The Acquisition of Syntax in Romance Languages (Language Acquisition and Language Disorders)

The Acquisition of Syntax in Romance Languages Language Acquisition & Language Disorders Volumes in this series pro...

Author: Vincent Torrens | Linda Escobar

299 downloads 1349 Views 5MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

The Acquisition of Syntax in Romance Languages

Language Acquisition & Language Disorders Volumes in this series provide a forum for research contributing to theories of language acquisition (first and second, child and adult), language learnability, language attrition and language disorders.

Series Editors Harald Clahsen

Lydia White

University of Essex

McGill University

Editorial Board Melissa F. Bowerman

Luigi Rizzi

Max Planck Institut für Psycholinguistik, Nijmegen

University of Siena

Katherine Demuth

Bonnie D. Schwartz

Brown University

University of Hawaii at Manao

Wolfgang U. Dressler

Antonella Sorace

Universität Wien

University of Edinburgh

Nina Hyams

Karin Stromswold

University of California at Los Angeles

Rutgers University

Jürgen M. Meisel

Jürgen Weissenborn

Universität Hamburg

Universität Potsdam

William O’Grady

Frank Wijnen

University of Hawaii

Utrecht University

Mabel Rice University of Kansas

Volume 41 The Acquisition of Syntax in Romance Languages Edited by Vincent Torrens and Linda Escobar

The Acquisition of Syntax in Romance Languages Edited by

Vincent Torrens Linda Escobar Psycholinguistic Institute, Madrid

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 The acquisition of syntax in Romance languages / edited by Vincent Torrens and Linda Escobar. p. cm. (Language Acquisition & Language Disorders, issn 0925–0123 ; v. 41) Includes bibliographical references and indexes. 1. Language acquisition. 2. Romance languages--Syntax. I. Torrens, Vincent. II. Escobar, Linda. III. Series. P118.A14267 2006 401’.93--dc22 isbn 90 272 5301 3 (Hb; alk. paper)

2006047727

© 2006 – 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

JB[v.20020404] Prn:20/06/2006; 16:34

F: LALD41CO.tex / p.1 (47-109)

Table of contents

Contributors The acquisition of syntax in Romance languages Editors: Vincent Torrens and Linda Escobar

vii ix

Part I. Clitics, determiners and pronouns The production of SE and SELF anaphors in Spanish and Dutch children Sergio Baauw, Marieke Kuipers, Esther Ruigendijk and Fernando Cuetos On the acquisition of ambiguous Valency-Marking Morphemes: Insights from the acquisition of French SE Isabelle Barrière and Marjorie Perlman Lorch

3

23

Definite and bare noun contrasts in child Catalan Anna Gavarró, Ana Teresa Pérez-Leroux and Thomas Roeper

51

Null arguments in monolingual children: A comparison of Italian and French Natascha Müller, Katrin Schmitz, Katja Cantone and Tanja Kupisch

69

Prenominal elements in French-Germanic bilingual first language acquisition: Evidence for cross-linguistic influence Maren Pannemann

95

Part II. Verbs, auxiliaries and inflection A cross-sectional study on the use of “be” in early Italian Claudia Caprin and Maria Teresa Guasti

117

Patterns of copula omission in Italian child language Elisa Franchi

135

Looking for the universal core of the RI stage Manola Salustri and Nina Hyams

159

The acquisition of experiencers in Spanish L1 and the external argument requirement hypothesis Vincent Torrens, Linda Escobar and Kenneth Wexler Early operators and late topic-drop/pro-drop Jacqueline van Kampen

183 203

JB[v.20020404] Prn:20/06/2006; 16:34



F: LALD41CO.tex / p.2 (109-136)

Table of contents

Part III. Movement and resumptive pronouns The acquisition of A- and A’-bound pronouns in Brazilian Portuguese Elaine Grolla Acquiring long-distance wh-questions in L1 Spanish: A longitudinal investigation María Junkal Gutiérrez Mangado Evidence from L1 acquisition for the syntax of wh-scope marking in French Magda Oiry and Hamida Demirdache

227

251 289

Part IV. Syntax/discourse interface Acquisition of focus marking in European Portuguese: Evidence for a unified approach to focus João Costa and Kriszta Szendröi Subject pronouns in bilinguals: Interference or maturation? Manuela Pinto

319 331

Part V. L2 acquisition Is the semantics/syntax interface vulnerable in L2 acquisition? Focus on mood distinctions clauses in L2 Spanish Claudia Borgonovo, Joyce Bruhn de Garavito and Philippe Prévost The development of the syntax-discourse interface: Greek learners of Spanish Cristóbal Lozano Beyond the syntax of the Null Subject Parameter: A look at the discourse-pragmatic distribution of null and overt subjects by L2 learners of Spanish Silvina Montrul and Celeste Rodríguez Louro Index

353 371

401 419

JB[v.20020404] Prn:27/06/2006; 10:57

F: LALD41CN.tex / p.1 (47-141)

Contributors

Sergio Baauw, Universiteit Utrecht, The Netherlands Isabelle Barrière, Johns Hopkins University, USA Claudia Borgonovo, Université Laval, Canada Joyce Bruhn de Garavito, University of Western Ontario, Canada Katja Cantone, Universität Hamburg, Germany Claudia Caprin, Università degli Studi di Milano-Bicocca, Italy João Costa, Universidade Nova de Lisboa, Portugal Fernando Cuetos, Universidad de Oviedo, Spain Hamida Demirdache, Université de Nantes, France Linda Escobar, Psycholinguistic Institute, Spain Elisa Franchi, Università di Venezia, Italy Anna Gavarró, Universitat Autònoma de Barcelona, Spain Elaine Grolla, University of Connecticut, USA Maria Teresa Guasti, Università degli Studi di Milano-Bicocca, Italy María Junkal Gutiérrez Mangado, Euskal Herriko Unibertsitatea, Spain Nina Hyams, University of California, Los Angeles, USA Marieke Kuipers, Universiteit Utrecht, The Netherlands Tanja Kupisch, Universität Hamburg, Germany Cristóbal Lozano, Universidad Autónoma de Madrid, Spain Silvina Montrul, University of Illinois, Urbana-Champaign, USA Natascha Müller, Universität Wuppertal, Germany Magda Oiry, Université de Nantes, France Maren Pannemann, Universiteit van Amsterdam, The Netherlands Marjorie Perlman Lorch, University of London, United Kingdom Ana Pérez-Leroux, University of Toronto, Canada Manuela Pinto, Universiteit Utrecht, The Netherlands Philippe Prévost, Université Laval, Canada Celeste Rodríguez Louro, University of Illinois, Urbana-Champaign, USA Thomas Roeper, University of Massachusetts, USA Esther Ruigendijk, Universiteit Utrecht, The Netherlands Manola Salustri, University of California, Los Angeles, USA Katrin Schmitz, Universität Wuppertal, Germany

JB[v.20020404] Prn:27/06/2006; 10:57

F: LALD41CN.tex / p.2 (141-153)

 Contributors

Krista Szendröi, Universiteit Utrecht, The Netherlands Vincent Torrens, Psycholinguistic Institute, Spain Jacqueline van Kampen, Universiteit Utrecht, The Netherlands Kenneth Wexler, Massachusetts Institute of Technology, USA

JB[v.20020404] Prn:2/06/2006; 11:35

F: LALD41ED.tex / p.1 (46-118)

The acquisition of syntax in Romance languages Editors: Vincent Torrens and Linda Escobar

Introduction This volume includes a selection of the papers delivered at the first and second language acquisition workshop ‘The Romance Turn’ which took place in Madrid on September 2004. The papers address a wide range of acquisition phenomena from different Romance languages and all share a common theoretical approach based on the Principles and Parameters theory. They favour, discuss and sometimes challenge traditional explanations of first and second language acquisition in terms of maturation of general principles universal to all languages. They all depart from the view that language acquisition can be explained in terms of learning language specific rules, constraints, or structures. The workshop intended to gather new acquisition research data in the Romance languages in the light of well known hypotheses about some linguistic phenomena. We have obtained papers on the acquisition of Catalan, French, Italian, Portuguese, Spanish and comparative papers with other languages. The papers are not only inherently interesting; they represent one way to look for concrete answers to questions that permeate linguistic science. We include the following topics on the fields of first, second language acquisition and bilingualism: null arguments, subject pronouns, root infinitives, wh-movement, verb movement, clitics, determiners and resumptive pronouns. The different parts into which this volume is organized reflect different approaches that current research has offered. The contributions in Part 1 deal with issues of development of reflexive pronouns, determiners and clitics in Catalan, French, Italian and Spanish. In the opening chapter Sergio Baauw, Marieke Kuipers, Esther Ruigendijk & Fernando Cuetos analyse reflexive clitics in Spanish in comparison with reflexive pronouns in many Germanic languages under the experimental condition that there are pragmatic differences between SE- and SELF anaphors in the line of Reuland (2001). In order to capture their results, they support Avrutin’s (1999) claim that children have early knowledge of the principles of syntax but often have problems with the use of syntax (feature checking, A-Chain formation) and structure information (formation of referential dependencies). Anna Gavarró, Ana Teresa Pérez-Leroux & Thomas

JB[v.20020404] Prn:2/06/2006; 11:35



F: LALD41ED.tex / p.2 (118-188)

Vincent Torrens and Linda Escobar

Roeper report on their work on acquisition of the bare noun/definite contrasts in English versus Catalan. They claim that children are sensitive to structural contrasts in the NP domain, i.e. in the object position, independent from parametric variation. The paper of Isabelle Barrière & Marjorie Perlman Lorch gives an account of the order of acquisition of different types of SE-constructions in French along with a wide range of research strategies, proposing a modified version of Borer & Wexler’s (1987) Maturation Hypothesis. Natascha Müller, Katrin Schmitz, Katja Cantone & Tanja Kupisch’s paper provides evidence that acquisition of object clitics in French and Italian presents language-specific differences in relation to the acquisition of verbs and the setting of the null subject parameter. The paper by Pannemann studies the cross-linguistic influence in bilingual language acquisition formulated by Müller & Hulk (2000) for the acquisition of determiners and adjectives. Part 2 focuses on the acquisition of verbs, auxiliaries, and the properties of Inflection. The first two papers deal with early Italian. Claudia Caprin & Maria Teresa Guasti focus on the use/omission of the copula and auxiliary BE. They claim that children omit auxiliaries more often than the copula because the computation in which the former are involved is more costly than the computation in which the latter is involved and also conclude, supporting Hirsch and Wexler’s (2004) claim (based on work by Embick 2004), that children’s passives are resultative passives, in which “be” selects a V(erbal) P(hrase). Elisa Franchi also investigates the acquisition and the developmental pattern of copular constructions. The pattern emerging from the data is analysed along the lines of the Truncation Hypothesis (Rizzi 1993, 1994). Manola Salustri & Nina Hyams’s paper compares early Germanic languages, on the one hand, with Spanish, Catalan, and Italian along with other null subject languages, on the other hand. It is argued that that there exists an analogue of the Root Infinitive (RI) stage in the latter languages: the imperative. According to this Imperative Analogue Hypothesis (IAH), they propose that what is universal about the RI stage is the mapping of irrealis mood onto a tenseless clausal structure. The last two papers in this part of the volume analyse acquisition of verbs with respect to presence/absence of subjects. Vincent Torrens, Linda Escobar & Ken Wexler examine early Spanish comprehension data from different experimental paradigms and give support to the External Argument Requirement Hypothesis, according to which children have trouble with base structures that don’t assign a subject/external argument. Finally, Jacqueline van Kampen examines early French, Spanish and Portuguese with respect to finite verbs that lack a subject. She proposes a four-stage process of acquisition. Part 3 contains three papers which deal with wh-movement transformations in child language. Elaine Grolla’s paper relates a number of configurations in Brazilian Portuguese along with English in which the following holds: when a derivation involves movement, the insertion of a pronoun is blocked. In this sense she proposes a unified account of the problems that have been reported so far in child language with respect to certain configurations containing A and A’ bound pronouns.

JB[v.20020404] Prn:30/03/2006; 15:00

F: LALD41P1.tex / p.1 (47-74)

 

Clitics, determiners and pronouns

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.1 (46-127)

The production of SE and SELF anaphors in Spanish and Dutch children Sergio Baauw*, Marieke Kuipers*, Esther Ruigendijk* and Fernando Cuetos** *Universiteit Utrecht / **Universidad de Oviedo

Both Dutch zich and Spanish se are considered Simple Expression or SE anaphors. The present study shows that, in spite of this classification, Spanish children perform much more adultlike on se than Dutch children do on zich. We argue that this developmental difference supports a different analysis of zich as a pronominal element, as opposed to se, which is best analyzed as a reflexive-marker. We also found that Spanish children perform highly non-adultlike on the complex anaphor sí mismo and other strong reflexives, unlike Dutch children who performed highly adultlike on zichzelf. We claim that Spanish children’s difficulties with strong reflexives are for the most part the result of their limited ability to construct integrated discourses.

.

Introduction

Many languages, including Dutch and Spanish, have two types of anaphoric elements: so-called Simple Expression or SE-anaphors, such as zich in Dutch and se in Spanish, and complex anaphors, called SELF-anaphors, such as zichzelf in Dutch and sí mismo in Spanish. These two types of anaphors differ from each other pragmatically. In general, SE-anaphors are the unmarked option (1a), whereas SELF-anaphors are used in contrastive situations (1b). (1) a.

El mago se levanta de la cama, y la bruja se viste. ‘The wizard gets out of bed, and the witch is dressing SE.’ b. La mujer viste a la niña y luego se viste a sí misma. ‘The woman is dressing the girl, and then she is dressing SELF.’

However, the difference between SE-anaphors and SELF-anaphors is not just pragmatic, but also reflects syntactically different ways of coding reflexivity. In addition, Dutch zich and Spanish se show some strikingly different properties, despite their shared status as SE-anaphors.

JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.2 (127-184)

Sergio Baauw et al.

In this paper we will present experimental results on Dutch and Spanish children’s production of SE- and SELF-anaphors. We will argue that the results support early knowledge of both the syntax and the pragmatics of SE and SELF-anaphors. At the same time it will be shown that the use of some of this knowledge is problematic. This affects Dutch children’s performance on the SE-anaphor zich, but not Spanish children’s performance on se, which confirms the different status of se and zich.

. Reflexivity in and outside narrow syntax According to Reuland (2001) SE- and SELF-anaphors reflect different levels of encoding reflexivity. In general lines, the proposal distinguishes referential dependencies formed in narrow syntax (a modular and independent computational system) from dependencies formed outside narrow syntax. Dependencies formed in narrow syntax involve A-Chain formation between two elements that are in a feature checking relation. A-Chain formation results from checking operations between SE-anaphors, such as Dutch zich, and their local subjects. According to Reuland (2001) this operation proceeds in the following way: the formal features of zich (third person, accusative case), move to INFL where they end up in a checking configuration with the subject in [Spec, IP]. (2) a.

Jan waste zich. ‘John washed himself.’ b. [Jan [INFL [Fzich Fwaste ]] [VP waste zich]] John washed SE

Zich checks the D-feature and the phi-features number and person of the subject. However, checking the subject’s person feature will lead to the elimination of the person feature on zich. Since the person feature of zich is interpretable, its elimination leads to loss of information (Chomsky 1995). To prevent this from happening, Reuland proposes that the subject Jan “recovers” the checked and deleted person feature of zich. This recovery operation has an important side effect: it creates a referential dependency – an A-Chain – between zich and the local subject. Note that third person pronouns, such as hem ‘him’or haar ‘her’ cannot be involved in A-Chain formation. Reuland argues that this is because they are specified for number, number being a feature that cannot be recovered after checking and deletion of this feature has taken place (see Reuland 2001 for a discussion).1 Referential dependencies can also be formed outside narrow syntax, at a level that we will call “information structure” (Vallduví 1992), but which other may call “linguistic discourse” or Conceptual-Intentional (C-I) interface (Reuland 2001). In the non-local domain, pronouns can be identified with their antecedents through variable binding or coreference. In the local domain SELF-anaphors can be used. The SELFpart of SELF-anaphors, -zelf in Dutch and mismo in Spanish, is identified with the local subject by projecting a “guise”, a mental representation of an object/individual

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.3 (184-242)

The production of SE and SELF anaphors in Spanish and Dutch children

with properties highly similar but not necessarily identical to the object/individual represented by the local subject (Heim 1982; Jackendoff 1992).2 Importantly, the fact that referential dependencies can be encoded at different levels may affect the acquisition of these elements. Avrutin (2004) argues that although five-year-old children have targetlike syntactic systems, they have problems with the use of narrow syntax to build information structure. Instead, they may often rely on extra-syntactic strategies (i.e., context) to structure information. Stated differently, problems are expected to arise whenever there is competition between narrow syntactic and extra-syntactic operations. This explains why children show significantly more difficulties with those aspects of grammar that play a role in building information structure, such as tense, determiners and pronouns, than with for instance agreement, which is present only to satisfy requirements of narrow syntax. Ruigendijk et al. (in press) showed in their study on the interpretation of pronouns that children (and agrammatics) exhibit problems with the use of narrow syntax to establish referential dependencies. If these claims are correct, children are expected to show problems with the production and interpretation of SE-anaphors, but not with SELF-anaphors. As argued above, SE-anaphors involve a purely syntactic way of establishing referential dependencies, whereas the anaphoric properties of SELF-anaphors rely basically on an extra-syntactic association of the guise introduced by the SELF-morpheme with the local subject. In other words, children could prefer the non-syntactic way to encode referential dependencies. Also, children may have problems with pragmatic or discourse principles, as argued by many authors (Chien & Wexler 1990; Kraemer 2000). If this is true they are predicted to show problems with the correct context of use of SE-anaphors and SELF-anaphors, using SELF-anaphors in non-contrastive contexts and vice versa. In the next section, we present an elicited production experiment that tested the different predictions with respect to the acquisition of SE- and SELF-anaphors by Dutch and Spanish children.

. Experiments . Method The Dutch and Spanish subjects were tested with a Story Elicitation Task. The aim of the task was to elicit short stories on the basis of three-picture-sequences. One experimenter, who was sitting opposite to the child and could not see which picture sequence was being described, had to guess which picture the child was describing. In the Dutch experiment, another experimenter sat next to the child, and acted as the child’s helper. The kind of help that was allowed consisted in a general description of the action depicted when the child misinterpreted the picture (e.g., “I think the story is about dressing”). In the Spanish experiment, when the child misinterpreted the pictures, the experimenter asked the child to show the backside of the



JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.4 (242-400)

Sergio Baauw et al.

picture sequence, on which the verb representing the main action of the sequence was printed. Then, the experimenter mentioned the verb (e.g., “Aha, I see the story is about dressing”), and the child was invited to tell the story again. The child picked up a picture sequence from the pile, and put the picture aside after telling the story. The stories were recorded with a DAT recorder (Dutch experiment)/tape recorder (Spanish experiment), and transcribed afterwards. The experiment consisted of two conditions of seven items each, eliciting either SE- or SELF-anaphors. Since two items of both conditions gave rise to many visual errors in both children and adults, we decided to exclude them from the analysis, limiting ourselves to five items per condition. The items differed from each other in the verb that was used. The following (Dutch/Spanish) verbs were used: wassen/lavar ‘wash’, aankleden/vestir ‘dress’, afdrogen/secar ‘dry’, schminken/pintar ‘make up’ and insmeren/untar ‘put oil on’. All these verbs allow both SE- and SELF-anaphors. The test items were intermingled with 20 filler items.3 The total number of items was 34. In the Dutch experiment the items were divided over two test versions of 17 items each. Each child received one version, which was administered to her in one session of 20 minutes. The adult controls were tested in a similar way, with the exception that the verb representing the action was written underneath the last picture of the three-picture-sequence, in order to avoid visual errors, and no second experimenter, acting as a helper was present. Moreover, the adults received the complete test (34 items) in one session. In the Spanish experiment, both children and adults received the complete test in one session of 30 minutes, with a brief break in between. Before the actual test started, some practice items were administered. In (3) we give an example of a SE-item and a SELF-item, together with a model response. (3) a.

SE-condition Dutch: Een heks en een tovenaar lagen in bed te slapen. Spanish: Una bruja y un mago estaban dormidos en la cama. ‘A witch and a wizard were sleeping in bed.’ Dutch: Toen stapte de tovenaar uit bed. Spanish: Entonces el mago salió de la cama. ‘Then the wizard got out of bed.’

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.5 (400-411)


Dutch: En daarna kleedde de heks zich aan. Spanish: Y luego la bruja se vistió. ‘And then the wich dressed SE.’

b. SELF-condition Dutch: Een vrouw maakte haar kindje wakker. Spanish: Una mujer despertó a su niño. ‘A woman woke up her child.’

Dutch: Toen kleedde ze het kind aan. Spanish: Entonces vistió al niño. ‘Then she dressed the child.’

Dutch: En daarna kleedde ze zichzelf aan. Spanish: Y luego se vistió a sí misma. ‘And then she dressed SELF.’

. Participants 19 Dutch-speaking children were tested, ranging from 5;4 to 6;7, with a mean age of 5;11. The children were tested individually, in a quite room at a primary school in Montfoort, Netherlands. In addition, 13 Dutch-speaking adults were tested. Also, 14



JB[v.20020404] Prn:2/06/2006; 12:01

Sergio Baauw et al.

80

percentages realized



F: LALD4101.tex / p.6 (411-462)

60 children

40

adults

20 0 zich

zichzelf

Figure 1. Zich versus zichzelf

Spanish-speaking children and 4 adults were tested. The children ranged from 5;3 to 6;1, with a mean age of 5;8. The children were tested individually, in a quite room at a primary school in Oviedo, Spain.

. Results . Dutch results In Figure 1 we present the results of the Dutch experiment. The results clearly indicate that Dutch children have considerably more problems with the production of the SEanaphor zich than with SELF-anaphors such as zichzelf. In fact, in the SE-condition they use zich only half as often as the adult controls did. A Mann-Whitney Test shows that children do not perform differently from adults on zichzelf (Z = –1.006, p = 0.314), but do perform differently on zich (Z = –3.265, p = 0.001). Children’s performance on zich differs significantly from their performance on zichzelf (Wilcoxon, Z = –2.524, p = 0.012). This difference was not significant in adults (Wilcoxon, Z = –1.235, p = 0.217). In order to avoid the use of zich, Dutch children frequently use alternative strategies to describe the final picture of a SE-trial, such as body-part construction (4a), and object omission (4b). Occasionally they used SELF-anaphors (4c) or pronouns (4d). (4) a.

De the b. De the c. De the d. De the

jongen boy jongen boy jongen boy jongen boy

waste washed waste. washed waste washed waste washed

zijn buik. his belly

zichzelf / hemself SELF hem. him

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.7 (462-584)


Among these strategies, (4a) is grammatical. (4b) may be grammatical, if no reflexive interpretation is intended. (4c) is ungrammatical if the pronominal reflexive hemzelf is used, and is just infelicitous if zichzelf is used. (4d) is plainly ungrammatical with a reflexive interpretation. Children’s most frequent errors on the SELF-condition involved the use of inalienable possession constructions (sometimes in combination with the possessive eigen ‘own’, like in The boy washed his own belly). It is important to note that Dutch children hardly ever used zich in the SELF-condition.

. Spanish results In Table 1 we present the results on the SE-condition. The total number of child responses was 70, but we excluded those responses that were the result from misinterpretation of the picture sequence. Consequently, the total number of responses that was considered was 58. As can be seen, Spanish children’s production of SE-anaphors is fully adultlike. The few non-targetlike responses that were produced (by both children and adults) mainly involved the use of strong pronoun constructions, with or without SELF-morpheme (Pron(Mismo) responses; see (5)), and a few object omissions. The results on sí mismo, on the other hand, show a different picture. As for the SE-condition, the total number of child responses was 70, but we only considered 45 responses. 25 responses that were the result of a misinterpretation of the picture sequence were excluded.4 As Tables 2a and 2b show, children hardly use sí mismo (a SíMismo response). However, the same table shows that this form is not very frequent in adults either. Table 1. Responses in SE-condition

Children Adults

SE

Pron(mismo)

Rest

48 (83%) 16 (80%)

6 (10%) 2 (10%)

4 (7%) 2 (10%)

Table 2a. Responses in SELF-condition

Children Adults

SíMismo

aPron

aPronMismo

Pron

PronMismo

1 (2%) 5 (25%)

10 (22%) 2 (10%)

3 (7%) 1 (5%)

5 (11%) 3 (15%)

3 (7%) 2 (10%)

Table 2b. Responses in SELF-condition

Children Adults

SeSolo

TambiénSe

Se

Rest

4 (9%) 3 (15%)

2 (4%) 2 (10%)

13 (29%) 0 (0%)

4 (9%) 2 (10%)



JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.8 (584-640)

Sergio Baauw et al.

Instead of sí mismo both children and adults often use alternative constructions, mostly constructions containing strong pronouns, such as those in (5): (5) a.

aPron La niña se the girl SE b. aPronMismo La niña se the girl SE c. Pron La niña se the girl SE d. PronMismo La niña se the girl SE

lavó a ella. washed acc her lavó a ella misma. washed acc her SELF lavó ella. washed she lavó ella misma. washed she SELF

Sometimes they also used constructions with solo ‘alone’, or combine the SE-anaphor with the adverb también ‘too’. (6) a.

SeSolo La niña se lavó (ella) sola. the girl SE washed she alone b. TambiénSe La niña también se lavó. the girl too SE washed

(5a, b) are constructions that involve a combination of a SE anaphor and an accusativemarked strong pronoun. In (5b) the strong pronoun is combined with a SELFmorpheme. These forms are used by some of our adult controls, and are judged grammatical by Torrego (1995). However, some native speakers consider them dubious. In (5c, d) the SE anaphor is combined with a nominative strong pronoun. The strong pronoun, which is combined with a SELF-morpheme in (5d), is in a postverbal, as it seems, right-dislocated position. This type of reflexive construction is also used by our adult controls. Finally, our adult controls sometimes used (6a, b), just like some children. The remaining forms produced by children and adults consist mainly of object omissions. Although children and adults are similar in their preference for pronominal forms, children differ from adults in their frequent use of se, without any pronominal form, as in (7). In producing these utterances children ignore, as it seems, the contrastive contexts of the SELF-experimental condition. (7) Se La niña se lavó. the girl SE washed

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.9 (640-680)

percentages realized


80 60 children 40

adults

20 0 SE

PRON-SELF

Figure 2. SE versus PRON-SELF

Except (6b), the constructions represented by (5) and (6a) can be seen as strongreflexive forms. Like sí mismo they are used in contrastive reflexive contexts and receive the main stress of the sentence. Therefore, these responses should be seen as target-like. In Figure 2 we present children and adults’ target-like performance on the SE- and SELF-conditions. Given the acceptability in the adult-language of pronominal forms, we rename the SELF-condition to PRON-SELF condition. Figure 2 shows that the contrast in performance between Spanish children and adults is less strong than the contrast between Dutch children and adults. As expected, the difference between adults and children was not significant for either SE (MannWhitney, Z = –0.277, p = 0.782) or PRON-SELF (Mann-Whitney, Z = –0.652, p = 0.515). On the other hand, Spanish children’s difference in performance between SE and PRON-SELF was marginally significant (Wilcoxon, Z = –1.926, p = 0.054).

. Discussion Summarizing, the results indicate that Dutch children exhibit problems with the use of the SE-anaphor zich, but not with the SELF-anaphor zichzelf. Spanish children, on the other hand, showed adult-like performance on the SE-anaphor se, but performed relatively poorly on strong reflexives (including SELF-anaphors). Both Spanish children and adults did not often use the SELF-anaphor sí mismo. We will first try to account for children’s performance on Dutch zichzelf and zich, and Spanish children’s performance on se. In Section 5.2 we will present an account on Spanish children’s poor performance on strong reflexives.

. Reflexivity and the production of se, zich and zichzelf In order to account for children’s performance on zich, zichzelf and se, some understanding of the reflexivisation process itself is required. We consider reflexivisation a lexical operation on the theta-grid of verbs that transforms transitive verbs (two-place



JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.10 (680-760)

Sergio Baauw et al.

predicates or relations) into intransitive verbs (one-place predicates or properties) by identifying the internal theta-role with the external theta-role: λx λy (xRy) → λx (xRx). However, transitive predicates cannot be interpreted reflexively across the board. Reinhart and Reuland (1993) propose that the reflexive interpretation of verbs is constrained by Principle B of their alternative Binding Theory. (8) Principle B A reflexive semantic predicate is reflexive-marked. Principle B can be interpreted as an interface filter on arity reduction in the syntax or at the interface (Reuland 2001). It basically states that a transitive verb (relation) cannot be reduced to a reflexive (i.e., intransitive) verb (property), unless the predicate is reflexive-marked. In some languages reflexive-marking takes place by adding a reflexive-marking morpheme to the verb or to an argument of the verb. In Dutch reflexive-marking is dependent on lexical-semantic properties of the verb. Some verbs, such as afdrogen ‘dry’ and wassen ‘wash’ have the correct lexical semantic properties to be reflexive-marked, others, such as haten ‘hate’ or bewonderen ‘admire’ do not (Everaert 1986). Although reflexive verbs are one-place predicates, Dutch requires the object position of these verbs to be occupied, perhaps to check a residual accusative case feature (Reinhart 1996).5 Generation of a pronoun or DP in object position would satisfy this requirement. However, the presence of such an element is problematic; reflexive verbs, being intransitive, can only assign one theta-role. If this role is assigned to the subject position, the direct object is left without a theta-role. However, Dutch has a solution to this problem: zich. As was argued in Section 2, zich establishes an A-Chain with the local subject, as a result of the feature checking relation that exists between these two elements. (9)

Since A-Chains count as one object in syntax, the A-Chain formed between zich and the local subject (Marie – zich) receives the single theta-role of the reflexive verb. Let us now turn to SELF-anaphors. As argued in Section 2, the kind of reflexivity that is encoded by SELF-anaphors is different from the reflexivity of SE-anaphors. In fact, we will argue that verbs that are marked by SELF-anaphors retain their status as two-place predicates. We assume that SELF-anaphors such as zichzelf are DPs that consist of a D zich and an NP -zelf . DP

(10) D zich

NP -zelf

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.11 (760-818)


Zich checks the features of the local subject in the same way as in (2), creating an AChain between zich and the subject. The nominal part -zelf incorporates into the verb and receives the theta-role assigned to the object (Reuland 2001). This means that no arity reduction takes place. But saving the arity of a predicate does not provide it with a reflexive meaning (on the contrary; it remains a two place predicate). The reflexive interpretation is the result of the semantic properties of -zelf. Zelf, which is nominal in nature, has the capacity to “stand for” the referent of the subject (or more precisely, the referent of the chain Jan – zich) by representing it through a “guise”. With a guise we mean a representation of an entity that shares properties with, but is not necessarily identical to this entity (Heim 1982; Jackendoff 1992).6 In formal terms, this means that the predicate in (11) has the form “x H f(x)”, with the referent of f(x) quite close to, but not identified with the referent of x (see also Postma 1997; Lidz 1997).7 In (11) this construal is represented by the identification between J, the mental representation of Jan – zich, and the highly similar, but not identical “guise” J. (11)

Since no arity reduction takes place in the case of SELF-anaphors, they can be used to assign a reflexive-like interpretation to verbs that do not have the appropriate lexical-semantic properties to be reflexive-marked, such as haten ‘hate’ and bewonderen ‘admire’. As shown in (12), the reflexive interpretation of these verbs can only be obtained with the help of SELF-anaphors. (12) a.

Jan John b. Jan John

haatte zichzelf/*zich. hated SELF / SE bewonderde zichzelf. admired SELF / SE

If arity reduction is applied to these verbs, a Principle B-violation will occur. Alternatively, if no arity reduction is applied to haten ‘hate’ or bewonderen ‘admire’, generation of zich in the object position will lead to a Theta Criterion violation; since the AChain Jan – zich will receive one of the two theta-roles, the other theta-role will remain unassigned. Summarizing, the capacity of -zelf to establish a semi-identity with the (subject) antecedent allows, on the one hand, the predicate to have a reflexive meaning, and on the other hand prevents a Principle B violation from occurring at information structure or C-I interface, the level at which binding relations are interpreted.8 The difference between zich and zichzelf with respect to the level at which reflexivity is encoded explains why Dutch children have problems with the use of zich, but not with zichzelf. Ruigendijk et al. (in press) have shown that children exhibit difficulties with the establishment of referential dependencies in narrow syntax. This follows from a more general problem that children have with the use of syntax to structure



JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.12 (818-888)

Sergio Baauw et al.

information, a problem that is the result of their more limited processing capacity (Avrutin 2004). The anaphoric properties of zich are the result of processes that take place in narrow syntax: feature checking and A-Chain formation. Therefore it is predicted to give rise to difficulties. Zichzelf, on the other hand, is identified with its local antecedent through processes that lay outside narrow syntax, and is therefore predicted to be relatively unproblematic. Interestingly, our experimental results are in line with spontaneous production data from Dutch; until the age of four, Dutch children hardly ever use zich. SELFanaphors, on the other hand, develop earlier (Van Kampen, personal communication).9 The results are also in line with Coopmans and Avrutin’s (1999) study on the interpretation of zich and zichzelf in four-to-six year old children. They found that children, who were tested with a Truth Value Judgment Task, accepted the identification of zich with the non-c-commanding antecedent of de princess/de beer ‘the princess/the bear’ 70% of the time in sentences such as (13). (13) a.

[De the b. De the

boerin [naast de princes]] wast zich. farmer’s-wife next-to the princess washes SE olifant [naast de beer]] houdt een paraplu boven zich. elephant next-to the bear holds an umbrella above SE

Children’s performance on similar sentences with zichzelf was far from adultlike, but considerably better.10 Finally, we would like to point out that although children show difficulties with the use of (narrow) syntax, this does not imply that they lack knowledge of the relevant (morpho)syntactic and pragmatic properties of zich. After all, children still produced zich 30.6% of the time in contexts that require it. More importantly, whenever they used zich, it was appropiate; they hardly ever used zich in the SELF-condition.11 Dutch children’s difficulties with the use of zich contrasts with Spanish children’s fully adultlike performance on the experimental condition testing the use of se. This might seem surprising, since both zich and se are claimed to be SE-anaphors. However, we argue that in spite of their superficial similarities se and zich are in fact different elements. Whereas zich is a pronominal element underspecified for [number], se is a reflexive-marking morpheme.12 This means that in Spanish reflexivemarking is not dependent on lexical-semantic properties of the verb, but instead on the presence of se in the functional structure of the sentence (Baauw & Delfitto 2005; Baauw 2002). This explains why the lexical restrictions on zich, as shown in (12), do not apply to Spanish. Se can be combined with practically any verb: (14) a.

Juan John b. Juan John

se SE se SE

odiaba. hated admiraba. admired

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.13 (888-940)


In Spanish any verb can undergo arity reduction, as long as se is added to the structure to reflexive-mark the verb and license its reflexivity. The status of se as a reflexivemarker is also confirmed by (15) and (16). (15) a.

Wassen is gezond. washInf is helathy ‘Washing oneself/someone is healthy’ b. Haten is slecht. hateInf is bad ‘Hating *oneself/someone is bad’

(16) a.

Lavar es saludable. washInf is healthy ‘Washing *oneself/someone is healthy’ b. Odiar es malo. hateInf is bad ‘Hating *oneself/someone is bad’

(15) supports our claim that in Dutch the reflexive interpretation of a verb depends on its lexical-semantic properties. In constructions such as (15a) zich can be absent, but the reflexive interpretation is still possible. In (15b) only the transitive interpretation is possible; unlike wassen ‘wash’, haten ‘hate’ does not have the right lexical-semantic properties to be reflexive-marked. (16) shows that in Spanish the absence of se leads to the loss of the reflexive interpretation. Both lavar ‘wash’ and odiar ‘hate’ can get a reflexive interpretation only by adding se to the structure (lavarse ‘wash-SE’, odiarse ‘hate-SE’). Children’s different performance on Dutch zich and Spanish se can be related in a straightforward fashion to the syntactic differences between these two elements. The crucial difference between zich and se is that zich involves an operation in narrow syntax, namely feature checking and the establishment of an A-Chain between zich and the local subject. Se, on the other hand, does not involve such an operation; it is not a pronominal element, but a reflexive-marking morpheme. Spanish, unlike Dutch, does not require the object position of reflexive verbs to be filled. Stated differently, in Spanish, reflexive constructions such as (14) only involve the lexical/morphological operations of arity reduction and reflexive-marking. These operations do not seem to be problematic to children. In Dutch, on the other hand, in addition to these two operations, reflexive constructions such as (2a) involve the narrow syntactic operation of feature checking/A-Chain formation. As we have argued, this additional syntactic operation often exceeds children’s limited processing abilities (Avrutin 2004).

. Spanish children’s production of strong reflexives The strong reflexive sí mismo is the Spanish counterparts of the Dutch SELF-anaphor zichzelf. We assume that they are DPs that consist of two morphemes, the D sí and the NP mismo. We further assume that sí forms an A-Chain with the local subject, whereas



JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.14 (940-1006)

Sergio Baauw et al.

mismo is identified with the local subject through its guise. One aspect of sí mismo may cast some doubt as to its role in the reflexive interpretation of the predicate; it is always doubled by SE, as in (17). (17) María *(se) lavó a sí misma. Mary SE washed acc SELF However, in Spanish strong object pronouns are always doubled by a clitic. Nonpronominal indirect object DPs, and in some dialects direct object DPs, are optionally doubled by a clitic. It is therefore reasonable to analyse se in (17) as a clitic that obligatorily doubles the “reflexive pronoun” sí mismo.13 In spite of the similarities between sí mismo and zichzelf, the results showed that sí mismo is almost absent in children’s responses on the SELF-condition. This should not come as a surprise, though. For many speakers sí mismo has a rather formal flavor, and is often replaced by alternative constructions, particularly constructions involving strong pronouns, such as those in (5).14 In fact, the adult controls used sí mismo only 25% of the time in the SELF-condition. We assume that children are not frequently exposed to sí mismo and will therefore acquire this form rather late. Spanish children’s overuse of se in constructions that require a strong reflexive cannot be explained by the input. Whereas children gave responses such as (7) 29% of the time, the adult controls never gave such responses in the SELF-condition. At first glance this might indicate a lack of knowledge about the difference between se and strong reflexives. However, this is not very likely. Children used se much more often in the SE-condition than in the SELF-condition. In fact, most of the time Spanish children used strong reflexives in the SELF-condition. Therefore, we will argue that the use of se in the SELF-condition does not indicate any problem specific to the use of strong reflexives, but is instead the result of an independent pragmatic factor. It is well known that children around the age of 5 often use a “deictic” strategy in story-telling tasks (Karmiloff-Smith 1980). Some children from our experiment were found to use such a strategy, in which the different pictures of a story were described as more or less independent events. It is clear that if, as a result of the deictic strategy, the child does not perceive the third picture of a SELF-item as contrastive with the second picture, the use of se to describe the final action follows naturally.15 Two examples of children using this deictic strategy are given in (18) and (19). (18) Fernando, 6;0 Había un señor que estaba en la piscina con una señora. Y la señora le estaba secando, y la señora se secaba. ‘There was a man in the swimming pool with a woman. And the woman was drying him, and the woman was drying SE.’ (19) Miguel, 5;3 Unos caminando en la playa. Luego echándole la madre a la niña crema. Luego la madre echándose crema.

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.15 (1006-1065)


‘Some walking on the beach. Then the mother greasing the girl. Then the mother greasing SE.’ It turns out then that children’s overuse of se in contexts were a strong reflexive is required can be explained as a side effect of children’s inmature ability to construct integrated discourses.

. Conclusions The results of this study indicate (i) that children have early knowledge of the main syntactic and pragmatic properties of SE- and SELF anaphors, (ii) that children have more difficulties with the establishment of referential dependencies that involve narrow syntactic operations than those that draw on operations outside narrow syntax, and (iii) that children have no difficulties with the lexical/morphological operations of reflexivisation and reflexive-marking. The first conclusion is supported by children’s overwelming use of SE- anaphors (Dutch zich, Spanish se) and strong reflexives (Dutch zichzelf, Spanish sí mismo and several constructions involving strong pronouns) in appropriate contexts. The second conclusion is supported by Dutch children’s frequent failure to produce zich, whose anaphoric properties are the by-product of a feature checking operation, and their highly adult-like performance on zichzelf. The third conclusion is supported by Spanish-speaking children’s adultlike performance on se, which unlike Dutch zich is not pronominal, but a reflexive-marking morpheme. Finally, Spanish children’s surprising non-adultlike performance on strong reflexives can be explained as the result of independent factors, namely the low frequency of the SELF-anaphor sí mismo in the input, and children’s difficulties with the construction of integrated discourses.

Acknowledgements This publication is supported by the project Comparative Psycholinguistics, which is funded by the Netherlands organization for Scientific Research (NWO). We are indebted to the children, their parents and teachers at the Openbare Basisschool de Hobbitstee in Montfoort, Netherlands, and the Colegio Público San Pedro de los Arcos in Oviedo, Spain. We also thank Frank Heinen, Pascalle Pennings, Veronique Pijnenburg, Lenny Hamersma and Rozemarijn Morreau for their help in running the experiment and the transcription of the data. Finally, we thank Sergey Avrutin, Ignacio Bosque, Jacqueline van Kampen, Joke de Lange, Nada Vasić and Shalom Zuckerman for their useful comments and suggestions.



JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.16 (1065-1130)

Sergio Baauw et al.

Notes . Reuland (2001) argues that pronouns such as hem ‘him’ and haar ‘her’ can be identified with local antecedents through variable binding or coreference. However, an economy rule blocks this possibility in favour of feature checking. This explains why (i) is ungrammatical with a reflexive reading. (i) *Jani waste hemi . John washed him Importantly, in languages that do not have pronominal elements underspecified for number (SE-anaphors), such as Frisian and Afrikaans, (i) is grammatical; if SE-anaphors are absent, feature checking/A-Chain formation is not an option, hence referential dependencies created outside narrow syntax, such as variable binding, become the most economic option. Obviously, non-local bound-variable or coreference relations between a pronoun and an antecedent are not blocked by economy, since in this case syntactic barriers (such as CP) block the creation of an A-Chain (Chomsky 1986). (ii) Johni thought that Mary liked himi . . In Spanish the SELF-anaphor is inflected for number and gender: sí mismo (sg.masc), sí misma (sg.fem), sí mismos (pl.masc), sí mismas (pl.fem). . The filler items belonged to a study that tested children’s use of definite and indefinite articles and pronouns. . The exclusion affected the total number of se-responses, which would reach almost 50% if all responses were considered. This was due to the fact that children sometimes misinterpreted the second and last picture as involving different types of actions. So, for example, a picture sequence representing a mother that first dressed her child and then herself, was interpreted by one child as representing a mother first dressing her child and then drying herself (misinterpeting the mother’s dress as a towel). If the last picture is interpreted as involving a different type of action, the use of a SE-anaphor becomes an appropriate response. Also, some children tended to “summarize” the picture sequences in one or two statements, ignoring the constrativeness of the second and third picture. One child sometimes described the pictures in the reverse order. Although we tried to disuade children from giving such responses, we were not always successful. Since these factors increased the number of se-responses, we decided to excluded them from our analysis. . Not all languages have this requirement. In English the object position of a reflexive-marked verb remains empty: (i)

John is washing (= John is washing himself).

. The ability of -zelf to be associated with the subject antecedent through a guise explains particular semantic properties of SELF-anaphors, such as their ability to refer to statues, pictures and other reprentations of the referent, as in (ia). SE-anaphors, on the other hand, do not have this ability (ib) (Rooryck & van den Wyngaerd 1997; Jackendoff 1992; Reuland 2001). (i)

Context: Ringo visiting a wax museum passing by a wax statue of himself. a. Plotseling begon Ringo zichzelf uit te kleden. [zichzelf = Ringo/statue] ‘All of a sudden, Ringo started undressing SELF.’

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.17 (1130-1208)


b.

Plotseling begon Ringo zich uit te kleden ‘All of a sudden, Ringo started undressing SE.’

