acquisition (“he play” for “he plays”). One strain argues that children have a weaker Checking requirement. For instance, Schütze and Wexler (1996) argue that children must satisfy either Tense or Agreement. Owen and Benedicto (2001) make a more refined argument on behalf of the view that children may satisfy just one Phifeature. Roeper and Rohrbacher (2000) argue that UG options play a role: children may have a Default grammar in which there is no Feature-checking requirement for Tense/Agreement. In effect, they choose an Asian grammar, which by hypothesis, does not require this node (Speas 1994). That account explains the presence of a number of examples like where go (but never *where goes) in early files and the fact that subject deletion co-occurs with Tense deletion. Schütze (2002) has done a close analysis of the naturalistic data and discovered that much of the deletion is sensitive to negation. He points out that there are numerous instances of missing Agr/Tense with negation, but virtually none without. For instance, statistics on Nina and Sarah show:

346 Tom Roeper

Table 1

Nina Sarah

he do

he don’t

1 34

65 299

Schütze argues that do attaches to stray affixes that require a root in order to appear. Unless the affix is phonologically real, the do does not get inserted. Those instances where the Agr feature is not represented by any phonological information, a null affix, requires no do-support and therefore children will say: (16) he play10

Now, however, we may ask whether these facts will submit to the analysis we have just advanced. In other words, does an economy factor conspire to promote deletion when do attaches to negation? A structural reason for Agr/Tense omission would not make it a random occurrence. In fact, as Guasti and Rizzi argue, Agr/Tns deletion is far from random. Their evidence poses another challenge which we consider below. The notion of cliticization is precisely to create a more complex object with greater internal structure. We argue that however one chooses to represent this fact, the presence of a Negative Clitic forces the Agr affix to a lower position in a tree inside TP: (17)

TP [+Agr] T NEG

T Do

VP

[+Agr]

n’t

[-es [+Agr]]

Precisely as we argued for buried [+past] in walked, the -es is now buried by virtue of the cliticized negative. Therefore if Checking must occur, then it must reach down below the cliticized Neg. Therefore we argue: if Seek Sublabel/Path is indirect, then choose an economical Default, which may avoid the computation altogether, namely Attract. If Attract cannot be satisfied, then deletion of [-es = Agr] is the only option to allow the derivation to succeed.

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 347

The most economical default is precisely the Asian option where there is no Agr to be satisfied. Note that this solution requires that Feature-checking occurs after cliticization. One variant of this approach would be to argue that in fact cliticization should occur after lexical insertion under Distributed Morphology (Halle and Marantz 1993). This analysis, however, depends upon putting the Agr within the TP and not an independent node. That is, if we had two nodes, under Guasti and Rizzi’s model, then it would be: ººAgrP

(18) do

A [+Agr]

TP

-es

T [+Tns,] ø

Neg n’t

VP V

In this system, the -es is not lower and therefore no reason exists for it to undergo deletion. Instead there is a separate raising operation for Neg, Tns, and Agr. Therefore our analysis provides direct acquisition support for Chomsky’s claim — based on interpretability — that TP should subsume an Agr feature rather than projecting an independent node. Putting AGR lower makes the path deeper, which blocks immediate Feature-checking. 3.1 SLI Roeper et al (2000) provides a similar argument from SLI grammars for the treatment of Agr as a Feature of TP rather than a separate node, namely, correlations between verbal (Me don’t know), nominal (me Daddy), and prepositional deletion (Me go beach not far away) varieties of Agreement all of which show a simultaneous deficit in a child who otherwise shows the capacity for wh-movement (That why them put a lot of sand in), quantification and raising over Neg (Me never take a shower), and reflexivization (her standing and

348 Tom Roeper

her see herself). It appears that other modules advance substantially where no Agreement is involved. Table 2do vs. does in negative and interrogative sentences

NEG INT

do

does

144 5

220 279

χ2 = 128.29, p < 0.005

4. The Inversion challenge This analysis must confront the facts uncovered by Guasti and Rizzi (2002). They show that the presence of Agr is far more likely under inversion. Their tabulations are shown in Table 2. This result clearly shows that Agreement is much more likely in the case of inversion. Our analysis has no straightforward way to account for these facts. Their analysis has no straightforward way to account for the impact of negation on the presence of Agr. In order to preserve our analysis, then, another account for why Agr is present in inversion must be found. We make a tentative proposal here, although any analysis that would allow Agr to be buried in Tense would accomplish the same goal. Suppose we simply argue that Subject–Auxiliary inversion is in reality: (19) Subject–Agreement inversion

This would assert that an Agr feature must be involved, but not assert that it must be an independent node. It is a requirement on the Feature content of the moved entity. As such it would require Feature-Checking in the new environment. It would be Agr+yes/no that motivates movement. (20) Subject inverts with Agr+yes/no

If this Feature bundle moved, then it would prefer Minimal Seek Sublabel/ Attract as well. In fact, there is evidence reported by Roeper (1992) and Guasti, Thornton, and Wexler (1995) to this effect. We find for instance:

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 349

(20) a.

do you don’t want to go outside too? (Roeper 1992) b. What did he didn’t want to bring to school (preference found for 10 children (Guasti, Thornton, and Wexler 1995)11

Why does the full negative auxiliary fail to move? Our account suggests that Attract prefers immediate satisfaction to Feature-checking under Seek Sublabel/Path. At an earlier stage, we find auxiliary copying, but it is clearly promoted by the presence of the negative (did I didn’t). If we assume that a yes/no feature exists in the CP, then it is minimally satisfied by Agr: (21)

CP [+yes/no, +AGR] C [+AGR] do do

TP AGR Spec [+AGR] you do

T do

Neg Agr

n’t

The Agr+yes/no Feature bundle can also explain the presence of bare do under inversion. Some Feature beyond Agr must force inversion in the creation of yes/no questions. This approach also can allow the Schütze account to cover cases of bare inverted do: (22) do it be colored

Children readily prepose do to form yes/no questions. We find a number of them in the examples above. If inverted do carries a more complex Feature bundle, then it would be used to support more than Agr. This then would require a revision in Schütze’s account to state that inversion makes a Feature syntactically visible if not phonologically visible, and that movement is sufficient to force do-insertion.12 Inversion in wh-questions reveals the same pattern. One of the earliest observations about wh-questions is that children resist inversion with negation (Klima and Bellugi 1966). This was initially explained in terms of number of transformations. In effect, we translate that account into one of tree depth.

350 Tom Roeper

Table 3.Auxiliairies inverted (for Adam)

Tensed (do, does, was, is, has, had) Modals Negatives

Inverted

Non-inverted

111 5 0

118 32 26

Recently, Rowland and Pine (2000) provide extensive evidence that children prefer to invert precisely those forms that involve Minimal Seek Sublabel (does, did) and also (is, was). The latter have been argued by Lasnik to be directly inserted under Tense, rather than moving from a lower V-position. This leads to precisely the prediction that a grammar that prefers Minimal Seek Sublabel/Attract will be inverted first by children. Questions with Modals and Negatives, each involving Seek Sublabel/Path are clearly avoided initially. They argue that it is because the modal lacks overt Tense. In our system, the raising of the Modal to the Spec of TP will have the impact of lowering the (invisible) Tense feature in the tree, thereby making Minimal Seek Sublabel/Path more difficult. In sum, we have proposed an avenue to meet the challenge posed by Guasti and Rizzi’s analysis. In essence, while their astute observation demonstrates the role of Agreement in inversion, it does not necessarily hinge upon the treestructure that they advocate. 4.1 Tense/Agr Predictions The theory of Schütze and Wexler (1996) seeks to account for the possibility ‘that either Tense or Agreement is deleted. That possibility is compatable with the SeekSublabel/Path Economy approach. However, we are able to make a further prediction: Agreement, if lower in the tree, should delete more readily than Tense when negation occurs. This prediction is difficult to evaluate statistically at this point, but it may be relevant as our data becomes more refined. (See Schütze (2002) whose evidence on this point suggests that Tense deletion is harder to find than Agreement-deletion. However, the argument depends upon situational estimates of where past tense should occur.) We also have a prediction about German which should be possible to evaluate. If it is the case, as argued by Zwart (1993), that Subject–Verb–Object in Dutch is an IP structure, with no V-2, but Object–Verb–Subject is true V-2 in CP,

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP?

then we make a prediction to the effect that Agr should be present more readily when the object is preposed, than when the subject is first. On the other hand, if Agr is equally present in the two cases, then it argues on behalf of the V-2 to CP everywhere view. Future work should reveal the answer.