[zich = Ringo/*statue]

. Many languages use body-part reflexives for this purpose. . If -zelf involved complete identity with the local subject, the resulting predicate would be represented as λx (x haatte x), which would amount to a Principle B violation. . Stojanović (2002) did not find any instances of zich and zichzelf in the spontaneous productions of three Dutch children up to 3;10. . The four-year-olds allowed reference of zichzelf to the non-c-commanding antecedent 57% of the time. The 5;0-to-6;5-year olds allowed it 28% of the time when zichzelf was the complement of a locative preposition (13b) and only 10% of the time when it was the argument of a verb (13a). A recent improved experiment showed an overall improvement of children’s performance on zichzelf. However, the children still showed a strong contrast between their performance on zichzelf in argument position and as complement of a locative peposition (Coopmans, Krul, Planting, Vlasveld, & Van Zoelen 2004). . Our account gives rise to a potential problem. We have argued that not only zich but also zichzelf involves a feature checking operation, since it contains a zich-component. If feature checking/chain formation is problematic, children could be expected to show difficulties with the use of SELF-anaphors. We propose that the reason for children’s relatively good performance performance on zichzelf is due to the fact that zich forms a morphological unity with the zelf part of the SELF-anaphor. This entails that whenever -zelf is used, this will facilitate the use of zich. . This does not necessarilly apply to all instantiations of se. Se in Spanish is notoriously multifunctional, and is associated with passive, middle, impersonal and other readings. In fact, in some constructions, such as ECM-constructions, se can be argued to be a pronominal clitic with properties highly similar to Dutch zich (see Baauw 2002). (i)

María se vio bailar en el espejo. Mary SE saw dance in the mirror ‘Mary saw herself dancing in the mirror.’

Also when se doubles sí mismo, we argue that it is not a reflexive-marker. See Section 5.2. . Torrego (1998) shows that sí mismo is the element responsible for the reflexive interpretation of (17), not se. (i)

a. *María Mary b. ? María Mary

se SE se SE

parece strikes parece strikes

inteligente. intelligent inteligente a sí misma. intelligent acc SELF

Although (ib) is marginal in Spanish (see Torrego 1998 for an account), it is much better than (ia). The contrast in grammaticality would not be expected if the reflexivity of (i) relied on se. Moreover, semantically sí mismo behaves similar to zichzelf in Dutch, giving rise to a “guise” interpretation (see Note 6). . Strong reflexive constructions such as those in (5) raise important questions with respect to the nature of se; it is not clear whether se should be considered a reflexive-marker, a clitic pronoun or some other element. Torrego (1995) discussed constructions such as (5a, b). Doubling structures such as (i) have been discussed by Piera (1987) and Fernández Soriano (1999), but so



JB[v.20020404] Prn:2/06/2006; 12:01



F: LALD4101.tex / p.18 (1208-1302)

Sergio Baauw et al.

far we have not come across any study dealing with the reflexive counterparts of (i) (5c, d), nor studies dealing with (6a). We leave the analysis of these structures for future research. (i)

Juan quire lavar él el coche. John wants wash he the car ‘John himself wants to wash his car/Johhn wants to wash his car by himself.’

. Note that this overuse of SE is not predicted for Dutch children, although they also may use a “deictic” strategy. Dutch zich is problematic for children, unlike Spanish se, and is therefore avoided.

References Avrutin, A. (2004). “Optionality in Child and aphasic speech”. Lingue e Linguaggio, 1, 67–89. Baauw, S. (2002). Grammatical Features and the Acquisition of Reference. A Comparative Study of Dutch and Spanish. Outstanding Dissertations in Linguistics. New York/London: Routledge. Baauw, S. & Delfitto, D. (2005). “New views on reflexivity: Delay effects in acquisition, crossmodular principle B and reflexive clitics in Romance”. Probus, 17(2), 145–184. Chien, Y.-C. & Wexler, K. (1990). “Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics”. Language Acquisition, 1, 225–295. Chomsky, N. (1986). Barriers. Cambridge, MA: MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Coopmans, P. & Avrutin, S. (1999). “A syntax-discourse perspective on the acquisition of reflexives in Dutch”. Paper presented at GALA, Potsdam. Coopmans, P., Krul, M., Planting, E., Vlasveld, I., & van Zoelen, A. (2004). “Dissolving a Dutch delay in the acquisition of syntactic and logophoric reflexives”. In A. Brugos, L. Micciulla, & C. E. Smith (Eds.), Proceedings BUCLD, 28, 108–119. Everaert, M. (1986). The Syntax of Reflexivisation. Dordrecht: Foris. Fernández-Soriano, O. (1999). “Los pronombres de la serie tónica”. In I. Bosque & V. Demonte (Eds.), Gramática Descriptiva de la Lengua Española. Madrid: Espasa-Calpe. Heim, I. (1982). The Semantics of Definite and Indefinite Noun Phrases. Doctoral Dissertation, Amherst, University of Massachusetts. Jackendoff, R. (1992). “Mme. Tussaud meets the binding theory.” Natural Language and Linguistic Theory, 10, 1–31. Karmiloff-Smith, A. (1980). “Psychological processes underlying pronominalization and nonpronominalization in children’s connected discourse”. In J. Kreiman & A. G. Ojeda (Eds.), Papers from the Parasession on Pronouns and Anaphora. Chicago, Il: Chicago Linguistic Society. Kraemer, I. (2000). Interpreting Indefinites. An Experimental Study of Children’s Language Comprehension. Doctoral dissertation, Universiteit Utrecht, MPI Series in Psycholinguistics 15. Lidz, J. (1997). “When is a reflexive not a reflexive?: Near reflexivity and condition B”. In K. Kusumoto (Ed.), Proceedings of the North Eastern Linguistic Society, 27. University of Massachusetts, GLSA, Amherst. Piera, C. (1987). “Sobre la estructura de las cláusulas de infinitivo”. In V. Demonte & M. F. Lagunilla (Eds.), Sintaxis de las Lenguas Románicas. Madrid: El Arquero.

JB[v.20020404] Prn:2/06/2006; 12:01

F: LALD4101.tex / p.19 (1302-1354)


Postma, G. (1997). “Logical entailment and the possessive nature of reflexive pronouns”. In H. Bennis, P. Pica, & J. Rooryck (Eds.), Atomism and Binding. Dordrecht: Foris. Reinhart, T. (1996). “Syntactic effects of lexical operations: Reflexives and unaccusatives”. UiL OTS Working Papers. Universiteit Utrecht. Reinhart, T. & Reuland, E. (1993). “Reflexivity”. Linguistic Inquiry, 24, 657–720. Reuland, E. (2001). “Primitives of binding”. Linguistic Inquiry, 32(3), 439–492. Rooryck, J. & van den Wyngaerd, A. (1997). “Self as the other. A minimalist approach to zich and zichzelf in Dutch”. Proceedings of the North Eastern Linguistic Society, 28. University of Massachusetts, GLSA, Amherst. Ruigendijk, E., Baauw, S. Zuckerman, S., Vasić, N., de Lange, J., & Avrutin, S. (in press). “A crosslinguistic study on the interpretation of pronouns by children and agrammatic speakers: Evidence from Dutch, Spanish, and Italian”. In E. Gibson & N. Pearlmutter (Eds.), The Processing and Acquisition of Reference. Stojanović, D. (2002). ““I shave just like man”: Intrinsic and extrinsic constructions in child English”. In S. Burelle & S. Somesfalean (Eds.), Proceedings of the 2002 Annual Canadian Lingustics Association Meeting. Torrego, E. (1995). “From argumental to non-argumental pronouns: Spanish doubled reflexives”. Probus, 7, 221–241. Torrego, E. (1998). The Dependencies of Objects. Linguistic Inquiry Monograph 34. Cambridge, MA: MIT Press. Vallduví, E. (1992). The Informational Component. New York: Garland.



JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.1 (47-110)

On the acquisition of ambiguous Valency-Marking Morphemes Insights from the acquisition of French SE Isabelle Barrière* and Marjorie Perlman Lorch**1 *Johns Hopkins University / **University of London

A modified version of the Maturation Hypothesis that considers the maturation of chains, the impact of ambiguity on acquisition and the use of overt and systematic morphological cues by children is shown to account for the acquisition findings reported on English, Hebrew, Inuktitut, Kiche, Russian and Sesotho. The Barriere Version of the Maturation hypothesis enables us to predict: (a) the order of acquisition of different SE-constructions (including reflexive, reciprocal, anticausative and middle-passive), (b) the order of acquisition of SE and related (short and long) passive constructions and (c) the manifestations of the overgeneralization of argument structure alternation by French-speaking children. These predictions are born out by the analysis of speech production and experimental data collected on 200 children.

.

Introduction

The study reported here focuses on an ambiguous Valency-Marking Morpheme, the French clitic SE that appears in Reflexive and Reciprocal (with animate subjects), Anticausative (no implied agent) and Middle-Passive (implied agent) (with inanimate subjects) constructions. On the basis of cross-linguistic findings on the acquisition of argument structure, a modified version of Borer and Wexler’s (1987) Maturation Hypothesis is proposed and enables us to test three predictions on (a) the order of acquisition of different types of SE-constructions, (b) the manifestations of overgeneralizations of argument structure alternations produced by children and (c) the order of acquisition of SE-constructions and be-passives. These hypotheses are tested using a range of research strategies, including (a) the analyses of a large corpus of speech production (3 diary studies, 2 CHILDES corpora, 2 cross sectional corpora of children aged between 2–4 and 6–7) collected on 200 children and (b) two experimental tasks: a comprehension task (act-out) using existing verbs and a grammaticality judgment

JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.2 (110-177)

Isabelle Barrière and Marjorie Perlman Lorch

task using nonsense verbs that were administered to 2 groups of children, 18 aged 3–4 and 18 aged 5–6, and to 10 adults. The results confirm the order of acquisition and the manifestations of overgeneralizations predicted by the Barrière Version of the Maturation Hypothesis. Current accounts of the acquisition of argument structure – including Bowerman’s (1974) Causativity Hypothesis and the Semantic (Pinker 1989) and the Syntactic (Landau & Gleitman 1985; Gleitman 1989) Bootstrapping hypotheses – are discussed. It is argued that the considerations of ambiguous Valency-Marking Morphemes is essential to hypotheses that claim to be compatible with what is known about the nature of the input and cross-linguistically valid. The findings that emerge from this study are shown to shed light on the interpretation of implied arguments and to necessitate principles on the learnability of syncretic morphemes.

. Hypotheses on the acquisition of argument structure Four hypotheses have attempted to account for children’s acquisition of verb argument structure. According to Bowerman’s (1974) Causativity Hypothesis, children have a strong tendency to add CAUSE to their semantic representation of verbs and therefore to their representation of the event that the verb expresses. Thus, if children are exposed to a new verb, e.g. the nonsense verb gorp, whether they hear it in the tiger gorps or the girl gorps the tiger they will infer that it refers to an event that involves an instigator, i.e. Picture set 1 below, rather than Picture set 2. Bowerman (1974) proposed this Hypothesis to account for children’s overgeneralizations in which some intransitives are used as transitive with a causative meaning (e.g. giggle) and some non-causative transitives are used as causative transitives (e.g. eat used to refer to feed).

©MariaSmall

Picture set 1

©MariaSmall

Picture set 2

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.3 (177-244)

On the acquisition of ambiguous Valency-Marking Morphemes

In contrast, according to the Semantic Bootstrapping Hypothesis (Pinker 1989) children have the ability to (a) focus on salient aspects of events, (b) categorize events into different types such as causative versus non causative and (c) infer the intention of the speaker when interpreting a sentence. Thus when exposed to Picture set 1 above and to the tiger is gorping, depending on the speaker’s focus, children may either interpret it as the tiger is drinking, in which case the thematic role assigned to tiger is Agent or as the tiger is being turned into a lion by the girl in which case the tiger is assigned the role of Theme. According to this Hypothesis, children’s overgeneralizations stem from one of three sources: (a) children’s semantic misrepresentations trigger overextensions; (b) the semantic restrictions on lexical causative alternations fail to be noticed by children; (c) or children entertain adult-like representations of the meaning and thematic roles associated with a verb but under discourse pressure they retrieve an inappropriate verb. This Hypothesis fails to explain the patterns of overgeneralizations produced by Kiche (Pye 1994) and Inuktitut (Allen 1996) speaking children: in these two morphologically ergative languages, the unadultlike use of transitive causative by children involves the production of inappropriate allomorphs rather than the retrieval of inappropriate verbs (see Barrière, Perlman Lorch, & Le Normand 1999, 2000, 2001). Both the Maturation Hypothesis (Borer & Wexler 1987) and the Syntactic Bootstrapping Hypothesis (Landau & Gleitman 1985; Gleitman 1989) differ from the Semantic Boostrapping Hypothesis in that they assume that the child applies an economy principle when hearing a verb and its arguments: a participant is relevant only when it is explicitly expressed. When they hear a sentence such as the tiger gorps, they assume that the only important participant is the tiger. They would therefore map this sentence onto Picture set 2 above, rather than Picture set 1, that requires the postulation of an implicit argument mapping onto a participant. The Maturation Hypothesis (Borer & Wexler 1987) also relies on the distinction between constructions that involve A-chains and those that do not. A-chain constructions result in the movement of the object – the Theme- to the subject position. In English this principle characterizes verbal as opposed to adjectival passives and the class of intransitive verbs – e.g. break, move – that can also be used as transitives (Burzio 1986). The Maturation Hypothesis proposes that before around age 4, children do not have access to A-chains, which explains (a) their lack of understanding of verbal passives (especially reversible ones – e.g. The cow was hit by the horse – given that semantic cues such as animacy cannot be used to interpret the sentences) and (b) their use of intransitives as transitives given their inability to distinguish between those intransitive that are underpinned by A-chains and those that are not (Borer & Wexler 1987). While this Hypothesis explains similar findings that emerge from the acquisition of verb argument structure by children acquiring Hebrew (Borer & Wexler 1987) and Russian (Babyonishev et al. 2001), it fails to explain the early acquisition of passives (much before 4) in Kiche (Pye 1994), Inuktitut (Allen 1996) and Sesotho (Demuth 1989, 1990).



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.4 (244-297)


In English the island was deserted has two possible interpretations: (a) an adjectival one referring to the properties of an endstate – which does not involve A-chains – and a verbal one that refers to an event and implies an agent (Wasow 1977; Fellbaum & ZribiHertz 1989). In contrast, in Inuktitut (Allen 1996) and Sesotho (Demuth 1989, 1990) the counterpart systematically implies the involvement of an agent. Secondly Kiche (Pye 1994) and Inuktitut (Allen 1996) are both morphologically ergative languages that make use of numerous, overt and systematic cues including verbal morphology and case-marking to indicate ‘who is doing what to whom’. Very young children acquiring Kiche are sensitive to the morphological markers of their language (Pye 1983) which they may use in order to interpret constructions involving A-chains before they have access to this linguistic principle, just like young English-speaking children make use of animacy features when interpreting irreversible passives, before they have access to A-chains. The Barrière Modified Version of the Maturation Hypothesis (henceforth BVMH) encompasses the two elements that make up Borer and Wexler’s (1987) Maturation Hypothesis – i.e. the economy principle and late access to A-chains – and also considers (a) the ambiguity of a construction (the more ambiguous the longer it will take to master) and (b) the fact that young children make use of overt and systematic cues when assigning roles to arguments.

. French SE-cliticization and argument structure alternations2 The French SE-clitic is an interesting test case for the BVMH. First, some SEconstructions involve A-chains while others do not. Secondly, French SE-constructions and short passives exhibit different degrees of ambiguity. According to Wehrli (1986), SE-constructions that do not involve A-chains include: Reflexives, Reciprocal, and Inherent SE (i.e. verbs that always require SE cliticizations and cannot be used as transitives). An example of each of them is presented below. (1) La fille se lave The girl SE-3rd-person-singular wash ‘The girl washes herself ’ (2) Les filles se battent The girls SE-3rd-person-singular fight ‘The girls fight with one another’ (3) a.

Marie s’évanouit Marie SE-3rd-person-singular faints ‘Marie faints’ b. *Pierre évanouit Marie *Pierre faints Marie *‘Pierre faints Mary’

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.5 (297-354)


Two other SE-constructions involve A-chains: anticausative SE (e.g. (4b)) and MiddlePassive SE (e.g. (6b)). In both of these constructions the object – to which the role of Theme is assigned – has moved to the subject position. (4) a.

J’ai réchauffé l’eau I have warmed the water ‘I have warmed up the water’ b. L’eau s’est réchauffée The water SE-3rd-person-singular has warmed ‘The water has warmed up’

(5) a.

J’ai refroidi l’eau I have cooled the water ‘I have cooled down the water’ b. L’eau a refroidi The water has cooled ‘The water has cooled down’

(6) a.

Jei mange le chocolat eni buvant du cognac I eat chocolate while drinking cognac ‘I eat chocolate accompanied with cognac’ b. Le chocolat sei buvant mange eni Chocolate SE-3rdi -person-singular eats while Cli drinking du cognac cognac. ‘Chocolate is best eaten accompanied with cognac’

The French Middle-Passive systematically implies the involvement of an animate agent, just like its English counterpart, e.g. This book reads well (Wehrli 1986; Fellbaum & Zribi-Hertz 1989). All French transitive verbs except cognitive verbs can be used in Middle-Passive constructions. French Middle-Passive constructions systematically involve the use of SE-cliticization (Wehrli 1986; Fellbaum & Zribi-Hertz 1989). In contrast, not all French transitive verbs can undergo the anticausative alternations (Zubizarreta 1982, 1985; Wehrli 1986; Fellbaum & Zribi-Hertz 1989). Among those that do, some of them require SE-cliticization (e.g. (4)) while others do not (e.g. (5)). Middle-Passive constructions tend to be used with non-punctual tenses while anticausatives are used with both punctual and non-punctual tenses. Although French short passives, like English short passives (Wasow 1977), may also give rise to ambiguity, such ambiguity characterizes fewer constructions: the derivation of French adjective in -é is less productive than that in -ed in English (Fellbaum & Zribi-Hertz 1989), thus less French past participles and adjectives share the same forms. The BVMH considers (a) the lack of access to A-chains in the early stages of the acquisition of argument structure, (b) the economy principle with respect to the assignment of thematic roles, (c) the fact that children pay attention to Valency-Marking



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.6 (354-484)


Table 1. Order of acquisition of be-passives and Middle-Passives as predicted by the Maturation Hypothesis Before access to A-chains

Adjectival be-passive Anticausative

Access to A-chains

Adjectival be-passive Verbal be-passive Anticausative Middle-Passive

Morphemes and (d) that the more ambiguous a construction, the longer it takes to be mastered. It enables us to formulate three predictions: Prediction I: Reflexive/Reciprocal, Inherent and Anticausative SE are acquired before Middle-Passive SE. Prediction II: Manifestations of overgeneralizations of argument structure alternations include three types of *Increase Valency, four types of *Maintained Valency, and S-cliticization/omission and two types of *Decreased Valency (see details in Barrière, Perlman Lorch, & Le Normand 1999, 2000, 2001). According to the Maturation Hypothesis (Borer & Wexler 1987) lack of access to Achains leads to the predictions outlined in Table 1. In contrast, the BVMH (which also considers the degree of overlap between homophonous constructions) enables us to formulate Prediction III: Prediction III: Short be-passive and Anticausative construction are acquired before the long be-passive, which in turn is acquired before the MiddlePassive.

. Method . Speech production corpora Two research strategies were used: analyses of a large corpus of speech production data and the development and administration of two experimental tasks. The essential features of each of these are briefly outlined below. Table 2 below presents the sources of spontaneous speech production data. Preliminary analyses of these corpora included (a) the identifications of contexts in which SE occurs – i.e. utterances that contain verbs, (b) identification of SE-constructions and (c) a detailed assessment of the productivity of SE according to conservative criteria developed for the purpose of this study. These included the considerations of appropriate versus inappropriate uses of SE, whether it exhibited agreement with the subject (given that lack of agreement is ungrammatical in the adult grammar and that different person and number exhibit different forms), the relation

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.7 (484-512)


Table 2. Sources of speech production data Corpus

Long. versus Recording of Transcription Number of cross-sectional data subjects (CS)

Age range

Grégoire (1939–47)

Long.

Note-taking

0 to 3;6 0 to 3;3

François (1978) Long.

Note-taking

Fondet (1979)

Long.

Leveillé Champaud Le Normand (1986; 1996)

Long. Long. CS

Note-taking (and some recordings) video-rec. video-rec. video-rec.

Mereasse-Polart CS (1969)

audio-rec

Phonetic (1st 2 2 Males years) – then Orthographic Phonetic and 1 Female Orthographic Phonetic and 1 Male Orthographic

0;10 to 3;6 0 to 6

Orthographic 1 Male Orthographic 1 Male Orthographic 360 (1/2 Male, 1/2 Female in each group)

2;1 to 3;3 1;9 to 2;5 2 to 4

Orthographic 72

6–7

of SE with its host (i.e. verbs) – including whether the same verbs appear with and without SE in the same speech sample, the placement of SE before the auxiliary and the verb (grammatical) or between the auxiliary and the verb (ungrammatical), the lexical innovations that involved the omission of SE with inherent SE-verbs and the use of SE with verbs that do not allow it – and self repairs (for details on this aspect of the study, see Barrière, Perlman Lorch, & Le Normand in press). The results show that the majority of children start using SE productively between 2;9 and 3;3 (see Barrière, Perlman Lorch, & Le Normand in press, for details).

. Experimental task The participants in the experimental task included two groups of children: 18 (9 boys and 9 girls) aged between 3 and 4 (i.e. an age at which not all children are expected to have access to A-chains) and 18 (9 boys, 9 girls) between 5 and 6 (i.e. an age at which most children are expected to have access to A-chains) and 10 adults, 5 of whom were parents of the children. Relevant aspects of each experiment are presented in each section below. A grammaticality task using nonsense verbs was developed for the purpose of this study (see details of this task in Barrière & Perlman Lorch 2000 and in press). The visual stimuli depicted animals turning into other animals, including a tiger into a lion, a rat into a rabbit, a horse into a zebra and a dog into a cat. The set of visual



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.8 (512-545)


stimuli depicting the tiger turning into a lion is reproduced below with the verbal stimuli illustrating the uses of one of the nonsense verbs.

©MariaSmall

Picture set 3

Puppet 1: Puppet 2:

la fille elle vile le tigre la fille elle vile le tigre The girl, she gorps the tiger

Puppet 1:

la fille elle le vile The girl, she IT gorps ‘The girl she gorps it’

Puppet 2:

*la fille elle lui vile The girl she TO IT gorps ‘The girl she gorps to it’

©MariaSmall

Picture set 4

Puppet 1:

Puppet 2:

le tigre il se vile The tiger it SE gorps The tiger it gorps itself *le tigre il vile The tiger it gorps

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.9 (545-592)


©MariaSmall

Picture set 5

Puppet 1:

Puppet 2:

les tigres ils se vilent The tigers they SE gorps ‘The tigers they gorp each other’ *les tigres ils vilent The tigers they gorp

©MariaSmall

Picture set 6

Puppet 1:

le tigre il se vile The tiger it SE gorps ‘The tiger it gorps itself ’

Puppet 2:

le tigre il vile The tiger it gorps

©MariaSmall

Picture set 3

Puppet 1: Puppet 2:

la fille elle vile le tigre la fille elle vile le tigre The girl, she gorps the tiger



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.10 (592-659)


©MariaSmall

Picture set 7

Puppet 1:

Puppet 2:

le tigre il se vile The tiger it SE gorps ‘The tiger it gorps itself ’ *le tigre il vile The tiger it gorps

The procedure used for this task is briefly outlined below.

a. Training phase for each picture set and verb For each nonce-verb, each subject was: 1. Exposed to a set of pictures (set 3 above) depicting a causative event and to an auditory stimulus which consisted of transitive constructions involving the use of lexically realized NP arguments; 2. Asked to describe this event: all children were able to do so; 3. Exposed to the same visual stimuli and two different constructions (Indirect Object clitics versus Direct Object clitics) produced by two different puppets; 4. Asked to select and repeat the appropriate construction and provided with feedback.

b. Experimental phase for each set of pictures and each verb For each nonce-verb, each participant was: 5. Exposed in turn to sets of pictures depicting the Reflexive, Reciprocal, Anticausative and Middle-passive versions of this event; 6. Asked to describe the set of pictures: all children demonstrated an appropriate understanding of the event and the role of the participants involved; 7. to SE- versus no SE-verb constructions produced by the puppets and asked to choose and repeat appropriate constructions. The order in which they heard the LE/LUI (training phase 2) and SE/non-SE (experimental phase) constructions and the puppet who produced this utterance were counterbalanced. The mapping between the set of pictures and the nonce verbs and the order in which picture sets were presented were randomized across subjects.

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.11 (659-730)


. Results . Order of acquisition of SE-constructions In the speech production corpora, all the SE-constructions that appear in the speech production samples were categorized (see Barrière, Perlman Lorch, & Le Normand in press, b, for details of the categorization). The data analysis reveals that ambiguous Anticausative/Middle-Passive SE-constructions are more likely to be identified when children’s use of SE-cliticization is productive than when it is not (see the summary of the results presented in Table 3 below). In addition it is only in the cross-sectional corpus collected on 6 year olds that one instance of unambiguous Middle-Passive SE was identified. However, three factors favor an Anticausative (no involvement of an agent) interpretation of the ambiguous Inanimate-SE-V constructions identified in children’s samples: (a) the absence of adverbials implying an agent, (b) the semantic category of the verbs that appear in the constructions that can typically have the Anticausative alternation and (c) the frequent use of punctual tense markers typically associated with Anticausative constructions.

Table 3. Categorization of SE-constructions that pertain to Prediction 1 across corpora (and in relation to SE-productivity) Corpora

Evidence of Ref/Rec Inherent (Ambiguous) Unambiguous productivity Anticausative/ Middle/ (none, weak, Middle Passive Passive strong)

DIARIES

strong

+

+

–

–

strong

+

+

–

–

strong

+

+

–

–

strong

+

+

+

–

strong

+

+

+

–

strong

+

+

+

–

none

+

–

–

–

weak

+

+

+

–

strong

+

+

+

–

(strong)*

+

+

+

+

Grégoire Ch. [from 2;4;31] Grégoire Ed. [from 3;0;30] François [from 3;0] Fondet [from 2;5] CHILDES Leveillé [from 2;1.26] Champaud [2;5] CROSSLN (n = 11) SECTIONAL [2–3;3] LN (n = 35) [2–4] LN (n = 73) [2–4] MP [6–7]



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.12 (730-789)


As predicted by both the Maturation Hypothesis (Borer & Wexler 1987) and the BVMH, Reflexive/Reciprocal and Inherent SE are acquired early and a number of factors indicate that ambiguous Anticausative/Middle-Passive SE-constructions identified in children’s speech samples are best interpreted as Anticausative. In order to test this Hypothesis further experimental evidence was collected. For the experimental procedure, two tasks were developed: a comprehension task and a grammaticality judgment task. The comprehension of Reflexive, Reciprocal and Middle-Passive SE-constructions by children and adults was tested using an act-out task. The context of the experiment encouraged a Middle-Passive interpretation of the Inanimate-SE V constructions. All groups of participants performed significantly above chance when acting out the Reflexive/Reciprocal stimuli and the ambiguous control stimuli (that involved the use of subject clitics and object clitics- rather than lexically realized arguments). In contrast, only the adults performed above chance (at ceiling) when enacting the Middle-Passive SE constructions. This result partly supports prediction I in that it demonstrates that Reflexive/Reciprocal SE is understood before Middle-Passive SE. However, it exhibits two limitations: (a) as it relies on the understanding of existing verbs the results may reflect the participants’ knowledge of these particular lexical items and the constructions in which they appear rather than general grammatical principles and (b) it did not enable us to determine whether children understand Anticausative before Middle-Passive SE. In order to address this limitation a grammaticality task using nonsense verbs was developed for the purpose of this study. An important characteristic of the stimuli that was taken into consideration in the analysis of the results below pertains to the grammaticality of the Anticausative with and without SE-cliticization. In contrast, in the adult grammar the other picture sets can only be mapped onto SE-constructions. Children’s performance was compared to that of adults: adults select SE-constructions 100% of the time for Reflexive, Reciprocal and Middle-Passive events. In contrast, the percentage of selection of SE-cliticization for the Anticausative constructions is lower, which is not suprising given that some existing verbs Anticausative alternation involve SE-cliticization while others do not: the adult participants selected SE 77.5% of the time. The performance of the children was compared to that of the adults. The performance of children was above chance on Reflexive and Reciprocal constructions. In contrast, children’s selection of SE-clitic in the context of the MiddlePassive event was below chance and significant differences were identified between the two groups of children and the 5–6 year olds and the adults. With respect to the Anticausative constructions, both the proportions of selection of SE and of correct replies (SE or no SE) were above chance. The results of this experiment demonstrate that SE-cliticization is relatively systematically matched to Reflexive and Reciprocal events by children, but not to MiddlePassive events. When exposed to Anticausative constructions, most children choose a morphosyntactic construction found in the adult language. Thus it seems that on the

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.13 (789-848)


whole the developmental pattern predicted with respect to the order of acquisition of SE constructions was born out. The analyses of speech production samples revealed that Inherent and Reflexive/Reciprocal SE constructions are produced by very young children. The results of the comprehension task show that at an age when they understand Reflexive and Reciprocal SE, French-speaking children do not understand Middle-Passive SE constructions. The grammaticality task further demonstrates that young children map SE-constructions to Reflexive/Reciprocal and Anticausative events more systematically than to Middle-Passive events. The order of acquisition predicted by the BVMH – Reflexive/Reciprocal, Inherent and Anticausative SE before Middle-Passive SE – is confirmed.

. Manifestations of overgeneralizations of ASA Instances of *Increased Valency, *Maintained Valency and *Decreased Valency were identified in the three sources of speech production data. Examples are briefly described and discussed below (Barrière, Perlman Lorch, & Le Normand 1999, 2000, 2001). The overgeneralizations identified in children’s speech confirm Prediction II in that they exhibit the three types of overgeneralizations which our account predicts. Overgeneralizations were found to occur in children’s speech samples that exhibit strong evidence of SE-productivity. Also, instances of *Increased Valency, *Decreased Valency and *Maintained Valency were identified and in most corpora more than one type of overgeneralizations was identified, although none of them were collected in order to investigate this issue. Therefore, prediction II on overgeneralizations is confirmed. Bowerman’s (1974) Causativity Hypothesis predicts only one type of overgeneralizations involving SE-cliticization, namely the omission of SE with inherent SE verbs that are used as causative transitives. It is not clear how Bowerman (1974) would also account for the other patterns of overgeneralizations involving SE-cliticization. On the basis of the Semantic Bootstrapping Hypothesis, as mentioned in section 5 above, the simultaneous acquisition of Anticausative and Middle-Passive would also be predicted, which is not confirmed by the outcome of the data analysis. In addition, three predictions can be formulated with respect to the children’s ASA overgeneralizations, namely: (a) children’s misrepresentation would trigger overextensions: in this case overgeneralizations of both Anticausative and Middle-Passives (applied to Inherent-SE verbs for instance so that a construction like l’oiseau s’envole ‘the bird INSE-flies away’ i.e. ‘the bird flies away’ would be interpreted as ‘the bird MPSE-flies away’, i.e ‘someone makes the bird fly away’) are expected; (b) the semantic restrictions on the lexical causative alternations fail to be noticed by children: in this case overgeneralizations of Anticausative SE to verbs which do not allow this alternation and transitivization of Inherent SE are expected; (c) or they entertain an adult-like representation but under discourse pressure the inappropriate verb is retrieved: this



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.14 (848-904)


is expected to give rise to lexical substitutions rather than *SE-cliticization and *SEomission. The analysis of overgeneralizations combined with the outcome of the experimental tasks reveal that overgeneralizations to Middle-Passives predicted by the first factor (a) mentioned above are not identified. With respect to the second prediction, while transitivization of Inherent SE are identified, overgeneralizations of Anticausative SE to verbs that do not permit this alternation are not found: all instances of overgeneralizations are found to occur with verbs that express a change of state or location, which the adult grammar allows. Anticausative constructions produced by children are sometimes inappropriately used with or without SE by children, which leads us to the last prediction (c) formulated on the basis of the semantic Bootstrapping Hypothesis (Pinker 1989). As in the case of Kiche (Pye 1994) and Inuktitut (Allen 1996), the overgeneralizations produced by French-speaking children do not involve lexical substitution but *SE-cliticization and *SE-omission.

. Order of acquisition of SE and be-passive The Maturation Hypothesis (Borer & Wexler 1987) and the BVMH make different predictions with respect to the order of acquisition of SE and be-passive constructions. While the Maturation Hypothesis predicts that both French long be-passives and Middle-Passive are acquired simultaneously, after Reflexive/Reciprocal, Anticausative and short be-passive, the BVMH (which also takes the ambiguity of constructions into account) predicts that Reflexive, Reciprocal, Anticausative and short be-passives are acquired before long passives, which in turn are acquired before Middle-Passives. For the speech production corpora, an analysis of all the speech production was carried out (see details reported in Barrière, Perlman Lorch, & Le Normand in press, b). Across corpora, no instance of long passives is reported, except in the oldest age group, and in 3 corpora (Fondet, Leveillé and Champaud) Anticausative SE and short passives are found to be used at a time when (a) no long passive occurs; (b) the linguistic context and/or semantic category of the verbs favour a Anticausative over a Middle-Passive interpretation of constructions involving Inanimate-SE+V. In the Le Normand corpus, all the groups of participants (including those who exhibit no evidence of SE-productivity) are found to produce short passives, including those whose sample does not contain Anticausative constructions. However, the proportions of both Anticausative SE and short passives are much higher in the group who exhibits strong evidence of SE-productivity: 26% increase in the number of subjects who produce Anticausative SE and 31% increase in the number of subjects who produce short passives. The aim of this section was to test Prediction III. Given that only one unambiguous SE-construction and one long passive occur in the corpus collected on the oldest children and that they are not produced by the same child, assuming that all the occurrences of ambiguous Anticausative/Middle-Passives identified are best analysed as Anticausative, the empirical evidence supports the following aspects of Prediction

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.15 (904-962)


III, namely: Short Passives, Anticausative SE are acquired before (reversible) Long be-Passives and Middle-Passives. In order to investigate whether long be-passives are acquired before MiddlePassives the results of the experimental task was studied. The experimental stimuli of the comprehension act-out task mentioned above included irreversible and reversible long be-passives, short be-passives and Middle-Passives. Given that the BVMH encompasses the Maturation Hypothesis, it assumes that children will make use of semantic cues such as animacy when interpreting constructions: thus it predicts that children’s performance on irreversible passives will be significantly better than on reversible passives in which animacy cues do not help booststrap into the adult-like interpretation. We focus here on the results obtained on long reversible be-passives, short be-passives, and Middle-Passives. The performance of the youngest age group is significantly better on short-be passives than on the reversible-be passives and Middle-Passives. On these two constructions, this group’s performance is below chance. Both the Borer and Wexler (1987) Maturation Hypothesis and the BVMH predict these results. In contrast, the performance of the 5 year olds is above chance for both the short be-passive and the long reversible be-passive but below chance on the Middle-Passives. This result is predicted by the BVMH but not by the Maturation Hypothesis, which predicts simultaneous acquisition of both long reversible be-passives and Middle-Passives. Neither Bowerman (1974) Causativity Hypothesis nor Pinker Semantic Bootstrapping Hypothesis would predict the late comprehension of Middle-Passives. Although Borer and Wexler’s (1987) Maturation Hypothesis predicts the late comprehension of Middle-Passives, also it predicts the simultaneous acquisition of long reversible passives and Middle-Passive, which is not borne out by our results. The only hypothesis that predicts the order of acquition of SE and be-passive constructions is the BVMH.

. Discussion The examination of the acquisition of an ambiguous valency-marking morpheme – the clitic SE – by a large population of children investigated using different research strategies, such as analyses of speech production samples, comprehension and grammaticality judgment tasks, cast doubt on the cross-linguistic validity of Bowerman’s (1974) Causativity Hypothesis, Pinker’s (1989) Semantic Boostrapping Hypothesis and Borer and Wexler’s (1987) Maturation Hypothesis. Like the Maturation Hypothesis, the BVMH was formulated within the P&P framework that encompasses the notion of A-chain. This section demonstrates that the consideration of ambiguous Valency-Marking Morpheme should be integrated to hypotheses on the acquisition of ASA, whatever the linguistic theory used. As emphasized by Pinker (1989), Landau and Gleitman (1985), Gleitman (1989), Ingham (1993) and Chiat (2000), two sources of information may be used by the child in order to acquire the Argument Structure of a verb, namely:



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.16 (962-1021)


a. The event that they observe b. The utterance that describes the event. Landau and Gleitman (1985) have demonstrated that the number, nature and distribution of lexically-realized arguments provide a more reliable source of information than the extra-linguistic context in which the utterance is produced. Gleitman (1989: 31) notes: the subcategorization frames, if their semantic values are known,3 can convey important information to the verb learner. This study of the acquisition of the French clitic SE has demonstrated that the child is also faced with Valency-Marking Morphemes which do not alternate with pronominal or nominal expressions to express the same meaning: in this sense Valency-Marking Morphemes are distinct from pronouns. These Valency-Marking Morphemes may be ambiguous in that they may or may not imply a thematic role (Inherent and Anticausative SE versus Middle-Passive SE) and in different constructions they involve distinct Thematic roles. The Syntactic Bootstrapping Hypothesis constitutes an attempt to formulate predictions that consider the nature of the input (Gleitman 1989). Such consideration is essential to all language acquisition hypotheses (Pinker 1989; Gleitman 1989 and Atkinson 1992). Although this attempt may be successful for the Syntactic Bootstrapping Hypothesis when focusing on English, it seems that this is not so when investigating the acquisition of other languages. The Syntactic Bootstrapping Hypothesis focuses on the number, nature and distribution of lexically realized arguments. However, in a functionalist perspective, Du Bois (1985, 1987) has demonstrated that in typologically different languages (including ergative-absolutive and nominative-accusative languages), the maximum number of arguments realized by an NP in corpora of spontaneous speech production is one, even in transitive constructions. The other arguments are expressed using pronominal expressions and Valency-Marking Morphemes. French is no exception (Ashby & Bentivoglio 1993). Thus the Syntactic Bootstrapping Hypothesis is untenable. French SE does not constitute an unusual case: the review of Hebrew (Berman 1994), the re-analyses of Kiche (Pye 1994) and Inuktitut (Allen 1996) and the Argument Structure Alternation overgeneralizations and SE-cliticization reported in other Romance languages demonstrate that the limitation raised here applies to a number of typologically different languages. It is hoped that the limitation mentioned above in addition to the proposal for a more accurate categorization of overgeneralizations of Argument Structure Alternation constitute convincing arguments in favor of a hypothesis that considers morphology in the acquisition of Argument Structure Alternation. In formulating the BVMH a number of assumptions that are relevant to a Morphological Bootstrapping Hypothesis for the acquisition of ASA were adopted, namely: (a) the conceptualization of the young child as a good morphologist; (b) the role of the animacy of the lexically realized arguments in the interpretation of the Valency-

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.17 (1021-1087)


Marking Morphemes; (c) the role of the degree of ambiguity of two constructions on the pattern of acquisition; (d) a default/unmarked assignment of thematic roles. In the generative framework, authors who adopt the Full Competence Hypothesis (Phillips 1995; Wexler 1998; Borer & Rohrbacher 2002) appeal to the notion of morphological complexity to explain the earlier use of morphological (including casemarking, finite verb forms, etc.) marking in morphologically rich languages (Morgan, Barrière, & Woll 2005). The notion of paradigmatic complexity is one of the three factors that according to Phillips has an impact on the emergence of functional categories (p. 10, p. 50). As Phillips (1995) himself admits this is reminiscent of Slobin (1985) and Pinker (1984) (among others) according to whom children “learning more complex inflectional systems do so faster” (Phillips 1995: 10). The within-individual crosslinguistic patterns of morphosyntactic development exhibited by German-English and Latvian-English simultaneous bilingual children are explained in similar terms by Sinka and Schelletter (1998). However Phillips (1995) and Sinka and Schelletter (1998) differ in that the former places paradigmatic complexity at the interface between Performance and Competence: according to this account children’s multi-word combinations in English or other languages does not provide evidence for a lack of knowledge (given the syntactic contexts in which the use of default versus adult-like forms occur) but for the inability to use the full morphological paradigms. In contrast, Sinka and Schelletter (1998) assume that the lack of use provides evidence for the lack of knowledge. Whether paradigmatic complexity is a performance or a competence issue is debated in morphological theory (e.g. Carstairs-McCarthy 1994, among others) although the authors mentioned above do not mention this literature. Secondly, morphological richness and paradigmatic complexity could refer to either or both (a) the number of contrasts in meaning expressed through morphological processes and (b) the number of sub-lexical morphological paradigms for a given lexical class and morph. One of the problems in attempting to disentangle these two factors is that in the languages that have been the focus of acquisition studies, they are typically confounded: Spanish, Latvian, German are morphologically richer than English with respect to both (a) and (b) mentioned above. Thirdly, it is unclear whether Phillips (1995) and Sinka and Schelletter (1998) consider that these factors have the same impact on comprehension and production. Finally, the roles of two other aspects of morphological processes are not discussed by Phillips (1995) and Sinka and Schelletter (1998), namely syncretism and fusion. Fusion is defined as “the morphological phenomenon in which a word consists of several morphemes but in which no one-to-one correspondance can be established between morphemes and morph” (Trask 1993: 112). While the findings to date suggest that the sub-lexical dimension of paradigmatic complexity does not seem to explain the pattern of acquisition of functional categories and Argument Structure Alternation in Kiche, Inuktitut and French, the impact of fusion on the acquisition of Basque versus Georgian morphology is still debated.