5. Economy and Multiple Grammars We have utilized an economic notion — Attract — to account for the preference for (24a) and the notion of Seek Sublabel/Path to account for the grammaticality of (24b). (24) a. who did you want a picture of b. a picture of whom did you want

Why should locality not choose one form? We have argued that the Lexical subcategorization properties are in a different module from wh-Movement. Satisfaction of subcategorization can therefore require an additional property equivalent to the old notion of Proper Government (= immediate domination) which presents a different range of factors to be satisfied by movement operations than those found in other forms of Feature-checking linked to Functional Categories. Therefore the competing constraints identified by Fitzpatrick (2002) may be independently needed when independent demands of various modules are honored. The same question can be asked about verb-raising itself: why should some grammars raise the verb overtly and others raise the verb covertly? We will sketch an approach to these questions that exploits the modularity of grammar. We propose that economic factors should be relativized to different modules: (25) Economy should be measured independently in each module (see Roeper 1999)

This is a natural consequence of the fact that the primitives in each module (case, thematic roles, chains) are different. Therefore one concept of simplicity (e.g. economy) cannot fit them uniformly. It is natural therefore that simplicity in one module generates complexity in another. This abstract point deserves some emphasis: We are asserting that it is inherently in the nature of biological objects whose parts contain smaller “organs” that one system of simplicity cannot apply across the whole system. This perspective has non-trivial consequences,

351

352 Tom Roeper

but this is not the place to explore them at the abstract level. Instead we will proceed with this perspective as a guiding concept and see how it leads to a theory of Multiple Grammars. What other module is subject to the impact of syntactic variation.? We argue that semantic economy is affected by syntactic choices, although it may itself contain subdomains (thematic roles, quantification). In a word, we propose that every shift in syntactic structure alters the micro-scopic semantics associated with it. The consequence is that syntactic simplicity can generate semantic complexity. It is important to observe the impact of the term microscopic: microscopic changes are those that have no impact in the bulk of sentences. We provide two examples and leave this notion as a suggestion. 5.1 The semantics of verb-raising13 It is often asserted that overt verb-raising has no semantic consequences. But it can hardly be an accident that in English the verb does not raise and that it has a generic reading that is not anchored in the present for sentences like: (26) John plays tennis

The generic interpretation of (28) has led people to say commonly that “English has no present tense”. What it actually lacks is a temporal-anchor: a situation link that is like a definite article which confers time-specificity, which we will illustrate in a moment. Some subtle (microscopic) examples are needed to grasp exactly what we mean. Moreover, this concept no doubt needs a careful articulation in light of concepts of aspect, sequence of tense, and other semantic factors. However we think an intuitive version provides an adequate link to the syntax. We claim: (27) Overt verb-raising provides a Tense-anchor that does not occur when raising is Covert.

The raised verb in German supposedly captures both the generic reading and the progressive reading. A close look, however, shows that the Present tense is available, but not the progressive. In English one can say: (28) John is killing Bill, but he should stop

but not: (29) *John kills Bill, but he should stop.

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 353

In other words, the present tense carries a [+telic] reading that prevents the added phrase “but he should stop”. The German translation of (31) has present tense force, but it is also ungrammatical: (30) *Er tötet Fritz, aber er soll aufhören

Stage directions in English carry the [+telic] reading: (31) King kills slave π king is killing slave. *king kills slave, but suddenly stops.

The raised verb therefore captures a slightly different kind of present than the be+progressive circumlocution used in English. In effect, then, there is a missing meaning in English: present telic verb. And in German there is no direct progressive in the verb system. The meaning is achieved by a nominalization: (32) Fritz ist am/beim Töten. ‘Fritz is killing.’

We arrive at the familiar observation that different grammars express the same meanings in slightly different ways. However, there is a stronger consequence from this argument: (33) Overt Feature-checking entails meaning shifts not found in Covert Feature-checking.14

Recent work by Sauerland and Elbourne (2002) makes a similar suggestion about covert Subject–Verb Agreement. They point out that (shifting to American English examples, since these are not acceptable in British English), one can say (34a) but not (34b): (34) a. b. c. d.

a group of men are here *there are a group of men here British: the committee are here *there are the committee here

They argue, in effect, that certain forms of Subject–Verb agreement are fixed at PF and therefore unable to undergo Covert movement and retain a lower interpretation. We argue that certain forms of anchoring, still just intuitively identified, require overt movement as well. Since this is not the primary topic of this paper, we leave this as a suggestion which calls for a more precise investigation.

354 Tom Roeper

5.2 Syntactic and semantic complexity Our general argument, at an intuitive level, is that two kinds of simplicity are involved: (35) Verb-raising a. semantic simplicity: raise the verb fi direct present tense telic b. syntactic complexity: Seek Sublabel/Path is indirect because the morphology must be found inside a V-projection.

Alternatively, if there is syntactic simplicity, then there is semantic obscurity: (36) Do-insertion: a. syntactic simplicity: do-insertion, no verb-raising overtly b. semantic complexity: indirect representation of present tense, or telicity involved.

This proposal invites the development of a semantic model in which an economy metric can be stated. So far we have just said that if one must combine two verbs to generate the notion “present”, then there seems to be a kind of semantic complexity present that is not required if one generates the present directly by movement. This is just an example and not a model. Note that it is very rare that the semantic distinctions which arise make a difference. In most situations: (37) John is playing baseball

would not have to have an implied telic result and therefore would mean the same as the raised German form: Er spielt Fussball. It is only in special environments that one can tease out a semantic difference. Nevertheless, it is microscopic differences which provide scientific insights and therefore they can be of great intellectual significance even if they are of very little communicative significance. Ultimately, a single “grammar” may favor either syntactic or semantic simplicity in different modules, or even in different lexical verb-classes. In effect, two kinds of “grammar” are involved and therefore in a sense, every speaker is a speaker of Multiple Grammars, if particular modules make particular choices. (See Roeper (1999) for an argument that the quotation “yes” said Bill involves Germanic V-2.) We put the term grammar in quotation because as our level of magnification increases, the term begins to lose its basic coherence. We might advance to a notion of modular simplicity.

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP?

5.3 Pied-piping and Scope If this approach to syntactic/semantic variation is correct, then we predict that there should be a distinction between: (38) a. a picture of whom do you want to see b. who do you want to see a picture of

We argue that (38a) favors the wide-scope reading of a picture and (38b) favors the narrow scope reading. It is a particular picture that one wants to see. Therefore we now have a reason why strict Attract is not always obeyed for wh-movement. These preferences are like the familiar ones with passive: (39) a. everyone in the room speaks two languages b. two languages are spoken by everyone in the room

In (39b) the wide scope (same two languages) is favored, but reconstruction is possible. Here we need to refine the notion of grammaticality judgement. Our system leads to the possibility that we sense a difference, a preference, for different readings in each case. The wide-scope reading for (38a) is more natural, a specific picture is involved, because it is semantically simple (Surface Structure = LF), while a narrow scope reading of (38b) is natural with a picture inside the VP. Now we can construct a more complex metric: (40) simple syntax + narrow scope fi ideal for (38b) simple syntax + wide scope fi one LF movement, more complex complex syntax + wide scope = (38a) no LF movement complex syntax + narrow scope = reconstruct a picture, more complex

The judgment of grammatical or ungrammatical is too crude to state this range of possibilities, but they are what arises if we imagine that different forms of simplicity in different modules are in conflict. What one can feel is an appropriate sense that the computational systems are distinct here. If one can factor in both systems, then we arrive at the view that the narrow scope reading of (38a) is the most difficult, and perhaps earns a“?” judgement, while the narrow scope reading of (38b) involves both syntactic and semantic simplicity and therefore should be the easiest. The other two possibilities lie in between.15 Other factors, like speech register, could also have an influence on the choice. For instance, to whom did you speak feels more formal than who did you speak to. Chomsky (1987) in fact suggested that pro-drop phenomena may be linked to speechregisters. Our proposal supports that conjecture by suggesting that grammarshift would be a natural signal of speech-register shift.

355



356 Tom Roeper

6. Conclusion We have argued on behalf of: A. the role of acquisition data in identifying and justifying fundamental operations in UG. In addition, we have made a proposal that B. refines the formulation of Feature-checking to include a metric that measures the complexity of Seek Sublabel/Path. We have argued that if economy is interpreted in terms of a variety of modules, C. then there may be no uniform simplicity across a grammar. The preferences found in individual modules are typical for certain grammar types. A V-2 grammar maximizes semantic simplicity, but creates syntactic complexity. A do-insertion grammar maximizes syntactic simplicity, but creates semantic complexity. Particular languages will exhibit different preferences for different modules, different speech registers, and different lexical classes. In that respect every language contains Multiple Grammars (Roeper (1999), Yang (2000).

Notes * Thanks to audiences in Purdue, Michigan State University, Leiden, the Summer school in Düsseldorf and to conversations with Elena Benedicto, Kyle Johnson, Larry Leonard, Alan Munn, Natascha Müller, Uli Sauerland, Christina Schmidt and for comments from two anonymous reviewers and Natascha Müller. I am particularly indebted to stimulating suggestions from Gereon Müller (see Müller (2002.). 1. Work by Lebeaux (2000), Roeper (1996), Powers (1996), Hoekstra and Jordens (1994), Drenhaus (2001), Schoeneberger and Penner (1995) is directly pertinent to these questions, and many other papers are indirectly relevant. 2. See Roeper (1999), Yang (2000), Kroch and Taylor (1997), Müller and Hulk (2001), Tracy (1991), Satterfield (1999). 3. There is a substantial tradition of work in Underspecification, from Roeper (1992, 1996), (especially) Hyams (1996) and recently Schütze (2002). We take the extreme form to be a node characterized by just a few Features, rather than an underspecified label. In principle there might be a difference: a CP node with only certain features, or just arguing that the node has [+tense] [+complement].

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 357

4. In between, lie many tantalizing possibilities. Hollebrandse and Roeper (1997) have argued that the Split-IP hypothesis reflects an acquisition operation. (See also Perez, Munn, and Schmidt (2002) for further arguments about node-splitting in acquisition). Children maximize the Features under a projection and then split them if evidence forces them to. Others have argued that children’s first projections are radically underspecified. Possibly, some nodes are initially underspecified and others are overspecified. The challenge for every technical proposal is to find empirical evidence that chooses between them. 5. Adger (p.c.) argues for a different technology where the affix is not generated in the lower position. This is a possible option which does not change the core argument that the TP/AgrP are not split. 6. Chomsky (2001), see also Roeper et al. (2000) for extension of Probe-Goal to complement including PP’s. 7. As Fitzpatrick (2002) points out, these formulations also reflect the Minimal Link Condition. See also Heck and Müller (2002) for a reformulation in terms of Phrase Impenetrability. It is not clear yet whether these formulations all make the same predictions when do-support is subjected to the same constraints. The claim here is that one should broaden the database before refining the principles. 8. Lasnik (1999) provides further arguments in behalf of the view that Feature-movement should apply. 9. Guasti and Cardinaletti (to appear) point to a variety of differences for Pied-piping with questions and relative clauses (which correlate with resumptives as well). In particular they point to Pied-piping in questions and non-Pied-piping in child relatives. If the conditions on how relatives are adjoined (a kind of weak subcategorization) involve a different kind of Feature-checking locality, as we saw with wonder, then it may not be surprising that sharper locality restrictions are imposed on relative clause wh-feature-checking than on interrogative feature-checking. As they suggest, this is just the kind of variation a Multiple Grammars approach would allow. 10. It is not clear under this explanation why there are 34 instances of he do for Sarah. 11. An anonymous reviewer suggests that n’t cliticization could be a PF operation. A very nice study of partial movement and negation in English (Abdul-Karim 2001) shows that n’t blocks LF movement, but not not. This result entails that cliticization requires a pre-existing Head and must occur in the syntax. 12. Schütze (2002) provides evidence of does and not do in non-inverted environments like Mark does too which is predictable under our Path account, but not by Guasti/Rizzi. In addition, he provides evidence that a subtle form of Mood marking may be involved where do is inserted. 13. See Zuckerman (2001) for an argument that optionality has motivation and for evidence that do-insertion has a meaning effect. 14. If one suggests that Covert-movement is not required, then the claim is simply that Overt-movement is required to anchor certain temporal effects. 15. This approach (see Roeper 1999) is compatible with the proposals in Sauerland and Elbourne (2002) and Fox (1999) where the possibility of movement is linked to whether an interpretive difference arises.

358 Tom Roeper

References Abdul-Karim, L. 2001 Complex Wh-Questions and Universal Grammars: New Evidence from the Acquisition of Negative Barriers PhD Dissertation, University of Massachusetts, Amherst. Bobalijk, J. 1995. Morphosyntax: The syntax of verbal Inflection. Ph D Dissertation, MIT. Chomsky, N. 1987. Knowledge of Language: Its nature, Origin, and Use. New York: Praeger Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass: MIT Press. Chomsky, N. 2001. “Beyond Explanatory Adequacy”. Ms., MIT. Chomsky, N. 2001. “Derivation by Phase”. In Ken Hale. A Life in Language, M. Kenstowicz (ed.), 1–52. Cambridge, MA: MIT Press. Drenhaus, H. 2000. “From head to adjunct: Evidence from the acquisition of German ditransitive verbs.” In Proceedings of the 24th Annual Boston University Conference on Language Development, S. C. Howell, S. A. Fish, and T. Keith Lucas, 276–285. Cascadilla Press. Erb, C. 2001. Finite Auxiliaries in German. PhD Dissertation, University of Tilburg. Fitzpatrick, J. 2002. “On minimalist approaches to the locality of movement” Linguistic Inquiry 33 (3): 443–463. Fox, D. 1999. “Economy and semantic interpretation”. Linguistic Inquiry Monographs 35, MIT Working Papers in Linguistics. Cambridge: MIT Press. Guasti, M. and Cardinaletti, A. (forthcoming). “Relative Clause Formation in Romance Child Production”. Probus. Guasti, M. and Rizzi, L. 2002. “Agreement and tense as distinct syntactic positions: Evidence from Acquisition”. In The Strcuture of DP and IP: The Cartography of Syntactic Structures, Vol. 1, G. Cinque (ed.), Oxford University Press. Guasti, M. T., Thornton, R. and Wexler, K. 1995. “Negation in children’s questions: The case of English”. In Proceedings of the 19th Boston Conference on Language Acquisition and Language Development, D. MacLaughlin and S. MacEwen (eds), 228–239. Somerville, Massachusetts: Cascadilla Press. Gavruseva, E. and Thornton, R. 2001. “Getting it right”. Language Acquisition 9 (3). 229–267. Halle, M. and A. Marantz. 1993. “Distributed Morphology and the Pieces of Inflection”, In View from Building 20, K. Hale and S. Keyser (eds), 111–176. Cambride, Massachusetts: MIT Press. Heck, F. and Müller, G. 2002. “Repair-driven movement and the local optimalization of derivations”. Ms., University of Tübingen. Hoekstra, T. and Jordens, P. 1994. “From Adjunct to Head”. In Language Acquisition Studies in Generative Grammar (= Language Acquisition & Language Disorders 8), T. Hoekstra and B. D. Schwartz, 119–149. Amsterdam: John Benjamins. Hoekstra, T. Koster, H. and Roeper, T. 1992. “Acquisition of the leftbranch condition”. Presentation at the 16th Annual Boston University Conference on Language Development. Boston, Mass. 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.

Multiple grammars, feature-attraction, pied-piping, and the question: Is Agr inside TP? 359

Holledbrandse, B. and T. Roeper. 1996. “The Concept of Do-insertion and the Theory of Infl in Acquisition”. In C. Koster and F.Wynen (eds), Proceedings of the Groningen Assembly on Language Acquisition, 261–272. Den Haag: CIP-Gegevens Konnklijke Bibliotheek. Klima, E. and U. Bellugi. 1966. “Syntactic regularities in the speech of children”. Psycholinguistic Papers, J. Lyons and R. Wales (eds), 183–207. Edinburgh University Press. Koster, J. 2000. “Pied Piping and The Word Orders of English and Dutch”. NELS 30. Seiten. Kroch, A. and A. Taylor. 1997. “Verb Movement in Old and Middle English: Dialect variation and and language contact.” In Parameters of Morphosyntactic Change, A. van Kemenade and N. Vincent (eds), 297–325. Cambridge: Cambridge University Press. Lasnik, H. 1999. “On feature strength: Three minimalist approaches to overt movement”. Linguistic Inquiry 30:197–217. Lebeaux, D. 2000. Language Acquisition and the Form of the Grammar. Amserdam: John Benjamins. Müller, G. 2002. Summer School Handouts, Düsseldorf. Müller, N. and Hulk, A. 2001. Cross-linguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition 4 (1): 1–21. Owen, A. and E. Benedicto. 2001. “Processing effects and use of nondistinct features in SLI: An error analysis of verb morphology”. Ms., Perdue University. Perez, A., A. Munn, and Schmidt, C. 2002. “Acquisition of generics”. Ms., University of Toronto. Powers, S. 1996. The Growth of the Phrase-marker: Evidence from Subjects. PhD Dissertation, University of Maryland. Roeper, T. 1992. “From the initial state to V-2: The architecture of acquisition”. In The Acquisition of Verb Placement: Functional Categories and V2 Phenomena in Language Development, J. M. Meisel (ed.), 333–370. Kluwer: Dordrecht. Roeper, T. 1996. “The role of merger and formal features in language acquisition”. In Generative Perspectives on Language Acquisition. H. Clahsen (ed.), 415–449. Amsterdam: John Benjamins. Roeper, T. 1999. “Universal Bilingualism”. Bilingualism: Language and Cognition 2 (3). 169–186. Roeper, T. and B. Rohrbacher. 2000. “Null subjects in Early Child English and the theory of economy of projection” In The Acquisition of Scrambling and Cliticiation, Studies in Theoretical Psycholinguistics 26, S. M. Powers and C. Hamann, 345–397, Kluwer Academic Publishers. Roeper, T., Ramos, E. Seymour, H., Abdul-Karim, L. 2000. “Language Disorders as a Window on Universal Grammar: An Abstract theory of Agreement for IP, DP and V-PP” Brain and Language 77 (3): 378–397. Rowland, C. F. and Pine, J. M. 2000. “Subject–auxiliary inversion errors and wh-question acquisition: What children do know?” Journal of Child Language 27: 157–181. Satterfield, T. 1999. Bilingual Selection of Syntactic Knowledge: Extending the Principles and Parameters Approach. Dordrecht: Kluwer Sauerland, U. and Elbourne, P. 2002. “Total Reconstruction, PF Movement, and Derivational Order”. Linguistic Inquiry 33 (2): 283–319.