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.18 (1087-1136)


When discussing the typically developing infant’s sensitivity to forms, Chiat (2001) and Naigles (2002) refer to specific phonological processes but they do not discuss the role of specific morphological processes. When infants’ sensitivity to morphology has been empirically examined (see Soderstrom 2002 and Santelman & Jusczyk 1998), the focus has been on a morphologically poor language and on morphological processes that exhibit relatively low paradigmatic complexity, while the literature on children’s production of languages that exhibits high paradigmatic complexity (with respect to case-marking and verb finiteness) demonstrates that this factor seems to have a facilitating effect, or at least does not delay acquisition despite the challenge it may place on the form-meaning mapping. Despite the fact that little is understood with respect to the facilitating versus delaying effect of specific morphological processes, the studies of infants’ sensitivity to grammatical morphemes in a morphologically poor language and the timing of acquisition of inflectional morphemes in languages in which many meaning contrasts are expressed on the basis of morphological processes and that exhibit high paradigmatic complexity suggest that children aged 2 pay attention to morphological factors. This consideration is conceptualized here as the first constraint that children apply when exposed to constructions that involved both lexically-realized arguments and ValencyMarking Morphemes. The reliance on morphological cues helps explain the findings that emerge on the relatively early acquisition of be-passive in morphologically ergative languages. A number of findings in typologically different languages confirm that children are better morphologists than semanticists. Between 2 and 3, at the age when children expand their vocabulary, including their verb lexicon, they make use of morphophonological cues, rather than semantic cues, at least in their acquisition of the French gender system (Karmiloff-Smith et al. 1997) and of the Bantu noun classes (Demuth 1992). According to Berman (1982, 1994), between 2 and 3, although children assign the appropriate number of arguments and thematic roles to verbs, they do not systematically use adult-like morphological patterns. It is around 3 that Hebrew-speaking children start to show evidence of productive use of Argument Structure Alternation, especially of non-causative/causative alternations. Between 4 and 5 the productive use of these morphological patterns to mark Argument Structure Alternation give rise to overgeneralizations and lexical innovations (Berman 1993, 1994). Interestingly, Berman (1994) also notes the fact that overgeneralization patterns are not based on semantic cues. The fact that children’s overgeneralizations do not seem to be driven by semantic cues was also noted with respect to the Inuktitut data (Barrière, Perlman Lorch, & Le Normand 2001). In this section, evidence that children are sensitive to morphology from as early as 3 in three different areas, i.e. gender and noun-class assignments and Argument Structure Alternation, and in typologically drastically different languages has been presented. In light of these findings, the results obtained on our analysis of the emergence and productivity of SE-cliticization are not surprising, as they show that around 2;6–3 the vast majority of children’s speech samples exhibit evidence of the productivity of

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.19 (1136-1199)


SE-cliticization. Although the 3 linguistic phenomena mentioned above are distinct, like SE-cliticization they all involve subcategorization features applied to lexical classes. In addition, (a) the acquisition of ASA in morphologically rich languages, (b) the acquisition of tense-marking (as a temporal or aspectual marker) (Behrend, Harris, & Cartwright 1995; Wagner 2002) suggest that Naigles’s (2002: 157) conclusions that “learning form is easy, but learning meaning, and especially linking meanings and forms, is hard” may be rephrased in the case of 2/2;6 children as “learning form is easy, but learning [all] [adult-like] meanings, and especially linking meanings and forms [in an adult-like way], is hard”. One of the assumptions in this study was the role of animacy features in children’s interpretation of constructions, which in the case of irreversible passives enables them to reach an adult interpretation at a time when they do not have access to the principles underlying these constructions. Findings that emerge from different aspects of morphosyntactic development and obtained by researchers with different theoretical approaches have demonstrated that young children are sensitive to animacy (Imedadze & Truite 1992; Forbes & Poulin-Dubois 1997; Dodson & Tomasello 1998; Wagner 2002), a semantic feature tied to volition and control (Corrigan 1988; Pinker 1989; Gropen 1990). The results obtained on the comprehension (act-out) task provide evidence of children’s use of animacy features in the interpretation of SE-constructions: it is only in 3 cases that Inanimate-SE-V constructions were interpreted as Reflexive and Reciprocal. One aspect of the status of animacy that may have been overlooked in the acquisition literature: In the case of SE constructions, children paying attention to the animate features of the subject do not interpret Animate-SE-V in the same way as InanimateSE-V. In doing so they apply a distinction that is found in the adult language. This is not only the case of the interpretation of French-SE constructions. The two uses of open in examples (7a) and (7b) will help illustrate our point: (7) a. John opened b. The door opened In the adult-grammar, the grammatical subject in (7a) is likely to be interpreted as an Agent while in (7b) it would typically be interpreted as a Theme. A young child interpreting these sentences paying attention to the animacy features of the subject will entertain the same interpretations as an adult. The Argument Structure Alternation illustrated by the constructions with open mentioned above is lexically restricted in English while in the cases of both the French Reflexive/Reciprocal and the MiddlePassive, this phenomenon applies to all transitive verbs. There is one pair of constructions for which this distinction does not hold and this is the case of irreversible and reversible passives, both in French and in English. In ergative languages, young children are able to appropriately use (Inuktitut) and understand (Kiche) passives. In these ergative languages, in the constructions (7a) and (7b), different case-markers (and different verbal forms) are used and provided children pay attention to morphological markers; they would not entertain the de-



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.20 (1199-1269)


fault interpretation that English and French-speaking children first entertain. Thus, the semantic feature animacy plays a crucial role in children’s verbal comprehension, and this feature is used as a default strategy when no relevant morphological cues are available. The main difference between Borer and Wexler’s (1987) Maturation Hypothesis and the BVMH lies in the consideration of the degree of ambiguity of constructions that plays a role in the latter but not in the former. First, it is interesting to note that while lexical ambiguity has been considered in the literature on the acquisition of the lexicon (Backsheider & Gelman 1995), it has been much less discussed in its own right in relation to morphological development, although, (a) this phenomenon has contributed to the current debates on children’s use and understanding of tense markers (e.g. Ingham 1998), and (b) it has given rise to research strategies that attempt to distinguish morphs expressed by a single form (Legendre et al. 2002) and (c) it has been assigned contradictory effects. In this section, hypotheses which have made specific developmental predictions on the basis of the relation between forms and meaning are presented and discussed. The relation between form and meaning may be characterized in three ways, namely: a. 1 form = 1 meaning b. > 1 form = 1 meaning c. 1 form > 1 meaning The issues of whether linguistic theory should say anything about morphological paradigms (i.e. whether they belong to the realm of Competence or Performance) and if it does, whether the relations characterized as (b) and (c) are random or constrained (e.g. syncretism is principled in the sense that it tends to apply to certain morphs, for instance or to specific cells in the paradigms) (Carstairs-McCarthy 1994) are debated in morphological theory. In the literature on acquisition, two hypotheses have been proposed. KarmiloffSmith (1979, 1983, 1986) proposed a 2-phase developmental pattern according to which between 3–5, relation (c) above is excluded and between 7–10, it is no longer excluded. With respect to morphosyntactic development, the findings that emerge from this study on French SE and former studies on the acquisition of Romance SE (Gathercole 1990; Jacubowicz 1991) do not support this Hypothesis. According to Jacubowicz (1991), the fact that SE exhibits syncretism with respect to case-marking would explain the earlier acquisition of this clitic, compared to object clitics LE and LUI that do not exhibit such syncretism. The problem with the account proposed by Jacubowicz (1991) is that syncretism does not constitute the only difference between SE and LE/LUI. The Principle of Contrast (Clark 1987, 1988, 1990, 1993): ‘every 2 forms contrast in meaning’ (Clark 1987: 2) is said to guide acquisition. It predicts that in the early stages of development, meaning-form relation characterized as (b) above is excluded. With respect to morphological development there are at least two ways to interpret the Principle of Contrast (Carstairs-McCarthy 1994). According to the first interpretation,

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.21 (1269-1312)


more than 1 form can be interpreted as more than one allophone/allomorph, regardless of the lexical item or lexical class. If the Principle of Contrast is to be interpreted in this way, the acquisition literature reports much evidence that does not support it: the findings that have emerged from both the acquisition of functional categories and ASA in morphologically rich languages demonstrate that (b) does not guide the morphological development of 2 year old children. In addition, this study which documents the acquisition of Anticausative constructions (an instance of (b) above, given that some verbs involve SE-cliticization, while others do not in this construction) before the acquisition of Middle-Passive (which systematically involves SE-cliticization), demonstrates that at least the first interpretation of this principle does not account for the acquisition pattern. The Principle of Contrast Hypothesis can also be captured as integrating the inflectional class membership as part of the meaning – “under tight conditions- namely if it unambiguously identifies the inflection class of the lexemes to which it attaches” (Carstairs-McCarthy 1994: 737). Carstairs-Mc Carthy (1994) does not discuss the application of his proposal to Valency-Marking Morphemes. However, given the importance the author assigns to class membership identifiers, it is unclear how the idiolectal and dialectal variations that French Anticausative exhibits would be dealt with by children: given that the same lexeme may or may not be cliticized SE in this construction depending on one’s dialect or idiolect, it should constitute a major challenge for the child (see discussion in Imedadze & Tuite 1992; Morgan, Barrière, & Woll 2005). The findings that emerge from the data analysis show that unstable Anticausative constructions are acquired before stable Middle-Passive constructions, suggesting that it does not constitute a major challenge for the child. While such idiolectal and dialectal variation with respect to the expression of Argument Structure Alternation, including Valency-Marking Morphemes, has been discussed in linguistic theory, it has not been addressed with respect to hypotheses on the acquisition of ASA and/or the characterization of the input (Lieven 1999; Bowerman 1999). Given (a) what is currently known about the roles of morphological processes in the acquisition of functional categories and subcategorization features and (b) the lack of explanatory power of and empirical support for hypotheses that mainly rely on the characterization of form-meaning relations, the consideration of the degree of ambiguity of constructions is relevant only once the other factors mentioned above and that include sensitivity to morphological patterns and animacy features of the lexically realized arguments do not permit the disambiguation of a construction involving a Valency-Marking Morphemes. This issue is relevant to the issue of positive evidence, the only type of evidence children are supposed to have access to according to Learnability Principles (Wexler & Culicover 1980; Pinker 1989). Given, as mentioned above, the lack of one to one correspondence between utterances and the events they refer to, the higher the degree of ambiguity between two utterances, the longer it will take to gain access to an alternative interpretation, when the two other principles mentioned above do not play a role in the disambiguation process.



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.22 (1312-1379)


The Thematic Inference Principle assumed in this study states that in the process of acquisition, a Thematic role is assumed by a learner only if it can be assigned to an appropriate phrase in the sentence (Borer & Wexler 1987). In a way, this assumption is compatible with the Syntactic Boostrapping Hypothesis (Landau & Gleitman 1985). While the Syntactic Bootstrapping Hypothesis makes predictions on the cues children use when first exposed to a verb and its arguments, both the MH and BVMH assume that children classify verbs into categories that differ from those of adults. Borer and Wexler (1987) convincingly invoke learnability theory to support the necessity of this assumption. In the context of what is known about (a) children’s vocabulary development, (b) the unsystematic mapping between utterances and the events they refer to and (c) the unpredictability of verbs (Fisher et al. 1994; Gillette et al. 1999), the implication of this assumption for our study deserves our attention. According to Fisher et al. (1994) and Gillette et al. (1999), the degree of predictability of nouns is much higher than that of verbs; and according to the Syntactic Bootstrapping Hypothesis, children pay attention to the position, the number and the semantic features of the argument(s). Two examples of SE-constructions are presented below with a nonce-verb, to illustrate the point: (8) a.

Jeanne elle se gabe Jeanne SE-nonce verb b. La voiture elle se gabe The car SE-nonce verb

The adoption of the assumptions outlined above on children’s sensitivity to morphological markers and animacy would predict that on hearing the construction exemplified by (8a), children are likely to interpret the subject as an Agent. In addition, given that the child may have been exposed to gabe + DO, and gabe on its own referring to events involving different participants (as they were in the grammaticality judgment task), and to other similar constructions with the same type of arguments used with a different verb, it is reasonable to assume that the child may gain access to the Reflexive interpretation (see also Borer & Wexler 1987 for a discussion on Reflexives). Findings reported by Dodson and Tomasello (1998) suggest that this assumption is correct: a Russian-speaking child exposed to a transitive construction involving the use of LRAs subsequently spontaneously produced a Reflexive form of this verb. This finding confirms Borer and Wexler (1987) in that it demonstrates the child’s ability to derive Reflexivization from (a) the transitivity of the verb and (b) the animacy of the subject. In contrast, in construction (8b), one of the semantic features that characterizes the lexically realized external argument is its inanimacy. However, unlike in the case of the Reflexive, at least two mappings are possible: with or without underlying agent. The findings that emerge from the experimental task suppport the consequence of the Thematic Inference Principle. The adoption of the most restricted interpretation with respect to the number of Thematic roles is in line with the learnability principle referred to as the Subset Principle discussed in the literature according to which it is

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.23 (1379-1437)


logically necessary to assume that children start with the most constrained possibility (Wexler & Culicover 1980; Pinker 1989). In addition, the consideration of Agent and Themes presented here is compatible with the empirical findings that emerge from Angiolillo and Goldin-Meadow (1982) regarding the existence of these roles in young children’s representations and with the principles derived from the Universal Thematic Assignment Hierarchy (see van Kampen 2002, among others). To summarize the sections above, Morphological Boostrapping is best captured as a hierarchy of principles that are predicted to apply to the acquisition of ValencyMarking Morphemes and according to which: i.

Young children pay attention to morphological cues including case-marking, and Valency-Marking Morphemes, (a) although the meaning they assign to these Valency-Marking Morphemes may not be adult-like in the early stages of acquisition and (b) the sublexical dimension of paradigmatic complexity does not delay the acquisition of Valency-Marking Morphemes; ii. Animacy features of lexically realized arguments in constructions that involve the use of Valency-Marking Morphemes play a role in constraining the assignment of thematic roles to lexically realized arguments and Valency-Marking Morphemes, when factors under (i) (i.e. morphological factors) do not provide systematic cues regarding Thematic Role assignment; iii. The ambiguity of Valency-Marking Morphemes only contributes to the acquisition patterns if factors (i) and (ii) above do not enable to disambiguate a construction; iv. In cases when Valency-Marking Morphemes are ambiguous in that it may or may not imply a Thematic Role, young children apply the Thematic Inference Principle as a default strategy.

Notes . This chapter is based on the first author’s PhD dissertation. The financial support of the Association for French Language studies, Birkbeck College, the British Federation of Women Graduates, the British Council, the Reeve Foundation, the Socrates-Erasmus exchange program and the University of London Research fund is gratefully acknowledged. Special thanks to: Dr. Malcolm Edwards, Birkbeck College for feedback on various drafts, Prof. Annette Karmiloff-Smith, Institute of Child Health for useful suggestions on the design of the experiments, Prof. Jean-Luc Nespoulous and all the members of the laboratory Jacques-Lordat, Université de Toulouse-Le Mirail, for their hospitality, Dr. Marie-Therèse Le Normand for providing access to a large fully transcribed cross-sectional corpus (research grant C.R.I. # 44 009B awarded by INSERM), Ms. Maria Small for drawing the wonderful pictures used in the grammaticality judgment task with nonce words (some of these are reproduced here, i.e. sets 1 to 7), the teachers, parents and children who participated in the experiments, all the colleagues with whom I have discussed some aspects of this work at different venues, including reading groups and conferences, to the people and pets who have welcomed me in their houses during my stays in France, and last, but not



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.24 (1437-1528)


least my nephew Alexis and my niece Esther who kept reminding me there is more to childern’s development than the acquisition of SE-cliticization. . Henceforth ASA. . Auhors’ emphasis.

References Allen, S. E. M. (1996). Aspects of Argument Structure Acquisition in Inuktitut. Amsterdam: John Benjamins. Angiolollo, C. J. & Goldin-Meadow, S. (1982). “Experimental evidence for agent-patient categories in child language”. Journal of Child Language, 9, 627–643. Ashby, W. J. & Bentivoglio, P. (1993). “Preferred argument structure in spoken French and Spanish”. Language Variation and Change, 5, 61–76. Atkinson, M. (1992). Children’s Syntax: An Introduction to Principles and Parameters Theory. Oxford: Blackwell. Babyonishev, M., Ganger, J., Pesetsky, D., & Wexler, K. (2001). “The maturation of grammatical principles: evidence from Russian unaccusatives”. Linguistic Inquiry, 32, 1–44. Backscheider, A. G. & Gelman, S. A. (1995). “Children’s understanding of homonyms”. Journal of Child Language, 22, 107–127. Barrière, I. & Perlman Lorch, M. (2000). “The morphosyntax/semantics interface in children’s grammar: evidence from the acquisition of French reflexive/reciprocal, neuter and middlepassive SE”. In I. Barrière, S. Chiat, G. Morgan, & B. Woll (Eds.), Child Language Seminar 1999 Proceedings – City University Department of Language and Human Communication Science Working Papers, Vol. 1 (pp. 39–49). Barrière, I. & Perlman Lorch, M. (in press). “The acquisition of Argument Structure Alternations by French-speaking children: Evidence from the production of different passive and SE constructions.” Barrière, I. & Perlman Lorch, M. (under review). “New perspectives on semantic and syntactic bootstrapping and on the acquisition of the passive: A view from the French middle.” Barrière, I., Perlman Lorch, M., & Le Normand, M. T. (1999). “On the overgeneralization of the Intransitive/Transitive alternation in children’s speech: A cross-linguistic account with new evidence from French”. The International Journal of Bilingualism, 3(4), 351–362. Barrière, I., Perlman Lorch, M., & Le Normand, M. T. (2000). “Argument structure alternation in French children’s speech”. In M. Perkins & S. Howard (Eds.), New Directions in Language Development and Disorders (pp. 139–148). New York: Kluwer Academic/Plenum Publishers. Barrière, I., Perlman Lorch, M., & Le Normand, M. T. (2001). “La cliticisation en SE dans le langage de l’enfant: surgénéralisation des affixations et des omissions”. In C. Muller (Ed.), Clitiques et cliticisation (pp. 251–265). Paris: Honoré Champion. Barrière, I., Perlman Lorch, M., & Le Normand, M. T. (in press). “The acquisition of Argument Structure Alternations by French-speaking children: Evidence from the production of different passive and SE constructions”. Applied Psycholinguistics. Barrière, I., Lorch, M. P., & Le Normand, M. T. (in press). “On the early production of argument structure alternations: Evidence from the acquisition of SE-cliticization by French-speaking children”. Journal of Child Language.

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.25 (1528-1641)


Behrend, D. A., Harris, L. L., & Cartwright, K. B. (1995). “Morphological cues to verb meaning: Verb inflections and the initial mapping of verb meanings”. Journal of Child Language, 22, 89–106. Berman, R. A. (1982). “Verb-pattern alternation: The interface of morphology, syntax and semantics in Hebrew child language”. Journal of Child Language, 9, 169–191. Berman, R. A. (1993). “Marking of verb transitivity by Hebrew-speaking children”. Journal of Child Language, 20, 641–669. Berman, R. A. (1994). “Developmental perspectives on transitivity: A confluence of cues”. In Y. Levy (Ed.), Other Children, Other Languages: Issues on the Theory of Language Acquisition. Mahwah, NJ: Lawrence Erlbaum. Borer, H. & Rohrbacher, B. (2002). “Minding the absent: Arguments for the Full Competence Hypothesis”. Language Acquisition, 10, 123–175. Borer, H. & Wexler, K. (1987). “The role of maturation in the explanation of linguistic development”. In T. Roeper & E. Williams (Eds.), Parameter Setting and Language Acquisition (pp. 123–169). Dordrecht: Reidel. Bowerman, M. (1974). “Learning the structure of causative verbs: A study in the relationship of cognitive, semantic and syntactic development”. Papers and Reports on Child Language Development, 8, 142–178. Bowerman, M. (1999). “Symposium on argument structure”. Oral Communication. 8th Conference of the International Association for the Study of Child Language, San Sebastián. Burzio, L. (1986). Italian Syntax: A Government-Binding Approach. Reidel: Dordrecht. Carstairs-Mc Carthy. (1994). “Inflection classes, gender and the Principle of Contrast”. Language, 70, 737–788. Chiat, S. (2000). Understanding Children with Language Problems. Cambridge: Cambridge University Press. Chiat, S. (2001). “Mapping theories of developmental language impairment: premises, predictions and evidence”. Language and Cognitive Processes, 16, 113–142. Clark, E. V. (1987). “The principle of contrast: A constraint on language acquisition”. In B. MacWhinney (Ed.), Mechanisms of Language Acquisition (pp. 1–33). Mahwah, NJ: Lawrence Erlbaum. Clark, E. V. (1988). “On the logic of contrast”. Journal of Child Language, 15, 317–335. Clark, E. V. (1990). “On the pragmatics of contrast”. Journal of Child Language, 17, 417–431. Clark, E. V. (1993). The Lexicon in Acquisition. Cambridge: Cambridge University Press. Corrigan, R. (1988). “Children’s identification of actors and patients in prototypical and nonprototypical sentence types”. Cognitive Development, 3, 285–297. Demuth, K. (1989). “Maturation and the acquisition of the Sesotho passive”. Language, 65(1), 56–80. Demuth, K. (1990). “Subject, topic and Sesotho passive”. Journal of Child Language, 17, 67–84. Demuth, K. (1992). “The acquisition of Sesotho”. In D. I. Slobin (Ed.), The Cross-Linguistic Study of Language Acquisition (pp. 557–638). Mahwah, NJ: Lawrence Erlbaum. Dodson, K. & Tomasello, M. (1998). “Acquiring the transitive constructions in English: The role of animacy and pronouns”. Journal of Child Language, 25, 605–622. DuBois, J. W. (1985). “Competing motivations”. In J. Haiman (Ed.), Iconicity and Syntax (pp. 343–365). Amsterdam: John Benjamins. DuBois, J. W. (1987). “The discourse basis of ergativity”. Language, 63, 805–855. Fellbaum, C. & Zribi-Hertz, A. (1989). The Middle Construction in French and English: A Comparative Study of its Syntax and Semantics. Bloomington: Indiana University Linguistics Club Publications.



JB[v.20020404] Prn:2/06/2006; 12:08



F: LALD4102.tex / p.26 (1641-1752)


Fisher, C., Hall, D. G., Rakowitz, S., & Gleitman, L. (1994). “When is it better to receive than to give: Syntactic and conceptual constraints on vocabulary growth”. Lingua, 92, 333–375. Fondet, C. (1979). Un enfant apprend à parler: Récit et analyses d’un apprentissage de la langue maternelle de la naissance à six ans. Dijon: les Presses de l’Imprimerie Universitaire. Forbes, J. N. & Poulin-Dubois, D. (1997). “Representational change in young children’s understanding of familiar verb meaning”. Journal of Child Language, 24, 389–406. François, F. (1978). “Syntaxe, lexique et contraintes formelles”. In F. François, D. François, E. Sabeau-Jouannet, & M. Sourdot (Eds.), La Syntaxe de l’Enfant avant 5 ans. Paris: Larousse. Gathercole, V. C. (1990). “Multifunctional SE; course of development in Spanish children”. Papers and Reports in Child Language Development, 29, 46–53. Gillette, J., Gleitman, H. Gleitman, L., & Lederer, A. (1999). “Human simulations of vocabulary learning”. Cognition, 73, 135–176. Gleitman, L. R. (1989 “The structural sources of verb meaning”. Papers and Reports on Child Language Development, 28, 1–48. Grégoire, A. (1939). L’apprentissage du langage, Vol. 1. Paris: Droz. Grégoire, A. (1947). L’apprentissage du langage, Vol. 2. Paris: Droz. Gropen, J. (1990). “Bootstrapping from agency: Early notions of agency according to Pinker’s semantic bootstrapping hypothesis”. Papers and Reports in Child Language Development, 29, 148–155. Imedadze, N. & Tuite, K. (1992). “The acquisition of Georgian”. In D. I. Slobin (Ed.), The Cross-linguistic-study of Language Acquisition, Vol. 3 (pp. 39–109). Mahwah, NJ: Lawrence Erlbaum. Ingham, R. (1993). “Critical influences on the acquisition of verb transitivity”. In D. Messer & G. Turner (Eds.), Critical Influences on Child Language Acquisition and Development (pp. 121–139). New York: St Martin’s Press. Ingham, R. (1998). “Tense without agreement in early clause structure”. Language Acquisition, 7, 51–81. Jacubowicz, C. (1991). “L’acquisition des anaphores et des pronoms lexicaux en français”. In J. Guéron & J. Y. Pollock (Eds.), Grammaire Générative et Syntaxe Comparée. Paris: CNRS. Karmiloff-Smith, A. (1979 “Micro- and macro-developmental changes in language acquisition and representational systems”. Cognitive Science, 3, 91–118. Karmiloff-Smith, A. (1983). “Language acquisition as a problem-solving process”. Papers and Reports on Child Language Development, 22. Karmiloff-Smith, A. (1986). “From meta-processes to conscious access: Evidence from children’s metalinguistic and repair data”. Cognition, 23, 95–147. Karmiloff-Smith, A., Grant, J., Berthoud, I., Davies, M., Howlin, P., & Udwin, O. (1997). “Language and Williams Syndrome: How intact is “intact”?”. Child Development, 68, 246– 262. Landau, B. & Gleitman, L. R. (1985). Language and Experience: Evidence from the Blind Child. Cambridge, MA: Harvard University Press. Legendre, G., Hagstrom, P., Vainikka, A., & Todorova, M. (2002). “Partial constraint ordering in Child French syntax”. Language Acquisition, 10, 189–227. Lieven, E. (1999). “Symposiun on Argument structure”. Oral Communication. 8th Conference of the International Association for the Study of Child Language, San Sebastián. Morgan, G., Barrière, I., & Woll, B. (2005, in press). “The roles of modality and typology in the acquisition of verb morphology in British Sign Language”. First Language. Naigles, L. R. (2002). “Form is easy, meaning is hard: resolving a paradox in early child language”. Cognition, 86, 157–199.

JB[v.20020404] Prn:2/06/2006; 12:08

F: LALD4102.tex / p.27 (1752-1838)


Phillips, C. (1995). “Syntax at age two: Cross-linguistic differences”. MIT Working Papers in Linguistics, 26, 37–93. Pinker, S. (1984). Language Learnability and Language Development. Cambridge: Harvard University Press. Pinker, S. (1989). Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, MA: MIT press. Pye, C. (1983). “Mayan telegraphese: Intonational determinants of inflectional development in Quiché Mayan”. Language, 59(3), 583–604. Pye, C. (1994). “A cross-linguistic approach to the causative alternation”. In Y. Levy (Ed.), Other Children, Other Languages: Issues in the Theory of Language Acquisition. Mahwah, NJ: Lawrence Erlbaum. Santelman, L. & Jusczyck, P. (1998). “Sensitivity to discontinuous dependencies in language learners: Evidence for limitations in processing space”. Cognition, 69, 105–134. Sinka, I. & Schelletter, C. (1998). “Morphosyntactic development in bilingual children”. International Journal of Bilingualism, 2, 301–326. Slobin, D. I. (1985). The Cross-linguistic Study of Language Acquisition. Mahwah, NJ: Lawrence Erlbaum. Soderstrom, M. (2002). The Acquisition of Inflection Morphology in Early Perceptual Knowledge of Syntax. Unpublished PhD dissertation, Johns Hopkins University. Trask, R. L. (1993). A Dictionary of Grammatical Terms in Linguistics. London: Routledge. van Kampen, J. (2002). “Language-specific bootstraps for UG principles”. Paper presented at BOOT-LA workshop, Indiana University, Bloomington. Wagner, L. (2002). “Understanding completion entailments in the absence of agency cues”. Journal of Child Language, 29, 109–125. Wasow, T. (1977). “Transformations and the lexicon”. In P. Culicover, T. Wasow, & A. Akmajian (Eds.), Formal Syntax (pp. 327–360). New York: Academic press. Wehrli, E. (1986). “On some properties of the French clitic SE”. In H. Borer (Ed.), The Syntax of Pronominal Clitics, Syntax and Semantics, Vol. 19. Orlando: Harcourt Brace Jovanovich. Wexler, K. (1998). “Very early parameter setting and the unique checking constraint: A new explanation of the optional infinitive Stage”. Lingua, 106, 23–76. Wexler, K. & Culicover, P. (1980). Formal Principles of Language Acquisition. Cambridge, MA: MIT press. Zubizarreta, M. L. (1982). On the Relationship of the Lexicon to Syntax. Dordrecht: Foris. Zubizarreta, M. L. (1985). “The relation between morphophonology and morphosyntax: The case of Romance causative”. Linguistic Inquiry, 16, 247–289.



JB[v.20020404] Prn:5/06/2006; 14:52

F: LALD4103.tex / p.1 (47-112)

Definite and bare noun contrasts in child Catalan Anna Gavarró*, Ana Teresa Pérez-Leroux** and Thomas Roeper# *Universitat Autònoma de Barcelona / **University of Toronto / # University of Massachusetts Amherst

Thirty-three children aged 3–5 were tested for comprehension of definite and bare noun (BN) contrasts in Catalan, a language with bare nouns in direct object but not in subject position. Catalan objects have the same semantics as English subjects: BNs map to generic and definites into specific referents. Global semantic parameters cannot explain the contrast in this type of language. The test of Catalan sentences equivalent to She needs shoes/She needs the shoes showed no discrimination at 3 and sensitivity by 4 . We argue under an economy approach that BNs have a default generic reading that can be shifted into an ‘instance-of-a-kind’ interpretation. Acquisition is driven by a principle of semantic contrast that restricts the initial overextension of BNs to specific readings.

.

Introduction

The purpose of this paper is to examine the acquisition of bare noun/DP contrasts in direct object contexts in child Catalan.1 The recent literature on the acquisition of determiners across languages indicates that they emerge early in languages where determiner insertion is generalized, as is the case with Italian, Catalan, and Greek (Chierchia et al. 1999; Guasti & Gavarró 2003; Marinis 2003). The literature debates whether differences in the rate of acquisition depend on prosodic factors (Lleó & Demuth 1999) or on syntactic/semantic factors (Chierchia et al. 1999). Despite robust findings that emergence of determiners is early in those languages, there is little evidence that evaluates the extent of the children’s mastery of the target meanings. Here we present an experiment on child comprehension of Catalan definite DP vs. bare nouns in direct object contexts. We argue that a theory of economy in language favors minimal structures as acquisition defaults. One natural hypothesis is that these minimal structures have automatic semantic mappings, and thus certain meanings should not require extensive language experience. If minimal default structures produce a de-

JB[v.20020404] Prn:5/06/2006; 14:52



F: LALD4103.tex / p.2 (112-183)

Anna Gavarró et al.

fault interpretation directly from UG, we expect children to immediately recognize the relevant meaning. The paper proceeds as follows. Section 2 characterizes the syntax and semantics of bare nouns in adult grammar. In Section 3 we present our basic assumptions and discuss some relevant acquisition literature on the development of NPs and DPs. Section 4 is devoted to the presentation of an original experiment on the interpretation of bare nouns and definite DPs in object position in child Catalan, and Section 5 presents our conclusions.

. The syntax and semantics of bare nouns . Bare objects Compare the distribution of definite/bare noun contrasts, which holds across all positions in English but exhibits a subject/object asymmetry in Catalan and other Romance languages, as shown in (1). (1)

English Subjects Tigers live in India. (Gen) The tigers live in India. (Spec) Objects I want shoes. (Gen) I want the shoes. (Spec)

Catalan (and Spanish, etc.) *Tigres viuen a l’Índia. Els tigres viuen a lÍndia. (Gen/Spec) Busco sabates. (Gen) Busco les sabates. (Spec)

The semantic literature (Carlson 1977; Diesing 1992; and many others) has extensively debated the nature of bare plural (BP) subjects. These have existential or kind reference, depending on the meaning of the predicate. With characterizing predicates such as in (2a), BPs take on a kind interpretation, referring to the class as a whole. In other contexts, the interpretation is existential: (2b) asserts that there are some firemen who happen to be available. (2) a. Firemen are nice. (Gen) b. Firemen are available. (∃) The case of bare objects is less clear. We note that a similar alternation in sense operates in the object domain as shown in (3a–c). With an extensional verb in (3a) BPs are clearly existential: the implication of (3a) is that there are some carrots that were eaten by me. Under an intensional verb such as that in (3b), there is no existential assertion: it is entirely possible that the object does not exist, as shown in (3c). (3) a. I ate carrots. b. I need shoes. c. I need unicorns/flying cars/etc. The semantic literature fails to single out this reading of the bare object. We will refer to these bare plural objects of intensional verbs as generic objects in the understanding that

JB[v.20020404] Prn:5/06/2006; 14:52

F: LALD4103.tex / p.3 (183-244)

Definite and bare noun contrasts in child Catalan

the reading is not a reference to a kind, but instead reference to a potential instance of the kind (see below the discussion on Chierchia’s formalization of this notion).2 What (3) and (4) have in common is the massification/lack of quantization of the concept. Carrots are treated as substance in (3a), and we see what is treated as a coerced mass reading in Gomeshi’s (2003) discussion of examples such as (4). (4) I saw lion today.

. Genericity across languages: the status of bare nouns There are good reasons to believe that the linguistic expression of generic reference represents a challenging learning problem for the child. Children seem to acquire generic knowledge quickly and efficiently from a single instance (see Gelman & Tardif 1998; Pappas & Gelman 1998; Hollander et al. 2002; Gelman, Star, & Flukes 2003). But how does the child figure out that a given statement has generic reference? No single morpheme directly encodes genericity in any language, and there is a wide range of crosslinguistic variation in how generic meanings are represented (Dayal 2004). The child’s task is simplified if one considers bare nominals to have a special status, as non-quantized elements that by default receive generic reference. If, as pointed out by Chierchia (1998) and Dayal (2004), the distribution and interpretation of bare nouns in a language depends on the existing lexical material, children should be able to restrict the semantic extension of the default bare noun with the acquisition of the specific elements in the functional skeleton of the DP. But precisely how does parametrization in the DP domain relate to BN interpretation? The canonical view of the syntax/semantics interface given in (5) posits that syntactic categories are mapped onto semantic types, determining denotations. Semantic type mismatches are resolved via a highly constrained universal inventory of available type-shifting operations. (5) NP → <e,t>, DP → e, or GQ (generalized quantifiers). Chierchia (1998) proposed a refinement of this canonical mapping to account for typological variation with regards to the distribution and interpretation of bare nouns. He proposed that the range of possibilities for the mapping of nominal projections is parametrized, and that languages vary depending on whether nominals are mapped directly as semantic arguments (type e), or semantic predicates (type <e,t>). His nominal mapping parameter (NMP) differentiates between three language types. In Type I, that of no-determiner languages (i.e., Chinese and other classifier languages which lack determiners and numbers) NPs are [+arg, –pred], that is they map directly as arguments. In Type III, there is generalized use of determiners, as NPs are [–arg, +pred], and the DP layer is required for the nominal to be a semantic argument; the Romance languages belong to this group. In Type II, to which the Germanic languages belong, bare nouns (NPs) can function as either arguments or predicates [+arg, +pred].



JB[v.20020404] Prn:5/06/2006; 14:52



F: LALD4103.tex / p.4 (244-299)


The [+arg, +pred] setting of the parameter allows the possibility of bare nouns in English, but individual lexical items may be associated with one feature value or the other. This provides a strictly lexical treatment for what can appear as bare noun in English: only mass nouns can appear in subject position in singular form (6a), and count nouns must be pluralized (6b). As Chierchia notes, the mass/count distinction is lexically arbitrary: change vs. coins have the same denotation but behave differently. (6) a. Gold is rare. b. Books are valuable. c. *Book is valuable. In Chierchia (1998), singular mass nouns are allowed in English because they can directly map as e, if they are marked in the lexicon as +arg. Bare plurals are possible because count nouns are of type <e,t>. This can be shifted up to a kind interpretation by the free type-shifting operator up (∩ ), which lifts the type to e. The plural feature is required because kinds are functions from world into pluralities, which are the sum of [typical] instances of the kind (Chierchia 1998: 349). Count singular subjects are ungrammatical because the type-shifter up (∩ ) applied to a singular count noun such as dog will not yield a proper kind (∩ dog), because the dog-kind includes the set of dogs, not an atom from that set. This approach follows Carlson (1977) in adopting a monosemous view of English bare plurals. Under this view, the generic reading is the fundamental part of the sense of bare nouns, and the existential reading illustrated above in (2b) is a derived sense. This sense results from a semantic operation, Derived Kind Predication (DKP), which introduces local existential closure. This yields the ‘instance of a kind’ reading. The operation of DKP becomes obvious when a species name produced along the word kind is used with a specific reading, as in (7a, b) (from Chierchia 1998). It is what underlies the specific plural indefinite reading of (7c). (7) a. That kind of animal is ruining my garden. b. [Pointing at the picture of a lion in a biology book] That kind is what you saw this morning at the zoo. c. He bought books.

. Bare nouns in Catalan Recall that bare NPs are generally disallowed in Catalan. We leave aside bare singular NPs in object position in lexicalised constructions (see Espinal 2001), and bare NPs in special registers (Guasti et al. 2004). According to Chierchia, bare subjects are excluded because Romance nouns are <e,t>, and NPs can only be arguments if D is projected. (8) a. *Dinosaures es van extingir. dinosaurs refl past-pl extinct ‘Dinosaurs became extinct.’

JB[v.20020404] Prn:5/06/2006; 14:52

F: LALD4103.tex / p.5 (299-356)


b. *Tigres són mamífers. tigers are mammals ‘Tigers are mammals.’ Bare NPs appear only in restricted conditions: in governed positions as objects of P and as direct objects, and as postverbal subjects of unaccusatives. The rationale for this object/subject asymmetry is provided by the putative presence of a null D head (δ), as shown in (10), which has the same semantic contribution as the type-shifter (∩ ) of raising the semantic type to e. (9) a.

Vindrem en tren. come-fut-3pl in train ‘We’ll come by train.’ b. Comprarem llibres. buy-fut-3pl books ‘We’ll buy books.’ c. Cauen rocs. fall-3pl rocks ‘Rocks fall.’

(10) [DP δ [NP llibres] ‘books’ Despite the differences evident in the subject domain, bare objects in Catalan share properties with both bare objects and bare subjects in English. In parallel to English, Catalan objects of intensional verbs display a contrast between generic and non-generic readings: bare NPs receive a generic object reading, overt full DPs a specific reading. (11) a.

Busco sabates. look-for-1s shoes ‘I look for shoes (generic).’ b. Busco les sabates. look-for-1s the shoes ‘I look for the shoes (specific).’

It is interesting that the same restrictions described for English subjects apply to both English objects and Catalan objects. The restrictions are both lexical (mass/count) and morphosyntactic (singular vs. plural). Bare nouns must either be plural or mass singular; see (6) and (12) for English, and (11) and (13) for Catalan. (12) a. *I am looking for shoe. b. I am looking for flour. (13) a. *Busco sabata. look-for-1s shoe



JB[v.20020404] Prn:5/06/2006; 14:52



F: LALD4103.tex / p.6 (356-418)


b. Busco farina. look-for-1s flour ‘I look for flour.’ These phenomena are not given a unified account under the system proposed by Chierchia. In Catalan bare objects are possible by virtue of the availability of the null D. In English the introduction of a bare mass noun is possible for lexical reasons, but in Catalan it is made possible by a type-shifting null determiner. The subject/object asymmetry in Romance receives two explanations, one syntactic (ECP regulated distribution of δ), and one semantic (definites, which are mapped as ι are allowed to intentionalize – become kind-referring –, in appropriate semantic contexts: individual level predicates and as subjects of generic sentences). The analysis is therefore problematic in two respects: it misses a generalization that can be established crosslinguistically, and creates a redundancy in the analysis Romance.

. Assumptions on acquisition We assume both the uniformity principle (Chomsky 1999: 2), which posits underlying uniformity in languages, with variety across languages restricted to easily detectable properties of utterances, and economy of acquisition, which proposes that children start with the minimal structure projected in a domain, and that the child avoids early commitment to functional representation (Lebeaux 1988; Vainikka 1993/1994; Roeper 1992, 1998). This leads us to expect that the initial mapping of nominals is always generic, and that children’s seeming specificity is actually the result of pragmatically induced typeshifting. Chierchia et al. (1999) arrive at a comparable conclusion.3 We also adopt the formal requirement of semantic contrast (i.e., Semantic Uniqueness): each element gets a semantic mapping, and formal elements do not fully overlap in features, although they may overlap in meaning (Chierchia 1998). These assumptions in the domain of the acquisition of NP/DP have several consequences: i.

The initial projection is NP. All initial representations of nominal reference in child language have NP type properties. ii. Bare nouns will always be primarily mapped as generic, and allowed to function semantically as either predicates or arguments. iii. Children will rely on the (local) interaction with other lexical options to determine the interpretation of definites. If the minimal child grammar allows mapping to a specific interpretation given bare NP syntax, thus meeting all the referential needs of the child, why should the child ever work towards a mature grammar? First, the child identifies articles in the adult input, and must assimilate them into the grammar. Second, more explicit (i.e., more mature) grammars are more efficient, in a pragmatic sense. The child starts out with

JB[v.20020404] Prn:5/06/2006; 14:52

F: LALD4103.tex / p.7 (418-469)


G2

G1

Graphic 1.

generic, non-quantized reference that can pragmatically adapt to other senses by typeshifting and later becomes capable of distinguishing generic from specific reference, by acquiring the language specific forms that introduce disambiguation. The null, most economical form retains the default meaning. Our approach is consistent with Chierchia’s answer to the question of acquisition, according to which a universal mechanism takes the form of an ‘elsewhere’ condition: the universal form is bled by language-specific mechanisms. More overt, specified morphology will subtract meanings from the default expressed by the zero form: once the child relates the presence of the definite article to specificity, that meaning is blocked from the bare form by virtue of uniqueness. Using this approach, development can be described as the sequence of stages G1 and G2: (14) a.

G1 stage – All NPs are initially mapped as generic – Specificity is initially supplied by universal mechanisms in the syntax/semantics interface [type-shifting] b. G2 stage – Overt morphemes bleed the reference space of the bare noun default

The size of G1 varies across languages, and G2 always contains G1 (see Roeper 1999; Lebeaux 1988, 2000). Thus, for development, semantic uniqueness has the following consequences: i.

Bare nouns initially have generic reference and, where retained, remain available for generic reference. Alternative possibilities are given in the languagespecific D system. ii. The introduction of overt determiners subtracts the existential and definite readings from the bare form. Other possibilities in nominal reference are introduced



JB[v.20020404] Prn:5/06/2006; 14:52



F: LALD4103.tex / p.8 (469-532)


explicitly as well, once the rest of the functional system comes in: demonstrative, partitive senses, etc. iii. The system will continue to readjust by virtue of semantic uniqueness until the distribution of the full inventory of categories involved in nominal reference is mapped. One specific feature of our proposal is that the distribution of bare/full DPs is regulated locally (at the syntactic level), rather than macro-parametrically (at the language level). This departs from the standard treatments that propose a language-level parameter for nominal reference in Romance vs. Germanic (Longobardi 1994; Chierchia 1998).

. The emergence of DP in child grammar Two kinds of research have been conducted on the emergence of DPs in child grammar: research on the emergence of articles based on data from corpus studies, and research on the interpretation of bare/DP contrasts.

.. Evidence for crosslinguistic variation in the emergence of D Research based on corpus studies has recently shown that definite articles cease to be omitted earlier in French and Italian than in English and Swedish (Chierchia et al. 1999), and earlier in Catalan than in Dutch (Guasti et al. 2004). While phonological factors may play a role in boosting article production in Romance (Lleó & Demuth 1999), these do not suffice to explain the difference in the achievement of the target system between Romance and Germanic. Despite phonological differences, Romance articles are produced as in the target earlier than in Germanic; Guasti et al. (2004) show that omission is homogeneous in Catalan, Dutch and Italian at a first stage of development (the stage at which children use up to a 100 words), but at the second stage (101 to 200 words) omission is already very low in both Catalan and Italian while it remains over 50% in Dutch. Child Greek seems to pattern with Romance, with early and extensive use of determiners (Marinis 2003). These overall differences between Romance and Germanic are consistent with NMP, showing an advantage in the emergence of D in Romance, but, given that these results come from overall use, they do not provide direct evidence about the nature of the semantic mappings. Therefore experimental testing is needed to evaluate whether the semantic mapping is indeed as in the target when syntactic production is virtually adult-like. This is the kind of evidence sought in the experiments detailed below. .. Evidence of knowledge of semantic mapping of nominals Sensitivity to bare/DP contrasts has been investigated by Burns and Soja (1997) by looking at the idiomatic features of the bare noun construction: (15) a. She is at church. (Institutional reading) b. She is at a church. (Location reading)

JB[v.20020404] Prn:5/06/2006; 14:52

F: LALD4103.tex / p.9 (532-624)


Children showed sensitivity to the contrast in (15): 4 year olds demonstrated substantial discrimination, selecting determiners 20% of the time for the institutional scenario versus 75% of the time for the location scenario. Pérez-Leroux and Roeper (1999) provide evidence that young children know the locality properties of bare nominal idioms in contrast with the properties of DPs illustrated in (16): (16) a.

Everyone went to his home? (referential/bound variable) Everyone went home? (bound variable only) b. Everybody hoped the Lion King would go home and he did. (local interpretation=LK only) Everybody hoped the Lion King would go to his home and he did. (long distance or local interpretation)

For (16a), children in that study gave high proportion of quantifier-bound readings for the BN (75–100%), but a much lower one with the possessor determiner (around 25%). For the contexts in (16b), non-local binding of bare home occurred in less than 10% of cases, while with possessors it occurred close to 40%. Children clearly understood the local, anaphoric nature of the bare nominal idiom, and used the structural contrast to build the idiomatic readings. In sum, the early detection of the contrasts in (15) and (16) shows fine-grained semantic distinctions to be present from early on in development. In what follows we present an experiment designed to detect the interpretation given by Catalan speaking children to BNs and full DPs.

. An experiment on the contrast between bare nouns and definite DPs . Methods We designed some short stories to evaluate the children’s sensitivity to the NP/DP distinction and their respective generic/definite interpretations in object position. Each story had two characters, one searching for something specific, and another with general needs. Children heard four stories of this type, with pictures left visible during the narration. Each story was followed by a question of the form Who needs X?, as in (17). Determiner type (Ø or definite) was counterbalanced per story across children. Figure 1 illustrates a sample story. (17) a.