360 Tom Roeper

Schönenberger, M. and Penner, Z. 1995. “Probing Swiss-German clause by means of the placement of verbal expletives: Tun ‘do’ insertion and verb-doubling”. In Topics in Swiss German Syntax, Z. Penner (ed.), 291–330. Bern: Lang. Schütze, K. 2002. “Some do’s and dont’s in child and adult grammar”. Handout, University of Massachusetts. Schütze, K. and K. Wexler. 1996. “Subject case licensing and English root infinitives”. In Proceedings of the 20th Annual Boston University Conference on Language Development, E. Hughes and A. Zukowski (eds), 670–682. Somerville, Massachusetts: Cascadilla Press. Speas, M. 1994. “Null arguments in a theory of economy of projection”. In Functional Projections, E. Benedicto and J. Runner (eds), 179–208. University of Massachusetts, Occasional Papers in Linguistics 17. Tracy, R. 1991. Sprachliche Strukturentwicklung. Tübingen: Narr. Van Kampen, J. 1996. “PF/LF convergence in acquisition.” In Proceedings of NELS 26, K. Kusumoto (ed.), GLSA, University of Massachusetts. Yang, C. 2000. Knowledge and Learning in Natural Language. PhD Dissertation, MIT: Zwart, J.-W. 1993. Dutch Syntax: A Minimalist Approach. PhD Dissertation, University of Groningen. Zuckerman, S. 2001. The Acquisition of “Optional” Movement. PhD Dissertation, University of Groningen.



Name index

A Aarsen, J. 261, 265 Abney, S. 3, 6, 37 Adams, M. 286, 291 Adamzik, K. 265 Aldridge, M. 74, 78 Alexiadou, A. 73, 74, 78, 79, 81, 134 Ambar, M. M. 133, 134 Antelmi, D. 48, 51, 57, 63, 78 Atkinson, M. 114, 134 Auer, P. xiii, 172, 189 B Backus, A. 260, 265 Bardel, C. 56, 74, 78 Barreña, A. 84, 87, 91, 92, 103–105 Bates, E. 198, 218 Becker, M. 38 Bel, A. xiii, 74, 84–86, 98, 103, 104 Bellugi, U. 349, 359 Benedicto, E. 345, 356, 359, 360 Bennis, H. 317, 332 Berkele, G. 10, 22, 37 Berman, R. 223, 266 Bernardini, P. v, viii, 11, 19, 39, 41, 42, 57, 74, 78, 145, 151, 153, 154, 159 Bernhardt, B. H. 194–195, 218 Bernstein, J. 44, 72, 74, 78 Bickerton, D. 114, 134, 272, 289, 291 Bley-Vroman, R. 303, 332 Bloom, P. 115, 134 Boersma, P. 198, 213, 218 Boeschoten, H. E. 260, 266 Bohnacker, U. xiii, 27, 36, 37, 48, 51, 78 Borsley, R. D. 78 Boser, K. 96, 103, 104

Bottari, P. xiii, 8, 16, 29, 37, 53, 78, 79 Brandt, M. 148, 158 Brannen, K. 215, 219 Brendemoen, B. 222, 260, 266 Brown, R. 7, 26, 37, 181, 189, 265, 269 Brünner, G. 266 C Cantone, K. vii, xiv, 35, 39, 79, 134–136, 332 Cardinaletti, A. 47, 62, 74, 78, 79, 110, 111, 134, 357, 358 Caruso, M. 283, 291 Caselli, M. 63, 80, 181, 190 Chan, C. Y. 274, 303, 332 Chierchia, G. 8, 16, 19, 23, 24, 26, 27, 31, 36, 37 Chilosi, A. M. 37, 78, 79 Chini, M. 63, 79 Chomsky, N. xiii, 84, 101, 104, 110, 133, 134, 271, 276, 278, 283, 288, 291, 297, 331, 332, 335, 336, 338, 347, 355, 357, 358 Cinque, G. 44, 45, 72, 74, 79, 331, 332, 358 Cipriani, P. 37, 53, 54, 57, 63, 66, 78, 79 Clahsen, H. xiv, 4, 25, 26, 37, 39, 49, 80, 100, 104–106, 124, 134, 303, 320, 321, 323, 332, 333, 358, 359 Clark, E. V. 117, 134 Clyne, M. 155, 158, 258, 266, 267 Cordes, J. 14, 35, 37 Costa, J. 122, 133–135 Crago, M. 49, 51, 80 Crain, S. 103, 104 Creunant, G. 78

362 Name index

D De Boysson-Barbies, B. 195, 218 De Houwer, A. 41, 79, 107, 135, 197, 218 Delattre, P. 193, 218 Delsing, L.-O. 48, 74, 79 Den Dikken, M. 332 Demuth, K. 28, 29, 31, 37, 38, 193, 195, 220 Deuchar, M. v, ix, x, 161, 162, 165, 166, 171, 178, 189 De Villiers, J. 157–158 Dik, S. C. 86, 263, 266 Dimitrova-Vulchanovica, M. 79 Döpke, S. 42, 79, 143, 152, 153, 157–159, 189 E Ehlich, K. 237, 240, 261, 266 Eisenbeiss, S. 25, 26, 30, 37 Elbourne, P. 353, 357, 359 Elordieta, A. 89, 102, 104 Emonds, J. 331, 332 Erdal, M. 242, 260, 265, 266 Eubank, L. 42, 79, 333 Everett, D. L. 111, 135 Ezeizabarrena, M. J. v, viii, 83–87, 91–94, 96, 98, 100–105 F Ferdinand, A. 309, 319, 323, 332 Fernández, B. 100, 102, 104, 105 Fikkert, P. 191, 216, 218 Fitzpatrick, J. 341, 351, 357, 358 Fodor, J. A. 271, 285, 292 Fritzenschaft, A. 156–158 G Galves, C. C. 113, 135 Garfield, J. L. 285, 292 Gawlitzek-Maiwald, I. v, ix, x, xiii, 19, 37, 42, 81, 139, 141, 145, 148, 153, 154, 156–159, 192, 199, 214, 218 Genesee, F. 12, 37, 42, 43, 79, 80, 107, 108, 135–137, 151, 153, 157–159,

162–164, 167, 173, 178, 179, 180, 182, 185, 189, 190, 197, 220 Gérard, C.-L. vii, xiii, 38 Gerken, L. A. 28, 37 Giorgi, A. 1, 3, 37, 45, 74, 79 Giusti, G. 6, 7, 34, 37, 38, 74, 79 Goad, H. 215, 219 Goldsmith, J. A. 197, 219 Goodz, N. 173, 189 Graefen, G. 266 Granfeldt, J. 8–10, 22, 26, 38, 48, 49, 51, 72, 74, 79 Greenberg, J. H. 297, 332 Grijzenhout, J. 195, 219 Grinstead, J. 84, 96, 101, 104, 105 Grosjean, F. 155, 158, 258, 267 Gualmini, A. 37 Guasti, M.-T. xiii, 23, 37–38, 98, 105, 319, 332, 337, 347–350, 357–358 Guilfoyle, E. 84, 105 Gumperz, J. 172, 189, 265, 267 Gut, U. 141, 158, 327 H Haider, H. 133, 135, 299, 332 Haig, G. 263, 267 Håkansson, G. 11, 39, 52, 81 Hall, T. A. 193, 219 Halle, M. 347, 358 Hamann, C. xiii, 4, 38, 124, 135, 332, 359 Harris, J. W. 193, 219 Hawkins, R. 303, 332, 334 Hayes, B. 198, 213, 218 Heck, F. 357, 358 Herkenrath, A. v, x, xi, 221, 238, 245, 258, 262, 264, 267, 268 Herschensohn, J. 286, 289, 291, 292 Hinzelin, M. v, ix, 107, 108, 134–136, 330–332 Hoekstra, T. 24, 25, 38, 86, 96, 98, 101, 105, 344, 356, 358 Hoffmann, L. 265, 267, 269 House, J. 47, 71, 88, 153, 260, 267, 278, 279, 282

Name index 363

Hulk, A. viii, xiii, 4, 38, 39, 42, 51, 73, 80, 98, 105, 143–145, 152, 155, 159, 281, 292, 309, 332, 356, 359 Huss, L. M. 42, 79 Hyams, N. 38, 84–86, 94, 96, 98, 101, 105, 115, 130, 135, 356, 358 Hyltenstam, K. 39, 80, 81, 136, 269, 296, 333 I Idiazabal, I. 104 Ingram, D. 195, 196, 219 J Jaeggli, O. 88, 105, 110, 135, 136, 290, 292 Jakobson, R. 215, 219, 261, 267 Jakubowicz, C. vii, xiii, 38 Johanson, L. 222, 241, 257, 260, 263, 265–267, 269 Johnson, J. 196, 199, 214, 218–220, 284, 292, 340, 356 Johnston, J. R. 198, 219 Jones, B. M. 78 Joppen, S. 195, 219 Jordens, P. 332, 356, 358 Juan-Garau, M. 108, 130, 132, 136, 164, 189 K Kaiser, G. 222, 268, 310, 333 Karakoç, B. v, x, xi, 221, 238, 245, 262, 264, 267, 268 Karmiloff-Smith, A. 26, 35, 38 Kato, M. v, xi, xii, 89, 101–103, 105, 110–113, 117, 133–137, 271, 274, 282, 289, 290, 292 Kaye, J. 197, 219 Kayne, R. 44, 79, 297, 330, 333 Kehoe, M. v, x, 132, 191, 212, 216, 219 Kennelly, S. D. 238, 268 Kim, K.-H.. 263, 268 Klima, E. 349, 359 Koehn, C. 10, 24, 38 Koopman, H. 297, 333

Köppe, R. xiii, 109, 132, 136, 167, 189, 304, 330, 333 Kornfilt, J. 238, 262, 263, 268 Korzen, I. 35, 38 Koster, H. 344, 345, 358 Koster, J. 345, 359 Krämer, I. 84, 87, 90, 94, 96, 98, 102, 103, 105 Kroch, A. 222, 268, 356, 359 Kupisch, T. v, vii, viii, xiii, xiv, 1, 14, 25, 31, 36, 38, 39, 63, 79, 102 L Labov, W. 101, 105, 265, 268 Laenzlinger, C. 45, 74, 80 Laka, I. 101, 105 Lamarche, J. 74, 80 Langacker, R. 2, 38 Lanza, E. xiii, 162–164, 167, 170–173, 180, 189, 266 Larrañaga, M. P. xiii, 84, 97, 106, 132 Lasnik, H. 134, 350, 357, 359 Lebeaux, D. 356, 359 Lehmann, C. 262, 268 Leopold, W. 195, 219 Lewis, G. L. 262, 268 Liceras, J. 87, 103, 105, 106 Lightfoot, D. 106, 222, 268 Lillo-Martin, D. 103, 104 Lindner, K. 38, 198, 219 Lleó, C. ii, v, x, 28, 29, 31, 38, 103, 106, 191, 195, 198, 201, 212, 215, 216, 219 Longobardi, G. 1, 3, 6, 33, 36–38, 45, 47, 74, 79, 80, 133, 136 Lust, B. 37, 104 M MacWhinney, B. 56, 60, 80, 136, 165, 171, 184, 190 Marantz, A. 347, 358 Maratsos, M. P. 26, 39 Matthews, S. J. 51, 81, 178, 179, 190 Meinhold, G. 193, 219

364 Name index

Meisel, J. M. ii, vi, xi, xii, 38, 41, 43, 80, 83, 92, 100, 104–108, 114–116, 124, 127, 132, 134, 136, 148, 156, 158, 222, 268, 275, 289, 292, 295, 303, 307, 320, 321, 323, 330–333, 359 Mills, A,E. 124, 131, 136 Milroy, J. 265, 268 Milroy, L. 158, 189, 265, 268 Möhring, A. vi, xi, xii, 108, 132, 135, 136, 295, 330–333 Müller, G. 356–359 Müller, N. ii, iii, iv, v, vii, viii, xiii, xiv, 4, 11, 20, 35, 38, 39, 42, 51, 52, 65, 72, 73, 79, 80, 98, 102, 105, 115, 136, 143–145, 152, 155–159, 264, 268, 281, 292, 303, 324, 333, 356, 359 Mundy, Ch. 241, 268 Muysken, P. 158, 189, 223, 261, 265, 268, 303, 320, 321, 323, 332

Pérez-Vidal, C. 108, 130, 132, 136, 164, 189 Perlmutter, D. M. 110, 137 Petersen, J. 42, 80 Pfaff, C. 260, 268 Pfanner, L. 37, 79 Pienemann, M. 100, 106, 303, 320, 321, 323, 332, 333 Pine, J. M. 84, 104, 350, 359 Pinker, S. 286, 292 Pizzuto, E. 63, 80, 181, 190 Platzack, Ch. vii, xiv, 1–3, 39, 51, 80 Pollock, J. Y. 84, 106, 301, 334 Powers, S. 332, 356, 359 Prince, A. S. 191, 215, 220 Prinz, M. 219

N Nash, L. vii, xiii, 38, 133, 137 Neeleman, A. 295–299, 301–303, 314, 317, 320, 330, 331, 333 Nespor, M. 74, 80, 332 Newport, E. 284, 292 Nicoladis, E. xiii, 12, 37, 42, 79, 80, 107, 135, 158, 162–164, 167, 173, 178–180, 185, 189, 190 Noonan, M. 84, 105, 268

R Radford, A. xiii, 23, 39, 83, 106, 114, 137, 227, 268, 275, 292 Raposo, E. 284, 292, 293 Rasetti, L. 94, 106 Redder, A. 231, 254, 260, 262, 264, 266, 268, 269 Rehbein, J. ii, v, x, xi, 221–223, 226–228, 232, 240, 241, 257, 258, 260, 261, 263, 266, 267, 269 Reis, M. 158 Rigaut, C. vii, xiii, 38 Rizzi, L. vii, xiv, 2, 3, 37, 39, 86, 88, 106, 110, 137, 275, 291, 292, 298, 334, 336, 337, 346, 347, 348, 350, 357, 358 Roberge, Y. 133, 137 Roeper, T. vi, vii, xii, xiv, 144, 159, 288, 291, 293, 335, 338, 343–345, 347–349, 351, 354, 356, 357–359 Rohrbacher, B. 345, 359 Ronjat, J. 304, 334 Rosengren, I. 158 Rowland, C. F. 350, 359

O Odriozoloa, J. C. 89, 106 Oliviera, M. de 117, 137 Ortiz de Urbina, J. 102, 106 Owen, A. 345, 359 P Paradis, J. xiii, 12, 37, 49, 51, 79, 80, 96, 106–108, 135, 137, 157–159, 178, 189, 195, 197, 219, 220 Parodi, T. xiii, 38, 49–51, 72, 80 Pater, J. 196, 199, 220 Penke, M. 37 Penner, Z. 7, 25, 38, 39, 343, 356, 360

Q Quay, S. 161, 162, 165, 166, 178, 189



Name index 365

S Safir, K. 88, 101, 105, 110, 135, 136, 290, 292 Santelmann, L. 104 Satterfield, T. 356, 359 Sauerland, U. 353, 356, 357, 359 Schaeffer, J. 314, 334 Schlyter, S. 9, 11, 19, 35, 39, 42, 43, 52, 57, 74, 78, 80, 81, 145, 151, 153, 154, 159, 260, 269 Schmitz, K. vii, xiii, xiv, 35, 39 Schönenberger, M. 343, 360 Schoorlemmer, M. 46, 81 Schütze, K. 344–346, 349, 350, 356, 357, 360 Schwartz, B. D. 42, 49, 80, 81, 358 Secco, G. 42, 80 Selkirk, E. 331, 334 Serratrice, L. 25, 39, 42, 80 Siloni, T. 3, 39 Silva-Corvalán, C. 281, 289, 293 Simões, L. J. 117, 137 Slobin, D. I. 134, 136, 223, 260, 266, 302, 334 Smith, N. 26, 35, 38, 295, 303, 334 Smolensky, P. 191, 197, 211, 215, 220 Snow, K. 11, 39, 56, 80 Soares, C. xiii, 133, 137, 155, 158 Sorace, A. vii, xiv, 78 Speas, M. 345, 360 Sprouse, R. 42, 81 Starke, M. 74, 79, 110, 111, 134 Stemberger, J. P. 194, 218 Stock, E. 193, 219 Stowell, T. 65, 81, 297, 334 Szabolsci, A. 39 T Taeschner, T. 48, 51, 81 Taylor, A. 356, 359 Thomason, S. G. 258, 269 Thornton, R. 345, 348, 349, 358 Towell, R. 303, 334