Qui necessita les sabates? who needs the shoes? [Specific answer: little Cinderella] b. Qui necessita sabates? who needs shoes? [Generic answer: the older sister]



JB[v.20020404] Prn:5/06/2006; 14:52



F: LALD4103.tex / p.10 (624-655)


. . .there is a little girl who wants to dress up as Cinderella and needs special shoes for that (point to little girl, princess-like). Her older sister (point to taller, barefoot girl) helped her look for them in the trunk, and finds them.

Figure 1. Sample story: Little girl is looking for the Cinderella shoes

Sessions were recorded, and data entered and coded as generic or specific. We had no instances of non-responses or anomalous responses, although many of the children offered their responses by pointing rather than by verbally identifying the referent.

. Participants Thirty-three Catalan speaking children aged 2;7–5;10, and 10 adults participated in the study. Children were divided into three age groups of comparable size. Most children were tested in schools, although some were recruited outside a school setting. (18) age group 3 year olds 4 year olds 5 year olds adults

# 11 11 12 10

age range 2;7–3;9 3;11–4;9 4;10–5;10

mean 3;4 4;4 5;3

. Results The results, in Figure 2, show the proportion of generic answers per group per condition. These figures include both the exclusive generic answers as in (17b), and the inclusive responses, in which speakers gave ‘both’ as an answer. Our intuition is that some speakers formulate the generic interpretation contrastively or exclusively (i.e., ‘X

JB[v.20020404] Prn:5/06/2006; 14:52

F: LALD4103.tex / p.11 (655-720)


Propotion Generic 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

3 years olds 4 years olds 5 years olds Adults

Definite

Bare

Figure 2. Proportion of generic answers per determiner in Catalan

only needs shoes in general’, not an actual set of shoes), or inclusively (with a broad inference, where if X needs specific shoes, it is true that X needs shoes in general). The results reveal clear patterns of development for both the definite and the bare noun condition. Consistently with our expectations, adults were uniformly willing to give generic responses to the bare noun phrase, and unwilling to give generic answers when the question was presented with a definite determiner. On average, the younger children gave around 1/3 of generic responses, and were generally unable to distinguish between definite and bare. All of the children’s generic responses were exclusive. Around the age of four, generic responses have doubled for BNs, resulting in a sizeable level of discrimination, although many children still accept the generic response for the definite condition. The older children, like adults, did not give generic responses to the definite, although they still fall quite short from the ceiling performance present in adults. The statistical analysis supports these observations: age was found to be significant (F 3,40 = 3.252, p = .0316), and both determiner and the determiner x age interaction were highly significant (F 1,40 = 55.572, p < .0001, and F 3,40 = 13.712, p 3 and used the respective structures productively. Figure 2 shows the results for objects as discussed in Schmitz and Müller (2004). The first French and Italian group (FR1 and IT1) differ with respect to the type of realisation and omission. Both French groups used more DPs in object position than Table 1. Monolingual Italian children in elicited production task (Tiedemann 1999), adapted from Schmitz and Müller (2004) Number

AGE

8

1;9,13–3;0,6

7

2;6,27–2;9,20

MLU Group 1 1.18–3.36 Group 2 3.40–6.25

Base3

S.D. (MLU)

56–246

0.38–1.62

54–320

1.5–4.68



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.4 (264-299)

Natascha Müller et al.

Table 2. Monolingual French children (Jakubowicz et al. 1996, 1997), adapted from Schmitz and Müller (2004) Number

AGE

MLU

5

2;0,13–2;7,3

7

2;3,22–2;7,0

Group 1 2.92–3.15 Group 2 3.22–4.95

Base

S.D.

87–195

1.53–1.84

105–451

1.68–2.92

their Italian peers and the French adults in Figure 1. The Italian group 1 (IT1) omitted objects twice as much as the French children in group 1 (FR1), a percentage which outranks the adults in Figure 1 4 times. Turning to clitic realisations, Italian children in both groups use more clitics than their French peers, and group 2 reaches the numbers shown by the adults. Summarizing, the possibilities of realizing or omitting the object revealed numerous differences between the two languages. This is unexpected considering the adults in Figure 1 and the acquisition literature. The differences between the languages are the following: – – –

MLU-mapped French (FR1, FR2) children use more DPs in object position than their Italian peers (IT1, IT2) MLU-mapped Italian children (IT1) omit more objects than their French peers (FR1) MLU-mapped Italian children (IT1, IT2) use more object clitics than their French peers (FR1, FR2)

The analysis of the cross-sectional studies we reported was based on the spontaneous speech of the recording sessions exclusively. In order to further investigate whether there is a difference between the two languages, we designed an elicited production 100

omission DP clitic strong

80 60 40 20 0 FR1

FR2

IT1

IT2

Figure 2. Realisation (in %) of object position in French and Italian child language, taken from Schmitz and Müller (2004)

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.5 (299-413)

Null arguments in monolingual children

task for monolingual French and Italian children. The test was conducted in France and Italy by the members of our research team.

. The cross-sectional study . The test This part of the paper presents the results of a production task designed to elicit pronominal verb arguments. Pictures with different types of actions serve as basis for questions of the interviewer who introduces the participants of the action to the discourse, e.g.: (1) Cosa fa il topo colla calzetta? ‘What does the mouse with-the sock?’ expected answer: Se la mette. Self it (she) puts ‘She puts it on her foot’ possible child answer: Si mette Ø/la calzetta. Self (she puts) Ø/the sock ‘She puts Ø/the sock’ Among the possible answers the option “Ø” is a target-deviant object omission, while the use of a DP object is a grammatically target-like but pragmatically deviant answer. To elicit the answers, we used cards with colored pictures taken from children’s books. The cards were prepared with small metal pieces so that they would stick to a fishing rod which was prepared with a magnet. The cards had to be fished from an aquarium (“fishing game”). The procedure ensures that the order of the pictures is always at random and cannot be remembered. In the following sections we present the results.

. Participants The test was first conducted with monolingual adult speakers (5 speakers of French and 4 speakers of Italian) to see whether it produces the desired results, namely that there is no difference between French and Italian speakers with respect to the realisation of the object position. The monolingual French and Italian children were recorded in France (kindergardens in Valenciennes and Niort) and Italy (kindergardens in Rome and Treviso). The test was designed for children under the age of 3 years. In Italy, we had the occasion to include children above this age, which allowed us to establish a third group of children which will, however, not be the focus of the present discussion. The following Tables 3 and 4 indicate the children’s age and MLU. Since the MLU was established based on



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.6 (413-441)


the spontaneous interaction part of the recordings, which was sometimes no longer than 10 minutes, MLU values are not always reliable. The children were divided into two groups according to a qualitative criterion in combination with a quantitative criterion. The omission of determiners in obligatory contexts was chosen as a qualitative criterion in order to distinguish two groups of children, a first and a second group, because determiners are among the earliest functional categories in the speech of children acquiring a Romance language. However, some children did not omit determiners but showed a weak performance in responding to the test questions. We therefore decided to use a second, quantitative criterion, namely the number of responses. If the performance in the test was weak, although the children had reached an adult-like performance with respect to determiner omissions, we decided to include these children in group 1. This results in the following group division: The first group of Italian and French children had a determiner omission rate of more than 10% in obligatory contexts; furthermore these children answered 1–5 questions. The second group was adult-like with respect to determiner omissions and answered 6–10 questions. Only some Italian children gave more responses (11–13). We decided on the basis of this quantitative criterion that they formed a third group. Tables 3 and 4 give a survey of the children, age and MLU. Group 1 French (10 children) ranged between 27 and 36 months in age (mean age 32 months, MLU 2.6), the Italian group 1 (6 children) between 25 and 34 months in age (mean age 25 months, Table 3. Italian children Name

AGE

MLU

FRA ROM CAR MAT BEA LUD MAR C

2;0,23 2;1,18 2;3,29 2;5,9 2;6,6 2;9,22

1.6 1.6 1.6 2.0 2.4 1.9

MAT GIU A SWA MAT

3;1,25 2;9,8 3;5,27 3;6,27

2.5 2.4 3.8 3.6

STE FRA OSP MAR L PAT MAR ALE GIU B VIR

3;5,17 3;6,2 3;6,18 3;7,13 3;11,4 3;3,22 3;0,10 3;1,1

4.5 3.4 3.6 3.6 3.4 2.7 2.2 2.7

No. of answers

D-omission (%)

5 2 5 3 1 2

20 50 47 0 47 0

9 8 8 10

5 2 0 4

12 12 13 12 13 11 12 13

0 0 2 0 0 0 5 5

Group 1

Group 2

Group 3

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.7 (441-534)


Table 4. French children Name

AGE

MLU

CHA M MAT KEV TOM B PHI TRI* AGA◦ THÉ N MIC PAU

2;3,0 2;5,9 2;6,27 2;9,30 2;3,22 2;8,0 2;9,24 2;10,20 2;11,16 2;11,21

3.0 3.0 1.9 1.8 2.5 – 4.5 2.5 1.5 3.0

MÉD SIM THÉ V◦ EVA ALE LUN* QUE B TOM R CHA V SER CHA*

2;6,29 2;7,23 2;11,1 2;6,19 2;8,4 2;11,14 2;11,19 2;11,23 3;0,7 3;1,14 3;1,15

2.3 2.9 3.8 4.8 4.7 – 3.7 2.3 3.5 2.2 –

No. of answers

D-omission (%)

5 4 4 3 3 1 1 5 4 5

25 26 16 43 0 0 0 14 11 4

7 8 9 9 10 9 8 6 9 9 7

4 3 3 0 1 0 2 0 2 0 1

Group 1

Group 2

* no spontaneous part

◦

MLU not representative

MLU 1.9). The second French group includes 11 children with an age range of 31–37 months (mean age 34 months, mean MLU 3.3), the second Italian group was constituted of 4 children with an age range of 33–43 months (mean age 39 months, MLU 3.1). The third Italian group had 8 children with an age range of 36–45 months (mean age 41 months, MLU 3.3). We now turn to the results of our analysis.

. Results Figure 3 shows the adults’ performance when the answer to the test question required an object. The results are expected and correspond to what we have said about the realisation of the object position in French and Italian adult grammar, namely that there are no differences between the two languages. There were no differences between the languages with respect to the realisation of the object position. In both languages, 51% of the objects were DPs and 46% were clitics. There were no omissions in Italian. In French, omissions amounted to 5%. This supports Schmitz and Müller’s (2004) findings with respect to spontaneous interaction, where they found a slightly higher amount of object omissions in French. Furthermore, the test elicited



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.8 (534-567)

Natascha Müller et al. 100

omission DP clitic strong

80 60 40 20 0 FR

IT

Figure 3. Realisation (in %) of direct objects in French and Italian adults

100

Omissions Realizations

80 60 40 20 0

IT1

FR1

IT2

FR2

IT3

Figure 4. Object omissions and realisations in all groups (percentages)

clitics, and was therefore successful. Figure 3 also shows that the test disfavoured the use of strong pronouns. Turning to the children, we will start with an overview of the results of all groups. We will then go into more detail with respect to each group. Figure 4 shows object omissions and realisations and reveals that the first and second French group do not differ. That is, the rate of omission and realisation does not improve with growing age, while there is a clear difference between the first and the second Italian group. The third Italian group differs only slightly from the second one. The overall omission rate of the second French group (25%) differs noticeably from the adult omission rate (9%) while the omission rate of the third Italian group (12%) is close to the target-value of 6%. In other words, the development in the French children is far from being completed.

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.9 (567-646)


In the next sections realisations and omissions of objects are distinguished in terms of different types: (1) reflexive and (2) non-reflexive clitics, (3) lexical objects (DP) and (4) strong object pronouns. In order to control the statistical relevance of the described differences, we conducted a non-parametric test, namely the Mann-Whitney-U-test (see Bortz 1999). This test is particularly suited for the analysis of small groups without normal distribution. This test involves a ranking of the observed values of each group member and calculates the difference of the rank means (U value) to which error probabilities (p values) are associated. We apply the usual significance levels: statistical relevance is given when p ≤ 0.05, weak significance when p ≤ 0.10. The first application of this test concerns the overall omissions and realisations of group 1 and 2 in the two languages: It turned out that in both group 1 and 2 the differences between the French and the Italian children are not significant (U = 60 in group 1, U = 44 in group 2, p > 0.05). We counter-checked this result by performing a chi-square test: The obtained values confirmed the result of the Mann-Whitney-U-test (Group 1: df = 1, p = 0.3537, group 2: df = 1, p = 0.3711). The lack of statistical significance may be attributed to the larger sample size in the French groups compared to the Italian groups, while, at the same time, the overall sample size is rather small. Still, we did observe different learner strategies. We discuss them in Section 4. In the following subsections, we will present the results in more detail.

.. Group 1 This section presents the results from the first groups of French and Italian children in more detail. Figure 5 shows that the Italian children omit objects very frequently (in 60% of all cases), while French children tend to realize objects by using DPs, omitting them only half as many times as the Italian children. Only the French group 1 already uses reflexive and non-reflexive clitics, although at a low rate of 20%, while adults use clitics in 40% of the respective contexts. The Italian children of group 1 do not yet produce clitics. Both French and Italian children do not use strong pronouns at all. The application of the Mann-Whitney-U-Test to the different types of omitted or realized elements (mainly DPs, reflexives and clitics) of this group does not reveal statistical relevance (U = 60 in all cases, p > 0.05). .. Group 2 Figure 6 shows that the French children of group 2 still omit objects in more than 20% of the obligatory contexts while the Italian children of group 2 omit them less frequently (rate below 20%, but not yet target-like). The French children continue to realize objects mostly by using DPs (50%), while they produce object clitics only in 20% of the relevant contexts. The Italian children use twice as many non-reflexive object clitics as the French children while the latter realize slightly more reflexive clitics than the Italian children. Strong pronouns are used in no more than 5% of all contexts, similarly to the adult data.



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.10 (646-685)

Natascha Müller et al. 100 IT

80

FR

60 40 20 0 Omission

DP

Reflexive

Clitic

Strong

Figure 5. Group 1 omissions and realisation types (percentages)

100 IT

80

FR 60 40 20 0 Omission

DP

Reflexive

Clitic

Strong

Figure 6. Group 2 omissions and realisations (percentages)

The application of the Mann-Whitney-U-Test to the different types of omitted or realized elements (mainly DPs, reflexives and clitics) of group 2 does not reveal statistical relevance (U = 44 in all cases, p > 0.05).

.. Italian group 3 Figure 7 shows that the children of the third group still omit objects in 12% of all obligatory contexts. The most frequent type of realisation is DP but the amount of reflexive and non-reflexive clitics together is target-like with 40%. Similar to the adults and the children’s group 1 and 2 the use of strong object pronouns is disfavored. The application of the Mann-Whitney-U-Test to the different types of omitted or realized elements DPs, reflexives and clitics) did not yield statistically significant results when comparing this group with the second Italian group (U = 32 in all cases, p > 0.05).

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.11 (685-719)

Null arguments in monolingual children 100 IT 80 60 40 20 0 Omission

DP

Reflexive

Clitic

Strong

Figure 7. Italian group 3 omissions and realisations (percentages)

. Discussion The main result for the younger groups of French and Italian children is that they behave differently with respect to the object domain: Whereas French children tend to repeat the direct object DP in contexts where a clitic would have been obligatory, Italian children omit the object altogether. This observation supports our hypothesis that the developmental stage during which children do not use object clitics should be described differently in French and Italian. The important finding for the second groups is that Italian children use more object clitics than their French peers. Furthermore, the differences between the two languages have become smaller. This is mainly due to the fact that the French group 2 “stagnates”. The percentages of the third Italian group indicate that even this group does not mirror the adults’ behaviour with respect to omissions. We will start by discussing the particular behavior of the first groups and address the question of whether empty objects are licensed and identified in the same way in French and Italian children. Intuitively speaking, we would like to defend the view that Italian children use a syntactic device in order to identify empty objects, similar to adult Brazilian Portuguese, whereas French children have recourse to a pragmatic device, as in languages like Japanese. Recently, Zushi (2003) has developed a theory of licensing null arguments. In essence, the approach tries to eliminate the difference between the pro-drop and the topic-drop parameter by assuming that the presence/absence of null arguments can be derived from particular properties of T. In her approach, all parametric differences can be reduced to morphological differences. Comparing English, Italian and Japanese, she argues that the occurrence of empty subjects in finite clauses is a consequence of the lexical, rather than the functional nature of T, which is the case in Italian and Japanese, but not in English. In English, a non-null-subject language, the EPP requires overt subjects in all finite clauses. Within minimalism, it is argued that T has an EPP feature that is obligatorily



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.12 (719-778)


eliminated before Spell-Out. Elimination can either be achieved by movement of a DP to the Spec of TP or by inserting an expletive through Merge. As Schmitz and Müller (2004) have argued for French subject clitics, nominative clitics may have the function of checking off the EPP feature in order to enforce the Spell-Out of a set of phi-features. In languages like English and French, a structure with pro in Spec of TP is excluded, since pro has no phonological content and “thus its movement to Spec of TP is unable to yield phonological contribution. Therefore, pro is not allowed in languages like English” (Zushi 2003: 586).4 For Italian, a (subject) pro-drop language, Zushi (2003) assumes that T is assigned an EPP feature that must be checked off before Spell-Out as well. The EPP feature of T is eliminated by agreement morphemes, which in turn makes Spec of TP unnecessary. Thus, the EPP feature of T can be checked off without movement or inserting of a DP into Spec of TP. The verbal complex V+v moves to T in Italian as well, which brings the agreement morpheme along with it. “This yields a configuration in which the agreement morpheme enters into a checking relation with the EPP feature of T via Head-to-Head agreement. Due to Head-to-Head agreement, the EPP feature of T is eliminated, and thus no other elements need to move to its Spec” (Zushi 2003: 587ff.). If the EPP feature can be checked off in Italian without moving or inserting a DP into Spec of TP it follows that overt DPs need not move to Spec of TP. Indeed, overt subjects in Italian appear in pre- or postverbal position.5 Finally, Zushi discusses so-called topic-drop languages like Japanese. For languages like Japanese which do not exhibit overt agreement morphology, it is assumed that functional categories with interpretable features that are relevant at LF (like T) are present in Japanese, but these categories do not induce Spec-Head agreement. If T has no Spec and the language lacks overt agreement morphology on the verb, T has no features which have to be checked off. Therefore, pro can freely occur in Spec of vP. Since nominal elements in Japanese have no features themselves to check off, it is reasonable to assume that we are dealing with a minimum pro in Japanese, a pro which lacks functional projections altogether. Thus, the language specific licensing options of Zushi’s approach to null subjects can be summarized as follows (Zushi 2003: 590): 1. Agreement morphemes play no role in feature checking 2. Agreement morphemes play a role in feature checking 3. No feature checking Null subjects are not allowed under the properties of T in (1), which represents the English case. When pro occurs in the Spec of TP, the Spec of the functional category will not be visible at PF, in other words, no phonological effect is produced and the derivation crashes. When pro occurs in Spec of vP, the EPP feature of T fails to be eliminated and the derivation will be cancelled. In the two other types of languages (2), Italian, and (3), Japanese, pro is allowed, but it is not responsible for checking off the EPP feature of T. In Italian, agreement morphemes have the ability to check the EPP feature. In Japanese, no checking takes place since T is more like a lexical category in that it does not have any features that induce Spec-Head agreement.

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.13 (778-836)


With respect to the identification of null arguments, Zushi suggests with Shlonsky (1997) that the external syntax of pro differs in the languages which allow for the empty pronominal. – – –

DP-pro requires AGR endowed with a person feature as its identifier ΦP-pro requires AGR endowed with Φ-feature, Number Bare-NP-pro does not require AGR to be identified, (base-generated) zero-topic binds null argument

The different structures of these pro-types are illustrated in (2): (2) Typology of pro’s DP D [person]

FP

FP

F [number]

NP

F

pro

pro

NP

NP pro

Italian subject pro is of the DP-type, a referential pro whose content is recovered through agreement. A non-referential pro lacks a referential D (a person feature is not specified). Its content is not recovered through agreement, but rather through binding by a discourse-identified topic (Huang 1984). The content of the ΦP-type of pro should be recoverable through number agreement. Müller, Crysmann and Kaiser (1996) have suggested an analysis of French object omissions in monolingual and bilingual children of the type discussed for adult Japanese. They relate two findings, namely object omissions in child French and the lack of constructions related to the C-system. Müller et al. (1996) assume that a structure like (3b) or (4b) underlies the children’s object drop constructions, in which PRO is adjoined to IP (for a detailed discussion of an analysis in terms of an IP-adjoined PRO see Müller et al. 1996). Structure (4b) can account for the fact that the children omit the object and the subject in the same construction, thus exhibiting multiple argument drop as in Chinese. (3) a.

Ivar répare ‘Ivar repares it.’ b. [IP PROj [IP Ivar répare tj ] (object-drop)

(4) a.

Verse ‘I pour it in.’ b. [IP PROj [IP PROi [IP ti Verse tj ] (multiple argument drop)

Moreover, they suggest that in the early stages of acquisition all children use a pragmatic strategy to license the empty element (PRO) via discourse (see Schaeffer 1997



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.14 (836-950)


and Hoekstra & Hyams 1998 for discourse related mechanisms in other domains of early child grammar). In what follows we would like to refine the approach by Müller et al. (1996) for French and make a new suggestion for Italian child language based on Zushi’s account. In essence, we will argue that Italian children do not assume the structure as proposed by Müller et al. for French object omissions. In fact, basing our analysis on Zushi’s licensing possibilities of empty pro’s, there is a huge difference between adult French and Italian with respect to agreement in the object domain which might play an important role in acquisition. One of the differences between French and Italian in the object domain is that French hardly exhibits audible past participle agreement, whereas Italian displays audible gender and number agreement. Past participle agreement is a phenomenon of written, but not of spoken French. Tanase (1976) reports for French that out of a total of 8000 French verbs, 7400 (92.5%) belong to the -er conjugation where agreement is not audible. Past participles of the -ir conjugation are similar in this respect and amount to 350 verbs (4.4%). Verbs of the third conjugation class amount to 232. And within this group, verbs with a variable past participle represent 42%, i.e. 99 verbs. With respect to the total of 8000 verbs, only 1.24% are verbs with variable past participles. However, these verbs belong to the most frequently used verbs in French (see Hunnius 1980). In spoken French, past participle agreement is often neglected, i.e. the masculine form is used in contexts where a feminine form would be required in written French. It is therefore quite plausible that French children learn past participle agreement lately in their development, probably at school (Müller 1999). Notice however that there is also a qualitative difference between past participle agreement in spoken French and Italian: If audible at all, French past participles have gender agreement, whereas in Italian, both number and gender agreement is marked overtly. Turning to child language, Schaeffer (1990), Antinucci and Miller (1976), Borer and Wexler (1992) have pointed out that agreement morphology on past participles is an extremely early phenomenon in the development of Italian by monolingual children and they discussed an acquisition stage (age span 2;0–2;6) which is characterized by an overgeneralization of object agreement morphology. The children they studied use agreement morphology on past participles also in cases where the adult language would exclude it, namely in cases of a phonetically realized postverbal object: ho chiusa la porta ‘(I) have closed the door’. In what follows, we will show that the Italian children of our study, even the very young children, used the correct agreement morphology on past participles: the list in (5) includes examples for gender and/or number agreement. In French, no example was found which contained a form different from the masculine form. Example (6) is representative of this situation. (5) Italian examples: a. adult: la lepre cosa fa qua? / ‘the rabbit what does it do there?’

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.15 (950-950)


FRA b. adult: FRA: adult: FRA: adult: FRA c.

adult: CAR: adult: BEA: adult: BEA

d. adult: MAT: e.

MAT:

f.

adult: MAT:

g. adult: h. ALE: adult:

deduta (= seduta) (FRA ROM 2;0,23) ‘seated’ quand´è che è pronta questa pappa ? / ‘when’s the baby food ready?’ (dopo) c’è / qua / qua / preso quello / ‘(after) there is / here / here / taken that’ hai preso quello? ‘(you) have taken this one?’ sì / ‘yes.’ e che cos‘è quello? / ‘and what is this?’ piattino / (FRA ROM 2;0,23) ‘saucer.’ la mamma sta gonfiando il salvagente / e la bimba? / ‘mummy is inflating the life vest / and the girl?’ vettita (= vestita) / (CAR 2;1,18) ‘dressed’ che belle scarpe che hai / ‘how nice the shoes you have’ (xx) comprate mamma / ‘bought mummy’ mh ? / mamma ha comprate / (BEA 2;5,6) ‘mummy has bought’ ciao mucca ciao / buon viaggio / ‘bye cow bye! Have a nice trip!’ caduto / (MAT 2;3,29) ‘fallen’ nagnato / (= bagnato, refers to the ground) (MAT 2;3,29) ‘wet’ hai visto isa stamattina? / ‘did you see isa this morning?’ ha fatto na-na mamma rotta poi è stata male (a) isa / (= la gamba) (MAT 3;1,25) ‘has done leg broken then isa has been mad’ qual’è il tuo tesoro ? / ‘which one is your treasure?’ scosso / ‘hidden.’ al corso? / ‘to the side?’



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.16 (950-1023)


ALE:

no scosso (= nascosto) ‘no, hidden’

(6) French examples (from group 1a and 1b): adult: et que font le nounours et l’cochon? / KEV: i’ (fait) pas dodo? / assis / ‘he doesn’t take a nap? / seated /’

(ALE 3;3,23)

(KEV 2;6,27)

Based on Zushi’s account, we are now in the position to refine the approach for French and develop a new analysis for Italian. We will restrict ourselves to objects. If the child pays attention to the (overt) agreement morphology in French (cf. Wexler, Gavarró, & Torrens 2004 on the role of past participle agreement), s/he will find no agreement at all, since past participles do not exhibit overt (audible) agreement morphology for most of the cases. The French child might then be led to assume a minimum pro in object position, an NP-pro, which is bound by a base-generated zero topic.6 The structure is illustrated in (7): (7) [IP [TOP ej ] [IP Jean a vu NP-proj ] (object-drop) ‘John has seen’ Even if the child is sometimes confronted with past participle agreement in French, agreement morphology is never audible with number. This might also lead to the assumption of a minimum pro. Our approach about the minimum pro in French object position would be further supported if we could show that French children omit objects to the same extent found in topic-drop languages like Chinese. Wang, LilloMartin, Best and Levitt (1992) report between 20% and 30% of object omissions for Chinese children. This percentage is low compared to the Italian children in our study and much nearer to the French group.7 Based on Zushi’s account, we can modify Müller and Hulk’s (2001) approach for Italian. The authors assume that the same structure as for French underlies early Italian child grammar.8 Again, we will restrict ourselves to objects. If it is plausible that the child looks into the (overt) agreement morphology in Italian, s/he will find number (and gender) agreement all over the place, since past participles do exhibit overt (audible) agreement morphology. The Italian child might then be led to assume a ΦP-pro in object position, which is identified by AGR endowed with the Φ-feature number, as illustrated in (8). (8) [IP Giulia ha vistaAGR ΦP-pro] (object-drop) ‘Giulia has seen’ The important difference between the structures (7) and (8) is that in French, identification occurs via discourse whereas in Italian, AGR identifies the content of pro. The question is what type of pro is present if the construction does not contain a past participle. Since we have argued that children only look into the domain of past participle agreement in order to decide on the type of pro, we will have to assume one pro for all constructions, independently of whether a participle is present. Our assump-

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.17 (1023-1085)


tion of a ΦP-pro in the Italian object position predicts no asymmetry between object and subject omissions in Italian child language. Both should be identified by AGR. In fact, Schmitz (2005) shows that Italian children omit subjects to a degree of approximately 65%. This confirms our prediction because these empty subjects are identified by subject-AGR in adult and presumably in child language and the amount of subject omissions in child language corresponds to the amount found for object omissions. It seems thus reasonable to us to assume the same type of empty element for both subject and object position in Italian child grammar. There is another observation about French and Italian which falls out rather naturally under the proposal made here. Crysmann and Müller (2000) and Müller (2005) have observed that French children (also monolingual) make auxiliary selection errors, using avoir ‘have’ in contexts which require être ‘be’. This error is typical for French children, not for Italian children and is mostly visible in constructions with the clitic se, which necessitates être in the adult language and is construed with avoir by the French, not by the Italian children. (9) *Il s’ a fait mal. he self has done bad ‘He hurt himself ’ If it is true that French child grammar has an “inactive” AGR object system, and that Italian child grammar has an active AGR object system, we would predict that child French patterns with adult Spanish in several respects, namely the absence of past participle agreement and the use of one auxiliary avoir. Further research will have to decide on this issue.9 Let us finally turn to the DP realisations of the French children. Déchaine and Wiltschko (2002) have argued that, from a semantic point of view, pro-NPs are constants, i.e. empty equivalents of overt epithets or definite descriptions (the guy in the following example, i.e. a DP which seem to act like a pronoun) which can directly pick up their referent in discourse: I tried to visit Johni last week, but I was unable to persuade the guyi to see me (Rizzi 1992: 107). Japanese and also Chinese have been described in the literature as languages which dispose of empty and non-empty epithets. Thus, if our hypothesis about the grammar of the French children is correct, we would predict the occurrence of overt epithets. Very speculatively, we would like to suggest that the DP realisations of the French children in the test have to be treated syntactically as overt epithets, i.e. DPs which act like pronouns. We now turn to the results of the second groups which we repeat in the following: –

–

The differences between the two languages have become smaller, namely that French children use more DPs in object position than their Italian peers and that Italian children omit more objects than their French peers Italian children, however, use more object clitics than their French peers

In what follows, we discuss the finding about the higher frequency of object clitics in Italian and try to define the children’s task in order to acquire object clitics. The crucial



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.18 (1085-1158)


question is the following: Why are object clitics used to a higher extent in Italian than in French? Notice that this finding again corresponds to what we have said about earlier studies in Section 1. There is evidence that clitics differ from strong pronouns in French and Italian in being deficient elements (Cardinaletti & Starke 1999, among others) and in displaying a similar degree of deficiency which, in the framework of Déchaine and Wiltschko (2002), would naturally be characterizable in terms of the syntactic category Φ.10 Furthermore, arguing within this framework, subject and non-reflexive object clitics are typical Binding Principle B pronouns, i.e. they must be free in their governing category and can function as bound variables (see the examples in (10), (10a) is taken from Déchaine & Wiltschko 2002: 429). (10) a.

[Chaque homme]i pense qu’ [il]i est intelligent. Every man thinks that he is intelligent ‘Everyone thinks that he is intelligent.’ b. [Chaque homme]i pense que Marie veut [l]i ’ épouser. ‘Every man thinks that Mary will mary him.’ c. [Ogni uomo]i pensa che María [lo]i voglia sposare. ‘Every man thinks that Mary will mary him.’

We observed that in all language groups, omissions decrease in favour of clitics. Therefore, we may assume that omissions in child speech are related to problems with object clitics. We have further argued that object omissions are of a different nature in the two Romance languages. – –

Italian child: ΦP-pro French child: NP-pro

If object clitics are of the type Φ in both Romance languages, as could be argued on the basis of Déchaine and Wiltschko (2002), the Italian child only has to spell out a syntactic category whose existence s/he has assumed from early on. We have argued for a ΦP-pro in Italian child grammar. All the Italian child has to recognize is that ΦP has a phonetically non-empty equivalent, namely the object clitic. Since object omissions are licit but lexically restricted in adult Italian (the verb class matters), we predict that it will take some time to figure out with which verb class it is possible to have an empty pronominal and with which class the object has to be spelled out as a clitic. Thus, the Italian child quickly revises her/his grammar with respect to whether – in Zushi’s framework – agreement morphemes play a role in feature checking. S/he starts out with a grammar for which agreement morphemes play a role in feature checking (group 1 children) and then passes onto a grammar for which agreement morphemes play no role in feature checking (group 2 children), with exceptions which are lexically defined. It is yet unclear to us what might trigger the reanalysis. The French child has to determine the verb class which allows for empty objects as well. In contrast to the Italian child, s/he starts out with a grammar for which no feature checking takes place (group 1 children, i.e. children who still learn that

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.19 (1158-1207)


determiners are obligatory in French) and then moves on to a grammar for which agreement morphemes play no role in feature checking (group 2 children), with exceptions which are lexically defined. In addition, starting with a Japanese setting these children will have to reanalyze the empty object in French as a ΦP-pro which has a phonetically non-empty equivalent, the clitic. Here, we might speculate that there is a trigger for the reanalysis. As Müller et al. (1996) have suggested, the French child might look into the C-domain in order to find out that French is not like Japanese which lacks feature checking in this domain as well. French children will recognize the agreeing complementizer, namely the que/qui alternation11 and may deduce from the morphological reflex that French is a language with feature checking. Since the lexical realisation of the C-domain is late, we predict a delay in French children which is absent in the Italian child.

. Conclusion The present paper has revealed a difference between French and Italian monolingual children with respect to realisations and omissions of direct objects in the course of acquisition. We have argued that young French children have a syntactic representation of object omissions which corresponds to adult Japanese. This means that French children use a pragmatic strategy in order to license empty objects. Italian children use a syntactic structure which is similar as in the adult grammar of Brazilian Portuguese, i.e. they license objects via AGR.12 With respect to object clitics, we have observed that Italian children converge quickly to the adult system while French children continue to use the pragmatic strategy. We have suggested that the Italian child, by using object clitics, spells out an empty category which was already part of the syntactic derivation during very early stages, namely ΦP-pro. The French child has to resolve the problem of conflicting evidence: French is a language which has object clitics but which lades overt agreement morphology to identify a pro. The resolution of this conflict is timeconsuming. As for French, we have proposed that children revise their first grammar by looking into the (agreeing) C-domain. Further research has to show what trigger drives the Italian child into the adult grammar. Our approach has the advantage of allowing predictions for grammatical domains of child grammar studied by other researchers, namely auxiliary selection errors in French and past participle agreement errors in Italian.

Notes * We are grateful to the grant given to Natascha Müller by the Deutsche Forschungsgemeinschaft. The scientific members of the project “Bilingualism in early childhood: Comparing Italian/German and French/German” were Katja Francesca Cantone, Tanja Kupisch, Katrin Schmitz. We hereby thank all children and their parents for allowing the recordings as well as



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.20 (1207-1284)


the heads of the Ecole maternelle Mathieu de Quinvignies, of the Ecole Ferdinand Buisson (Niort), of the kindergarden La maisonnette (Rome) and of the kindergarden of Ospedaletto (Treviso) for supporting our work. Special thanks go to Mme Rousseau and Mme Boullet as well as Teresa Genova, Leila Morgia and the family of Katia Damini. We are grateful to those who participated in data collection and transcription: Estelle Leray, Moïra Berger, Melanie Blome, Karin Campagnoli, Katia Damini. . All data come from monolingual adults who interacted with children aged 3 to 4 years. We suppose that the adults do not adapt their speech to the children at that age. For French, 2168 finite sentences with a subject and 1072 finite sentences with an object were analyzed. In Italian, the number of analyzed sentences for subjects amounts to 1637, for objects to 1327. Unintelligible or interrupted utterances as well as imperative clauses were not counted. . Unfortunately, the data from spontaneous interaction and elicited production have not been clearly separated in the Italian study. . The base refers to the number of utterances analyzed for the MLU. . Analyzing subject omissions in adult spoken French, Schmitz and Müller (2004) found that expletive il is the only type of omitted subject. From what we said thus far, pro should be excluded because it does not have phonological content. However, one possibility (which is excluded by Zushi 2003 for English) would be to merge an empty expletive in Spec of T, which in turn checks off the EPP feature of T. That French might have such empty expletives is indicated by a well-known difference between English and French, namely agreement morphology in expletive constructions. Whereas the finite verb agrees with the expletive (empty or non-empty) in French (ilMASC/.SG estSG venu trois fillesFEM/.PLUR , ilMASC/.SG y aSG trois filles dans la bibliothèque), it agrees with the “subject” in English (there have arrived tree girls, there are three girls in the library). It may well be the case that French agreement is default agreement (3rd person singular), since no element has been moved or inserted into Spec of T which can produce a phonological effect. . Kato (1999) analyses the null subject property as a consequence of a pronominal agreement. Instead of a pro in the specifier of the subject position, she analyses Agreement as an independent D element in the numeration. This D element contains case and phi-features which are merged with verbs inflected for tense. Case and phi-features are checked in T. As a consequence, Spec of T is not projected. Thus, in this approach T in Italian does not have an EPP feature. We think that Zushi’s analysis is simpler since it assumes the presence of an EPP feature in both French and Italian. We will therefore base our analysis on Zushi’s approach. Notice that the acquisition data are neutral with respect to the two approaches. . In a fine structured left periphery à la Rizzi (1997), the empty topic would be generated in TOP (see Grewendorf 2002). . The French children in the elicited production task omitted objects at a rate of 44%. Notice however that Jakubowicz et al. (1997) already mentioned that elicited production had a negative effect on the children in the sense that they tended to omit more objects in elicited production if compared with spontaneous interaction. The children tested in Wang et al. (1992) were presented with quasi-spontaneous tasks, like retelling a story. . Notice however that Müller and Hulk have analyzed bilingual children and have claimed that the Germanic language – German or Dutch – influences the Romance language – French or Italian. The fact that the authors have assumed the same structure for early French and Italian might still be correct for bilingual children, as an effect of cross-linguistic influence.

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.21 (1284-1340)

Null arguments in monolingual children . Wexler, Gavarró and Torrens (2004) present an alternative account for clitic omissions, based on Wexler’s Unique Checking Constraint. The UCC (Wexler 1998) allows the D-feature of DP to be checked only against one functional category in child grammar. The predictions of the approach are that children acquiring languages with participle agreement like Italian omit clitics (since objects need to double check a D-feature with two functional categories), while children acquiring languages without participle agreement like Spanish do not omit clitics (since objects do not need to double check a D-feature). Wexler et al. (2004) find indeed that Spanish children (even 2-year olds) omit clitics rarely (in 8% of all contexts). However, de la Mora et al. (2004) observe in a group of children with an average age of 3;9 still 40% of omissions. The Spanish data situation is thus far from clear. Furthermore, a theoretical objection has to be made. If the UCC is responsible for overt clitic realisation in simple present tense constructions, which do not contain a past participle and which do not exhibit object agreement, we have to ask why the UCC cannot guarantee that the object clitic is lexically realized in the present tense in Italian child speech. Italian children have been reported to leave out the clitic in constructions with present tense verbs. . See Gabriel and Müller (2005) for a refined approach which distinguishes among the different clitics, subject, object and reflexive clitics with respect to the internal structure of these Φ-elements. . French qui is often analyzed as que + Agr, a morphological reflex of specifier-head agreement between a trace and the head C: (i) L’hommei que je crois [ti qui [ti viendra]] (ii) *L’hommei que je crois [ti que /*qui [Jean connaît ti ] (object) (iii) L’hommei que je pense [ti que/*qui [Jean croit [ti qui [ti viendra]]]] (extraction of embedded subject) Agr in C is identical (coindexed) with Agr in I in French (and in English); cf. Rizzi (1990). Recently, Bošković and Lasnik (2003) have argued that qui is a C with an EPP feature, while que lacks this feature in French. . Our hypothesis is based on parameter setting. Although we were not able to formulate the trigger for Italian children, we have assumed that French children may revise their intermediate grammar by looking into the C-domain. Notice that maturational accounts of language acquisition data are faced with the same problem as approaches based on parameter setting. Let us take the Unique Checking Constraint by Wexler (1998). This constraint will have to fade out of grammar during development and this has to be explained. Assuming that this is due to maturation might be the correct analysis but notice that maturation is too powerful in the sense that it can always be invoked to explain changes in development.

References Antinucci, F. & Miller, R. (1976). “How Children Talk about What Happened”. Journal of Child Language, 3, 167–189. Borer, H. & Wexler, K. (1992). “Bi-unique relations and the Maturation of Grammatical Principles”. Natural Language and Linguistic Theory, 10, 147–189. Bortz, J. (1999). Statistics for social scientists. Berlin: Springer.



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.22 (1340-1446)


Bošković, Z. & Lasnik, H. (2003). “On the Distribution of Null Complementizers”. Linguistic Inquiry, 34(4), 527–546. Cardinaletti, A. & Starke, M. (1999). “The Typology of Structural Deficiency: A Case Study of the Three Classes of Pronouns”. In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 145–233). Berlin: Mouton de Gruyter. Cipriani, P., Chilosi, A. M., Bottari, P., & Pfanner, L. (1993). L’acquisizione della morfosintassi in italiano: fasi e processi. Padua: UniPress. Clark, E. (1986). “Acquisition of Romance, With Special Reference to French”. In D. I. Slobin (Ed.), The Crosslinguistic Study of Language Acquisition, Vol. I, The Data (pp. 687–782). Mahwah, NJ: Lawrence Erlbaum. Crysmann, B. & Müller, N. (2000). “On the Non-Parallelism in the Acquisition of Reflexive and Non-Reflexive Object Clitics”. In C. Hamann & S. Powers (Eds.), The Acquisition of Scrambling and Cliticization (pp. 207–236). Dordrecht: Kluwer. Cummins, S. & Roberge, Y. (2004). “Null Objects in French and English”. In J. Auger, C. Clements, & B. Vance (Eds.), Contemporary Approaches to Romance Linguistics (pp. 121– 138). Amsterdam: John Benjamins. Cummins, S. & Roberge, Y. (2005). “A Modular Account of Null Objects in French”. Syntax. To appear. Déchaine, R.-M. & Wiltschko, M. (2002). “Decomposing Pronouns”. Linguistic Inquiry, 33(3), 409–442. De la Mora, J., Paradis, J., Grinstead, J., Flores, B., & Cantu, M. (2004). Object clitics in Spanishspeaking children with and without SLI. Poster presented at the 25th Annual Symposium on Research in Child Language Disorders, University of Wisconsin-Madison, WI. Fónagy, I. (1985). “J’aime, je connais: Verbes transitifs à objet latent”. Revue Romane, 20(1), 3–35. Gabriel, C. & Müller, N. (2005). “Zu den romanischen Pronominalklitika: Kategorialer Status und syntaktische Derivation”. In G. Kaiser (Ed.), Deutsche Romanistik – Generativ. Tübingen: Narr. To appear. Grondin, N. & White, L. (1996). “Functional Categories in Child L2 Acquisition of French”. Language Acquisition, 5, 1–34. Guasti, M. T. (1993/1994). “Verb Syntax in Italian Child Grammar. Finite and Nonfinite Verbs”. Language Acquisition, 3, 1–40. Grewendorf, G. (2002). Minimalistische Syntax. Tübingen/Basel: A. Francke Verlag. Hamann, C., Rizzi, L., & Frauenfelder, U. (1994). “On the Acquisition of the Pronominal System in French”. Geneva Generative Papers, 2(2), 91–103. Hamann, C. (2002). From Syntax to Discourse. Pronominal Clitics, Null Subjects and Infinitives in Child Language. Dordrecht: Kluwer. Hamann, C., Rizzi, L., & Frauenfelder, U. (1994). “On the Acquisition of the Pronominal System in French”. Geneva Generative Papers, 2, 91–103. Hamann, C., Rizzi, L., & Frauenfelder, U. (1996). “On the Acquisition of Subject and Object Clitics in French”. In H. Clahsen (Ed.), Generative Perspectives on Language Acquisition (pp. 309–334). Amsterdam: John Benjamins. Hamann, C. et al. (2002). Aspects of Grammatical Development in Young French Children with SLI. Manuscript, University of Geneva, Centre Hospilatier Universitaire de Lausanne, University of Siena. Hoekstra, T. & Hyams, N. (1998). “Aspects of Root Infinitives”. Lingua, 106, 81–112. Huang, C.-T. J. (1984). “On the Distribution and Reference of Empty Pronouns”. Linguistic Inquiry, 15, 531–574.

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.23 (1446-1542)


Hulk, A. (1997). “The Acquisition of French Object Pronouns by a Dutch/French Bilingual Child”. In A. Sorace, C. Heycock, & R. Shillcock (Eds.), Language Acquisition: Knowledge, Representation and Processing. Proceedings of the GALA 1997 conference on language acquisition (pp. 521–526). Edinburgh: University of Edinburgh. Hunnius, K. (1980). “Sprachnorm und Sprachgebrauch. Zur Geschichte der Kongruenzregeln des participe passé”. In H. Bork, A. Greive, & D. Woll (Eds.), Romanica Europaea et Americana. Festschrift für Harri Meier zum 8. Januar 1980 (pp. 244–253). Bouvier: Bonn. Jakubowicz, C., Müller, N., Kang, O.-K., Riemer, B., & Rigaut, C. (1996). “On the Acquisition of the Pronominal System in French and German”. In A. Springfellow, D. Cahana-Amitay, E. Hughes, & A. Zukowski (Eds.), Proceedings of the 20th Annual Boston University Conference on Language Development (pp. 374–385). Somerville, MA: Cascadilla Press. Jakubowicz, C., Müller, N., Riemer, B., & Rigaut, C. (1997). “The Case of Subject and Object Omissions in French and German”. In E. Hughes, M. Hughes, & A. Greenhill (Eds.), Proceedings of the 21st annual Boston university conference on language development (pp. 331–342). Somerville, MA: Cascadilla Press. Jakubowicz, C., Nash, L., Rigaut, C., & Gérard, Ch.-L. (1998). “Determiners and Clitic Pronouns in French-Speaking Children with SLI”. Language Acquisition, 7, 113–160. Jakubowicz, C. & Rigaut, C. (2000). “L’acquisition des clitiques nominatifs et des clitiques objects en français”. Canadian Journal of Linguistics, 45(1–2), 119–157. Kaiser, G. (1994). “More about INFL-ection: the Acquisition of Personal Pronouns in French”. In J. M. Meisel (Ed.), Bilingual first language acquisition. French and German grammatical development (pp. 131–159). Amsterdam: John Benjamins. Kato, M. A. (1999). “Strong and Weak Pronominals in the Null Subject Parameter”. Probus, 11, 1–37. Lambrecht, K. (1981). Topic, Antitopic and Verb Agreement in Non-standard French. Amsterdam: John Benjamins. Lambrecht, K. & Lemoine, K. (1996). “Vers une grammaire des compléments zéro en français parlé”. In J. Chuquet & M. Fryd (Eds.), Absence de marques et représentation de l’absence 1 – travaux linguistique du Cerlico 9 (pp. 279–310). Rennes: Presses Université de Rennes. McKee, C. & Emiliani, M. (1992). “Il clitico: c’è ma non si vede”. Natural Language and Linguistic Theory, 10, 415–437. Müller, N. (1999). Past Participle Agreement in Romance. Unpublished Habilitationsschrift, University of Hamburg. Müller, N. (2002). “Cross-linguistic Influence in Early Child Bilingualism: Italian/German”. In S. Foster-Cohen, T. Ruthenberg, & M.-L. Poschen (Eds.), EUROSLA Yearbook 2 (pp. 135– 154). Amsterdam: John Benjamins. Müller, N. (2004). “Null-arguments in Bilingual Children: French Topics”. In J. Paradis & P. Prévost (Eds.), The Acquisition of French in Different Contexts. Focus on Functional Categories (pp. 275–304). Amsterdam: John Benjamins. Müller, N. (2005). “Erwerb von SE durch bilingual deutsch-französische Kinder”. In C. Sinner & G. Veldré (Eds.), Beiträge zum Aachener Romanistentag. Frankfurt am Main: Lang, to appear. Müller, N., Cantone, K., Kupisch, T., & Schmitz, K. (2002). “Zum Spracheneinfluss im bilingualen Erstspracherwerb: Italienisch – Deutsch”. Linguistische Berichte, 190, 157–206. Müller N., Crysmann, B., & Kaiser, G. (1996). “Interactions between the Acquisition of French Object Drop and the Development of the C-system”. Language Acquisition, 5(1), 35–63.



JB[v.20020404] Prn:5/06/2006; 14:57



F: LALD4104.tex / p.24 (1542-1643)


Müller, N. & Hulk, A. (2000). Crosslinguistic Influence in Bilingual Children: Object Omissions and Root Infinitives. In C. Howell, S. A. Fish, & T. Keith-Lucas (Eds.), Proceedings of the 24th Annual Boston University Conference on Language Development (pp. 546–557). Somerville, MA: Cascadilla Press. Müller, N. & Hulk, A. (2001). “Crosslinguistic Influence in Bilingual Language Acquisition: Italian and French as Recipient Languages”. Bilingualism: Language and Cognition, 4(1), 1–21. Müller, N., Hulk, A., & Jakubowicz C. (1999). “Object Omissions in Bilingual Children: Evidence for Crosslinguistic Influence”. In A. Greenhill, H. Littlefield, & C. Tano (Eds.), Proceedings of the 23rd Annual Boston University Conference on Language Development (pp. 482–494). Somerville, MA: Cascadilla Press. Müller, N. & Kupisch, T. (2003). “Zum simultanen Erwerb des Deutschen und des Französischen bei (un)ausgeglichen bilingualen Kindern”. Vox Romanica, 62, 145–169. Müller, N. & Kupisch, T. (2005). “The French DP in Bilingual First and Second Language Development: Proficiency versus Age as Windows to Language Acquisition”. In S. Haberzettl (Ed.). Berlin: Mouton de Gruyter, to appear. Noailly, M. (1997). “Les mystères de la transitivité invisible”. Langages, 127, 96–109. Paradis, J., Crago, M., & Genesee, F. (2003). “Object Clitics as a Clinical Marker of SLI in French: Evidence from French-English Bilingual Children”. In B. Beachley et al. (Eds.), Proceedings of the 27th annual Boston university conference on language development (pp. 638–649). Somerville, MA: Cascadilla Press. Prévost, P. (1997). Truncation in Second Language Acquisition. Ph.D. Dissertation, McGill University, Montréal, Québec. Rizzi, L. (1990). Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, L. (1992). “Early null subjects & root null subjects”. Geneva Generative Papers, 0(1–2), 102–115. Rizzi, L. (1997). “The Fine Structure of the Left Periphery”. In L. Haegeman (Ed.), Elements of Grammar (pp. 281–337). Dordrecht: Kluwer. Schaeffer, J. (1990). The Syntax of the Subject in Child’s Language. Master’s thesis, State University of Utrecht. Schaeffer, J. (1996). “Clitic Omission and Participle Agreement in Italian Child Language”. In K. Zagona (Ed.), Grammatical Theory and Romance Languages (pp. 219–228). Amsterdam: John Benjamins. Schaeffer, J. (Ed.). (1997). “The Interpretation of Root Infinitives and Bare Nouns in Child Language”. MIT Occasional Papers in Linguistics, 12. Schaeffer, J. (2000). The Acquisition of Direct Object Scrambling and Clitic Placement: Syntax and Pragmatics. Amsterdam: John Benjamins. Schmitz, K. (2005). Vulnerable Subjects?. Manuscript, Bergische Universität Wuppertal. Schmitz, K. & Müller, N. (2004). “Strong and Clitic Pronouns in Monolingual and Bilingual Acquisition of French and Italian”. Submitted. Shlonsky, U. (1997). Clause Structure and Word Order in Hebrew and Arabic. Oxford: Oxford University Press. Tanase, E. (1976). “L’accord du participe passé dans le français oral”. In M. Boudreault & F. Moehren (Eds.), Actes du XIIIe Congrès International de Linguistique et Philologie Romanes (pp. 475–482). Québec. Tiedemann, C. (1999). Erwerb des italienischen Pronominalsystems. Unpublished master’s thesis, University of Hamburg. Tuller, L. A. (2000). Null Objects in Deaf French. Manuscript, University of Tours.

JB[v.20020404] Prn:5/06/2006; 14:57

F: LALD4104.tex / p.25 (1643-1667)


Wang, Q., Lillo-Martin, D., Best, C. T., & Levitt, A. (1992). “Null Subject versus Null Object: Some Evidence from the Acquisition of Chinese and English”. Language Acquisition, 2(3), 221–254. Wexler, K. (1998). “Very Early Parameter Setting and the Unique Checking Constraint”. Lingua, 106, 23–79. Wexler, K., Gavarró, A., & Torrens, V. (2004). “Feature Checking and Object Clitic Omission in Child Catalan and Spanish”. In R. Bok-Bennema, B. Hollebrandse, B. Kampers-Mahne, & P. Sleeman (Eds.), Romance Languages and Linguistic Theory. Amsterdam: John Benjamins. White, L. (1996). “Clitics in L2 French”. In H. Clahsen (Ed.), Generative Perspectives on Language Acquisition (pp. 335–368). Amsterdam: John Benjamins. Zushi, M. (2003). “Null Arguments: The Case of Japanese and Romance”. Lingua, 113, 559–604.



JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.1 (47-119)

Prenominal elements in French-Germanic bilingual first language acquisition Evidence for cross-linguistic influence* Maren Pannemann Universiteit van Amsterdam, ACLC

This article investigates interference in the determiner phrase in French-Germanic bilingual first language acquisition. Hulk and Müller (2000) propose that cross-linguistic interaction takes place at the syntax/pragmatics interface given a surface overlap between simultaneously acquired languages. We tested this prediction in two domains: (a) the presence of the determiner in adjective-noun combinations and (b) the position of the attributive adjective. Hulk and Müller’s model predicts no interference in either domain. With respect to (a), the results show a delay in the acquisition of the determiner, but it is questionable whether this is caused by interference. Regarding (b), there is actually evidence for interference, suggesting that not only the syntax/pragmatics interface but also narrow syntax can be affected by cross-linguistic interaction.

.

Introduction

Previous studies on bilingual first language acquisition have shown that even though bilingual children develop two autonomous language systems from the very beginning, interaction between the two systems is possible (Hulk & van der Linden 1996; Gawlitzek-Maiwald & Tracy 1996). Hulk and Müller (2000) have argued that this cross-linguistic interaction is restricted by two conditions: First, there has to be a surface overlap between the two languages. Second, the overlapping structures must involve the interface between syntax and pragmatics. The present study is designed to investigate this proposal. The focus of the study is the development of prenominal elements in monolingual French and French-Germanic bilingual children. The French DP contrasts with the Germanic DP in a number of aspects, which makes the DP a suitable domain to investigate cross-linguistic interference. The contrasts investigated in this paper are the presence of the determiner and the position of the attributive adjective. Hulk and Müller’s model predicts that in French-Germanic bilingual acquisition, the development of the determiner in French is not affected by cross-linguistic

JB[v.20020404] Prn:5/06/2006; 14:59



F: LALD4105.tex / p.2 (119-170)

Maren Pannemann

interaction from the Germanic language. Furthermore, this model predicts that the acquisition of the position of the attributive adjective in French is not influenced by the simultaneous acquisition of the Germanic language. Concerning the first prediction it will be shown that bilingual children show a delay in the acquisition of the determiner in A-N contexts. However, as will be argued, it is not clear whether this delay is caused by interference or whether it is due to other factors. In contradiction to the second prediction, there is actually evidence for interaction in the position of the adjective. These results are unexpected in the model of Hulk and Müller (2000). I will discuss the consequences of these observations for a model of cross-linguistic interaction, with special attention to conditions and criteria for cross-linguistic interaction. This article is organised as follows: As I argue in Section 2, the domain of the Determiner Phrase (DP) in French and Germanic bilingual first language acquisition fulfils the conditions for cross-linguistic interference as proposed by Hulk and Müller (2000). Therefore, the domain focused on in this study is the acquisition of the DP in monolingual and bilingual French, more specifically the stage where nouns start to combine with other elements. Section 3 contains the methodology used in this study. As the present study shows, bilingual acquisition of the DP in a French-Germanic setting deviates from monolingual acquisition of French. Section 4 illustrates that with respect to the acquisition of determiners, there is a quantitative difference between monolingual and bilingual children. Section 5 shows that a qualitative difference can be observed in the acquisition of attributive adjectives. Section 6 discusses the implications of these observations for the model proposed by Müller and Hulk (2000) and presents the conclusions of this study.

. Possibility of cross-linguistic interference . Adult system This article focuses on the acquisition of the DP in children who acquire French simultaneously with a Germanic language. A look at DP properties in French and Germanic shows that their DPs contrast with respect to the syntax of prenominal elements. First, in French the presence of the determiner is obligatory in most contexts (for exceptions see Roodenburg 2004) while in Germanic languages determiner omission is more frequent. The presence of the determiner depends on aspectual properties of the verb and concerns plural count nouns with a generic interpretation (cf. (1)) and singular mass nouns (cf. (2)), illustrated here with German, Swedish and Dutch. The examples show that also when these nouns are preceded by an adjective, they appear without a preceding determiner. In contrast, mass nouns and plural count nouns must always be preceded by a determiner in French (cf. (3)). This also holds for adjective-noun combinations. In contrast to Germanic languages, French has an indefinite plural article: des. Note that when a plural noun is preceded by an adjective, de is used instead of des (cf. (3c)).

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.3 (170-248)

Prenominal elements in French-Germanic bilingual first language acquisition

(1) a. (kleine) Zebras sind niedlich b. (små) sebror är söta c. (kleine) zebra’s zijn schattig small zebras are cute

German Swedish Dutch

(2) a. (kaltes) Wasser b. (kallt) vatten c. (koud) water cold water

German Swedish Dutch

(3) a. *petits zèbres sont mignons small zebras are cute b. je vois des zèbres I see indef.pl zebras ‘I see zebras’ c. de petits zèbres indef.pl small zebras d. *(le) bon vin the good wine ‘(the) good wine’

French

A second contrast between French and Germanic DPs is the position of the attributive adjective. In Germanic languages, adjectives are only allowed in the prenominal position (cf. (4), illustrated with German) whereas in French, both prenominal and postnominal adjectives are possible (cf. (5a/b)). There is a small set of highly frequent adjectives that appear in prenominal position though the majority of French adjectives follow the noun. (4) a.

ein großes/ grünes a big/ green ‘a big/green house’ b. *ein Haus großes/ a house big/

(5) a.

Haus house

German

grünes green

une grande maison a big house ‘a big house’ b. une maison verte a house green ‘a green house’

French

We find that the syntactic distribution of prenominal elements in French and Germanic differs with respect to the presence of the determiner and the position of the attributive adjective. As a result, children who acquire French simultaneously with a Germanic language receive contradictory information, making the DP a potential domain for cross- linguistic interaction.



JB[v.20020404] Prn:5/06/2006; 14:59



F: LALD4105.tex / p.4 (248-301)

Maren Pannemann

. Hulk and Müller (2000): Conditions for cross-linguistic influence Hulk and Müller (2000, cf. also Müller & Hulk 2000, 2001) argue that interaction between two languages acquired simultaneously takes place under the following conditions: First, the interface between pragmatics and syntax has to be involved. Second, there must be a structural overlap between the two languages at the surface level. If these conditions are met, the interaction takes place as follows: “language A has a construction which may seem to the child to have more than one structural analysis (one of which is not correct) and that language B reinforces one of these two structural analyses (the “wrong” one – leading to problems – or the “right” one – leading to facilitation)” (Hulk & Müller 2000: 228). This means that if language A displays an ambiguity with respect to a particular structure and language B has a superficially similar structure, there will be interference from language B to language A, given that the syntax-pragmatics interface is involved. Section 2.1 illustrated two differences between French and Germanic DPs: The presence of the determiner (in adjective-noun combinations) and the position of the attributive adjective. The model by Hulk and Müller predicts that in French-Germanic bilingual acquisition, the French DP is not affected in any of these two aspects. With respect to the first aspect, the status of the determiner, the model predicts that the acquisition of the French DP is not influenced by the Germanic language acquired simultaneously. In fact, considering the conditions for cross-linguistic influence as proposed by Hulk and Müller, we expect that there is cross-linguistic interference from French to the Germanic language. This is motivated as follows: Hulk and Müller argue that cross-linguistic influence takes place at the pragmatic-syntactic interface domain only. An example for this is the C-domain, i.e., the left periphery of the clause as this is an interface domain connecting the syntactic and the cognitive level. As Kupisch (2003) points out, the left periphery of the DP also qualifies as an interface domain. Based on the Configurational Hypothesis by Giorgi and Longobardi (1991), who argue that DP structure is parallel to clause structure, Kupisch draws a comparison between the DP and the C-domain. She points out that the left periphery of the DP also has an interface character as the determiner connects the proposition and the discourse. This makes it potentially vulnerable for cross-linguistic influence, and thus the first condition for cross-linguistic interaction is met. The second condition for cross-linguistic influence also applies to the determiner: There is an overlap between French and Germanic in the sense that both have determiners. Note that this overlap concerns an overlap only on the surface. In French, the determiner is obligatory in most contexts whereas Germanic allows for determiner omission. However, the presence of the determiner in Germanic is not optional but, as illustrated in (1) and (2), depends on semantic properties of the noun (and aspectual properties of the verb) and also affects the interpretation of the noun. An interesting question is how this overlap is perceived by the child. However, Hulk and Müller point out that the overlap concerns “the surface level” (Hulk & Müller 2000: 227), suggesting that differences in interpretation are not relevant here. If we apply Hulk and Müller’s model to the acquisition of

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.5 (301-355)


the determiner, we find that the evidence from the Germanic language is ambiguous: The Germanic languages display both omission and presence of the determiner. As a consequence, the child might start out with more than one analysis for the Germanic language. In contrast, the evidence that the child gets for French is far less ambiguous as the determiner is virtually always present. Hulk and Müller’s model predicts then that French reinforces one of the possible analyses for the Germanic language, namely the one that favours the presence of the determiner. Therefore, we expect that the acquisition of determiners in Germanic is affected by French with the consequence that the bilinguals might use overt determiners more frequently than monolingual children. In contrast, the acquisition of the French DP, which is the focus of this article, should not be affected. The acquisition of the D in determiner-noun combinations has been extensively investigated for French L1 acquisition (cf. Chierchia et al. 1999; van den Berg 2001; Granfeldt 2000a/b; Kupisch 2003, 2004a/b and others). The subsequent stage where the DP becomes more complex and adjective-noun combinations are preceded by a determiner has been less studied so far. Therefore, the present article focuses in particular on the stage where determiners start to combine with adjective-noun combinations. The application of the model for cross-linguistic interference leads to the following prediction: Prediction 1: In French-Germanic bilingual first language acquisition, the acquisition of the determiner in adjective-noun combinations in French is not affected by the simultaneous acquisition of the Germanic language. With respect to the second contrast between French and Germanic, the position of the attributive adjective, the model by Hulk and Müller predicts once more that the French DP is not affected in French-Germanic bilingual acquisition. Notice that again there is an overlap between French and Germanic: Both languages have prenominal adjectives. Only French allows for a second possibility: Adjectives also appear in the postnominal position. However, in spite of this structural overlap, Hulk and Müller’s model predicts that no cross-linguistic interaction takes place. The position of the attributive adjective does not involve the interface between pragmatics and syntax and thus the first condition for cross-linguistic interference is not met. Therefore, even though there is a superficial overlap between French and Germanic attributive adjectives, we do not expect cross-linguistic interference in the acquisition of the position of the attributive adjective. Note that with respect to prediction 1, Hulk and Müller’s model predicts influence only from French to Germanic, which is not the focus of the present study. With respect to the position of the attributive adjective, the model actually predicts no cross-linguistic interaction at all. Prediction 2: In French-Germanic bilingual first language acquisition, the acquisition of the syntax of attributive adjectives in French is not affected by the simultaneous acquisition of the Germanic language.



JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.6 (355-468)

 Maren Pannemann

In sum, we expect that neither the acquisition of the French determiner nor the acquisition of the syntax of French attributive adjectives is affected by the simultaneous acquisition of a Germanic language. The two predictions made in this section are investigated in Section 4 and 5 respectively. If no cross-linguistic influence has taken place, we expect that there are no differences between monolingual and bilingual French. If there has been cross-linguistic influence, we expect monolingual and bilingual French to differ in the aspects investigated here. The next section contains an overview of the data and methodology used.

. Methodology In order to trace possible cross-linguistic effects from Germanic, it is necessary to investigate whether the acquisition of the French DP in monolingual children differs from the acquisition of the DP in French-Germanic bilingual children. In the present contribution, I investigated the acquisition of the DP in the monolingual French child Daniel. The data consist of a longitudinal corpus of spontaneous speech recordings of parent-child interaction. For Daniel, 5 recordings are available in the age period between 1;8.7 and 1;11.7 (MLU 1.5-2.45).1 A discussion in L1 research concerns the relatively small number of children investigated. It is, however, common practice in longitudinal L1 research to rely on case studies as this is all we have at the present. To compensate for this problem, I included results of previous research on monolingual French acquisition. The aim of this paper is to compare the acquisition of the French DP in monolingual and bilingual children. For the bilingual French-Germanic children, I again conducted not only a case study but also included results of earlier studies on bilingual French acquisition in order to get a better idea of the development of more children. The French-Dutch bilingual child investigated in the case study is Anouk. The data was collected in a research project directed by Aafke Hulk and Elisabeth van der Linden at the Amsterdam Center for Language and Communication, University of Amsterdam. I selected 6 files between 2;3.13 and 2;11.27 (MLU 1.38–3.47). These files represent a cross-section of the corpus; sessions recorded as close to one-month intervals as possible were analysed. Even though Daniel’s data is only available until the age of 1;11.7 (MLU 2.45), I decided to include the recordings of Anouk until the age of 2;11.27 (MLU 3.47) because, interestingly, at this age a change can be observed as we will see in Section 4.2. Furthermore, the development of the monolingual Philippe (CHILDES) will also serve as a reference point for the later files in Anouk since he was recorded until the age of 3;3 (MLU 4.53). Table 1 gives an overview of the studies on the acquisition of the French DP in monolingual and bilingual contexts that I refer to in this contribution. All these studies are based on longitudinal data. The criteria for counting determiner contexts in first language acquisition data often vary among researchers, potentially leading to problems with respect to the comparability of different studies. Therefore, I followed Granfeldt’s (2000b), Hulk’s

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.7 (468-480)

Prenominal elements in French-Germanic bilingual first language acquisition 

Table 1. A selection of previous studies on the acquisition of prenominal elements in French Müller (1994): Granfeldt (2000a, b): Hulk (2004): Berg, van den (2001): Palmer (2000): FrenchFrench-Swedish French-Dutch French-Dutch Monolingual German French Ivar (1;6–5;10) Caroline (1:5–5;0)

Jean (1;10–3;9) Anne (2;3–4;0) Mimi (2;0–4;2) Granfeldt (2000b): Monolingual French Grégoire (1;9–2;5)

Anouk (2;3–3;10)

Anouk (2;3–3;10) Annick (2;5.10–3;5.5

Monolingual French Philippe (2;1.–3;3) Grégoire (1;9–2;5)

Daniel (1;8–1;11) Philippe (2;1.–3;3) Grégoire (1;9–2;5)

(2004) and Kupisch’s (2004b) counting criteria, i.e., I excluded proper names, repetitions of the same noun unless interrupted by either an utterance of the parent or the child itself, imitation with or without determiner directly following a parent’s utterance, completion, incomprehensible utterances of the type ‘DET xxx’ and language mixed utterances. Like Granfeldt, Hulk and Kupisch, I included possessives, numerals, quantifiers, partitives and demonstratives. Furthermore, following Kupisch, I counted protodeterminers of the type l neige (‘the (protodeterminer definite article) snow’) and e wou wou (‘a (protodeterminer indefinite article) dog’). In contrast to Pannemann (in press), I excluded un autre (‘an other’), un peu (‘a bit’) and (encore) une fois (‘again a time’) as determiner-noun combinations because their status as noun is questionable. Both their morphological and distributional properties are atypical for nouns. un peu cannot carry a plural marker, and it cannot combine with an adjective other than petit with which it forms a fixed expression (cf.(6)).2 As illustrated in (7), une fois can be pluralized, but it can combine only with ordinal numbers and not with adjectives. un autre is likely to be a substantivised adjective but in some cases it was impossible to decide whether autre was a noun or an adjective. Only in utterances as un-autre-N did I include autre as adjective. (6) a.

un peu/ * deux peus a bit/ two bits b. un petit peu a little bit ‘a little bit’

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.8 (480-555)

 Maren Pannemann

c.

(7) a.

un *bon/ *mauvais peu/ *rouge a good/ bad bit red ‘a good/bad/red bit’

une a b. une a

fois/ deux fois time/ two times deuxième/ *grande/ *petite/ *bonne fois second/ big/ small/ good time

. The acquisition of the determiner in adjective-noun combinations Studies on the acquisition of the French DP have led to the following picture of the acquisition of French prenominal elements. In first language acquisition of the French DP the child passes through several stages. In the one-word stage, the first DP elements that emerge are bare nouns. The first adjectives also appear when the child is still in the one-word stage. Around the age of 18 months, children start to use twoword utterances. In this stage, the first combinations of nouns with one other element can be observed. The next step in the development of the internal DP structure is the extension of the DP such that two prenominal positions are possible and the noun can be preceded both by a determiner and a prenominal adjective. Let us turn to the first prediction, concerning the presence of determiners in adjective-noun combinations.

. Monolingual acquisition Table 2 shows the distribution of determiner-adjective-noun utterances (henceforth D-A-N) and A-N combinations in Daniel’s data. In the first two files, there are already D-A-N combinations. In the third, fourth and fifth files (1;9.21–1;11.7), D-A-N and A-N utterances co-exist. Notice that even though the determiner omission rate in obligatory contexts is still relatively high (between 66% and 68%), Daniel nevertheless produces D-A-N combinations from early on. 36% of all A-N sequences are preceded by a determiner. This tendency to realise the determiner in A-N combinations indicates his beginning insight in the obligatory presence of the determiner in French.3 A similar observation has been made for the monolingual French child Grégoire. The data of Grégoire is available in the CHILDES database (cf. MacWhinney & Snow Table 2. Acquisition of prenominal elements in Daniel Age

MLU

Total # of nouns

D missing %

D-A-N

A-N

D (1) 1;8.7 D (2) 1;8.21 D (3) 1;9.21 D (4) 1;10.7 D (5) 1;11.7

1.5 1.52 1.69 2.25 2.45

173 94 174 164 112

68 67 65 66 56

3 5 6 9 10

2 1 11 27 18

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.9 (555-659)

Prenominal elements in French-Germanic bilingual first language acquisition 

Table 3. Acquisition of prenominal elements in Grégoire4 Age

MLU

Total # of nouns

D missing %

D-A-N

A-N

G (1) 1;9.18 G (2) 1;10.3 G (3) 1;10.24 G (4) 1;11.21 G (5) 2;0.5 G (6) 2;1.24 G (7) 2;3.0 G (8) 2;5.1

1.9 1.9 1.7 1.9 2.1 2.3 2.5 3.2

33 210 139 84 131 60 142 73

85 53 46 62 33 12 17 5

0 0 0 0 1 0 29 10

0 0 0 0 0 1 7 2

Table 4. Acquisition of prenominal elements in Anouk Age

MLU

Total # of nouns

D missing %

D-A-N

A-N

Ao 2;3.13 Ao 2;4.17 Ao 2;5.20 Ao 2;6.1 Ao 2;7.5 Ao 2;8.3 Ao 2;8.22 Ao 2;11.27

1.38 2.0 2.1 2.5 2.38 2.35 2.48 3.47

16 17 39 32 46 65 51 54

94 94 77 63 52 63 41 11

0 1 0 1 1 3 1 11

0 2 5 3 7 15 8 2

1990). Until the age of 2;1.24, Grégoire hardly produces adjective-noun contexts. Then, from 2;3.0 onwards he produces a relatively high number of A-N combinations. The majority of these is preceded by a determiner (cf. Table 3).

. Bilingual acquisition The picture is different for bilingual children. Anouk has a tendency to omit the determiner in A-N combinations. This tendency to omit the determiner is stronger than in Daniel (cf. Table 2 vs. Table 4). After the first prenominal adjectives emerge at 2;4.17, she leaves out the determiner in the majority of the A-N combinations until 2;8.22. From 2;11.27, the picture changes, and the majority of A-N combinations is preceded by a determiner (cf. Table 4). This observation is confirmed by Müller (1994) who observes that the FrenchGerman bilingual children Ivar and Caroline pass through a stage (until 2;5) where determiners and adjectives are mutually exclusive in prenominal position while DA-N utterances are still absent. In a subsequent stage, D-A-N combinations become possible. As her paper focussed on gender acquisition, no number specifications are available. A similar development can be observed in Jean, Anne and Mimi, three FrenchSwedish bilingual children investigated by Granfeldt (2000a). Jean, whose stronger language is Swedish, has a strong tendency to omit the determiner in A-N combi-

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.10 (659-726)

 Maren Pannemann

Table 5. Acquisition of prenominal elements in Jean5 Age

MLU

Total # of nouns

D missing %

D-A-N

A-N

J(1) 1;10 J(2) 2;0 J(3) 2;2 J(4) 2;4 J(5) 2;6 J(6) 2;9 J(7) 2;11 J(8) 3;1 J(9) 3;3 J(10) 3;5 J(11) 3;7 J(12) 3;9

1.6 1.3 1.5 1.3 2.0 3.5 2.8 3.0 3.0 3.5 3.6 4.3

33 89 35 46 24 49 35 49 38 54 35 70

51 86 89 76 42 14 34 20 32 13 20 5

0 0 1 0 1 2 13 16 5 7 6 5

0 0 10 3 3 0 1 4 2 2 1 1

nations until the age of 2;6 (MLU 2.0). He omits the determiner in 18 out of 20 A-N contexts (cf. Table 5). From 2;9 onwards, there is a change, and he produces the determiner in the majority of A-N contexts. Similarly, Anne and Mimi pass through a stage where the determiner is omitted in A-N combinations. Interestingly, this stage does not last as long as in Jean. Anne produces the determiner in the majority of A-N contexts from 2;6 onwards (MLU 1.9). Between the age of 2;6 and 4;0, she omits the determiner only in 3 out of 24 A-N contexts. Likewise, Mimi consistently produces D-A-N combinations from 2;2 onwards (MLU 3.2). In the period from 2;2 till 3;7, the determiner is omitted only in 1 out of 25 A-N contexts. Note that both Anne’s and Mimi’s stronger language is French. This might indicate that language dominance has an influence on the realisation of determiners in A-N contexts.

. Discussion The comparison between monolingual and bilingual French shows that the monolingual children Daniel and Grégoire have a strong tendency to realise the determiner in A-N combinations even though the determiner is still omitted sometimes. D-A-N combinations are present from early on in Daniel’s production. Similarly, Grégoire produces D-A-N combinations consistently from 2;3 onwards. In contrast, the bilingual children pass through a stage where the determiner is absent in most A-N combinations. Interestingly the bilingual children Anne and Mimi, whose stronger language is French, pass through this stage faster than Anouk and Jean. In Section 2.2, based on the model of cross-linguistic interaction by Hulk and Müller (2000), the following prediction was made: “In French-Germanic bilingual first language acquisition, the acquisition of the determiner in adjective-noun combinations in French is not affected by the simultaneous acquisition of the Germanic language.” The results of the comparison show that the bilingual children have a stronger tendency to omit the determiner in French A-N contexts than the monolin-

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.11 (726-793)

Prenominal elements in French-Germanic bilingual first language acquisition 

gual children. The observed difference suggests that cross-linguistic influence might have taken place. This influence is unexpected according to the model by Hulk and Müller (2000). The bilingual children omit determiners in their French, which suggests that there has been influence from Germanic where determiner omission in A-N combinations is common in the input. This observation is in line with previous studies on French first language acquisition of the lexical determiner (Van den Berg 2001; Kupisch 2003; Hulk 2004). These studies agree that bilingual children who acquire French simultaneously with a Germanic language need more time to acquire lexical determiners than monolingual French children. For example, Van den Berg (2001) observes that the monolingual Grégoire reaches the 90% criterion for the correct use of determiners at 2;5.1 (MLU 3.26 ) and Philippe at 2;3.0 (MLU 3.67 ). In contrast, the bilingual children Anouk and Annick reach the 90% criterion at 3;3.17 (MLU 3.6) and 3;5.8 (MLU 3.1), respectively. This delay in the acquisition of the DP might indicate that there has been cross-linguistic influence from the Germanic language. The interference would then explain why the bilingual children investigated in the present study stay in a stage where they omit the determiner in A-N combinations longer than the monolingual children. However, a different interpretation of the delay is also conceivable. The observation that bilingual children need more time to reach the next step of complexity within the DP could be caused by the fact that they have to acquire two languages at the same time, which costs more effort than learning only a single language. Van den Berg (2001) found that Anouk and Annick are not only delayed with respect to the development of determiners in French. Similarly, they reach the 90% criterion for overt determiners in Dutch at 3;4.10 (Annick) and 3;8 (Anouk) whereas the monolingual Dutch child Hein reaches the 90% criterion at 2;8.14. Still, the monolingual Dutch child Abel reaches the 90% criterion at 3;4.1. This makes it difficult to decide whether there is a general delay in bilingual determiner acquisition. To conclude this section, it seems to be the case that French-Germanic bilingual children generally need more time to acquire the determiner in A-N combinations than French monolingual children. This difference between monolingual and bilingual children could be caused by interference from Dutch. Nonetheless we cannot exclude the possible influence of other factors on this delay. The implications of this observation for the model of cross-linguistic interference by Hulk and Müller (2000) are discussed in Section 6.

. The acquisition of the attributive adjective This section investigates the acquisition of the attributive adjective in monolingual and bilingual French. Like in the preceding section, results of previous studies are used in addition to the monolingual and bilingual case studies.

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.12 (793-873)

 Maren Pannemann

Table 6. Acquisition of attributive adjectives in Daniel Age

D-A-N

A-N

D-N-A

N-A

(D1) 1;8.7 (D2) 1;8.21 (D3) 1;9.21 (D4) 1;10.7 (D5) 1;11.7

3 5 6 9 10

2 1 11 27 18

0 4 1 0 1

0 1 0 4 1

Table 7. First emergence of prenominal and postnominal adjectives in monolingual French

Daniel Grégoire Philippe

prenominal

postnominal

1;8.7 (MLU 1.5) 2;0,5 (MLU 2.1) 2;1.19 (MLU 3.1)

1;8.21 (MLU 1.52) 2;5.1 (MLU 3.2) 2;2.3 (MLU 3.5)

. Monolingual acquisition In his first recording, at 1;8.7, Daniel produces five prenominal adjectives but no postnominal adjective (cf. Table 6). However, the first postnominal adjectives can be observed already in the second recording two weeks later at 1;8.21: tracteur vert and l’eau froide. It is likely that l’eau is a fixed D-N unit as eau does not appear independently in the same recording, but l’eau froide cannot be characterised as a fixed unit since both l’eau and froide appear independently in the same recording. After the first use of postnominal adjectives, the majority of adjectives appear in the prenominal position. Notice that even though prenominal adjectives dominate, there is evidence for postnominal adjectives from early on in Daniel. Granfeldt (2000b) obverved that the monolingual French child Grégoire produces the first prenominal adjectives at 2;0,5 (MLU 2.1) (cf. also Table 3). The first postnominal adjective emerges 5 months later at 2;5.1 (MLU 3.2). Palmer (2001), who investigated the acquisition of adjectives in French monolingual children, found that Philippe (2;1–3;3), whose data are also available in the CHILDES database, produces prenominal adjectives from the first recordings available (2;1.19). Very soon after, postnominal adjectives can also be observed in Philippe (cf. Table 7). Even though it is uncertain whether there might have been instances of pre- and postnominal adjectives before the beginning of the recording, it seems to be the case that prenominal adjectives emerge slightly earlier than postnominal adjectives and prenominal adjectives dominate for a while in the consecutive files. However, there are no instances where postnominal adjectives appear incorrectly in the prenominal position. All adjectives produced by Daniel between 1;8.7 and 1;11.7 appear in the correct position. None of the 98 adjectives that appear in the prenominal position (cf. Table 2 and Appendix) is a postnominal adjective in adult French. The same holds for Philippe and Grégoire: Palmer (2001: 38) observes that at the moment Philippe and Grégoire start to use descriptive adjectives, they put them in the correct postnomi-

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.13 (873-965)

Prenominal elements in French-Germanic bilingual first language acquisition 

Table 8. Acquisition of attributive adjectives in Anouk Age

D-A-N

A-N

D-N-A

N-A

Ao 2;3.13 Ao 2;4.17 Ao 2;5.20 Ao 2;6.1 Ao 2;7.5 Ao 2;8.3 Ao 2;8.22 Ao 2;11.27

0 1 0 1 1 3 1 11

0 2 5 3 7 15 8 2

0 0 0 0 0 0 0 0

0 0 0 1 0 0 0 0

Table 9. First emergence of prenominal and postnominal adjectives in bilingual French

Jean Anne Mimi Anouk

prenominal

postnominal

2;2 (MLU 1.6) 2;6 (MLU 1.9) 2;0 (MLU 2.1) 2;4 (MLU 2.0)

3;9 (MLU 3.5) 4;2 (MLU 3.5) 2;6 (MLU 3.5) 2;6 (MLU 2.5)*

* Note that there is only instance of a postnominal adjectives until the age of 2;11.27 (cf. Table 8)

nal position. There is only one exception to this: sales mains (‘dirty hands’) where the adjective sale should be in postnominal position (Grégoire, 2;3.0). Note that sale can occur in prenominal position in adult French under certain conditions; it evokes a pejorative interpretation then.

. Bilingual acquisition In the first file available for Anouk (2;3.13), there are no combinations of adjectives with nouns (cf. Table 8). At 2;4.17, the first prenominal adjectives can be observed. Then, at 2;6.1, there is one instance of a postnominal adjective. This instance remains an exception; there are no other postnominal adjectives until the age of 2;11.27. Granfeldt (2000a/b) observes that the child Jean produces his first postnominal adjectives at 3;9 and the child Anne at age 4;2 (cf. Table 6). Both children have an MLU of 3.5 at this age. Mimi starts to use postnominal adjectives already at 2;6; she also has an MLU of 3.5 at this point. For the bilingual children in Müller (1994), there are no specifications available about the first use of adjectives. In addition, both Granfeldt (2000b: 82) and Hulk (2004: 245) observe instances where postnominal adjectives incorrectly appear in prenominal position (cf. (8) and (9)). Notice that Mimi has already started to produce postnominal adjectives before this age (cf. Table 9). (8) a.

une rose papillon a pink butterfly

(Mimi 3;7)

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.14 (965-1033)

 Maren Pannemann

b. l’orange ballon the orange balloon (9) a.

jaune serviette yellow napkin b. une verte échappe a green scarf

(Mimi 4;2) (Anouk 2;8.22) (Anouk 2;11.13)

All these cases concern colour adjectives, which can under no circumstances occur in the prenominal position in adult French. Anouk produces a total of 60 A-N sequences (cf. Table 4 and Table 8). Even though there are only two instances where the postnominal adjective is incorrectly preposed, this observation is quite striking as it cannot be observed in the monolingual data.

. Discussion The comparison between monolingual and bilingual children shows that monolingual children have a preference for prenominal adjectives, but, nevertheless, pre- and postnominal adjectives can be observed from the first files available. In contrast, postnominal adjectives appear relatively late in the bilingual children Jean and Anne. Similarly, Anouk produces just a single postnominal adjective until the age of 2;11. Only Mimi starts to use postnominal adjectives at about the same age and MLU as her monolingual peers Grégoire and Philippe. The difference between the monolingual and bilingual children lies not only in the moment of emergence but is also evident with respect to the position of the attributive adjective. As pointed out above, monolingual children produce postnominal adjectives in the correct position (cf. also Appendix). In contrast, the bilingual children Anouk and Mimi use postnominal adjectives incorrectly in the prenominal position. One could argue that these are instances of topicalisation. The sequence produced by Anouk in (9a), jaune serviette would then be interpreted as elle est jaune, cette serviette (‘it is yellow, this napkin’). However, this interpretation is excluded for the instances in (8a/b) and (9b) because in these cases the A-N combination is preceded by a determiner. As soon as there is a determiner involved, topicalisation is clearly excluded. A second argument for a difference between monolingual and bilingual children is that the only instance where Grégoire produces a postnominal adjective incorrectly in the prenominal position occurs at 2;3.0, i.e., before he acquired postnominal adjectives. As soon as he starts to produce adjectives in the postnominal position (2;5.1), he no longer place them incorrectly in the prenominal position. On the contrary, the bilingual children still make these errors after they have already produced adjectives in the postnominal position (cf. Table 9). Furthermore, the adjective sale (‘dirty’) that Grégoire incorrectly produces in prenominal position can actually precede the noun under specific circumstances in adult language. In contrast, the bilingual children use colour adjectives in prenominal position, which never precede the noun in adult French.

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.15 (1033-1088)

Prenominal elements in French-Germanic bilingual first language acquisition 

Now let us return to the second prediction made in Section 2.2: “In FrenchGermanic bilingual first language acquisition, the acquisition of the syntax of attributive adjectives in French is not affected by the simultaneous acquisition of the Germanic language.” The observed differences between monolingual and bilingual acquisition indicate that there is cross-linguistic interaction between French and Germanic. The incorrect placement of the adjective suggests that there is interference from Germanic. Mimi and Anouk acquire Swedish respectively Dutch simultaneously with French. As pointed out earlier, in both Swedish and Dutch, the attributive adjective can only appear in the prenominal position (cf. (10) and (11) vs. (8) and (9)). (10) a.

en rosa fjäril a pink butterfly b. den orange ballongen the orange balloon gele servetten yellow napkin b. een groene sjaal a green scarf

Swedish

(11) a.

Dutch

To conclude, prediction 2 must be also rejected. Notice that the observation that there is interference is not predicted by the model for cross-linguistic interference as proposed by Hulk and Müller (2000) because the syntax-pragmatics interface is not involved here. In the following section, the consequences of this observation will be discussed.

. Conclusion The observations made in our two case studies and additional observations found in previous studies in French DP acquisition have consequences for our view on crosslinguistic interaction and the model as proposed by Hulk and Müller (2000). The first prediction made in Section 2 was the following: Prediction 1: In French-Germanic bilingual first language acquisition, the acquisition of the determiner in adjective-noun combinations in French is not affected by the simultaneous acquisition of the Germanic language. We have seen in Section 4 that French-Germanic bilingual children tend to omit the determiner in A-N combinations longer than monolingual children. At first sight this suggests that in the acquisition of the French DP, there has been cross-linguistic interference from Germanic as the Germanic languages allow for determiner omission. However, does this delay in the acquisition of the determiner imply that crosslinguistic interference has taken place? Kupisch (2003) has argued that the quantitative difference between monolingual and bilingual children, i.e., the delay in the acquisition of the determiner, cannot be taken as evidence for the vulnerability of the DP as

JB[v.20020404] Prn:5/06/2006; 14:59



F: LALD4105.tex / p.16 (1088-1156)

Maren Pannemann

no qualitative differences can be observed between the monolingual and bilingual children. Another argument against cross-linguistic influence is that there seems to be a delay not only in the acquisition of French but also in acquisition of Dutch, as observed by Van den Berg in the bilinguals Anouk and Annick. To conclude, we cannot decide whether the difference between monolingual and bilingual children with respect to the acquisition of the determiner in general and the determiner in adjective-noun combinations in particular is an effect of cross-linguistic influence or whether it is just a quantitative difference caused by the additional effort that the children have to make when acquiring two languages simultaneously. In order to decide between these options, future research should concern the acquisition of the determiner in children acquiring two Romance languages simultaneously. In Romance languages, the determiner is more consistently present in the input than in Germanic languages and therefore should not affect determiner omission in the language acquired simultaneously. If we still find the same delay effect, cross-linguistic influence from the Germanic language in the French-Germanic bilingual children becomes less of an option. The second prediction investigated in this study can shed more light on the question whether the DP is vulnerable to cross-linguistic influence: We have seen that there is a difference between monolingual and bilingual French first language acquisition with respect to the position of the adjective. In contrast to French-Germanic bilingual children, monolingual children do not incorrectly produce postnominal adjectives in prenominal position. As a consequence, the second prediction made in Section 2.2 is not correct. Prediction 2: In French-Germanic bilingual first language acquisition, the acquisition of the syntax of attributive adjectives in French is not affected by the simultaneous acquisition of the Germanic language. This qualitative difference between monolingual and bilingual children with respect to the position of the attributive adjective suggests that cross-linguistic influence has taken place. This is unexpected under the model of cross-linguistic interference as proposed by Hulk and Müller (2000), which predicts interference only at the interface between pragmatics and syntax. The position of the attributive adjective does not involve this interface but concerns narrow syntax. The observation that there is nevertheless interference raises the question whether cross-linguistic influence is actually restricted to the syntax-pragmatics interface. It is conceivable that aspects of narrow syntax can also be affected by cross-linguistic interaction. This is in line with the proposal by Argyri (2005), who investigated whatembedded interrogatives with a subject and the placement of object pronouns in declaratives in Greek-English bilingual children. She found that the use of object pronouns was target-like in both languages. But the English-dominant bilinguals overused preverbal subjects in the what-embedded interrogatives in Greek. This sugests that grammatical aspects anchored in narrow syntax can also be vulnerable to cross-linguistic interference.