Tracy, R. xiii, 19, 37, 42, 79, 81, 139, 141, 145, 148, 154–159, 192, 199, 214, 218, 292, 356, 360 Travis, L. 297, 334 Tsimpli, I.-M. 295, 303, 334 U Underhill, R. 262, 269 V Vainikka, A. 42, 81 Valian, V. 115, 117, 137 Valois, D. 3, 39, 74, 81 van der Linden, E. xiii, 309, 332 Vihman, M. 186, 189, 190, 195, 218 Volterra, V. 48, 51, 81 W Wanner, D. 133, 137 Watanabe, A. 232, 269 Weerman, F. xiii, 295–299, 301–303, 314, 317, 320, 330, 331, 333 Weinreich, U. 258, 269, 281 Weissenborn, J. 7, 25, 39, 332 Werker, J. 283, 293 Wetter, M. 49, 51, 81 Wexler, K. vii, xiv, 85, 86, 91, 102, 106, 345, 348–350, 358, 360 Wiese, R. 193, 220, 269 Wijnen, F. 90, 98, 106 Withman, J. 104 Y Yip, V. 51, 81, 178, 179, 190 Young-Scholten, M. 42 Z Zabala, I. 89, 106 Zamparelli, R. 74, 81 Zimmermann, I. 158 Zubizarreta, M. L. 133, 137 Zwart, J.-W. 350, 360



Subject index

π-element 227, 233 φ-features 109 2L1 42, 43, 48, 52, 53, 73, 108, 295, 296, 304, 324, 329 A absolutive 87, 88, 94, 99, 100, 102, 103 acceleration 197–199, 213–215 acquisition of linguistic procedures 223 address forms 273, 274, 276 adjective viii, 3, 26, 31, 35, 43–47, 49, 50, 52, 53, 60–63, 65–75, 77, 78, 80, 263, 291 adoption of procedures 254, 255 adpositional components 262 adpositions 225, 263 adult second language acquisition xi, 42, 296, 302, 320 AGR vi, xiii, 88, 89, 92, 144, 149, 335, 336, 345–351 Agreement xi, 3, 26, 84, 87–89, 92, 93, 95, 99–106, 108, 110, 111, 113, 115, 117, 123, 124, 136, 142, 148, 150, 204, 236, 238, 279, 282, 283, 286–288, 290–292, 307, 333, 336, 345, 347, 348, 350, 353, 358, 359 agreement affixes 110 anchoring 2, 353 article(s) 1, 2, 4–10, 15–18, 20–22, 24–32, 34–39, 45, 47, 48, 50, 61, 63, 66, 73, 74, 79, 81, 97, 115, 154, 198, 200, 201, 224, 282–284, 330, 352 asynchronic development ix, 147 Autonomous Development Hypothesis 107, 108, 132 Aux-to-Comp movement 287, 288, 291

B Basque v, viii, ix, 83, 84, 86–106, 263 bi-procedural qualities 228 bilingual bootstrapping xiii, 37, 139–141, 143, 145, 146, 151, 153, 154, 156–158, 199, 214, 218 bilingual children v, viii–xii, 1, 4, 8, 10, 11, 19, 22, 35, 39, 43, 49, 51–53, 55–57, 58–61, 63, 68–73, 77, 80, 81, 87, 91–94, 98, 107–109, 113, 131, 132, 135–137, 139, 143–145, 153, 154, 156, 158, 178, 190–192, 194–196, 200, 206–210, 213–215, 220, 221, 223–225, 243, 249–251, 257, 260, 265, 281, 296, 302, 304, 309, 311, 319, 323, 328–330, 332 bilingual discourse 254–257 bilingual first language acquisition viii–x, 38, 39, 79, 80, 105, 106, 136, 159, 189, 222, 264, 268, 292, 301, 333 bilingual mode of speaking 258 bilingualism vi, ix, xi, xiii, 11, 35, 38, 39, 79–81, 104, 105, 132, 136, 145, 152, 154, 158, 159, 189, 190, 213, 215, 218, 219, 221–223, 258, 267, 268, 292, 293, 295, 329, 330, 333, 359 bilinguals viii, xi, 22, 39, 101, 136, 145, 151, 152, 156, 158, 159, 161, 163, 173, 179, 186, 200, 206–209, 214, 218, 223, 249, 251, 258–260, 281, 285, 296 BN-stage 8, 16, 18, 24, 27, 36 branching rhyme parameter 191 C C-command 339, 341

368 Subject index

case marking 83, 89, 152, 236, 238, 242, 249, 253, 255 case marking suffix 238 case morphemes 225, 277, 290 CHILDES 10, 53, 54, 60, 74, 80, 165, 171, 184, 190, 344 clustering effect 111, 314, 330 co-construction 227, 237, 256, 262 coda condition 197 codas x, 191–202, 204–207, 206–216, 219 code-mixing v, x, 42, 78, 161, 166, 167, 169–171, 173–175, 177–180, 185, 186, 190, 265, 268 code-switching 80, 158, 189, 258, 267, 275, 289 complementizer 112, 147, 148, 150, 224, 225, 230, 231, 233, 243, 245, 247, 253–255, 258, 264, 265, 281, 324 Configurational Hypothesis 1, 3 connective 223, 230, 237, 238, 248, 258, 260 connectivity xi, 221, 223, 224, 230, 238, 243, 250, 255, 257, 258, 260, 263 connector 231, 232, 254 consciousness 272, 283, 285 construction-final position 230 construction-initial (position) xi, 230, 234, 245, 251, 253, 264 contact Turkish 225, 242, 244, 253, 255, 259 contact-induced mechanisms of language change 258, 259 Continuity Hypothesis 23, 102, 104 control 148, 152 converbial constructions 234 copresence of the speaker and hearer 258 copying 335, 349 core grammar 112, 288 correlation ix, x, 55, 163, 164, 169, 170, 175–177, 180, 184–186, 208, 209, 216 CP xiii, 1–4, 25, 34, 51, 104, 127, 130, 144, 145, 147, 148, 150, 338, 341, 349, 350, 351, 356

critical age xii, 276, 281, 283, 286, 289–291 critical period 292, 302 cross-linguistic ix, x, 1, 2, 8, 23, 24, 26, 27, 29–32, 34, 38, 39, 51, 52, 79, 108, 135, 145, 158, 181, 189, 215, 221, 224–227, 266, 332, 345, 359 cross-linguistic basis 225, 226 cross-linguistic configuration 226 cross-linguistic influence ix, x, 38, 39, 51, 52, 108, 145, 221, 359 cross-over phenomena 275 cue strength 143, 152 D default grammar 345 deficient pronouns 110 deictic field 224 delay 19, 22, 206, 207, 210, 213 delete 114, 335, 336, 345, 350 deletion xiii, 98, 103, 345–347, 350 determiner(s) viii, xiii, 1, 2, 4–7, 9, 10, 15–28, 30–38, 45–48, 50, 54, 61–64, 63, 78, 79, 81, 83, 97 devrik cümle construction 242 Differentiation Hypothesis 107 discourse 2, 13, 25, 34, 36, 111, 141, 162–164, 173, 174, 189, 190, 222–224, 226, 227, 231, 237, 240–242, 253–261, 263, 265, 267–269, 279 discourse pragmatics 231, 237 discourse-licensed 111 Distributed Morphology 347, 358 Do-insertion xiii, 342–344, 349, 354, 356, 357, 359 dominance x, 11, 41, 42, 49, 51–53, 57, 105, 145, 151, 152, 157, 161, 178, 179, 180, 182–186, 188, 190, 199 dominance and code-mixing 179 dominant language x, 21, 22, 39, 141, 145, 152, 154, 159, 178, 179, 181, 183, 275 DP v, viii, 1, 3, 4, 6, 7, 10, 17, 19, 22, 26, 30, 31, 34, 37–39, 41, 43, 44, 46, 48,