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.17 (1156-1218)


The present study shows that the acquisition of the French DP in a FrenchGermanic bilingual setting is different from the monolingual acquisition of the French DP. With respect to the first aspect investigated, it is problematic to decide whether the observed delay in the determiner acquisition is caused by interference or by other factors. In the research of bilingual language acquisition, criteria for cross-linguistic interference are needed, and it is necessary to specify the notion of ‘interference’. Quantitative differences alone are not enough to decide whether interference has taken place. Second, the differences with respect to the position of the attributive adjective are unexpected in the model of cross-linguistic interference as proposed by Hulk and Müller (2000). We have to reconsider the conditions for cross-linguistic influence. As argued above, this study suggests that cross-linguistic interference does not only take place at the pragmatics-syntax interface but can take place at narrow syntax, too. Furthermore, studies on bilingual acquisition have commonly assumed that there are only quantitative differences but no qualitative differences between monolingual and bilingual acquisition. The present study supports the view that there actually are qualitative differences between monolingual and bilingual acquisition.

Notes * I would like to thank Elisabeth van der Linden, Fred Weerman, Holger Hopp, Aafke Hulk, Tanja Kupisch, Hedde Zeijlstra and two anonymous reviewers for discussion and comments on earlier versions of this paper. I thank Robert Cloutier for correcting my non-native English. All remaining errors are mine. . I would like to thank Patsy Lightbown for the data. . For arguments supporting the view that peu is a noun, I refer to Kayne (2005). . One could argue that A-N combinations at this stage are fixed units. However, this cannot be confirmed as there is a considerable amount of variation among the adjectives in Daniel (cf. Appendix). . Number taken from Granfeldt (2000b: 67). . Numbers taken from Granfeldt (2000a: 272). . All MLU values for Grégoire are taken from Granfeldt (2000b). . All MLU values for Philippe are taken from Einbond (1997).

References Argyri, E. (2005). Transfer in Greek/English Bilingual Acquisition. Paper presented at ISB5 (5th International Symposium on Bilingualism), 20–23 March 2005, Barcelona. Berg, van den M. (2001). L’acquisition de l’article par des enfants francophones, néerlandophones et bilingues. MA thesis, University of Amsterdam. Chierchia, G., Guasti, M. T., & Gualmini, A. (1999). Nouns and Articles in Child Grammar and the Syntax / Semantics Map. Paper presented at GALA (Generative Approaches to Language Acquisition), 1999, Potsdam.



JB[v.20020404] Prn:5/06/2006; 14:59



F: LALD4105.tex / p.18 (1218-1415)

Maren Pannemann

Einbond, J. (1997). The Development of Children’s Responses to Questions – a Comparison of English and French. Harvard Graduate School of Education. Gawlitzek-Maiwald, I. & Tracy, R. (1996). “Bilingual bootstrapping”. Linguistics, 34, 901–926. Giorgi, A. & Longobardi, G. (1991). The Syntax of Noun Phrases. Cambridge: Cambridge University Press. Granfeldt, J. (2000a). “The acquisition of the determiners phrase in bilingual and second language French.” Bilingualism: Language and Cognition, 3(3), 263–280. Granfeldt, J. (2000b). “Le développement morphosyntaxique du syntagme nominal chez des enfants et des adultes – approche générativiste”. PERLES 9, Lund University. Hulk, A. (2004). “The acquisition of the French DP in a bilingual context.” In P. Prévost & J. Paradis (Eds.), The Acquisition of French in Different Contexts. Focus on functional categories [Language Acquisition and Language Disorders 32] (pp. 243–274). Amsterdam: John Benjamins. Hulk, A. & Linden, E. van der (1996). “Language mixing in a French-Dutch bilingual child”. Toegepaste Taalwetenschap in Artikelen, 55(2), 89–103. Hulk, A. & Müller, N. (2000). “Bilingual first language acquisition at the interface between syntax and pragmatics”. Bilingualism: Language and Cognition, 3(3), 227–244. Kayne, R. S. (2005). “Some notes on comparative syntax, with special reference to English and French”. In G. Cinque & R. S. Kayne (Eds.), The Oxford Handbook of Comparative Syntax (pp. 3–69). Oxford: Oxford University Press. Kupisch, T. (2003). “The DP, a vulnerable domain?”. In N. Müller (Ed.), (In)vulnerable Domains in Multilingualism (pp. 1–39). Amsterdam: John Benjamins. Kupisch, T. (2004a). The Acquisition of Determiners in Bilingual German-Italian and GermanFrench Children. Dissertation, University of Hamburg. Kupisch, T. (2004b). “On the relation between input frequency and acquisition patterns from a cross-linguistic perspective”. In J. van Kampen & S. Baauw (Eds.), Proceedings of GALA (Generative Approaches to Language Acquisition) LOT Occasional Series 3 (pp. 281–292). Utrecht University. MacWhinney, B. & Snow, C. (1990). “The child language data exchange system: An update”. Journal of Child Language, 17, 457–472. Müller, N. (1994). “Gender and number agreement within D”. In J. M. Meisel (Ed.), Bilingual First Language Acquisition. French and German Grammatical Development (pp. 53–88). Amsterdam: John Benjamins. Müller, N. & Hulk, A. (2000). “Cross-linguistic influence in bilingual children: Object omission and root infinitives”. In C. Howell, S. A. Fish, & T. Keith-Lucas (Eds.), Proceeding of the 24th Annual Boston University Conference on Language Development (pp. 546–557). Somerville, MA: Cascadilla. Müller, N. & Hulk, A. (2001). “Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages”. Bilingualism: Language and Cognition, 4(1), 1–21. Palmer, E. (2001). Un chapeau vert et un vert chapeau. L’acquisition du système adjectival de la langue française: une comparaison entre les enfants français et les locuteurs du néerlandais. MA thesis, University of Amsterdam. Pannemann, M. (in press). “More Variability in French L1: Consequences for theories of DPacquisition”. In M. Vliegen (Ed.), Akten des 39. Linguistischen Kolloquiums Amsterdam. Amsterdam: Lang. Roodenburg, J. (2004). Pour une approche scalaire de la déficience nominale: la position du français dans une théorie des ‘noms nus’. Dissertation, University of Amsterdam.

JB[v.20020404] Prn:5/06/2006; 14:59

F: LALD4105.tex / p.19 (1415-1470)


Appendix Prenominal and postnominal adjectives in Daniel

A-N

D-A-N

N-A

D-N-A

Daniel (1)

grand monsieur petit garçon

0

0

Daniel (2)

beau joujou petite fille

tracteur vert

l’eau froide l’eau froide l’eau froide l’eau froide

Daniel (3)

petit garçon petit chien petit tracteur petit tricycle petit train petit train petit garçon petit garçon petit garçon bon bain petit pieds petit pied petit pression petit garçon petit garçon petit fille petit garçon petite tasse petit garçon petit garçon bo choses petit bébé petit machin petite fille petit rouge petit train petit lampe petit baignoire petit bois petit bois gros bateau gros toît

un beau livre un autre café un autre-tartine le petit tracteur un petit tracteur un beau joujou un autre gâteau un autre couvercle un gros loup un gros loup un petit bébé un autre couteau un autre couteau une autre cuillière

0

l’eau chaude

robinet chaude robinet chaude robinet chaude robinet chaude

0

Daniel (4)

un petit garçon un petit joujou un grand auto le petit garçon la petit cloche une autre tasse un autre but un gros toit un autre coussin



JB[v.20020404] Prn:5/06/2006; 14:59



F: LALD4105.tex / p.20 (1470-1470)

Maren Pannemann

(continued)

Daniel (5)

A-N

D-A-N

N-A

D-N-A

petit bébé petit bébé grand lai grand verre petit lait grand lait grand lait petite fille petit bébé petite cuillère grand lait grand lait grand lait grand dame grand dame grand monsieur

un autre joujou la petit lait un tout petit bébé de tout petite main la gros tracteur un autre joujou un autre joujou un autre joujou un petit lait un petit quelque chose jaune

cheveux tout rouge

0

cheveux rouge petit quelque chose jaune

JB[v.20020404] Prn:31/03/2006; 9:09

F: LALD41P2.tex / p.1 (47-74)

 

Verbs, auxiliaries and inflection

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.1 (46-122)

A cross-sectional study on the use of “be” in early Italian Claudia Caprin and Maria Teresa Guasti Università degli Studi di Milano-Bicocca

This work aims at studying the acquisition of copula and the auxiliary BE in Italian children. It will be shown that from the earliest stages the copula and the auxiliary BE are omitted at different rate, suggesting that children distinguish between the two grammatical categories. Moreover, on the basis of an analysis of ambiguous contexts (adjectival or verbal passives), it will be claimed that earliest passive are verbal and not adjectival passives. The data come from the spontaneous productions of 59 children (ranging in age from 22 to 35 months) involved in a cross-sectional study on language acquisition.

.

Introduction*

In the earliest stages of language acquisition children typically omit function morphemes, such as articles, pronouns, copula, auxiliaries, prepositions. Such omissions have received great attention in the literature and have been the object of various analyses. However, the cause of such omissions is still matter of debate. In this paper, we will focus on the use-omission of “be” in early Italian. This morpheme, in Italian as in other languages, can serve as a copula or as an auxiliary. The copula is followed by a nominal, adjectival or prepositional predicate, while the auxiliary form selects a verbal predicate, as in passive clauses and in compound tenses with reflexive and unaccusative verbs. The other auxiliary verb is “have” that is also used in compound tenses with transitive and unergative verbs. Both “be” and “have” are fully inflected verbs with an irregular paradigm. In previous studies on early Italian (Pizzuto & Caselli 1992; Bottari et al. 1993/1994; Antelmi 1997; Nelli 1998; Leonard et al. 2002; Franchi 2004), it has been found that generally copular clauses are among the first kinds of verbal utterances produced by children, documented since early combinatorial speech (D’Odorico & Carubbi 1997). Some authors argued that schwa-like vocalic segments in combination with adjectives might be considered as antecedents of copula (Bottari et al. 1993/1994; Nelli 1998). Although copula appears early in children’s vocabulary, its full mastery requires some time and copula is frequently omitted, a phenomenon observed in other

JB[v.20020404] Prn:2/06/2006; 14:52



F: LALD4106.tex / p.2 (122-169)

Claudia Caprin and Maria Teresa Guasti

early languages, as for example English (Brown 1973; de Villiers & de Villiers 1973; Becker 2000). Other errors concern number agreement and the most common error is the substitution of the third plural form with the third singular person form (Antelmi 1997; Bortolini et al. 1997; Leonard et al. 2002). Generally, the auxiliary “be” emerges later than the copula (Antelmi 1997; Nelli 1998), but this might be a side effect of the fact that auxiliaries are mostly used in compound past tenses and these forms emerge later than simple tenses. The different auxiliary forms “be” and “have” might not appear simultaneously, as sometimes “be” emerges earlier than “have” (Antelmi 1997; Nelli 1998). When children start to produce compound tenses, substitution errors of the auxiliaries “be” and “have” may occur, even if it is quite rare (Pizzuto & Caselli 1992; Antelmi 1997; Nelli 1998; Lorusso 2003). This study aims at exploring the development of the morpheme “be” in early Italian on a larger population than the ones that have been studied so far. Unlike previous studies, we used a cross-sectional method that has the advantage of allowing us to investigate the speech of several subjects. We examined the spontaneous speech of 59 Italian speaking children between the age of 22 to 35 months. Although there are differences among the subjects, some very interesting patterns emerge in the use and omission of the morpheme “be”. First, we show that children clearly distinguish between copula and auxiliary “be” and treat this last on a par with the auxiliary “have”. This is evident in the frequency of omission of the morpheme “be”. Children omit the copula less than the auxiliary “be”, but omit the latter as much as the auxiliary “have”. This fact, we argue, is evidence that children pay attention to the syntactic distribution of elements more than to phonological similarities. Children’s acquisition of the copula is gradual, but nevertheless is subject to some syntactic constraint. As already found by Franchi (2004) on a smaller population of children, also in our study, the copula is omitted more often in main declarative clauses than in wh-questions, that is, omissions are sensitive to the syntactic structure. Taken together, these facts suggest that children are biased to find out the regularities of their language through the lentils of the syntactic structure. In other words, children are not looking for whatever kind of regularities, but only those that are syntactically relevant. Similarly, children omit in a way that is sensitive to the syntactic computation in which elements participate. Inspired by Jakubowicz and Nash (2001), we claim that children omit auxiliaries more often than the copula because the computation in which the former are involved is more costly than the computation in which the latter is involved. In most cases, the grammatical category of “be” (copula or auxiliary) can be unambiguously established. However, in some contexts “be” is ambiguous, as in short passive sentences (e.g., la porta è chiusa, lit. “the door is closed”). By looking separately at the omissions of “be” in these contexts, we were able to show that children treat it as an auxiliary. In turn, this means that children are unlikely to treat short passives as adjectival passives in which “be” selects an A(djectival) P(hrase). Rather, “be” is an auxiliary that selects a verbal phrase. Our conclusion is in contrast with Borer and Wexler’s (1987) claim that for children verbal passives are adjectival passives. However, it is compatible with Hirsch and Wexler’s (2004) claim, based on work by Embick (2004), that children’s passives

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.3 (169-230)

A cross-sectional study on the use of “be” in early Italian

are resultative passives, in which “be” selects a V(erbal) P(hrase). Although our data come from production, we think that this does not make them less dispensable and as we will show more extensively it may have interesting consequences for the analysis of children’s difficulties in comprehending passive sentences and may shed light on competing theories.

. Method . Subjects and linguistic corpora Fifty-nine children (25 males and 34 females), ranging in age from 22 to 35 months (Mean age = 28.92, sd = 2.8), were both videoteped and audiotaped in a semi-natural setting with the caregivers at their nursery school. The same set of toys including various animals, a barn and some props was provided to the caregivers and was used to set up the interactions with the child. Thus, the discussion was about what was happening or what the animals were doing. Each recording session lasted about 20–25 minutes. All the children were Italian and the socio-economic status of their parents was identified as low (29%), medium (47%), and high (29%), taking into account the educational and professional level of both parents (Entlewisle & Astone 1994; Miller 1991; Hauser 1994). Each tape was transcribed by one listener and checked by another; the doubtful portions were listened again by both coders together, and when an agreement between the two was not reached, the portions were labelled as unintelligible. Conventional Italian orthography was used. CHAT format of CHILDES coding system was adopted for all the transcriptions (MacWhinney 1997).

. Group composition In order to compare their linguistic development, children were divided into three groups based on MLU-w values: from 1.0 to 1.49 (G1 N = 15); from 1.5 to 1.99 (G2 N = 19); from 2.0 to 3.1 (G3 N = 25). MLU computed in words was used, instead of a computation based on morphemes which is very problematic in Italian because there is a lot of syncretism (Bates 1976; Slobin & Bever 1982; Pizzuto & Caselli 1992). Age was discarded as a criterion for grouping because at the ages considered there is a lot of individual variations (Caselli & Casadio 1995).

. Criteria of analysis We examined all clauses that included or should have included the morpheme “be”. Despite their allomorphy, copula and auxiliary “be” are two distinct morphemes and their distributional pattern was investigated separately. We also analysed the use of the auxiliary “have” in the compound tense Passato prossimo (the only compound tense used in our corpus). We identified: (a) obligatory contexts, (b) morpheme pro-



JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.4 (230-293)

 Claudia Caprin and Maria Teresa Guasti

ductions, without distinguishing between “filled forms” and “phonetic reductions” (including “phonetic approximations” and “Monosyllabic Place Holders” MPHs – see Bottari et al. 1993/1994); (c) errors (wrong number or person agreements; substitutions of the auxiliaries forms). Concerning the copula, Franchi (2004) pointed out that sometimes the identification of obligatory contexts for the copula is difficult, since Italian is a null-subject language and allows elliptic structures. For example, the phrase il gatto (the cat) can result from copula omission or be a genuine elliptical form in which the copula is not required. The question What is this? can have the following three grammatical answers: (1) Questo è il gatto ‘This is the cat’ (2) è il gatto is the cat ‘(it) is the cat’ (3) il gatto the cat The clause in (1) includes “Subject + copula + Nominal Predicate”, the one in (2) is a partial structure with a null subject: “copula + NP” and the clause in (3) is an elliptic structure, in which the copula is not required. We did not include (3) within omission cases, like unless the preceding context made it clear that it was a case of omission of the copula. Cases like (1) and (2) were both counted as cases of use of the copula. Given that (3) is sometimes ambiguous in that it could result from omission of the copula from a sentence like (2) or it could be an elliptical sentence and the fact that we did not include these ambiguous cases, our results may be skewed toward copula use. Copula use-omission was counted separately in declarative and interrogative sentences. The latter were further divided between “WH-questions” and “Yes/Noquestions”. “Yes/No-questions” in Italian have the same surface order as declarative clauses, differing only prosodically, as shown in (4); by contrast, “WH-questions” have a different word order, as shown in (5). (4) a.

Anna mangia la mela ‘Anna is eating the apple’ b. Anna mangia la mela? Anna is eating the apple? ‘Is Anna eating the apple?’

(5) Che cosa mangia Anna? What is eating Anna? ‘What is Anna eating?’ Auxiliaries “be” and “have” are used in compound tenses, such as the Passato Prossimo, as in Gianni ha mangiato (John has eaten). In this example, “have” expresses finiteness and tense, specifically present tense, and the resultative meaning of the past action is

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.5 (293-391)


expressed by the Past Participle of the lexical verb. “Have” is selected by transitive and unergative verbs, while “be” is requested by unaccusative verbs and is also used with passives and reflexive verbs. In this study, we examined the present form of the copula (the only tense present), the auxiliary “have” and “be” in the Passato Prossimo, where they are also used in the present tense, and the auxiliary “be” in passive constructions (also expressed as a present tense). It is important to point out that in Italian the past participle of transitive verbs in clauses such as pulita (cleaned) in (6) can either be analysed as a verb or as an adjective, that is, sentence (6) can either be an instance of adjectival passive or a short verbal passive. In the former case, “be” is a copula, as in “John is happy”, while in the latter it is an auxiliary. In our study, ambiguous forms were categorized separately. (6) la cucina è pulita ogni sabato ‘the kitchen is cleaned every Saturday’ In summary, we separately analysed copula “be” and auxiliary “be” (in the Passato Prossimo of unaccusative and reflexive verbs and in clauses that are ambiguous between verbal or adjectival passives); we also examined the distribution of the auxiliary “have” in the Passato Prossimo. Proportions of occurrence-omission of “be” and “have” in the different contexts were calculated for every child with the following formula: frequency of morpheme production on the total number of obligatory contexts of the morpheme. These data were then used for statistical analysis, computed by SPSS-10.1 automatic package. Unless otherwise specified, only morphemes for which there were at least 3 obligatory contexts of use were included in the analysis, following standard practice (see e.g., de Villiers & de Villiers 1973; Pizzuto & Caselli 1992; Cipriani et al. 1993 for a discussion about this kind of criterion).1 This is a measure to control for productivity of a given morpheme and to discard the possibility that what the child is producing are formulaic expressions. In longitudinal studies the point of acquisition of each morpheme is usually defined by its presence in at least 90% (Cazden 1968; Brown 1973; Pizzuto & Caselli 1992) or in at least 75% of obligatory contexts (Cipriani et al. 1993; Nelli 1998) for three consecutive speech samples. These values are conventional (Stromswold 1996) and they should be matched by measures concerning variability of morpheme use (Guasti 1993/1994; Baumgartner et al. 2000). In our study a morpheme was considered acquired when it was present in at least 75% of the obligatory contexts for each child (remember that in consequence of the cross-sectional study design we had only one speech sample for every child) and, as said earlier, in at least 3 contexts of use. When necessary, we also required variety of contexts of use.

. Results Let us start by examining the use-omission of copula “be”. Table 1 reports the correct use of the copula by the three groups of children. We may notice that the most frequent form is the 3rd person singular form already used in G1. The 3rd person plural appears



JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.6 (391-463)

 Claudia Caprin and Maria Teresa Guasti

Table 1. Correct use of agreement between the copula and its subject (for each inflectional form we report the percentage of correct use, the number of correct forms/total number of forms produced) GROUP

Ia p.sg

IIa p.sg

G1

N.C. 0/0 N.C. 0/0 100 2/2 100 2/2

N.C. 0/0 N.C. 0/0 100 2/2 100 2/2

G2 G3 Total

COPULA IIIa p.sg 100 49/49 96.8 181/187 97.65 374/383 97.58 604/619

Ia p.pl

IIa p.pl

IIIa p.pl

N.C. 0/0 N.C. 0/0 N.C. 0/0 N.C. 0/0

N.C. 0/0 N.C. 0/0 N.C. 0/0 N.C. 0/0

100 1/1 100 10/10 96.72 59/61 97.22 70/72

Table 2. Copula productions N G1 G2 G3 Total

11 17 24 52

Singular forms Mean SD .588 .733 .853 .758

.223 .181 .183 .214

N 3 6 15 24

Plural forms Mean DS 1.00 .916 .850 .556

0.00 .204 .239 .207

N 12 17 24 53

Total forms Mean DS .556 .738 .846 .746

.207 .152 .169 .204

in G2 and increases in G3. The other forms are almost absent, likely due to the fact that the interaction was such that it did not invite the use of these forms. Agreement errors between the copula and its subject seldom occur (between 2– 3%). We find substitution of 3rd pl with the 3rd sg morpheme. When copula is used, it is correctly used most of the time, i.e., with the correct inflection. However, the copula is initially often omitted and its use increases with MLU-w. Table 2 summarizes the main results. For each group, we reported the number of subjects who presented copula in at least three contexts of use (N), the mean (M) and the standard deviation (SD). The mean is calculated by first computing the child’s proportion of use of the morpheme in obligatory contexts and then by averaging in the group. These data were provided for singular and for plural forms separately and grouped together under the column “Total forms”. The analysis reveals that there is a significant difference between groups when we consider singular and plural forms together (Kruskal-Wallis N = 53, χ2 = 17.304, df = 2, P < 0.001) and when we consider only singular forms (Kruskal-Wallis N = 52, χ2 = 16.139, df = 2, P < 0.001). No difference is found between groups when only plural forms are examined. Furthermore, we computed the number of children who produced the copula over 75% of the time in obligatory contexts and found a significant association between the variable Group (the three groups of children) and copula acquisition (two categories: acquired versus not-acquired) (χ2 = 7.583, N = 47, df = 2,

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.7 (463-539)

A cross-sectional study on the use of “be” in early Italian 

Table 3. Copula production in declarative and WH-question contexts Group

N

G1 G2 G3 Total

10 17 24 51

Declarative sentences Mean SD .466 .617 .800 .674

.241 .225 .197 .249

N 2 10 17 29

WH-questions Mean SD 1.00 1.00 .944 .967

0.00 0.00 .137 .107

P = 0.023). This result indicates that the number of children who reached 75% of use in obligatory contexts varies as a function of MLU-w. Let us turn to the analysis of the copula use in declarative and wh-questions. Data are reported in Table 3, where the number of children who presented at least one context of use (N), the mean value (M) and the standard deviation (SD) for declarative and WH-question contexts are displayed. The mean is calculated by first computing the child’s proportion of use of the morpheme in obligatory contexts and then by averaging in the group. Analysis (where no group distinction was made) showed that copula production is greater in WH-questions than in declaratives contexts (Wilcoxon N = 30, z = –4.485, P < 0.001), as also shown by Franchi (2004) on a different corpus.2 To ensure that this result is sound and that copula is productively used in WHcontexts or alternatively that sequences wh-word+copula are not formulaic expressions, we repeated the analysis selecting only data coming from children who used the wh-words with two distinct inflected verbs, say a third person singular form of “be” and another singular or plural person form of “be” or of a lexical verb. Again a higher omission of copula in declarative contexts was found (Wilcoxon N = 18, z = –3.114, P = 0.002). Examples of copula use and omission are given in (7) and (8). (7) a.

c’è il coccodrillo lì ‘there is the crocodile there’ b. è grosso quel gioco is big that toy ‘That toy is big’ c. senti, quelle sono finestre? listen, those are windows? ‘listen, are those windows?’

(8) a.

non lo so (è) dentro don’t it-acc know (is) inside ‘(I) don’t know it, (it) (is) inside’ b. (sono) in quel sacchetto gli animali (are) in that bag the animals ‘The animals (are) in that bag’ c. quella (è) mucca that (is) cow

(Brando) (Claudio)

(Guido)

(Alessandro)

(Chiara)

(Andrea)

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.8 (539-647)

 Claudia Caprin and Maria Teresa Guasti

In summary, copula is used more often as children’s MLU increases, especially in the singular forms. Plural forms emerge later than singular ones, as it is generally the case with plurals. Omissions are sensitive to the structural environment: more omissions are found in declarative than in wh-questions, even when one controls for productivity. Let us turn our attention to auxiliaries “be” and “have”, whose uses and omissions are exemplified in (9): (9) a.

è caduta (she) is fallen ‘(she) fell’ b. hanno mangiato (they) have eaten ‘(they) ate (it)’ c. no ha fatto un pastiscio [= pasticcio] No (he) has made a mess ‘no (he) made a mess’

(10) a.

dopo (sono) andati tutti in spiaggia eh! After (they) (were) gone all to beach eh! ‘then all (they) went to the beach eh!’ b. sì si (è) svegliato yes SI (is) woken up ‘yes (he) woke’ c. ecco (la) (ho) messa There, (it-acc-fem-sg) (have) put-fem-sg ‘There (I) put (it)’

(Alessia)

(Anna)

(Edoardo)

(Carlo)

(Brando)

(Giulio)

As in the case of the copula, errors are rare (3%). However, omissions are often found. Table 4 presents a summary of the use of the auxiliary “be” and “have”.3 Notice that for the morpheme “be”, we kept separate cases in which the auxiliary is used in the passato prossimo of unaccusative and reflexive verbs and cases in which it is used in short passives (that are ambiguous between a verbal or an adjectival passive). We reported the number of children who present at least one contexts of use (N), the mean (M), and the standard deviation (SD). The mean is calculated by first computing the child’s proportion of use of the morpheme in obligatory contexts and then by averaging in the group. Since the requirement of at least three contexts of obligatory use of the auxiliary “be” was not met in the speech of several children in G1 and G2, we collapsed data from these two groups together and compared them with data from G3. The analysis showed no significant difference among groups, i.e., children in G3 did not use auxiliary “be” more than children in G1 or G2. When we compared the use of “be” as copula and as an auxiliary (no distinction among MLU-w groups was made), we found that the proportion of copular forms in obligatory contexts is higher than the corresponding proportion of the auxiliary form (Mann-Whitney N = 28, z = –1.367,

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.9 (647-689)

A cross-sectional study on the use of “be” in early Italian 

Table 4. Production of auxiliary “be”, of the morpheme “be” in ambiguous contexts (short passives) and of auxiliary “have”

G1 G2 G3 Total

Auxiliary “be” in unambiguous contexts N Mean SD

Morpheme “be” in ambiguous contexts* N Mean SD

Auxiliary “have”* N

Mean

SD

3 6 19 28

4 8 14 26

3 10 23 36

.333 .586 .645 .602

.577 .416 .344 .381

.000 .618 .659 .579

.000 .424 .272 .353

.500 .535 .648 .590

.577 .387 .365 .394

* Since few children showed at least three contexts of use of auxiliary BE in ambiguous contexts and of auxiliary HAVE, all children data without any restriction concerning the number of obligatory contexts are reported

P = 0.019). Finally, copula use is also higher than auxiliary “have” (Wilcoxon N = 36, z = –2.530, P = 0.011), but the use of the auxiliary “be” and “have” do not differ from each other.4 These results are very compelling and suggest that children distinguish between the two morphemes “be”, despite their allomorphy, but consider the auxiliaries “be” and “have” as members of the same class (auxiliary), despite their phonetic difference. Finally, we examined ambiguous contexts, i.e., short passives, in which “be” could be considered a copula or an auxiliary. In the former case, the sentence would be an instance of the adjectival passive and in the latter an instance of the verbal passive. Some examples are given in (11) and (12). (11) a.

questi sono spaventati ‘these are frightened’ b. è cotta però (it) is cooked though ‘(it) is cooked though’

(12) a.

adesso (è) finito ‘now (it) (is) finished’ b. (sono) morti (are) dead ‘(they) (are) dead’ or ‘(they) died’

(Beatrice) (Alessia)

(Giulio) (Francesca)

Since only few children satisfied the requirement of having at least three contexts of use, we did not apply this criterion and collapsed all the data together. Analysis showed that the production of “be” in ambiguous contexts was lower than the production of the copula (Wilcoxon N = 26, z = –2.087, P = 0.037), but did not differ from the production of auxiliary “be” (Wilcoxon N = 24, z = .654, P = 0.513) and of auxiliary “have” (Wilcoxon N = 19, z = .035, P = 0.972). These results are very interesting because they suggest that in ambiguous contexts “be” is a member of the auxiliary class; since auxiliary “be” is found in verbal passives, these ambiguous sentences are to be taken as instances of verbal passives.

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.10 (689-756)

 Claudia Caprin and Maria Teresa Guasti

. Discussion In this section, we are going to discuss the main results of our study, i.e. the discrepancy in the acquisition of singular and plural forms of the copula; that omission of “be” is subject to structural constraints; that the pattern of “be” omission is sensitive to the syntactic function; finally, that ambiguous contexts in which “be” could either be the copula or the auxiliary pattern like unambiguous contexts in which the auxiliary is used. Copula production increases as does the number of children who acquired the morpheme, reaching the 75% of use in obligatory contexts, from G1 to G3, i.e., as a function of MLU-w. Interestingly, some differences between singular and plural forms were observed: singular forms increase gradually among the three groups, while plural forms appear at the first MLU-w group rarely, in the other groups their production is very high. In other words, although the plural form emerges later, when it does, it is used consistently. Since lack of plural forms initially can hardly be taken as absence of knowledge of these forms, especially because errors are rare, we concentrate on the discrepancy in the development of the singular and plural forms. Children initially use singular forms and do this almost errorless, but often omit them for reasons we will discuss later; omission gradually decreases. By the time children start to use plural forms, omission has decreased and therefore we get the impression that the use of plural forms is more consistent than the singular one, but this is just a side effect of the fact that plural forms are employed later (Caprin 2003). At first sight, this result is at odds with a previous finding based on a morpheme elicitation task with 2 to 7 years old Italian children (Leonard et al. 2002). Leonard et al. found that the production of the 3rd person plural form was more difficult than the production of the 3rd person singular, as children omitted or substituted it more often. The discrepancy between our result and Leonard et al.’s can be understood by noticing that in elicitation tasks children are bound to use a specific morpheme, as they are prompted to retrieve the specific form when required by the experimenter, while in spontaneous speech this retrieval is under children control. It is known in the processing literature that retrieving a form may be difficult and is subject to a variety of factors, such as frequency or age of acquisition (see, e.g. Morrison, Ellis, & Quinlan 1992). Since singular forms are employed earlier than plural forms, as seen before (see also Pizzutto & Caselli 1992), it is not surprising that their retrieval in tasks in which this is not under children’s control is more difficult and is likely subject to age of acquisition effect. Thus, in our view, the discrepancy between Leonard’s et al. and our results is a consequence of the different task demands. Although we found that the use of the copula is gradually acquired, nonetheless its omission is a systematic phenomenon. By comparing the production of the copula in declaratives and in wh-questions, we were able to show that omission was more frequent in the former case than in the latter. This result still holds when we control for the formulaic use of these wh-questions with the copula: by limiting the analysis to those children who showed a productive use of the copula, the discrepancy between the two syntactic contexts is still present. This result confirms what was found

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.11 (756-811)

A cross-sectional study on the use of “be” in early Italian 

in Franchi (2004) on a longitudinal corpus and suggests that omission of the copula is subject to structural constraints. Following Franchi, we propose that the asymmetry in copula omission between declarative and wh-questions can be explained in terms of Rizzi’s Truncation Model (1993/1994, 2000). Rizzi argues that, while in the adult grammar CP is the root of all the clauses, so that all the clauses (both finite and infinitive) have a uniform syntactic representation, in child grammar this constraint is not fully operative or is applied optionally (for example, it does not apply when children computational resources are already absorbed by the elaboration of other aspects of the clause). As a consequence, children might generate both CP clauses, but also clauses that are structurally reduced, i.e. that lack the IP layer. The truncation mechanism operates at the top level of the structural hierarchy of the syntactic tree and deletes the projections above its site, but it cannot remove projections from the inside of the syntactic tree: every projection dominating the truncation site is missing, while every projection dominated by the truncation site must be present. Copula expresses tense and agreement and clauses missing it can be analysed as reduced structures with truncation at the IP layer. This is what happens in declarative clauses missing the copula. Thus, under this analysis, the structure of a sentence like (8c) repeated below is (13): (8) c.

(Andrea)

quella (è) mucca that (is) cow

(13)

IP Spec

I’ I è

NP Spec quella

N’ mucca

Truncation of the IP layer cannot apply in wh-questions because the CP layer needs to be projected to host the wh-operator and thus the full structure below CP is also projected. Since IP is present it must be filled, and thus the copula is produced to be inserted into the IP head (see also Franchi 2004).5 “Be” is used in copular constructions and as an auxiliary in compound tense. Despite the allomorphy between the two forms of “be”, their use in children’s speech is not the same. Auxiliary “be” is omitted more than copula “be”, but at the same rate as auxiliary “have”. In other words, children treat copula “be” as distinct from auxiliary “be” and the latter as similar to auxiliary “have”. This suggests that Italian children distinguish from early on between two grammatical functions of the morpheme “be” and that syntactic category is more salient than phonetic form, despite the fact that syntactic category is a more abstract notion. A greater production of copula with respect to auxiliary “be” was also observed in longitudinal studies. Examining the corpus of an Italian-speaking child, Nelli (1998) found a temporal discrepancy between the ac-

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.12 (811-842)

 Claudia Caprin and Maria Teresa Guasti

quisition of the copula and the auxiliary “be”: copula was acquired (present in 75% of obligatory contexts) at about 23 months, while the auxiliary “be” wasn’t productive before 25 months yet. Cipriani et al. (2001), who examined the longitudinal spontaneous speech of two young children and of eleven subjects with SLI, found a greater omission of the auxiliary “be”. Moreover, they showed that SLI children omit the auxiliary “be”, but not the copula, for a long time. The discrepancy between the auxiliary “be” and the copula is also attested both in normal children and in children with SLI in other early languages, such as English (see Joseph, Serratrice, & Conti-Ramsden 2002). It was also observed in elicited production tasks by Cleave and Rice (1997) for English and by Jakubowicz and Nash (2001) for French. This robust finding, replicated in several early languages, suggests that phonemic similarity does not help children in the generalization of the use of “be” and thus that phonological reasons can hardly be invoked to explain the omission of “be”. Children operate under the hypothesis that there are two distinct categories, copula and auxiliary (“have” and “be”), which undergo distinct computations. Both copula and auxiliaries are omitted, but at different rates. For the copula, omission has been explained by invoking the truncation hypothesis. This hypothesis can be extended to auxiliaries. Thus utterances like (10a), repeated below, are analyzed as truncated structure at the level of VP (this is a simplification that is enough for our current purposes), as displayed in (14).6 (10) a.

dopo (sono) andati tutti in spiaggia eh! After (they) (are) gone all to beach eh! ‘then (they) all went to the beach eh!’

(14)

(Carlo)

IP Spec

I’

I (Sono)

VP Spec

V’ V andati

Truncation does not explain the different rates of omission, however. To explain this we must examine the difference between auxiliaries and copula. Auxiliaries are used in compound tenses, such as Passato Prossimo, and participate in the expression of tense and agreement features in conjunction with the past participle. This means that tense is expressed by two distinct items in compound tenses. (e.g., “has” is just present, but “has eaten” is past). The copula, instead, directly expresses tense and agreement feature in simple tenses. In this respect, the copula is like a lexical verb in simple tenses. This difference in the expression of tense (and agreement) features is responsible for the difference between copula and auxiliaries in the speech of children. Based on the

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.13 (842-933)

A cross-sectional study on the use of “be” in early Italian 

insight by Jakubowicz and Nash (2001), we claim that children omit auxiliaries more often than the copula because of the different complexity of the syntactic computation of tense (and agreement) feature in compound tenses. As these features are expressed by two items in compound tenses (the auxiliary and the past participle), more computational resources are required. Thus, the computation of tense (and agreement) features is more costly in compound tense, and this led children to drop the auxiliary more often than the copula. One final finding that deserves discussion is the behaviour of “be” in ambiguous contexts, that is, in the case of short passives (lacking the by-phrase expressing the agent of the event). The sentence in (6), repeated below, is an example. (6) la cucina è pulita ogni sabato ‘the kitchen is cleaned every Saturday’ This sentence can be taken as an instance of verbal passive, as it is the case for long passives including the by-phrase, or as an instance of adjectival passive. In the former case, its analysis is the one of verbal passive, i.e., the auxiliary “be” located in I selects a verbal phrase hosting the past participle and movement to the preverbal position of the theme NP “la cucina” occurs in the syntax. In the latter case, “be” is the copula and selects an adjectival phrase headed by the past participle. Adjectival passives are formed in the lexicon through an operation that changes the grammatical category of the past participle from V(erb) to A(adjective), suppresses the external argument that can no longer be projected in the syntax and externalize the internal argument (here la cucina) that is directly projected into the subject position (Wasow 1977). Thus, adjectival passives do not involve movement and are similar to copular construction, in which the predicate is an adjective, as in Gianni è felice (Gianni is happy). Our analysis revealed that omission of “be” in these ambiguous contexts behave like omission of the auxiliaries “be” (with unaccusative and reflexives) and “have” and is different from omission of the copula, i.e., in these contexts omission is higher than copula omission, but is not different from omission of both auxiliaries “be” and “have”. This suggests that “be” in ambiguous contexts is categorized as an auxiliary by Italian children. Since auxiliary “be” selects a verbal phrase, this piece of evidence suggests that short passives are taken to be instances of verbal passives and not of adjectival passives. This conclusion has consequences for the proper analysis of children’s passives. It has long been observed that children tend to omit the by-phrase that is usually present in verbal passives and tend to produce passives that have a stative meaning (Horgan 1978); second, in languages in which adjectival and verbal passives are morphologically distinct, the adjectival passives are produced earlier than verbal passives (Berman & Sagi 1981). These features of children’s passives led Borer and Wexler (1987) to suggest that the adult analysis of verbal passives was unavailable to children, as it involves the formation of an A-chain to connect the internal argument position to the subject position (Spec, IP). Since children cannot form A-chains, they cannot represent verbal passives and have to analyze these with the closest UG-approximation, i.e., they have to analyze verbal passives as adjectival passives (which do not involve A-chains). The adjectival

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.14 (933-957)

 Claudia Caprin and Maria Teresa Guasti

account of children’s passives is compatible with the properties of children’s passives discussed above. As a matter of fact, adjectival passives do not take the by-phrase and have a stative reading. As pointed in Hirsh and Wexler (2004) the adjectival account of children’s passives has theoretical drawbacks. For example, adjectival passives convey a stative interpretation. Although this accords well with Horgan’s finding, it is not always the case that children’s passive report just states. In addition, Pinker et al. (1987) showed that children produce passives that do not have a stative reading. Our data show that the adjectival account is also empirically inadequate, as children do not treat short passives as adjectival passives. Remember that adjectival passives are like copular constructions as in “John is sick” and thus that “be” is a copula that selects an AP. If short passives were analyzed as adjectival passives we would have expected omission of “be” in these constructions to pattern with omission of the copula and not with omission of the auxiliary, but this was not the case. Thus, we conclude that children’s short passives are not adjectival passives. This conclusion fares well with Hirsch and Wexler’s recent observation and call for a new account of children’s passives. Hirsch and Wexler propose an alternative analysis of children’s passives, couched in minimalist terms, based on Embick (2004). According to this author there are three kind of passives rather than two: adjectival passives conveying a stative reading (e.g., the door is open), resultative passives (e.g., the door is opened) and eventive passives (e.g., the door is opened by John). Resultative passives convey the meaning of something being in a state as a result of some process. The linguistic representation of resultative passives includes an event whose occurrence results in a target state (the door results in the state of being opened). Interestingly, in resultative passives, as in verbal passives, “be” selects a VP or in minimalist terms a vP, that is, “be” is an auxiliary. Hirsch and Wexler propose that children’s passives are analyzed as resultative passives. This proposal accounts for the fact that children’s passive involve an eventive component and it also accounts for our result that children’s passives are not adjectival passives and that their representation must include a VP phrase. Interestingly, resultative passives differ from eventive passives and, simplifying a bit, they are similar to adjectival passives in that the internal argument is projected to the external argument position, rather than as sister to the verb. In this way, resultative passives do not involve the kind of A-chain that is present in eventive passive and is problematic for children. In conclusion, through the analysis of the morpheme “be” in the semi-natural production data of 59 Italian speaking children, we were able to show that the omission of the copula decreases as MLU-w increases. Omission, however, is not random, but is governed by structural restrictions and by the categorical status of the morpheme “be”. First, omission is more frequent in declaratives than in wh-questions, a fact that can be accounted for within the Truncation approach. Second, children omit copula less than auxiliary “be”, but they omit auxiliary “be” as much as auxiliary “have”. This is an indication that omission is constrained by the grammatical status of the object and it is determined by the computational complexity needed to derive the grammatical representation associated with the given grammatical object. Finally, we found a striking similarity in the pattern of omission of “be” in ambiguous short passives

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.15 (957-1028)


and non-ambiguous auxiliary contexts; we interpreted this similarity as evidence that short passives could not be analyzed as adjectival passives, as Borer and Wexler’s (1987) analysis held. Our results are compatible with and add evidence to Hirsch and Wexler’s (2004) proposal, based on Embick (2004), that children’s passives are analyzed as resultative passives. In so doing, we have shown that early production data, when properly analyzed, may offer evidence to test the different prediction of competing grammatical theories of children’s acquisition.

Notes * Although the article has been jointly written and the analysis were conducted jointly, Claudia Caprin takes responsibility of Sections 1, 2, 3 and Maria Teresa Guasti takes responsibility of Section 4. . Some authors, included those cited in the text, recommend to count only morphemes for which 5 obligatory contexts of use can be identified. The figures (3 or 5) are arbitrary, but the rationale behind them is to impose a minimal requirement of productivity. . Yes/No questions were not included since they were rare. . Although in G1 there are only two children who produce wh-questions, recall that these are the linguistically less developed children (MLU between 1–1,5). The asymmetry between whquestions and declarative is clearly evident when we consider G2 and G3, which include many more children who produce both types of sentences. . Given that few children showed at least three contexts of use of auxiliary “have”, we included all children data in the analysis without any restriction concerning the number of obligatory contexts. . A prosodic explanation of the asymmetry in copula omission can be rejected on the following ground. While it could be argued that in wh-question the copula can form a (trochaic) foot with the wh-element, this could also happen in declarative clauses as in Lia è bella (Lia is nice).

References Antelmi, D. (1997). La prima grammatica dell’italiano. Bologna: Il Mulino. Bates, E. (1976). Language and Context: The Acquisition of Pragmatics. New York: Academic Press. Baumgartner, E., Devescovi, A., & D’Amico, A. (2000). Lessico psicologico nell’infanzia. Roma: Carocci. Becker, M. K. (2000). “The acquisition of the English Copula”. In S. C. Howell et al. (Eds.), BUCLD 24 Proceedings (pp. 104–115). Somerville, MA: Cascadilla Press. Berman R. & Sagi, I. (1981). “On word formation and word innovation in early age”. Balshanut Ivrit Xofshit 18. Borer, H. & Wexler, K. (1987). “The maturation of syntax”. In T. Roeper & E. Williams (Eds.), Parameter Setting. Dordrecht: Reidel.



JB[v.20020404] Prn:2/06/2006; 14:52



F: LALD4106.tex / p.16 (1028-1135)

Claudia Caprin and Maria Teresa Guasti

Bortolini, U., Caselli, M. C., & Leonard, L. B. (1997). “Grammatical deficits in Italian-speaking children with specific language impairment”. Journal of Speech, Language and Hearing Research, 40, 809–820. Bottari, P., Cipriani, P., & Chilosi, A. M. (1993/1994). “Protosyntactic devices in the acquisition of Italian free morphology”. Language Acquisition, 3, 327–369. Brown, R. (1973). A First Language. The Early Stages. Cambridge, MA: Harvard University Press. Caprin, C. (2003). La grammatica nei primi discorsi – Uno studio trasversale sull’acquisizione della morfo-sintassi in Italiano. Unpublished PhD Dissertation. University of Pavia, Italy. Cazden, C. B. (1968). “The acquisition of noun and verb inflexions”. Child Development, 39, 443–448. Caselli, M. C. & Casadio, P. (1995). Il primo vocabolario del bambino: Guida all’uso del questionario MacArthur per la valutazione della comunicazione e del linguaggio nei primi anni di vita. Milano: Franco Angeli. Cipriani, P., Chilosi, A. M., Bottari, P., & Pfanner, L. (1993). L’acquisizione della morfosintassi: Fasi e processi. Padova: Unipress. Cipriani, P., Bottari, P., Chilosi, A. M., & Pfanner, L. (2001). “A longitudinal perspective on the study of specific language impairment: The long term follow-up of an Italian child”. International journal of language and communication disorders, 7, 245–280. Cleave, P. L. & Rice, M. L. (1997). “An examination of the morpheme BE in children with specific language impairment: The role of contractibility and grammatical form class”. Journal of Speech, Language and Hearing Research, 40, 480–492. de Villiers, J. & de Villiers, P. (1973). “A cross-sectional study of the acquisition of grammatical morphemes in child speech”. Journal of Psycholinguistic Research, 3, 267–378. D’Odorico, L. & Carubbi, S. (1997). “Dalle espressioni di una sola parola alle prime combinazioni di parole. Forme di transizione linguistica nel processo di acquisizione della lingua italiana”. Età evolutiva, 26–39. Embick, D. (2004). “On the syntax of resultative participles in English”. Linguistic Inquiry, 25, 355–39. Entwisle, D. & Aston, N. M. (1994). “Some practical guidelines for measuring youth’s race, ethnicity and socioeconomic status”. Child Development, 65, 1521–1540. Franchi, E. (2004). “Copula omission as evidence for truncation in Italian child language”. Paper presented at Second Lisbon Meeting on Language Acquisition 1–4 June, Lisbon, Portugal. Guasti, M. T. (1993/1994). “Verb syntax in Italian child grammar: Finite and nonfinite verbs”. Language Acquisition, 3, 1–40. Hauser, R. M. (1994). “Measuring socioeconomic status in studies of child development”. Child Development, 65, 1541–1545. Hirsch, C. & Wexler, K. (2004). “Children’s passive and their resultative interpretation.” Manuscript. MIT. To appear in Proceedings of GALANA 2004. Horgan, D. M. (1978). “The development of the full passive”. Journal of Child Language, 5, 65–80. Jakubowicz, C. & Nash, L. (2001). “Functional categories and syntactic operations in (ab)normal language acquisition”. Brain and Language, 77, 321–339. Joseph, K., Serratrice, L., & Conti-Ramsden, G. (2002). “Development of copula and auxiliary BE in children with specific language impairment and younger unaffected controls”. First Language, 22, 137–172. Leonard, L., Caselli, M. C., & Devescovi, A. (2002). “Italian children’s use of verb and noun morphology during the preschool years”. First Language, 22, 287–304. Lorusso, P. (2003). L’acquisizione dei verbi inaccusativi. Studio sulla produzione spontanea dei parlanti. Unpublished MA Dissertation. University of Siena (Italy).

JB[v.20020404] Prn:2/06/2006; 14:52

F: LALD4106.tex / p.17 (1135-1188)


Mac Whinney, B. (1997). Il progetto CHILDES. Pisa: Edizioni del Cerro. Morrison, C. M., Ellis, A. W., & Quinlan, P. T. (1992). “Age of acquisition, not word frequency, affects object naming, not object recognition”. Memory and Cognition, 20, 705–714. Miller, D. (1991). Handbook of Research Design and Social Measurement. Sage Publications. The International Professional Publishers. London: Newbury Park. Nelli, C. (1998). “L’acquisizione della morfologia libera italiana. Fasi di un percorso evolutivo”. Studi di Grammatica, 17, 229–362. Pinker, S., Lebeaux, D. S., & Frost, L. A. (1987). “Productivity and constraints in the study of passive”. Cognition, 26, 195–267. Pizzuto, E. & Caselli, M. C. (1992). “The acquisition of Italian morphology. Implications for models of language development”. Journal of Child Language, 19, 491–557. Rizzi, L. (1993/1994). “Some notes on linguistic theory and language development: The case of root infinitives”. Language Acquisition, 3, 371–393. Rizzi, L. (2000). “Remarks on early null subjects”. In E. Friedmann & L. Rizzi (Eds.), The Acquisition of Syntax (pp. 269–292). London: Longman. Slobin, D. H. & Bever, T. G. (1982). “Children use canonical sentence schemas: A cross-linguistic study of word order and inflections”. Cognition, 12, 229–265. Stromswold, K. (1996). “Analysing children’s spontaneous speech”. In D. McDaniel, C. McKee, & H. Smith Cairns (Eds.), Methods for Assessing Children’s Syntax (pp. 23–53). Cambridge, MA: The MIT Press. Wasow, T. (1977). “Transformations and the lexicon”. In P. Culicover, T. Wasow, & A. Akmajian (Eds.), Formal Syntax. New York: Academic Press.



JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.1 (47-124)

Patterns of copula omission in Italian child language Elisa Franchi Università di Venezia

This paper concerns the production of the copula Essere and of the auxiliaries Essere and Avere in Italian child language. A relevant omission phenomenon emerges from the data, robust in declarative contexts but absent in Wh interrogative contexts. This pattern indicates the operativity of a strong syntactic constraint and perfectly matches the distribution of Root Infinitives as analysed in Rizzi (1993/1994). It then seems that Child Italian copular data can be accounted for in terms of Rizzi’s Truncation hypothesis, thus supporting the idea that child grammar is a full fledged UG constrained system. Data on morphological selection within the copular domain further support this hypothesis, showing that even the choice of “non adult like” morphological forms is syntactically constrained.

.

Introduction

The presence of functional projections in child grammar is currently under debate. It is uncontroversial that young children do not speak like adults and that they tend to omit functional elements. Despite those facts, it seems possible to assume (different degrees of) continuity in language development. Under continuity assumptions, child language should comply with UG Principles as much as the adult language, despite the fact that a single child language might apparently differ under significant respects from the target adult language. All the facts of language development that might differ from the target adult language are thus expected to be grammatically based. The aim of this work is to present and analyse data concerning the production of the copula Essere and of the auxiliaries Essere and Avere in Italian child language, in order to detect and quantify omission of those functional elements. In a continuity framework, omissions are not expected to occur at random, and syntactically constrained patterns are expected to emerge from the data.1 Copula omission will therefore be assessed and patterns in its distribution will be detected. The pattern emerging from the data will be analysed along the lines of a specific account on the nature of child grammar known as the Truncation hypothesis.2 This

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.2 (124-175)

 Elisa Franchi

hypothesis, originally formulated in Rizzi (1993/1994) to account for child language Root Infinitives, predicts the presence of functional verbs omission in child language and a specific pattern in its distribution. In Rizzi’s most recent analysis the missetting of the Root Parameter is interpreted as an useful tool for reducing the computational load in the earlier stages of child development by drastically reducing the syntactic structure.3 If Truncation is a syntactically driven option helpful for computational purposes in the early stages of child development, it should apply cross-linguistically, while its use should always be syntactically regulated.4 The aim of the present study is to test whether such a syntactically constrained option for reducing structure is available in Child Italian. The second section of the paper briefly summarises the Truncation Hypothesis main assumptions and expectations regarding the distribution of functional verb omission. Copular data and auxiliary data are presented in the third section: copula omission is shown to be a relevant phenomenon in Child Italian, both considering the general picture (Section 3.1) and analysing different stages of development (Section 3.1.1). In Section 3.1.2 the distribution of copula omission across different sentence types will be analysed, comparing declarative and Wh-interrogative data within the copular domain. In Section 3.2.1 copular data will be compared with auxiliary data, in order to check whether the omission phenomenon is limited to the copular domain or emerges in the domain of other functional verbs. The presence of different patterns of omission in declarative versus Wh-interrogative data will be checked for in auxiliary contexts in Section 3.2.2. In the domain of functional verbs the data indicate a robust phenomenon of omission constrained to specific syntactic contexts. A syntactically based account of the facts being observed will be provided in Section 4.1. Section 4.2 integrates predictions made by the Truncation hypothesis and by the Full Competence hypothesis with regard to children’s ability to project a full ”adult like” structure and to select the appropriate member of the morphological paradigm of the copula.

. The Truncation hypothesis The Truncation hypothesis is a specific hypothesis on the structure of child grammar that seems to hold cross-linguistically.5 It was originally formulated to account for child language Root Infinitives (RI), that is for the presence of non finite verbal forms in children’s main clauses.6 (1) a.

Maman faire boum sur le camion Mum DO bum on the truck b. Aller dedans moi GO inside me

(Philippe 2;01 – Child French) (Grégoire 2;03 – Child French)

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.3 (175-240)

Patterns of copula omission in Italian child language 

c.

Pas marcher toboggan Not WORK toboggan7

(Philippe 2;02 – Child French)

In Rizzi’s (2004) proposal it is possible to analyse data in (1) as the result of the missetting of the Root Parameter which is supposed to regulate the choice of the main projection of the root. In many adult languages the root is the highest node of the CP field – ForceP – and a missetting of that parameter will allow the choice of any node lower than ForceP as the root. Crucially, if a node lower than T is selected as the root, the possibility arises of producing a main clause with an untensed verb, a RI. RIs are typically limited to the domain of lexical verbs. In the domain of functional verbs RIs are not expected; if truncation applies below Tense, omission of functional verbs should be detected. The Truncation hypothesis is consistent with a continuity framework. Children are allowed to reduce the structure under restricted circumstances, i.e. never violating existing pieces of their grammatical knowledge.8 Therefore they are always expected to respect UG constraints. Crucially, they should be able to project the whole syntactic structure (Full Competence hypothesis) and should do so when this is required by syntactic constraints (Continuity). More specifically, in Rizzi’s terms, the presence of any element in the CP field should force the projection of a full “adult like” clausal structure and the production of full inflected forms, even at earlier stages of acquisition, when truncation is widely attested. As a consequence of that, truncation will never apply when an element in the CP field, such as a preposed Wh element, requires the whole structure to be there. This prediction is borne out in Wh-preposed contexts, where RIs almost disappear. Truncation phenomena appear to be regulated both by UG constraints and by language specific constraints. As for Child Italian, given the nature of Italian verb syntax (rich morphology and V-to-I movement for all tenses), truncation is expected not to apply in the case of lexical verbs. As shown in Guasti (1993/1994), RIs are almost never attested in Italian child language. The present study will analyse functional verbs data in Child Italian in order to detect functional verbs omission as an indicator of the missetting of the Root Parameter. The first aim of the study is to detect functional verbs omission and to assess its robustness. Subsequently, distribution patterns in functional verbs omission are detected and described. More specifically, if omission is the result of truncation, it is expected to be banned in contexts of preposed Wh-elements. Declarative contexts will be contrasted against Wh-contexts in order to evaluate this prediction. A third part of the study aims at assessing children’s Full Competence, in the presence of Truncation. If children do master the full syntactic structure, as claimed by the hypothesis here assumed, they will be expected to always choose the correct copular form, when they are forced to use it.

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.4 (240-306)

 Elisa Franchi

. The data . Copula omission The data taken into consideration for the present analysis were drawn from the CHILDES corpus.9 Production data were compiled from three monolingual Italian children, their ages ranging from 1 year and 7 months (1;7) to 3 years and 3 months (3;3). 1587 utterances were coded that could be analysed as copular constructions. The most relevant non adult-like phenomenon detected was omission of copula. In (2) through (4) examples are provided of copula omission in Realised Subject contexts, identified as Subject-Predicate sequences (SP) and in Null Subject contexts, identified as bare Predicate sequences (P). Martina (1;08.02) (2) (SP) quello pezzo demonstr-masc.-sing ØBe piece-masc.-sing (P) un pezzo Null Subject ØBe det-masc.-sing piece-masc.-sing (3) (SP) Pallo butto Paolo ØBe ugly-masc.-sing (P) glossa Null Subject ØBe big-fem.-sing (4) (SP) ette bee demonstr-fem.-plur ØBe sheep (P) zucchero! Null Subject ØBe sugar

Raffaello (1;11.25)

Rosa (2;10.14)

As the Null Subject examples in (2) through (4) show, bare predicates such as un pezzo or zucchero counted as instances of copula omission, despite the obvious fact that they could count as legitimate ellipses in the target adult language. A first problem that emerged from the data was how to quantify the relevance and the role of bare predicates. In a system like Child Italian where both copula drop and subject drop are allowed, bare predicates have an ambiguous reading, since they can be analysed both as instances of omission and as ordinary ellipses (as in the adult language). (5) Interaction between Realised vs. Null subject and copula omission. (SCP: Subject-Copula-Predicate) a. Il cappotto è verde. The coat is green (SP: Subject-Predicate) b. Il cappotto verde. The coat ØBe green. (CP: Copula-Predicate) c. pro è verde. pro is green. (P: Predicate) d. pro verde. pro ØBe green.

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.5 (306-442)

Patterns of copula omission in Italian child language 

Table 1. Number of omissions in Realised Subject contexts (SP/S(C)P), in Null Subject contexts (P/(C)P) per child and in total

Martina 1;7–2;7 Raffaello 1;7–2;11 Rosa 1;7–3;3 Total

SP/S(C)P

P/(C)P

n. total

68/200 30/113 136/382 234/895

71/167 92/218 310/507 473/892

139/367 122/331 446/889 707/1587

Table 2. Omission rate in Realised Subject contexts (SP/S(C)P), in Null Subject contexts (P/(C)P) per child and in total

Martina 1;7–2;7 Raffaello 1;7–2;11 Rosa 1;7–3;3 Total

%SP/S(C)P

% P/(C)P

% total

34% 27% 36% 34%

43% 42% 61% 53%

38% 37% 50% 45%

Given that bare predicates like verde are legitimate fragments in standard adult Italian in some contexts such as question-answer pairs, we can be sure that the copula has been omitted in child productions only with sequences such as (5b). However, given the ambiguous reading of bare predicates in Child Italian, contexts such as (5d) were not excluded from the analysis. The omission rate was thus assessed by dividing the number of omissions in realised subject contexts by the number of all realised subject contexts requiring a copula. As for null subject contexts, in order to asses the percentage of omission, the number of omissions in null subject contexts was divided by the number of all null subject contexts requiring a copula (the relevant data are presented in Table 1 and in Table 2). In order to draw a complete picture of the phenomenon, omissions both in Realised Subject and Null Subject contexts were considered. More specifically, data from the two contexts as a whole were analysed in order to get the general picture, but only Realised Subject contexts (where only omissions are certainly found) were considered for a more detailed analysis. Data in Table 1 and Table 2 show that omission of copula is relevant in Italian child language, considering all data and selecting only omissions in Realised Subject contexts.

.. Stages of acquisition The analysis of development along the temporal line was based exclusively on Realised Subject contexts, that is genuine omission data.10 The analysis of the longitudinal data for each child indicates a non-linear decrease of omissions. Two stages can be identified in the development of copula production. As the graphs in Figures 1 through 6 indicate, the beginning of the second stage of development emerged at different ages in the three children and was marked by a sharp and sudden decrease in the omission rate. In the present study the beginning of the 2nd stage is marked by the first file

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.6 (442-442)

 Elisa Franchi

Martina: I stage % Omission Real. Subject

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

82% 70%

70% 60%

33%

0% 1;7.18

1;8.2

1;8.17

1;9.1

1;10.29

28%

1;11.2

1;11.20

Figure 1. Rate of copula omission in Realised Subject contexts in Martina’s first stage of acquisition

Martina: II stage % Omission Real. Subject

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

27%

27%

27%

18% 2;1.12

2;3.1

2;3.22

2;4.13

0% 2;5.21

5% 2;7.15

Figure 2. Rate of copula omission in Realised Subject contexts in Martina’s second stage of acquisition

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.7 (442-442)

Patterns of copula omission in Italian child language

Raffaello: I stage % Omission Real. Subject

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

100%

100% 80% 67% 50%

1;7.7

4;10.20

50%

1;11.25

2;0.10

0% 2;0.28

2;3.14

2;4.29

Figure 3. Rate of copula omission in Realised Subject contexts in Raffaello’s first stage of acquisition

Raffaello: II stage % Omission Real. Subject

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

28%

28% 14% 0% 2;5.13

2;6.13

0% 2;7.14

2;8.15

14% 0% 2;9.6

2;11.14

18%

2;11.9 2;11.20

Figure 4. Rate of copula omission in Realised Subject contexts in Raffaello’s second stage of acquisition



JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.8 (442-442)

 Elisa Franchi

Rosa: I stage % Omission Real. Subject

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

100% 100% 100% 100% 100%

100%

100% 77%

60%

54% 33%

1;7.13 1;9.11 1;10.81;11.24 2;0.7 2;1.14 2;1.29 2;2.11 2;4.29 2;4.23 2;5.25

Figure 5. Rate of copula omission in Realised Subject contexts in Rosa’s first stage of acquisition

Rosa: II stage % Omission Real. Subject

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

45% 32% 17%

26% 16%

39% 26%

24% 10%

16%

2;6.26 2;7.26 2;9.4 2;9.24 2;10.142;11.12 2;11.30 3;0.24 3;1.29 3;3.23

Figure 6. Rate of copula omission in Realised Subject contexts in Rosa’s second stage of acquisition

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.9 (442-556)

Patterns of copula omission in Italian child language 

Table 3. Age span of the 1st stage of acquisition with number of omissions and omission rate in Realised Subject contexts I stage

Age range

n. Ø

%Ø

Martina Raffaello Rosa

1;7–1;11 1;7–2;4 1;7–2;5

52/106 15/23 54/67

49% 65% 81%

Table 4. Age span of the 2nd stage of acquisition with number of omissions and omission rate in Realised Subject contexts II stage

Age range

n. Ø

%Ø

Martina Raffaello Rosa

2;1–2;7 2;5–2;11 2;6–3;3

16/94 15/90 82/315

17% 17% 26%

where the omission rate is below 50% if and only if it also remains under 50% in all subsequent files.11 The 1st stage of development that emerges from the analysis of longitudinal data indicates that copula omission in Italian child language is even more relevant than it seems from the overall picture (Table 3 and Table 4). The relevance of the phenomenon in the first stage of acquisition for all the three children and its sharp decrease over time, seem to suggest that omitted copula sentences are indeed grammatical in early stages of Child Italian. The data seem to indicate the availability of a grammatical option, no longer available in the adult grammar.

. The Wh-constraint The possible influence on the omission pattern of declarative versus interrogative Wh-contexts is now taken into account. 419 Wh-contexts were isolated in the three corpora. Only two cases of omission were found, with an omission rate of 0,48% (Table 5). The percentage of omissions in Wh-copular contexts thus differs significantly from that found in declarative copular contexts. In the declarative data some variation was found among the three children in the omission rate (see Tables 1 and 2) and developmental changes along the temporal line were detected. In the Wh-data, as shown in Table 5, neither inter individual variation nor evolution along the temporal line emerges. Table 5. Number of Wh-contexts and omission rate per child and in total Copular Wh

n. contexts

n. omissions

% omissions

Martina (1;7–2;7) Raffaello (1;7–2;11) Rosa (1;7–3;3) Total

48 79 290 419

1 0 1 2

2,00% 0,00% 0,34% 0,48%

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.10 (556-661)

 Elisa Franchi

The distinction between declarative and interrogative Wh-contexts within the copular domain appears to be a relevant one. One possible way to explain the facts just described could be to correlate them to a peculiar phonological property of the Italian sequence Wh+3rd person singular copula. Italian sequences such as Dove è? (Where is) or Cosa è? (What is) are normally pronounced as Dov’è? and Cos’è. The elision of the final vowel of the Wh-element thus makes the two elements sound like a single phonological unit. Given this fact, one might suspect that at this early stage of development the children have not yet been able to parse the speech stream into separate words. If this was the case, the copula that appears in Wh-sentences like those reported in (6) through (8) could be a non-analysed form, clustering as such with the Wh-element and forming with it – in the early grammar – one single word. (6) Chi è? Who BE-3rd-sing ‘Who is it?’

Martina (1;11.20)

(7) Cos’ è là? What BE-3rd-sing there ‘What is it there?’

Raffaello (2;3.14)

Rosa (2;7.26)

(8) Ov’ è? Where BE-3rd-sing ‘Where is it?’

The hypothesis of a non-analysed Wh+copula cluster could explain the presence of the 3rd person singular copula in Wh-contexts but, crucially, it would also predict absence of productive use of the two elements separately. The distribution of Wh-elements with the copula, the auxiliary and lexical verbs was therefore considered. Relevant data are summarised in Tables 6 through 8. In the 1st stage of development the number of Wh-elements in copular sentences equals that of Wh-elements in sentences headed by a lexical verb. Their use seems therefore productive at this early stage of development even outside the possible non-analysed Wh+copula cluster. Table 6. Distribution of Wh-elements in copular, auxiliary and lexical verb contexts, in the 1st and 2nd stage of acquisition: Martina12 Martina

WH+Copula

WH+Aux

WH+V Lex

I stage II stage

1/12 0/36

0 0

10 9

Table 7. Distribution of Wh-elements in copular, auxiliary and lexical verb contexts, in the 1st and 2nd stage of acquisition: Raffaello Raffaello

WH+Copula

WH+Aux

WH+V Lex

I stage II stage

0/7 0/72

0 1/10

8 46

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.11 (661-720)

Patterns of copula omission in Italian child language 

Table 8. Distribution of Wh-elements in copular, auxiliary and lexical verb contexts, in the 1st and 2nd stage of acquisition: Rosa Rosa

WH+Copula

WH+Aux

WH+V Lex

I stage II stage

1/30 0/260

0/1 0/2

33 93

Evidence for the independent use of Wh-elements is also indicated by the examples in (9) through (12), where Wh-elements are productively used outside copular contexts, in ellipses (9), with lexical verbs (10) and with auxiliaries (11) and are also used with other forms of the copula that do not create a phonological unit with them (12). (9) Bare Wh-elements (ellipses) a. cosa? what? b. penchè? why? (10) Wh-elements with lexical verbs a. cosa fa? what DO-3rd-sing ‘What does she do?’ b. chi chiama mamma? who CALL-2nd-sing mum ‘Who is mum calling?’ c. cosa mangiate? what EAT-2nd-plur ‘What are you eating?’ d. basta uno come? suffices one-masc-sing how (11) Wh-elements with auxiliaries a. cosa hai fatto? what HAVE-2nd-sing done ‘What have you done?’ b. cosa ha detto? what HAVE-3rd-sing said ‘What has she said?’

Martina (1;08.02) (1st stage) Raffaello (2;04.29) (1st stage)

Martina (1;11.20) (1st stage)

Raffaello (2;01.15) (1st stage)

Raffaello (2;04.29) (1st stage)

Rosa (2;01.29) (1st stage)

Raffaello (2;07.14)

Rosa (1;10.08)

(12) Wh-elements with non 3rd person singular forms of Essere Raffaello (2;11.14) a. dove sei Pongo where Be-2nd-sing Pongo ‘Where are you, Pongo?’

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.12 (720-813)

 Elisa Franchi

Raffaello (2;11.14)

b. dove sono? where Be-3rd-plur ‘Where are they?’ c. che sono? what Be-3rd-plur ‘What are they?’ d. dove sei? where Be-2nd-sing ‘Where are you?’

Rosa (2;11.30)

Rosa (2;11.30)

A more detailed picture of the phenomenon under discussion comes from the comparative analysis of the evolution over time of Wh and declarative copular contexts with realised subject in longitudinal data.13 The developmental pattern of Wh-copular sentences sharply contrasts with the pattern emerging from longitudinal declarative data. When longitudinal Wh-data are considered, no developmental changes along the temporal line are detected. Two out of the three children produce one Wh-clause with copula omission each in the first stage of development. The omission rate detected in this case (8,3% for Martina and 3,3% for Rosa) sharply contrasts with the omission rate detected in declarative contexts, which rises as much as 49% for Martina and 81% for Rosa. On the other hand, if only the 2nd stage of acquisition was considered relevant for the present discussion, still the 100% rate of realised copula in the three children would contrast with the average 77% rate of realised copula that emerges at that stage in declarative contexts. As the longitudinal data presented in Figures 7 through 12 indicate, when the frequency of declarative and Wh-copular sentences is not drastically different, a sig-

Martina I stg. 30 25 20 15 10 5 0 1;7

Declar Ø

1;8

1;8

Declar Real

1;9

1;10

Wh Ø

1;11

1;11

Wh Real

Figure 7. Number of declarative and Wh-contexts with omitted and realised copula: longitudinal data from Martina’s 1st stage14

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.13 (813-813)

Patterns of copula omission in Italian child language 

Martina II stg. 30 25 20 15 10 5 0 2;1

2;3

Declar Ø

2;3

2;4

Declar Real

2;5

2;7

Wh Ø

Wh Real

Figure 8. Number of declarative and Wh-contexts with omitted and realised copula: longitudinal data from Martina’s 2nd stage

Raf. I stg. 30 25 20 15 10 5 0 1;7

Declar Ø

1;9

1;10

1;11

Declar Real

2;0

2;0

Wh Ø

2;1

2;3

2;4

Wh Real

Figure 9. Number of declarative and Wh-contexts with omitted and realised copula: longitudinal data from Raffaello’s 1st stage

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.14 (813-813)

 Elisa Franchi

Raf. II stg 30 25 20 15 10 5 0 2;5

2;6

Declar Ø

2;7

2;8

2;9

Declar Real

2;11

2;11

Wh Ø

2;11

Wh Real

Figure 10. Number of declarative and Wh-contexts with omitted and realised copula: longitudinal data from Raffaello’s 2nd stage

Rosa I stg. 55 50 45 40 35 30 25 20 15 10 5 0 1;7

Declar Ø

1;9

1;10

1;11

2;0

Declar Real

2;1

2;1

2;2

Wh Ø

2;4

2;4

2;5

Wh Real

Figure 11. Number of declarative and Wh-contexts with omitted and realised copula: longitudinal data from Rosa’s 1st stage

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.15 (813-826)

Patterns of copula omission in Italian child language 

Rosa II stg. 55 50 45 40 35 30 25 20 15 10 5 0 2;6

Declar Ø

2;7

2;9

2;9

2;10

Declar Real

2;11

2;11

Wh Ø

3;0

3;11

3;3

Wh Real

Figure 12. Number of declarative and Wh-contexts with omitted and realised copula: longitudinal data from Rosa’s 2nd stage

nificant difference in the omission rate still emerges. In the 1st stage of development, when both declarative and Wh-structures are relatively rare in the data, the pattern of omission in the two contexts differs significantly. Later on, when the overall frequency of Wh-contexts rises significantly, the frequency of omission in Wh-copular contexts remains zero. Even when the number of Wh-contexts equals or exceeds that of declarative contexts, still no omission is detected.15 Despite the fact that individual differences emerge from the data with respect to the number of Wh-contexts produced, no such differences emerge with respect to omission in Wh-contexts. In the 2nd stage of development the number of Wh-elements is significant enough to assume them to be productively used, but no omission is detected at that stage, even though it is still relevant in declarative contexts.16 A sharp and relevant difference between declarative and Wh-interrogative contexts seems then to emerge from the copular data. Apparently, the very same grammar that allows a null copula in declarative contexts at a given time of development, seems to force a realised copula with Wh-elements. The strong correlation observed between Wh-elements and realised copula seems to express a syntactic constraint. The presence of omission in declarative contexts and its absence in Wh-contexts, as well as the different course of development of the omission phenomenon in the two conditions,

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.16 (826-963)

 Elisa Franchi

Table 9. Omitted auxiliaries: Absolute numbers per child and in total

Martina Raffaello Rosa Total

Essere Aux n. contexts

Avere Aux n. contexts

2/9 9/30 28/70 39/109

26/48 23/69 33/87 82/204

Table 10. Omitted auxiliaries: Omission rate per child and in total


Essere Aux % omission

Avere Aux % omission

22% 30% 40% 36%

54% 33% 38% 40%

seem to suggest that omission is generally allowed in Child Italian by a grammatical option that is banned in specific syntactic contexts.

. Auxiliary data The auxiliary acquisition pattern was analysed using data gathered from the same files considered for the copular data. Auxiliary constructions were taken into account both as a whole, grouping together the two Italian auxiliaries, Essere and Avere, and separating the two auxiliaries, in order to draw a direct comparison between copula and auxiliary Essere.

.. Auxiliary omission Data gathered from the three children concerning both Avere and Essere auxiliary indicate that their omission is also a relevant phenomenon in Italian child language (see Table 9 and Table 10).17 A direct comparison between the omission rate of copula and auxiliary Essere (see Figure 13) indicates a fairly homogeneous pattern of omission in the two contexts.18 The omission phenomenon detected in the copular domain belongs then to a more general phenomenon in Child Italian that marks the whole domain of functional verbs. .. The Wh-constraint In order to draw a complete comparison between the omission pattern in the auxiliary and in the copular domain, auxiliary omission in Wh-contexts is now considered. Wh-auxiliary contexts are very rare in the three corpora, not only with respect to the copular data, but also in absolute numbers.

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.17 (963-963)


Omission Rate with Copula & Auxiliary Essere

40%

30%

20% 36%

34% 10%

27%

40%

30%

Copular Essere Auxiliary Essere

22%

0% Martina

Raffaello

Rosa

Figure 13. Omission rate with Essere copula and auxiliary

One child, Martina, never produces Wh-auxiliary constructions in the time period here considered. Data from the other two children are reported in (13) and (14). (13) Raffaello a. o@p ha fatto? what have-3rd-sing done ‘What has s/he done?’ b. o@p sei andata? where be-2nd-sing gone ‘Where are you gone?’ c. penchè ha usato lui. because have-3rd-sing used he ‘Because he has used (it)’ d. cosa hai fatto? what have-2nd-sing done? ‘What have you done?’ e. chi suonato? who Øaux rang ‘Who has rang?’ f. sà dove stai andare. yes where be-2nd-sing going ‘Where are you going?’

(2;5.13)

(2;5.13)

(2;6.13)

(2;7.14)

(2;8.15)

(2;11.14)



JB[v.20020404] Prn:5/06/2006; 15:00



F: LALD4107.tex / p.18 (963-1027)

Elisa Franchi

g. e quando è ritornata? When be-3rd-sing returned ‘When has she returned?’ h. dove sei sconde, dove sei sconde mamma! Where be-2nd-sing hidden mum ‘Where are you hidden, mum?’ i. chi è stato? Who be-3rd-sing been ‘Who has done it?’ (14) Rosa a. cosa ha detto? what HAVE 3rd-sing said ‘What has s/he said?’ b. pecchè sei nato così cinino e quetta è grande. because BE 2nd-sing born small ‘because you were born small’ c. pecchè hai fatto caccare? why HAVE 2nd-sing made drop ‘Why have you dropped (it)?’

(2;11.9)

(2;11.14)

(2;11.14)

(1;10.08)

(2;9.24)

(3;3.23)

As the examples in (13) and (14) show, only 12 cases of Wh-auxiliary contexts are found in the three corpora and auxiliary omission is virtually absent. Some instances of Wh-elements are produced well into the 1st stage of development in Rosa (see (14a)) or at the very beginning of the 2nd stage of development in Raffaello (see the examples in (13a) and (13b)). Thus, even in the earlier stages of their development, when they are allowed to omit the auxiliary, children seem already able to produce realised forms of the auxiliary in specific syntactic contexts. The specific pattern of omission found in the case of copular constructions does not seem to be restricted to copular Essere. Indeed, the same pattern of significant omission in declarative sentences versus no omission in Wh-contexts, seems to emerge both with the copula and with the auxiliary. The data thus seem to suggest the presence of a strong syntactic constraint regulating the omission of functional verbs in Italian child language.

Table 11. Absolute numbers of copula and auxiliary omissions in Wh-contexts Structure

Martina

Raffaello

Rosa

Copular WH Auxillary WH

1/49 –

0/79 1/9

1/290 0/3

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.19 (1027-1098)


. Discussion . Omission as evidence for truncation Two main facts emerge from the data that must be accounted for: on the one hand, functional verbs omission seems to be a grammatical option allowed in Italian child language; on the other hand, the very same grammar does not seem to allow random omission of functional verbs, which is banned in Wh-contexts. Such a phenomenon reminds very closely well known facts of child language such as Root Infinitives and Early Null Subjects, where a similar restriction to Wh-contexts emerges from the data. The copular data presented so far will therefore be discussed following the predictions made by the Truncation hypothesis, independently formulated to account for those well known phenomena.19 Within the Truncation hypothesis it seems possible to predict the omission of functional verbs and specific restrictions on its distribution. If the auxiliary can be analysed as being base generated in Tense, its omission should be expected whenever truncation applies below Tense.20 Along similar lines, if the copula can be analysed as the spell-out of Tense features, it will require the presence of Tense to be realised, and truncation below Tense will lead to its omission. Copula and auxiliary omission can then be interpreted as the correlate of Root Infinitives for functional verbs, the result of a truncation phenomenon allowed in natural languages by the (missetting of the) Root Parameter. Copula (and auxiliary) omissions in Italian child language occur indeed at the same age as RIs, they are robust at early stages, undergo a sharp and abrupt decrease over time and, crucially, never occur with preposed Wh-elements. Under the Truncation hypothesis, it is possible to account for all those facts in a unified way. The possibility of generating truncated structures would account for both omission of functional verbs and its restrictions, crucially predicting obligatory presence of a realised copula (and auxiliary) in cases of Wh-preposing. Child Italian data in the copular domain seem to pattern well with those predictions.

. Truncation & full competence The data considered so far seem to strongly argue for the children’s full competence. Children have the option of generating a reduced structure under highly restricted conditions, and they show perfect mastery of the whole syntactic structure in specific contexts. Thus, the full forms children produce in Wh-contexts are evidence not only that they are sensitive to a syntactic constraint, but also that they can project the full syntactic structure when required to do so. Given the crucial role played by the Full Competence hypothesis in the analysis of the data discussed so far, it seems now worth considering how complete the children’s competence might be in the copular domain by looking at the morphological forms they select when they do not omit the copula.



JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.20 (1098-1190)

 Elisa Franchi

Table 12. Agreement errors in copula selection


Errors/Total

%

1/387 3/331 13/889 17/1587

0,27% 0,91% 1,46% 1,07%

Table 13. Plural context errors within the copular domain per child and in total


Errors/Plural

%

1/26 3/43 13/101 17/770

3,85% 6,98% 12,87% 10,00%

The copular data considered for this study suggest indeed a full competence in the morphological domain as well as in the syntactic domain. Children use the correct morphological form 99% of the time (Table 12). A more detailed analysis of the few errors detected indicates that those are all restricted to plural contexts, as shown in Table 13. More specifically, as the data in (15) show, the few errors detected are all related with the production of the 3rd person plural form of the copula. In fact, the only incorrect form children produce is a 3rd person singular form È with a 3rd person plural subject, when this is post verbal. (15) Plural forms errors within the copular domain: examples. Martina (2;03.22) a. s’è i gatti there is the cats Raffaello (2;11.09) b. cos’è quelle? what is those Rosa (2;9.04) c. tuo c’è e@p macchine intanto yours there is the cars meanwhile The “errors” detected in the three corpora do not seem to emerge as randomly chosen incorrect forms, but display a highly constrained distribution. If errors were randomly emerging in plural contexts, one could suspect the children not to control the use of the plural forms of the copular paradigm. In contrast, the fact that plural forms are indeed correct 90% of the time shows that children do master the full morphological paradigm of the copula. Moreover, the distribution of the non-target forms in post verbal subject contexts suggests the presence of a syntactically regulated option made available by UG. A cross-linguistic analysis on post-verbal subjects with respect to morphological agreement seems indeed to suggest the special status of this structure. As shown in Guasti and Rizzi (2002), a post verbal subject does not always require morpho-

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.21 (1190-1243)


logical agreement to be expressed on the verbal form. Some examples are presented under (16). (16) Agreeing and non-agreeing forms with post verbal subject cross linguistically: a. Three girls are in the garden. a’. There are three girls in the garden b. Trois filles sont arrivées. Three girls-fem.plur are arrived-fem.plur b’. Il est arrivé trois filles. It is arrived-masc.sing three girls-fem.plur c. Questo, i bambini lo fanno sempre This, the-masc.plur kids-masc.plur it-acc do-plur always This, the kids do it all the time. c’. Questo, lo fa sempre i bambini21 This, it-acc do-sing always the-masc.plur kids-masc.plur d. ci sono troppi soldi there are too much-masc.plur money-masc.plur d’. c’è troppi soldi22 there is too much-masc.plur money-masc.plur In order to explain data like those in (16), Guasti and Rizzi (2002) assume that “If a feature is checked in the overt syntax, then it is expressed in the morphology”, the implication being that when the Φ-Features are checked at LF, UG leaves the choice of morphological agreement open to parametric variation. As for the Italian case, if the subject is base generated within the VP Shell, the post verbal subject can be assumed to remain in situ and to check its agreement features at LF. In that case, it would be both possible to morphologically express agreement and not to do so. In the latter case, a default non-agreeing verbal form will be selected. Given this analysis, although the non-agreeing form is ungrammatical in the standard version of the adult language the children are exposed to, it still seems to be a legitimate option within UG. Then, the agreement “errors” found in copular constructions in the morphological domain are not errors at all, but instances of an option made available by UG and exploited both in child and in adult language.

. Conclusions A systematic omission of the copula has emerged from the present analysis of Child Italian copular data. Although the phenomenon appears to be significant in all the three corpora considered, it seems limited to declarative contexts and it is virtually absent in Wh-interrogative contexts. The sharp contrast between systematic omission in declarative contexts and absence of omission in Wh-interrogative contexts emerges also in the domain of auxiliaries. Then, functional verb omission seems to



JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.22 (1243-1285)

 Elisa Franchi

be a grammatical option in Child Italian, but it also appears not to be allowed in cases of Wh-preposing. Such a range of facts closely reminds analogous patterns emerging in other domains of child language, namely as in the Root Infinitives phenomenon. The Truncation hypothesis, independently formulated to account for child language Root Infinitives, makes precise predictions with regard to the domain of functional verbs. Those are born out in detail in the copular data presented here. Child Italian copular data seem thus to suggest that it is possible to give a unified account of various child language facts on a structural basis. The phenomenon does not occur at random but appears to be syntactically regulated, highly constrained and therefore predictable within the theory. Indeed, the data once more indicate that children do not ”make errors”. Rather, they explore specific possibilities made available by UG. They do so when they apply truncation only in the domain of declarative copular structures and when they choose not to express morphological agreement in specific syntactic contexts. According to the line of reasoning adopted here, such a predictable behaviour should indicate that the children are exploiting a syntax driven option and that child grammar is highly structured and UG constrained as any other full fledged adult grammar.

Notes . See among others Borer and Rohrbacher (2003). . Rizzi (1993/1994) and Rizzi (2004). . Rizzi (2004). . On theoretical grounds, even if it is not (yet) possible to test the prediction with data from all Early Grammars, truncation is expected to hold cross-linguistically. Indeed, if, as suggested in Rizzi (2004), the missetting of the Root Parameter in the early stages of development helps to reduce the computational load, one should expect children to exploit that possibility at different degrees and with different limitations according to the different languages they are acquiring. . See Rizzi (1993/1994) but also Rizzi (2004) for further refinements of that proposal. . As the few examples in (1) indicate, RIs are compatible with both pre- and post-verbal subjects and compatible with non-nominative subjects and occupy a position lower than pas in French. In Rizzi’s (1993/1994) terms these facts are expected if the non-finite forms the children produce are analysed as genuine non-finite forms, rather than finite forms lacking the appropriate morphology. . Data from Rizzi (1993/1994). . See Rizzi (2004). . The Childes Database, November 2002, but see also MacWinney and Snow (1985). . The evolution along the temporal line of Realised Subject omissions should accurately describe how the omission phenomenon decreases and disappears. In contrast, such a clear picture could neither emerge from the bare predicate data nor from the data as a whole, since in both cases the picture would be blurred by the presence of ellipses, whose frequency cannot be predicted.

JB[v.20020404] Prn:5/06/2006; 15:00

F: LALD4107.tex / p.23 (1285-1351)

Patterns of copula omission in Italian child language . Thus file 1;10.29 in Martina is still included in the first stage of acquisition. The omission rate is below 50% in this file and in the following one, but it rises again over 50% later on. On the other hand, file 2;1.12 seems to genuinely mark the beginning of a new stage of development, since the omission rate is constantly under 50% from then on. . In Tables 6 through 8 the number to the left of the slash indicates omissions, where the number to its right indicates the total number of Wh-contexts. . Realised Subject declarative contexts have been selected for the comparison because they can genuinely indicate the evolution over time of the omission phenomenon in declarative contexts. . Data in Figures 7 through 12 are in absolute numbers. . Moreover, a cluster hypothesis would predict a bell shaped curve to emerge from the analysis of longitudinal data: no omission should be expected in the “non analysed cluster stage” but then presence of omission should emerge in the 2nd stage of acquisition, at least at the same rate as in declaratives, that is around 20%, and then a slow decline towards zero. The absence of any development along the temporal line rules out this hypothesis. . This fact is especially visible in the case of Rosa. She is the one in which the omission phenomenon lasts longer and with the higher rate. Still her behaviour with respect to omission in Wh-contexts perfectly matches that of the other two children. No omission is found in Whcontexts in her 2nd stage, despite the high frequency of Wh-contexts and the persistence of a high rate of omissions in declarative contexts. . The omission rate with the two Auxiliaries seems fairly homogeneous. Only one child, Martina, shows a relevant difference in the omission rate between Essere and Avere auxiliaries, but the fact needs further investigation, given the very few forms of Auxiliary Essere found in her corpus. . The auxiliary data is here compared to the Realised Subject copular data. . Early Null Subjects data will not be mentioned here, but see Rizzi (1992) and Rizzi (2000a, 2000b) for relevant discussion. . That would hold even if the auxiliary was base generated in a dedicated position related to Tense. See Rizzi (1993/1994) for relevant discussion. . Examples from Guasti and Rizzi (2002). . Langhe dialect (Piedmont) (cf. Revelli 1977). The structure is in fact very common in many other local varieties of Italian.

References Borer, H. & Rohrbacher, B. (2003). “Minding the absent: Arguments for the Full Competence Hypothesis”. Language Acquisition, 10(2). Friedemann, M. A. & Rizzi, L. (2000). The Acquisition of Syntax. London: Longman. Guasti, M. T. (1993/1994). “Verb syntax in italian child grammar: Finite and Non-finite verbs”. Language Acquisition, 3, 1–40. Guasti, M. T. & Rizzi, L. (2002). “Agreement and Tense as distinct syntactic positions: Evidence from acquisition”. In G. Cinque (Ed.), The Structure of IP. The Cartography of Syntactic Structures, Vol. 1. Oxford: Oxford University Press.