Subject index 369

49, 50–52, 60, 61, 63, 66, 72, 73, 78–81, 84, 97, 106, 152, 268, 298, 300, 301, 331, 338, 358, 359 DP-Hypothesis 3 drift 226, 227, 229, 233, 234, 240, 262 E ECM 152 economy xii, xiii, 104, 106, 216, 335, 338, 340, 342, 343, 345, 346, 350–352, 354, 356, 358–360 embedded xi, 133, 148, 221, 226, 227, 230, 232–234, 237, 238, 242, 243, 247, 253–256, 263, 299, 300, 302, 311–314, 319, 321, 324, 329, 330, 338 English ix–xii, 2, 8, 23, 26, 27, 29, 31, 34, 35, 37, 39, 42, 45, 49, 51, 52, 74, 79, 81, 83, 84, 86, 96, 98, 101, 103–106, 108, 115, 130, 132, 136, 139, 140–143, 146–154, 156–159, 161–169, 171, 174–181, 182–185, 187, 188, 190, 194–196, 198, 199, 215, 216, 218, 219, 232, 244, 260, 261, 268, 271–273, 281, 284–289, 291, 292, 295, 299, 301, 302, 320, 331, 332, 343, 352, 353, 357–360 ergative 87, 88, 99, 100, 102, 103, 106 exothetic reference 232 expansion of function 255 expletive(s) 7, 25, 28, 30, 31, 111, 112, 116, 130, 133, 284, 290, 360 F faithfulness 197, 198, 211, 212, 218 feature vi, xiii, 2, 6, 24, 35, 47, 92, 109, 111, 114, 117, 128, 133, 137, 195, 230, 240, 259, 280, 288, 291, 335–351, 353, 356, 357, 359 Feature-attraction vi, 335 Feature-checking 100, 113, 335–338, 341, 343, 345, 347–349, 351, 353, 356, 357 field 3, 224, 231, 232, 242, 251, 253, 255–257, 263

finite verb form 103, 134, 238 finiteness ix, 2, 3, 56, 92, 104, 106, 108, 109, 111, 113–118, 120, 123–125, 127–129, 131–133, 136, 143, 229, 237, 240, 241, 253, 256, 292, 307, 317, 331, 333 fossilization xi, 275, 276 free subject inversion 109, 111, 112, 116, 120 French v, viii, xii, xiii, 1, 2, 4, 5, 8–12, 11–14, 17–22, 24–32, 31–36, 38, 39, 41, 42, 44, 45, 48, 49, 51, 52, 74, 78–81, 84, 85, 87, 96, 104, 106, 124, 127, 134, 136, 144, 145, 154, 157, 159, 163, 179, 218, 265, 268, 269, 291, 292, 295, 296, 299, 300, 299, 301, 302, 304, 305, 307–309, 313, 316, 317, 319, 320, 323, 331–333, 359 frequency viii, x, 1, 11, 12, 26, 29, 31, 50, 51, 63, 87, 93, 95, 97, 108, 112, 113, 129, 132, 157, 172, 192, 194, 195, 198, 199, 213, 214, 221, 224, 249, 287, 309 full DP 6 functional categories 9, 22, 23, 26–28, 31, 35, 37–39, 73, 78, 80, 83, 84, 102, 105, 136, 137, 158, 159, 261, 275, 332, 351, 358, 359 functional elements 6, 27, 154, 223, 224 functional expansion 222, 253, 258 functional pragmatics 224 functional projection(s) 6, 25, 26, 45, 120, 127, 130, 145, 360 functional skeleton 275, 289 G generative xiv, 2, 3, 6, 7, 22, 28, 33, 34, 37, 39, 80, 84, 104, 105, 196, 220, 222, 223, 232, 297, 358, 359 German v, viii–xiii, 1, 2, 4, 7, 10–14, 20, 21, 24–35, 37–39, 41, 42, 49–51, 79–81, 84, 85, 87, 96, 98, 103, 104, 106, 107–116, 123–125, 127–129, 130–132, 134–136, 139–144,

370 Subject index

146–154, 156–159, 191–195, 197–200, 202–225, 228–232, 234, 237, 242, 243, 249, 253, 254, 256, 258, 259–263, 267, 273, 290, 295, 296, 299, 300–302, 304–308, 311, 312, 314, 315, 317–324, 326–333, 343, 350, 352–354, 358, 360 German discourse 254, 258, 263 H hearer 223–228, 232, 237, 240–242, 256, 258, 259, 261–264, 274 I I-language 271, 285 illocution 226–234, 237, 240, 249, 254, 255, 261, 263, 264, 341 in situ 133, 134, 232, 233, 237, 238, 245, 251, 264 Indo-European 230, 255, 261, 262 infinitival construction 142, 148, 150 infinitive xiv, 83, 85–87, 89, 91, 92, 94, 95, 97–103, 106, 133, 134, 148, 150 inflectional morphology 111, 114, 198 inflectional paradigm 110, 111 influence ix–xi, xiii, 19–22, 38, 39, 42, 43, 50–53, 56, 68, 72, 80, 105–108, 132, 139, 145, 153, 158, 159, 161, 192, 213–215, 221, 281, 292, 309, 314, 355, 359 innovative constructions 221, 249, 257 innovative word structure 259 Interdependent Development Hypothesis 107, 108 interlingual discourse 222, 258–260 interpretable feature(s) 114, 336 interrogative v, xi, 2, 17, 124, 221, 223–234, 237, 238, 241, 242, 247, 248, 249–255, 257, 261–265, 348, 357 interrogative element(s) v, 221, 223, 257, 261 interrogative illocution 226, 233, 237, 264 interrogative mode 226

inversion 109, 111–113, 116, 120, 122, 133, 137, 142, 147, 348–350, 359 IP 2, 42, 84, 102, 106, 114–116, 120, 127, 130, 140–143, 147, 148, 150, 154, 268, 334, 336, 350, 357–359 K knowledge 35, 43, 50, 108, 223–234, 237, 240, 242, 254, 256, 258, 259, 261–263, 271, 283–285, 289, 291, 292, 303, 358–360 knowledge deficit 225–227, 229–233, 237, 242, 256, 261 knowledge domain 226, 227, 229, 232–234, 259, 261, 262 knowledge of H 226, 227 knowledge of the hearer 226 L L1 xi, xii, xiv, 3, 8, 41–43, 48–52, 58, 63, 72, 73, 105, 106, 271–273, 275–277, 280, 281, 283–291, 296, 303, 314, 320, 324, 330, 333 L2 v, xi, xii, xiv, 3, 9, 41–44, 48–53, 55, 71–74, 77, 79–81, 100, 105, 106, 108, 271–273, 275–277, 281–286, 288–292, 295, 296, 303, 320, 324, 326, 329, 330, 333 lack of knowledge 223–225, 228, 259 language change 134, 221, 222, 258–260, 268 language choice 161, 162, 166, 171, 172, 186, 187, 189 language contact 22, 42, 43, 189, 221–223, 253, 258, 266, 269, 359 language differentiation 37, 43, 79, 135, 158, 189, 190 language domains 222 language dominance 11, 41, 42, 49, 51, 52, 57, 105, 145, 157, 178, 180, 190 language interaction 42, 43, 197 language mixing 139, 159, 163, 169, 189, 190, 332 language proficiency 12

Subject index

lexical learning 25, 26, 33, 37, 273, 275, 276 Lexical learning hypothesis 25 lexical subject 133 light verb 275 linguistic field 224, 231 linguistic procedure(s) 223–225, 262 linkage 230, 255, 258, 265 M markedness 191, 192, 197–199, 211, 218 mass nouns 10, 19–21, 26, 36, 37 matrix xi, 227, 229, 231, 232, 237, 238, 241–245, 247, 249–251, 253–256, 262–264 matrix verb(s) 238, 241, 242, 244, 247, 251, 264 Maturation Hypothesis 23 meteorological verb 130 minimal grasp strategy 172 Minimal Word 199, 214 mixing v, ix, x, 13, 20, 39, 42, 78, 139–143, 145–150, 152, 153, 157, 159, 161–164, 166–171, 173–180, 184–190, 265, 268, 276, 332 Modelling Hypothesis 173, 174, 177 monolingual children viii, x, 8, 22, 26, 41–43, 48, 50–52, 54–56, 59–62, 64, 65, 67–71, 73, 75, 91, 140, 144–146, 151, 152, 155, 196, 199, 200, 202, 206, 210, 214–216, 223, 234, 251, 307, 309, 310, 319 monosyllables 196, 199, 205, 204–206, 212–214 morphological subordination suffixes 247 movement operations 287, 351 multifunctional central system 271, 285 Multiple Grammars vi, xii, xiii, 335, 342, 351, 352, 354, 356, 357 N N-raising parameter 52, 73 negation 56, 79, 105, 259, 333, 342, 345, 346, 348–350, 357, 358

neurolinguistic maturation 302 new variety of Turkish xi, 257 NoCoda 191, 197–199, 211–214 Nominal Mapping Parameter 26 nominalized form 236, 244 non-finiteness 237, 241 non-interrogative 221, 223–228, 230–234, 237, 238, 241, 242, 250, 249–255, 262–265 non-interrogative usage 226, 234, 254 non-interrogative use(s) 227, 228, 230–232, 234, 264, 238, 265 non-interrogative wh-constructions 224, 231, 241, 242, 249–251, 255, 263 non-null subject language 108, 132, 279 non-rhematic 242, 251, 257 null auxiliary ix, 83, 90, 96, 98, 102 null element 96 null expletive 111, 112 Null Modal Hypothesis 87, 91, 96 null referential subject 112 null subject 84, 105, 108–115, 131–137, 277, 279–281, 290, 292 Null Subject Parameter 105, 108–111, 113–115, 131, 132, 134–137, 277, 292 O object agreement 95, 100 operative field 224, 257 operative procedure(s) 224, 227, 230, 231, 233, 254, 255, 257 Optimality Theory 29, 191, 194, 196, 211, 213, 215, 218, 220 optional infinitive xiv, 85–87, 102, 106 OV/VO parameter 295–299, 301–305, 307, 308, 311, 313, 314, 317, 319, 320, 324, 326, 327, 329, 331 P para-operative field transposition 224 parameter setting xiv, 39, 42, 84, 106, 109, 111, 297, 303, 319 parameter value 41, 302, 313

371

372 Subject index

parameters xii, 22, 23, 42, 80, 106, 111, 135, 136, 191, 219, 268, 271, 276, 277, 286, 288, 292, 297, 303, 330, 334, 359 Parental Discourse Hypothesis 162–164, 173, 174 participle 89, 91, 103, 234, 238, 249, 259 particle(s) xii, 140, 224, 238, 243, 247, 250, 251, 259, 261, 263, 264, 274, 295, 299, 300, 317–319, 328–332 particle construction 328 Path Quantification 341 pauses 141 performance 3, 27, 91, 100, 102, 115, 129, 283 periphery 34, 39, 288, 289 personal pronouns ix, 74, 88, 97, 109, 111–118, 123–125, 131, 132, 278 phi-feature 345 Pied-Piping vi, xiii, 335, 336, 338, 340, 345, 355, 357 positioning of viii, 43, 48, 50, 52, 53, 61, 65–71 possessive 26, 30, 45–48, 51, 53, 61–64, 66–69, 71–73, 78, 157, 236, 238, 249, 253, 255, 259, 290 post-verbal position 232 post-verbal subjects 113, 120, 133 postposition 249, 253, 259 postpositional phrases 234 pre-pre-field 263 pre-predicative focus position 232, 233 precursors 147, 148, 150 present tense 89, 120, 259, 352–354 principles and parameters approach/model 111, 219, 276, 359 pro-drop 84–88, 91, 94, 96, 101, 102, 105, 108, 113, 115, 135, 136, 290, 355 procedural compositionality 221, 257 procedure(s) 223–225, 227, 229–231, 233, 241, 245, 249, 254, 255, 257, 261, 262, 264 processing limitations 115, 137

productive use 108, 109, 111, 113–116, 118, 121, 124, 125, 128, 132, 223, 307, 309, 317 progressive 352, 353 pronominal paradigm 112, 113, 116, 123 pronoun(s) v, ix, xiii, 9, 17, 24, 30, 38, 74, 79, 88, 89, 96, 97, 102, 103, 107–118, 120, 123–125, 128–134, 157, 224, 273, 274, 276, 278, 280, 282, 284, 290–292, 333 proper nouns 80, 114–118, 124, 125 propositional knowledge 227, 254 proto-determiners 16, 17 Q quantifier(s) 2, 6, 32, 61, 74, 290 quantitative analysis 153, 163, 173, 188 question word(s) 223, 225, 226, 233, 256, 261, 262, 265 question-answer pattern 226, 240 questions 90, 109, 139, 142, 147, 154, 155, 178, 221, 226, 233, 261–263, 267, 288, 336, 341, 349–351, 356–358 quotation particle 264 quotative particle 238, 247, 250, 251 quote particle 263 R referential features 6, 26 referential null subject 111, 112, 133 rhematic xi, 231–233, 242, 245, 251, 253, 255, 257 rheme position 231 root infinitive(s) viii, ix, xiv, 38, 83–87, 90–92, 93–96, 98–106, 292, 360 S Scope 224, 229, 233, 307, 355 Seek Sublabel xiii, 338–344, 346, 348–351, 354, 356 Spanish v, viii–x, 29, 34, 38, 50, 84–87, 89, 91, 93, 96, 101, 104–106, 110, 133, 161, 162, 164–169, 171, 175, 174–177, 180–185, 187, 188,

Subject index 373

191–202, 204–216, 218, 219, 281, 296, 320, 324, 327 speaker(s) xi, xiii, xiv, 3, 7, 28, 31, 56, 91, 101, 151, 154, 155, 157, 158, 164, 189, 199, 200, 223, 224, 226, 227, 231–233, 237, 240, 254–259, 261, 263–265, 272–275, 280–284, 288, 338, 354 speech action(s) 224, 226, 227, 229, 231, 240 speech action sequence 226 strong pronouns 89, 102, 103, 110, 115, 117, 133, 292 stronger language viii, 8, 12, 13, 43, 52, 73, 145, 154 subject(s) v–vii, ix, xi, 3, 24, 33, 83, 84, 87, 89, 93–102, 104–118, 120–125, 127–137, 139, 148, 157, 161, 224, 238, 274, 277, 279–281, 284, 290–292, 309, 310, 317, 323, 332, 335, 342, 345, 348, 350–353, 359–361, 363 subject agreement 89, 93, 238 subject inversion 109, 111, 112, 116, 120 subject omission 108, 129 subject pronoun 108, 112 subordinate xii, 51, 144, 147, 148, 152, 156, 157, 230–232, 234, 238, 241, 242, 244, 245, 247–249, 251, 254–256, 258, 262, 263, 264, 312, 325, 326 subordinating morphology 242, 251, 253, 255 subordination 54, 227, 230, 234, 236, 240, 242, 247, 253, 262, 263, 267 superordinate 226, 227, 230–232, 234, 250, 253–256, 264 surface phenomena xii, 111, 295, 297, 299 Swedish viii, 1, 3, 8–10, 22, 27, 34, 39, 41, 43–45, 47–49, 51–53, 55–57, 56–61, 67, 68, 71–74, 77, 80, 81, 145, 154, 269 syllable structure x, 192–195, 197, 199, 213, 215, 218, 219

syntactic variation 352 T Tense 2, 9, 83, 84, 89, 92, 99, 103, 106, 120, 136, 143, 212, 216, 259, 291, 292, 333, 336, 343, 345, 346, 348, 350, 352–354, 358 theme position 231 topic construction 280 topic prominent language 279 topic-drop 111, 124, 126–129, 131 topicalized pre-field 263 transcript 166, 167, 174, 180, 182, 184, 228, 237, 262 transfer 21, 41, 49, 50, 52, 72, 80, 81, 98, 144, 152, 153, 157, 159, 190, 197, 199, 258, 268, 279, 281, 284, 287, 291 Turkish v, xi, 50, 221–225, 231–234, 237, 238, 240–245, 247–249, 253–263, 265–269 Turkish discourse 258, 259 Turkish post-field 242 typological characteristics 226 typological perspective 232 typological reasons 225 typology 79, 134, 191, 215, 221, 230, 268 U unconscious knowledge 283 underspecification 24, 33, 37, 38, 83, 84, 96, 98, 105, 218, 356, 358 uninterpretable feature 114, 336 Universal Determiner Requirement 6, 23 Universal Grammar 79, 106, 276, 292, 295, 302, 332–334, 359 utterance(s) ix, 8, 11–13, 12–14, 17, 18, 20, 21, 25, 32, 36, 42, 54, 57, 59, 61, 109, 116, 117, 118, 120, 123–125, 131, 139–141, 143, 145, 146, 149–157, 161–167, 170–175, 178, 180–183, 186–189, 200, 213, 214, 223, 224, 226–228, 230, 232–234, 236–238, 244, 245, 249, 254, 256,



374 Subject index

261, 263, 264, 284, 307, 309, 311, 312, 314, 317, 323, 324, 326, 328 V V-to-I movement 277, 287 V2 52, 104, 108, 123, 127, 131, 136, 141, 147, 152, 158, 228, 229, 321, 359 variety xi, 25, 33, 87, 102, 110, 112, 113, 133, 165, 196–198, 222, 256–259, 265, 268, 281, 311, 333, 341, 344, 345, 347, 356, 357 verb movement 104, 106, 113, 115, 127, 130, 131, 333, 334, 342–344, 359 verb placement 79, 108, 130, 131, 130, 131, 136, 158, 359 verb-raising 351, 352, 354 verba sentiendi and dicendi 238 verbal morphology 110, 117, 148 verbal planning 255 W weak pronouns 89, 110, 111, 117, 133 weaker language viii, ix, 1, 4, 9, 11, 13, 14, 35, 38, 39, 41–43, 52, 53, 59, 69, 73, 78, 80, 81, 141, 145, 154, 159, 269

wh-construction(s) xi, 224–228, 231–234, 237, 238, 240–244, 249–257, 261–264 wh-element(s) xi, 98, 221, 223–227, 229, 230–233, 237, 238, 240–243, 245, 247, 249, 251–256, 261, 262, 264 wh-movement 133, 341, 347, 351, 355 wh-word(s) 225–234, 237, 238, 253–255, 261, 262, 264, 265, 342, 345 word class structure 257 word order v, viii, 39, 41, 43, 44, 48, 51, 52, 60, 61, 72–74, 78, 81, 89, 108, 131, 134, 137, 143, 144, 148, 156, 224, 228, 232, 241, 256, 263, 266, 268, 277, 279, 295, 297, 299, 303, 304, 309, 312–314, 320, 324, 326, 327, 329, 330, 332–334 X X-bar schema 277