JB[v.20020404] Prn:5/06/2006; 15:00



F: LALD4107.tex / p.24 (1351-1383)

Elisa Franchi

Mac Whinney, B. & Snow, C. (1985). “The child language data exchange system”, Journal of Child Language, 12, 271–296. Revelli, N. (1977). Il mondo dei vinti: Testimonianze di vita contadina. Torino: Einaudi. Rizzi, L. (1992). “Early null subjects and root null subjects”. Geneva Generative Papers, 1–2, 102– 114. Rizzi, L. (1993/1994). “Some notes on linguistic theory and language development: The case of root infinitives”. Language Acquisition, 3, 371–395. Rizzi, L. (2000a). Comparative syntax and language acquisition. London: Routledge. Rizzi, L. (2000b). “Remarks on early null subjects”. In M. A. Friedemann & L. Rizzi (Eds.), The Acquisition of Syntax. London: Longman. Rizzi, L. (2004). “On the grammatical basis of language development: A case study”. In G. Cinque & R. Kayne (Eds.), Handbook of Comparative Syntax. Oxford: Oxford University Press.

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.1 (47-116)

Looking for the universal core of the RI stage Manola Salustri and Nina Hyams University of California, Los Angeles

As it is well known, child Italian does not show a typical Root Infinitive (RI) stage (Guasti 1994), in contrast to German, Dutch, etc. Salustri and Hyams (2003) provided evidence from 3 monolingual Italian children and 1 bilingual German-Italian child that there exists an analogue of the RI stage in this language. We argued that the RI analogue in Italian is the imperative. In this paper, we present data from 4 additional null subject languages, Spanish, Catalan, Slovenian and Hungarian – in support of the Imperative Analogue Hypothesis (IAH). We also present relevant data from Icelandic and Dutch, both RI languages, which further support the IAH. Finally we will evaluate two competing theories of the IAH against these data.

.

Introduction

Much recent acquisition research within the principles and parameters framework has focused on early morphosyntactic development. To the extent that parametric variation is tied to functional categories (cf. Borer 1984; Chomsky 1981), we expect that this will be a locus of learning for children and of variation among different child languages. Indeed, there is substantial evidence that children acquire the specific morphosyntax of the target very quickly. Parameters such as V to I, V2, and so on are set very early. Early English is essentially English, early German is essentially German, and so on. These findings of “early morphosyntactic convergence” (Hoekstra & Hyams (H&H) 1998) seem to fly in the face of stage theories, which purport to show that there are universal syntactic stages in acquisition. Yet, there is also substantial evidence that children acquiring different languages show similar developmental effects. Root infinitives (RIs), such as those in (1), are a case in point. (1) a.

Papa schoenen wassen. Daddy shoes wash-inf b. Auch Teddy fenster gucken also Teddy window look-inf c. Jag också hoppa där å där. I also hop-inf there and there

Dutch German Swedish

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.2 (116-167)

 Manola Salustri and Nina Hyams

d. Kisa finna dúkkinni cat find-inf the-doll

Icelandic

In the various languages that show an RI stage, the phenomenon is quite robust and relates to a number of other salient properties of early language. For example, RIs tend to occur with null subjects and they occur in declarative clauses but not in whquestions or topicalized structures, for example, in Dutch and German (cf. Hoekstra & Hyams 1998 for review of relevant literature). In terms of their semantics, RIs often have a modal or irrealis meaning (either of volition or necessity) and are largely restricted to eventive predicates (Wijnen 1997; Hoekstra & Hyams 1998; Ferdinand 1996). These various properties and contingencies provide strong evidence that RIs are not the result of production limitations or other performance factors, but are a genuine grammatical effect. That said, they nevertheless fall short of universality. There are various child languages that do not show an RI stage, notably the Romance null subject languages, Italian, Spanish, Portuguese, Catalan, and this despite the fact that the adult languages have an infinitival form (cf. Sano & Hyams 1994; Rhee & Wexler 1995). The question we consider in this paper is: Is there some sense in which the RI stage is universal (and if so, what is the appropriate level of grammar at which to capture this universality)? In this paper we will argue that that there exists an analogue of the RI stage in the null subject languages. We argue that the RI analogue in these languages is the imperative. We refer to this as the Imperative Analogue Hypothesis (IAH). We will propose that what is universal about the RI stage is the mapping of irrealis mood onto a tenseless clausal structure. The paper is organized as follows. We begin by reviewing some of the salient properties of RIs in Section 2. In Section 3 we first compare German and Italian with respect to the Root Infinitive (RI) phenomenon. We provide evidence from 3 monolingual Italian children and a bilingual Italian-German child that supports the IAH. Next, we expand the empirical base of our hypothesis by considering data from 4 more null subject languages, Spanish, Catalan, Slovenian and Hungarian. We also present relevant data from Icelandic and Dutch, both RI languages, which further support the IAH. Finally, in Section 4, we discuss possible analyses for the observed imperative analogue effects.

. RIs As a point of departure, we focus on three salient properties of RIs. First, and most obviously, RIs are non-finite, non-agreeing forms that occur in root contexts. In addition, they are characterized by the two important semantic properties noted above. First, RIs typically have a modal (or irrealis) interpretation expressing desires, intentions and needs. Examples are provided in (2).

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.3 (167-230)

Looking for the universal core of the RI stage

(Leo 2;1, Berger-Morales & Salustri 2003) Enzo malen Enzo draw-inf Context: Enzo wants to draw (Peter 2;1, Blom 2003) b. Peter bal pakken Peter ball get-inf Context: Peter wants to get the ball (Josse 2;8, Blom 2003) c. Op kist zitten on box sit-inf Context: Josse wants his mother to sit on the box

(2) a.

This modal reference effect (MRE – Hoekstra & Hyams 1998) has been described for Dutch (Hoekstra & Jordens 1994; Kramer 1993; German: Ingram & Thompson 1996; Becker & Hyams 1999; Lasser 1997; Behrens 1993; Swedish: Plunkett & Strömqvist 1990) among other languages. Wijnen (1997), for example, reports that 86% of Dutch RIs have a modal meaning (compared to 3% of finite verbs that have modal meaning) and Blom (2003) finds that modal RIs constitute about 73% of all RIs in Dutch. Hoekstra and Hyams (1998) proposed that the modal meaning associated with RIs comes from the infinitival morphology, which carries an irrealis feature. More recently, however, Hyams (2005b) has proposed that RIs optionally contain a null non-finite modal that gives rise to an irrealis interpretation.1 We will assume some version of the null modal analysis of RIs here though nothing really hinges on this. Turning now to the eventivity constraint (EC – H&H 1998), Wijnen (1997) (cf. also Blom 2003; Ferdinand 1996) reports that 95% of Dutch RIs are eventive, while finite verbs are evenly split between eventive and stative verbs. The EC is also reported for German (Becker & Hyams 1999) and French (Ferdinand 1996) and Russian (Brun & Babyonyshev 2004). The robustness of the RI stage and the strength of associated effects such as the MRE and the EC suggest that there is some universal basis for the phenomenon. It is therefore interesting to see if an RI analogue can also be found in those languages that seem at first blush not to have one, specifically the null subject languages.2

. The Imperative Analogue Hypothesis We propose that there is such an analogue – the imperative. Prima facie, the imperative is a good candidate because it shares the essential RI properties: Imperatives have modal meaning, that is, they express obligation or volition with respect to some eventuality; they are restricted to eventive predicates; and they are tenseless (and arguably non-agreeing). We adopt Han’s (2001) description of the imperative as a form that is marked with an ‘irrealis’ feature that contributes an unrealized mood interpretation and a ‘directive’ feature encoding directive illocutionary force. In terms of syntax, we assume that the imperative verb checks the irrealis feature in MoodP and the directive feature in ForceP, part of the extended left periphery (Rizzi 1997). Again following



JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.4 (230-324)

 Manola Salustri and Nina Hyams

Han (2001), we further assume that infinitives/RIs check an irrealis mood feature (cf. Note 1) but can have directive force (cf. (2c)) through pragmatic inference. It is thus possible to cancel the directive illocutionary force of infinitives and to generate some other force with unrealized interpretation (cf. (2a, b)). Imperatives are fully grammatical in adult language – in contrast to RIs – and so the simple appearance of imperatives in the child’s language is not remarkable. If, however, the imperative in child null subject languages does represent an RI analogue, we expect it to have distinguishing properties. Two possibilities suggest themselves. First, we expect that the imperative will occur more frequently in the child language than in adult language, and second, we would expect it to occur more frequently in child null subject languages than in the RI languages. These two predictions are stated in (3). (3) Predictions of the IAH (i) In null subject languages imperatives will occur significantly more often in child language than in adult language. (ii) In child language imperatives will occur significantly more often in the null subject languages than in the RI languages. The logic behind (3i) is that over time some of the imperatives will be replaced by modals, just as RIs trade-off with modals in Dutch (cf. Blom 2003). A similar reasoning gives rise to (3ii): if RIs and imperatives both express irrealis mood, then RIs might bleed imperatives in languages such as German and Dutch where RIs are known to have a high rate of occurrence. Before turning to our data, we wish to clarify one point. We are not claiming that imperatives and RIs are functionally equivalent in early grammar. RIs express various modal meanings including volition, future/intention, and obligation. Imperatives express only the latter. Rather, our claim is that RIs and imperative verbs are formally related in that they both have an irrealis feature (cf. also Han 2001), that is, they denote eventualities that are unrealized at speech time, which we take to be defining characteristic of the irrealis class.

. Italian and German To test the predictions in (3) we compared imperatives in child and adult Italian and imperative and RI frequency in Italian and German child language, in both monolingual and bilingual development. In Table 1 we provide information about the children whose files we examined. We also examined adult data from two sources: naturally occurring adult-directed language from several Italian adults engaged in informal discourse3 (transcripts provided by Sandro Duranti), as well as child-directed adult language taken from the adult tiers of two of the Childes transcripts we examined (diana5 and diana1) and from Leo’s data.

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.5 (324-379)

Looking for the universal core of the RI stage 

Table 1. Subject information Child

Age-range

Language

Source

Francesco Denis Martina Diana Viola Caroline Kerstin Simone Leo

1;5–1;8 1;5–2;2 1;10–2;7 1;8–2;6 2;1–2;7 1;3–2;6 2;0 2;0–2;7 2;0–2;7

Italian Italian Italian Italian Italian German German German Italian-German

Childes, Roma corpus Leonini 2002 Childes, Calambrone corpus Childes, Calambrone corpus Childes,Calambrone corpus Childes, Nijmegen corpus Childes, Nijmegen corpus Childes, Nijmegen corpus Berger-Morales & Salustri 2003

Table 2. Italian imperative and 3rd person indicative forms -are -ere -ire mangi-are ‘eat’ prend-ere ‘take’ dorm-ire ‘sleep’ Imperative mangia! Indicative mangia

prendi! prende

dormi! dorme

To test the predictions of the IAH given in (3) we searched the child and adult data for all occurrences of imperatives. In Italian the imperative is sometimes homophonous with the indicative. Thus, in the first conjugation class (-are) the 2nd person singular imperative is homophonous with the 3rd person indicative form, as shown in (4). (4) a.

Mario mangia una mela. Mario eat-ind.3per.sing an apple ‘Mario eats an apple.’ b. Mangia! eat-imp ‘eat!’

In the 2nd (-ere) and 3rd (-ire) conjugation classes, however, this is not the case. The imperative has a distinct form. Table 2 presents the indicative and imperative forms in the 3 conjugation classes.4 Although imperatives in Italian sometimes have the same morphological form as indicatives, they have a special syntax. Consider, for instance, the position of the clitics. In Italian, clitics immediately precede finite verbs in indicative mood, as in (5a), but they follow the imperative, as in (5b). (5) a.

La mangi, la mela it-acc.cl-eat, the apple ‘(you) are eating the apple’

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.6 (379-455)

 Manola Salustri and Nina Hyams

b. Mangiala ora! eat-it-acc.cl now ‘Eat it now!’ German imperatives have a similar morphosyntax to Italian imperatives in that they raise to a Force position in the left periphery. For the majority of verbs the imperative form is identical to the stem. The only overt morphological marking of the imperative in German is a stem vowel change, and this occurs with only a few verbs, such as geben (give) and nehmen (take), whose imperative forms are gib and nimm, respectively. As in Italian, the German imperative is also identifiable by its syntactic position. It occurs in first position, in contrast to indicatives, which typically occur in second position (when there is no topic drop). The German imperative is illustrated in (6). (6) a.

Kauf das Buch! buy-imp the book ‘Buy the book!’ b. Er kauft das Buch he buy-3per.sing the book ‘He buys the book.’ c. Komm bald nach Hause! come-imp soon to home ‘Come home soon!’ d. Er kommt nach Hause He come.3per.sing to home ‘He comes home.’

Thus, imperatives in the child and adult data were identified by the context of use, their morphology (where distinguishable) and syntax, including the position of clitics in Italian and verb position in German. Turning now to our results, the first prediction of the IAH (cf. (3i)) is that imperatives will occur less frequently in adult language than in child language. Table 3 shows the rate of imperatives in Italian for adults in both adult-directed and child-directed language.5 As we might expect, imperatives are more frequent in child-directed language (15%) than in adult-directed language (5%). We will use the higher child-directed rate as our adult norm. Turning to the Italian child data, imperatives are among the first verbal forms used. For example, between the ages of 1;5 to 1;8, 40% of Francesco’s verbs are imTable 3. Frequency of imperatives in adult Italian (all forms) Imperatives Tokens %

Total verbs

Discourse context Adult-directed Child-directed

36 82

950 550

5.6 14.9

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.7 (455-520)

Looking for the universal core of the RI stage 

60 50 Viola Martina Diana Denis

40 30 20 10 0 2

2.1 2.2 2.3 2.4 2.5 2.6 2.7

Figure 1. Frequency of imperatives in 4 Italian children (ages 2;0 to 2;7)6

perative. Similarly, in Denis’ corpus at 1;5 to 1;8, 78% of his verbs are imperatives. However, at this young age there are few utterances overall and it is likely that many of the imperatives such Guarda! (Look!) are rote forms. We therefore focus on the older children (Viola, Martina, Diana, and an older Denis) whose data is presented in Figure 1. We see that even in the older children the proportion of imperatives is still quite high as compared to 15% adult norm. The child rates peak at about 40% somewhere between the ages of 2;0 and 2;4 and then drop to adult-like levels by about age 2;6 or 2;7. Thus, as predicted (cf. (3i)), the rate of imperatives is considerably higher for children than for adults, even in child-directed language. Two further points are worth noting. First, the child frequencies do not at all mirror the adult frequencies despite the salience of imperatives in the input (e.g. special prosody, first position) (Newport, Gleitman, & Gleitman 1972). This argues against san account of these effects in terms of statistical learning or input matching. This will be clear in the German data as well. The second point is that the rate of imperatives in Italian child language is higher than in adult language despite the fact that adults have the full imperative paradigm (cf. Note 4) while the children use only the 2nd person singular imperative. Thus, if we were to compare only the singular imperative forms in the two groups, the difference in frequencies would be even greater. The proportion of imperatives in the Italian child data should be compared to the very low rate of RIs, as shown in Table 4. Table 4. Frequency of imperatives and RIs in early Italian Child

Age

% RI

% Imp

Total verbs

Denis Martina Diana Viola

2:0–2:7 2:1–2:7 2:0–2:7 2.1–2:7

2.8 0 0 0.2

31.1 17.5 16.4 30

318 513 863 198

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.8 (520-609)

 Manola Salustri and Nina Hyams

Table 5. Frequency of imperatives in adult German7 Files (from Childes)

Imperatives

Total verbs

K2010613.cha KE720905.cha KE020005.cha Total

189 (61%) 118 (32%) 300 (36%) 538 (36%)

309 366 836 1511

We note, finally, that the “overuse” of imperatives happens at the same age as the RI stage in the RI languages (roughly between the ages of 2 and 2;6), further suggesting an analogue analysis. Let us now turn to the German data. Table 5 presents the frequency of imperatives in child-directed adult German. We see that this proportion is quite high, about 36% on average. It is therefore remarkable to observe that German-speaking children use imperatives very infrequently. Figure 2 presents the proportion of imperatives for 3 Germanspeaking children (Caroline, Kerstin, & Simone). For the purposes of this analysis we counted as imperative, verbs (i.e. stems) that had clearly raised, as in (7). (7) a.

Mach auf! Open prefix b. Beiss mal Bite particle c. Trink Milch! Drink (the) milk

In (7a) the verb has raised across the prefix auf; in (7b) the verb across the particle mal; and in (7c) across the object. We did not count as imperative stems that clearly had not raised to a higher position such as those in (8). (8) a. Aufmach b. Mal beiss c. Milch trink Single word utterances were counted only when it was clear from context that they were imperative.8 Figure 2 shows the frequency of imperatives for the German-speaking children between the ages of 1;6 and 2;7. We note first that, as predicted, the overall rate of imperatives is quite low as compared to German adults and as compared to Italian children. Also, the frequency of imperatives remains fairly constant at around 10% across all data points. We do not find the same peak during the first half of the 3rd year as we do for the Italian children. Thus, both the predictions in (3) are confirmed by the monolingual acquisition data, consistent with the hypothesis that the imperative is the RI analogue in languages such as Italian.

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.9 (609-679)

Looking for the universal core of the RI stage 

60 50 40 30 20 10 0

2 2. 2 2. 4 2. 6

1.

6 1. 8 1. 1

Caroline Simone Kerstin

%

Figure 2. Frequency of imperatives in 3 German children (ages 1;6-2;7)

70 60 50 40 30 20 10 0

Italian Geman

2;0

2;3

2;7

age

Figure 3. Frequency of imperatives bilingual Italian-German

We also examined the frequency of imperatives and RIs in a bilingual GermanItalian child, Leo. The bilingual child is the perfect matched pair (de Houwer 1995). Leo is the child of a German mother and Italian father living in Florence, Italy. He was equally exposed to both languages. As shown in Table 6, there are very few imperatives in Leo’s German during the relevant period, (1.5%–6.5%) while in his Italian corpus between 34% and 56% of the verbs are imperative. Figure 3 illustrates that Leo is also like the monolingual Italian-speaking children in showing a peak in imperatives at around age 2;3.9 Table 7 shows that Leo only uses RIs in German and that the frequency is comparable to monolingual German-speaking children (cf. Figure 2).10 Table 6. Frequency of imperatives (bilingual child) Imperatives Stage

German Tokens %

Italian Tokens %

2;0–2;4 2;6–2;7

1/63 3/46

25/45 10/29

1.5 6.5

55.5 34.4

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.10 (679-783)

 Manola Salustri and Nina Hyams

Table 7. Frequency of root infinitives (bilingual child) RI Stage

German Tokens %

Italian Tokens

%

2;0–2;4 2;6–2;7

51/63 28/46

1/45 2/29

2 7

81 61

Table 8. Percentage of RIs in Dutch: 6 children, from Blom 200311 Age

RI

1;7–2;1 2;1–2;6

572/810 (70.6%) 993/4921 (20.1%)

Table 9. Percentage of RIs and Imperatives in Icelandic: Eva Age

RI

IMP

1;1–1;7 1;8–2;4

1254/1927 (65%) 620/2565 (24.1%)

25/1927 (1%) 30/2565 (1%)

Thus, Leo perfectly mirrors the cross-linguistic differences that we observe in the monolinguals. His data also clearly support the hypothesis that bilingual children develop separate grammars for each language (Meisel 1990).

. Dutch and Icelandic German is not the only RI language to show a low rate of imperatives. Blom (2003) observes that the rate of imperatives in Dutch child language is under 10% for the 6 children she studied. The average rate of RIs across children is reported in Table 8. Similarly, Sigurjónsdóttir (2004 and p.c.) calculates the rate of imperatives and RIs in an Icelandic-speaking child, Eva, between the ages of 1;1 and 2;4. The Icelandic results are reported in Table 9. The data presented thus far are thus consistent with our hypothesis that the imperative is an RI analogue in Italian. In the following section, we extend the empirical base of the IAH by discussing several other null subject languages. We begin with two Romance languages, Spanish and Catalan, and then turn to Hungarian and Slovenian, which are also null subject languages.

. Spanish and Catalan An RI stage is not attested in the acquisition of Spanish and Catalan (Grinstead 1998; Bel 2001; and Montrul 2003). Bel (2001) analyzed the language of 3 Catalan and 3 (continental) Spanish-speaking children (ages range 1;7–3;0). Table 10 (based on Bel 2001) shows that the incidence of RIs on Catalan and Spanish is very low, under 4% for each child.

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.11 (783-876)

Looking for the universal core of the RI stage 

Table 10. Frequency of RIs in Catalan and Spanish-speaking children Catalan % RI Gisela Pep Julia

0/627 (0%) 25/1248 (2%) 22/720 (3%)

Spanish % RI Maria Emilio Juan

39/1956 (2%) 0/1588 (0%) 6/335 (2%)

Table 11. Percentage of imperatives in child Catalan Child

Age

% Imp

Age

% Imp

Laura Pep Gisela Guillem

1;7–1;8 1;3–2;0 1;0–1;9 1;0–1;9

41% 41% 25% 56%

2;4-3;0 2;1-2;7 2;2-3;0 1;11-2;9

31% 26% 22% 33%

On the other hand, as noticed by Grinstead and Montrul, imperatives are attested from the very onset of language acquisition and they occur very frequently. Some examples of imperatives produced by the Catalan-speaking children are given in (9). These examples are particularly noteworthy because they show that the children correctly position the verb with respect to clitics, that is, the imperative verb has raised to a position to the left of the clitic.12 (9) a.

Ajuda’m help-imp me-acc.cl ‘Help me!’ b. Dame Give-imp me-acc.cl ‘Give me!’

The percentage of imperatives is very high initially and decreases with age. Table 10 (based on Grinstead 1998) shows the percentage of imperatives at earlier and later data points for several Catalan-speaking children. Two of these children, Gisela and Pep, are among the children studied by Bel and whose RI data are presented in Table 11. To determine the rate of imperatives in Spanish child language we turn to data presented in Arbisi-Kelm (2005). Arbisi-Kelm analyzed the verbal system of two Spanishspeaking children, Maria (Childes, Lopez Ornat 1994), Emilio (Childes, Vila 1985). Imperatives constituted a large percentage of verbal utterances in the children’s corpora. This percentage decreased over time approaching adult frequency of 17%.13 These results are presented in Table 12. Note that at this stage modals are unattested in both Emilio’s and Maria’s data, consistent with our hypothesis that modals eventually drive out some of the imperatives. RIs are also absent from Emilio’s data. Maria, on the other hand, shows a relatively high RI rate of 20% between the ages of 1;7 and 1;9, which decreased rapidly to under 5% after that point.14

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.12 (876-945)

 Manola Salustri and Nina Hyams

Table 12. Percentage of imperatives in Spanish (based on Arbisi-Kelm 2005)

Emilio Maria

Age

%Imp

2;1–2;3 2;5–2;9 1;7–1;11 2;1–2;2

108/265 (41%) 149/482 (30%) 149/524 (28%) 138/687 (20%)

Table 13. Percentage of RIs in Hungarian Language

Child

Age

% RI

Hungarian

Miki Andi 15 children

2;1–2;4 2;1–2;5 1;5–2;0

0% 1% 1%

Slovenian

Table 14. Imperatives in Hungarian Child

Age

% Imp

Miki

2;1–2;3 2;4–2;7 2;1

26% (27/104) 10% (19/191) 21% (40/121)

Andi

Summing up the discussion thus far, an RI stage is not attested in the acquisition of Spanish and Catalan (but cf. Note 13). On the other hand, imperatives are quite frequent in the early stage and decrease over time. The generalization that emerges is that imperatives are very frequent in languages without an RI stage but very infrequent in the RI languages. These data support the IAH which suggest a complementary relation between RIs and imperatives. In the next section, we discuss data from two non-Romance languages null subject languages, Hungarian and Slovenian.

. Hungarian and Slovenian Like the Romance languages just discussed, RIs are virtually unattested in child Hungarian and Slovenian, as reported by Londe (2004) and Rus (2004), respectively. Data from two Hungarian-speaking children (Londe 2004) and 15 Slovenian-speaking children (Rus 2004) are presented in Table 13. Given the rarity of RIs in Hungarian, the IAH leads us to expect a high percentage of imperatives in child Hungarian and Slovenian. This prediction is confirmed. As shown in the Hungarian data in Table 14, imperatives start out high and decrease with age (Miki). Note also that the percentage of imperatives in Hungarian in the adult language (child-directed) is around 8% (Londe 2004). Similar results hold for Slovenian: in the corpora of the 15 children (age 1;5–2;0) studied by Rus (2004) over half of the verbal utterances (56%) are imperatives. As no data from older children or adults is reported, we do not know if there is a decrease

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.13 (945-1006)


in the use of imperatives, but we strongly suspect that the proportion of imperatives is lower that 56%. Examples of imperatives in child Slovenian are given in (10) (from Rus 2004) and the Hungarian examples in (11) (from Londe, p.c.). (10) a.

(Po)krov dej! cover-acc give-2sg.imp ‘Give me the cover!’ b. Tuki glaj. here look-2sg.imp ‘Look here!’

(11) a.

Te is játsszál! you too play-2sg.imp ‘You play too!’ b. Anya, nevessé (l) ! mom laugh-2sg.imp ‘Mom, laugh!’

(Lenart, 1;9)

(Vesna, 1;7)

(Miki 2;7)

The empirical data are quite clear. In languages with a robust RI stage (e.g. German, Dutch, Icelandic), imperatives occur infrequently. In non-RI languages, including the Romance null subject languages as well as Hungarian and Slovenian, imperatives occur at an extraordinarily high rate during the early stage of acquisition, both as compared to adult frequencies and to the frequency of imperatives in the RI languages.

. Is the RI analogue really an imperative form? We have maintained that in the non-RI languages the imperative is an RI analogue. It has been suggested, however, by Grinstead (1998) and others, that in the Romance languages the 3rd person singular indicative is a default non-finite form for children. Along similar lines, Joao Costa (p.c.) suggests that what we have identified as an imperative is not in fact an imperative, but rather the default indicative form. In the following section, we will consider this hypothesis in more detail. We will also consider the hypothesis that the imperative form constitutes an underspecified form in child language in the distributed morphology sense, along the lines proposed by Wexler et al. (2004) for the Dutch RI and English bare verb. We turn first to the ‘3rd person default hypothesis’ or 3D Hypothesis, for short.

. The 3D Hypothesis Grinstead (1998) observes that some imperatives are homophonous with the 3rd person indicative. As we noted earlier, in Italian, for example, this is true of first conjugation (-are) verbs, the most common class, as illustrated in (5), repeated below in (12a, b).



JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.14 (1006-1101)

 Manola Salustri and Nina Hyams

Table 15. Percentage of imperatives from 3 verb classes: Diana (1;8–2;6) Verb class

Imperative form

Frequency

-are (mangiare) -ere (prendere) -ire (dormire)

Mangia! Prendi! Dormi!

114 (66%) 43 (25%) 14 (8%)

(12) a.

Mario mangia una mela. Mario eat-ind.3ps an apple ‘Mario eats an apple.’ b. Mangia! eat-imp ‘eat!’

Similarly, in Spanish the imperative is homophonous with the 3rd person singular indicative in all conjugation classes. An example is provided in (13). (13) a.

Juan come una naranja. Juan eat-ind.3ps an apple ‘Juan eats an apple.’ b. Come! eat-imp ‘Eat!’

Thus, in principle, in these cases, what we have identified as the imperative could be a default indicative form. However, as was shown in Table 2 in the Italian 2nd (-ere) and 3rd (-ire) conjugation classes, the imperative is not homophonous with the 3rd person indicative. Thus, if the 3D Hypothesis is correct, we should not find imperatives from these conjugation classes because they are morphologically distinct from the 3rd person indicative. This prediction is not confirmed. Table 15 provides a breakdown of the imperatives in Diana’s corpus. We see that she uses imperatives from all 3 verb classes. In particular, she uses imperatives of the -ere and -ire classes, which are not homophonous with the 3rd person indicative. This finding is inconsistent with the hypothesis that the RI analogue is a default indicative form. However, Table 15 also shows that the majority of Diana’s imperatives (114/171 or 66%) are from the 1st conjugation class. This might be construed as support for the 3D Hypothesis, since this is the hypothesized default form. But there are independent reasons why the majority of Diana’s imperatives are from this class. The -are conjugation is in fact the most productive class in the language as evidenced by the fact that loan words are typically assigned to this class, for example, cliccare ‘to click’, zippare ‘to zip’ and scannerizzare ‘to scan’ (L. Brunetti, p.c.). Also, this class is the most frequently occurring in the language, as shown in Table 15. The figures in Table 16 (from Albright 2002) are based on a calculation of verb frequency in a spoken corpus of half a million words (de Mauro et al. 1993).

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.15 (1101-1172)

Looking for the universal core of the RI stage 

Table 16. Relative distribution of Italian verb classes15 Verb Class

Frequency

-are -ere -ire

1463 (72%) 323 (15%) 197 (10%)

Table 17. Clitic position in imperative and indicative clauses: Diana 1;8–2;6

Imperative Indicative

Enclitic

Proclitic

27 0

1 28

Comparing Tables 15 and 16, we see that the distribution of Diana’s imperatives is very close to the overall distribution of the 3 verb classes in the (adult) spoken language. Diane’s greater frequency of imperatives from the -are class are therefore expected for independent reasons and do not support the 3D Hypothesis. Another source of evidence against the 3D Hypothesis is provided by the distribution of clitics in early language. As shown earlier in example (5), the position of the clitic in Italian differs in indicative and imperative clauses. In the indicative (subjunctive and conditional) the clitic occupies a position immediately preceding the verb, while affirmative imperatives precede the clitic. Thus, another prediction of the 3D Hypothesis is that children will not correctly position the clitic with respect to the imperative verb, but rather treat it like an indicative. This prediction is also not confirmed in our data. The results in Table 17, from Diana, show that she is able to distinguish imperatives and indicatives and correctly position the clitic in both cases (cf. also Guasti 1993/1994). Some examples are provided in (14). (14) a.

Portamelo! bring-imp it-acc.cl to me-dat.cl ‘Bring it to me!’ b. Io ti lavo i capelli I to you-dat.cl wash the hairs ‘I (will) wash your hair’ c. Mettilo! put-imp it-acc.cl ‘Put it!’ d. Ti metto le cappe to you-acc.cl put-1per.sing the shoes ‘I (will) put your shoes on’

(Diana 2;6)

(Diana 2;6)

(Diana 2;1)

(Diana 2;6)

In short, the predictions of the 3D Hypothesis are not supported by our data. Let us turn now to a second hypothesis, that the imperative form is actually an underspecifed form, in the distributive morphology sense.

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.16 (1172-1256)

 Manola Salustri and Nina Hyams

Table 18. Dutch verbal paradigm

1p 2p 3p

Sing.

Pl.

0 -t -t

-en -en -en

Inf. -en Imp. 0

Table 19. English verbal paradigm

1p 2p 3p

Sing.

Pl.

0 0 -s

0 0 0

Inf. 0 Imp. 0

. The Underspecification (DM) Hypothesis Based on the theory of distributed morphology (Halle & Marantz 1993), Wexler et al. (2004) propose that the RI in Dutch, as in (15a) and the English bare verb, as in (15b), represent overgeneralizations of underspecified form. (15) a.

Papa schoenen wassen Daddy shoes wash-inf b. Eve sit floor.

The Dutch -en form of the verb appears in multiple positions in the verbal paradigm – on infinitives, and in all the plural persons, as shown in Table 18. The morpheme -en is thus underspecified with respect to tense and also person. As is well known, during the RI stage in Dutch, children use the -en form of the verb in contexts that would be ungrammatical in the adult language, viz. with singular subjects. It is possible, therefore, to think of the child’s -en form, not as an RI per se, but rather as an overgeneralized underspecified form. Similarly, in English, the bare form occupies most slots in the verbal paradigm – all except 3rd person singular, as shown in Table 19. The bare verb is an underspecified form as well – underspecified for tense, number, person, etc. Thus, as suggested for the Dutch RI, we can think of the English bare verb that occurs in 3rd person contexts in early language not as an RI per se (as originally proposed in Wexler 1994), but rather as an overgeneralized underspecified form. Following the logic of this approach, we might hypothesize that the imperative is an underspecified form and that the high frequency of imperatives in the non-RI languages is due to overgeneralization of an underspecified form. The Underspecification Hypothesis (UH) leads to two predictions. First, we expect a higher frequency of imperatives in languages in which the imperative is homophonous with other forms in the paradigm. The Italian, Catalan and Hungarian data are consistent with this hypothesis since in these languages the imperative is not a unique form.16 Conversely, languages with a specific form for imperatives, that is, a

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.17 (1256-1382)


Table 20. Slovenian imperative and indicative paradigm (from Rus 2004)

present indicative Imperative

2sg.

1dual

2dual

1pl.

2pl.

delas delaj

delava delajva

delata delajta

delamo delajmo

delate delajte

Table 21. Spanish irregular imperatives Imperative Underspecified/default form tener ‘have/hold’ venir ‘come’ ir ‘go’ poner ‘put’ hacer ‘make/do’

ten ven Ve(te) pon haga

tiene viene va pone hace

form that is not homophonous with other forms in the paradigm, should not show imperative analogue effects. In other words, we do not expect a high frequency of imperatives. However, the Slovenian data are a counterexample to the UH since the forms of the imperative paradigm (singular, dual, plural) are not homophonous with other forms in the language. The Slovenian indicative and imperative paradigms are given in Table 20. A final prediction of the Underspecification Hypothesis, as well as the 3D Hypothesis, concerns the use of irregular imperative forms. As illustrated earlier in (14), Spanish imperatives are typically homophonous with the 3rd person indicative. However, there is a small set of verbs (e.g. tener ‘have/hold, venir ‘come’, etc.) that have an irregular imperative (e.g. ten!, ven!) that is distinct from all other verbal forms in the language and hence also distinct from the 3rd person indicative (e.g. tiene, viene). If imperative analogue effects are due to an overgeneralization of an underspecified form, we do not expect to find irregular imperatives in their early language since these are specific forms. Rather, we would expect an overgeneralization of the least specified form in the paradigm, which is the 3rd person indicative (cf. Pertsova & Salustri 2004). The UH and the 3D Hypothesis converge with respect to this prediction. In Table 21 we list the irregular imperative and 3rd person indicative form of the verbs with an irregular imperative that are likely to be used by children. All imperative contexts were examined in the files of 3 Spanish-speaking children studied by Arbisi-Kelm (2005) (Maria, Emilio, Koki) for occurrences of these irregular imperatives as well as the hypothesized underspecified/default forms. These results are in Table 22. It is quite clear from these results that children use the correct form of the irregular imperative and do not overgeneralize the default/underspecified form. This argues strongly that the early imperative is indeed an imperative.



JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.18 (1382-1408)

 Manola Salustri and Nina Hyams

Table 22. Spanish irregular imperatives vs. overgeneralized forms

Maria Emilio Koki Total

Imperative

Overgeneralized form

18 33 43 94

1 0 2 3

. Concluding remarks We have seen that the predictions of the IAH are confirmed by the Italian, Catalan, Spanish, Hungarian, and Slovenian-speaking children considered here. We have also shown that the imperatives attested in those child languages are real imperative forms and not default indicative or underspecified forms. The obvious question for the IAH is: Why do some child languages avail themselves of the imperative as an RI analogue, while others are true RI languages? One possibility, suggested in van Kampen (2004), is that the difference is input-driven, that is, it is an effect of the frequency of difference structures in the input to children acquiring RI and non-RI languages. On its face, this is problematic since German-speaking children, for example, receive plenty of imperatives in the input (the rate of imperatives in the adult input is 35%), but nonetheless adopt the RI option and fail to produce imperatives in any significant proportion. Van Kampen, however, proposed that Dutch children hear many imperatives of the form given in (16a), in which there is a “light” imperative auxiliary followed by an infinitive. If children simply ignore the auxiliary, what they are left with is a root infinitive with imperative force. Italian children, on the other hand, mainly hear simple imperatives of the form in (16b) so that is what they produce. (16) a.

Ga ‘ns eten jij! Go part eat-inf you ‘Go ahead and eat! (roughly) b. Mangia! Eat!

While this hypothesis is more nuanced, hence more appealing, it still leaves unexplained why the Dutch (and German) children ignore the simple imperatives in their input such as (17). (17) a.

Ruim je speelgoed pick-imp your toys ‘Pick up your toys’ b. Was je handen Wash-imp your hands

(‘ns) op (part) up ‘ns even part

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.19 (1408-1464)

Looking for the universal core of the RI stage 

Why don’t we find both RIs and imperatives in these child languages? While it is possible that differences in the input of the sort suggested by van Kampen play some role, this cannot be the whole story. The complementary of the two forms suggest a competition, consistent with an approach based on economy considerations. Salustri and Hyams (2003) suggested that children choose the most economical option available to them given the constraints of the early grammar. Earlier we proposed that both the imperative and the RI check an irrealis feature. An assumption we make, which we think is relatively uncontroversial, is that the derivation of an RI is more economical than an imperative derivation in that imperatives involve verb movement (to Mood and Force) while the RI does not. We would thus expect that in a language in which RIs are a grammatical option (e.g. Dutch, German, etc.), they would win out over imperatives as the expression of irrealis mood. Imperatives, on the other hand, are realized when the more economical RI option is blocked. Following ideas of Guasti (1992) and Rizzi (1993/1994) (based on Belletti 1990), we propose that RIs are blocked in Italian because Italian infinitives have (abstract) Agr features that must be checked. The specification of Agr entails the specification of Tense (either because they are part of the same node, or because AgrP is above TP). Notice that the competition between imperatives and (root) infinitives is not restricted to child language. There are many adult languages in which infinitives have an imperative or imperative-like function. This occurs when the true imperative is blocked out for some reason (cf. for example, Zanuttini 1997). Italian negative imperatives are a case in point. Second person imperatives cannot be negated in Italian (and many other languages). Negative imperatives are thus formed with the infinitive, as illustrated in (18). (18) a.

Mangia la pasta! Eat-imp the pasta b. *Non mangia la pasta! Not eat-imp the pasta c. Non mangiare la pasta! Not eat-inf the pasta ‘Don’t eat the pasta’

Han (2001) has suggested that infinitives can function as imperatives precisely because imperatives and infinitives (and subjunctives) share an irrealis feature. Imperatives differ from infinitives in that they are also grammatically marked as ‘directive’ and this is responsible for their imperative syntax, i.e. driving the verb to ForceP. Infinitives, on the other hand, when they are directive, get their directive force through pragmatic inference, as noted earlier. This analysis extends quite naturally to the child language facts under discussion, assuming that in the early grammar as in the adult grammar, ForceP must be specified in all root clauses (e.g. as directive, interrogative, assertive, negative) and second, that the specification of Force must be compatible with the Mood specification of the clause. The irrealis feature of the RI is compatible with ‘directive’ Force

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.20 (1464-1518)

 Manola Salustri and Nina Hyams

(as in adult imperatives), hence we find ‘directive’ RIs (cf. (2c)). The irrealis feature is also compatible with assertive force when this involves an expression of desire or intent, in other words, hence the volitional and intentional RIs (cf. (2a, b)). Summing up, we assume that RIs are the most economical expression of irrealis mood in child languages in which the grammar allows this option. The irrealis feature on the RI is compatible with directive illocutionary force as well as volitional and intensional meanings. Where RIs are precluded, as in the ‘rich Agr/null subject’ languages, imperatives check irrealis mood features, and also a directive force feature. This means that from a functional point of view RIs and imperatives overlap to a fair degree, but not completely in that imperatives have a more restrictive semantic specification. Thus, economy considerations together with language specific morphosyntactic factors give rise to the imperative/RI split among languages. From this perspective, we would say that what the universal core of the RI stage is the mapping of irrealis mood onto a tenseless structure.17

Acknowledgements Parts of this paper were presented at the Boston University Conference on Language Development 27 in November 2003 and at GALANA I (Generative Approaches to Language Acquisition – North America) at the University of Hawaii in December 2004. Sections of this paper appear in the proceedings of those conferences (Salustri & Hyams 2003, 2005). We wish to thank Tim Arbisi, Dominik Rus, Zsu Zsa Londe, Sigga Sigurjónsidóttir for providing us with the Spanish, Slovenian, Hungarian and Icelandic data and Zahra Khalili for editorial and other assistance. This research was partially funded by a UCLA Faculty Senate Grant to Nina Hyams.

Notes . The null AUX/modal hypothesis is originally due to Boser et al. (1992). Hyams (2005b) departs from Boser et al. in several respects. Most importantly, she assumes that the null modal is non-finite. . Greek 2-year olds use a “bare perfective” form (ungrammatical in the adult language), as in (i), that shares the central properties of RIs. (i)

Ego katiti I sit-perf.3sing ‘I am going to/want to sit’

(Spiros 1;9, Childes, Stephany 1997)

The bare perfective is arguably non-finite, it expresses modal meaning, and it is overwhelmingly eventive. Greek is a null subject language but it does not have an infinitival form, hence it is distinct from the languages discussed in this paper. Hyams (2002, 2005a) has proposed that the bare perfective is the Greek RI analogue. For reasons of space we will not discuss the Greek bare

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.21 (1518-1588)

Looking for the universal core of the RI stage 

perfective here, but see Hyams (2005a) for extensive discussion and comparison with “true” RI languages. . These transcripts were kindly provided by Sandro Duranti. A total of 1232 verbal utterances were coded in the analysis of the adult transcripts. . There is also a 1st person plural imperative, e.g. Andiamo ‘Let’s go’ and a polite imperative e.g. Vada pure! (Go ahead and eat (roughly)’, the latter form is homophonous with the 2nd person indicative form in the 1st conjugation. Since the children in our study produced only the 2nd person singular familiar imperative shown in Table 2, we do not discuss these other types of imperatives. . This and subsequent frequencies are calculated the proportion of imperatives (or RIs) out of all verbal utterances. . We do not have figures for every data point for each child. . These figures are based on the adult tiers in the Kerstin corpus. . German has more potential for indeterminacy than Italian because the bare stem in one word utterances can be an infinitive lacking -en, an imperative, or a 1st person singular. Only one child shows a relatively high percentage of indeterminate forms. . For this analysis we did not include the files in which the total number of verbs is less than 10. . Like monolingual children (cf. Section 2), Leo’s RIs show both a modal reference effect: 89% of his RI have a future/modal meaning, and the eventivity constraint: 100% of Leo’s RIs were eventive while his finite verbs verb split between eventive and stative (cf. Berger-Morales, Salustri, & Gilkerson 2003). . The children are Abel, Daan, Josse, Laura, Matthijs and Peter. These findings are in line with Wynen (1997) who found an overall RI rate of 73% for the children he studied. . As we will discuss below, Italian children also position clitics correctly in imperative clauses. . The adult imperative rate was calculated on the basis of the adult input in the Childes files. . This very early use of RIs has been reported for other children acquiring Romance languages. Bel (2001) observes that both Maria (Spanish) and Julia (Catalan) showed a relatively high rate of RIs before age 1;9, after which time they decrease dramatically (cf. also Torrens 1995 and 2002; Davidson & Goldrick 2003; Schaeffer 1990). Children acquiring “real” RI languages do not show this pattern. Their rate of RIs is higher and continues well into the third year. It is an intriguing possibility that all children may in fact pass through an RI stage, which, however, ends much earlier for children acquiring null subject languages. In this case, we would still want to know why the two language types show distinct patterns and why imperatives come into play in the null subject languages but not the “RI” languages. . Although Albright distinguished 4 verb classes, -are, -ire, -ere and -ére, we make use of the traditional 3 way classification, collapsing Albright’s -ere and -ére classes, in order to compare the adult and child data. This does not affect the results. . In Hungarian the imperative forms are homophonous with the subjunctive and indicative forms in the 3rd person singular and plural forms (e.g. adja, adjuk, adjatol, adjak, while the infinitive is a specific form, adni (to give). . Another avenue to explore is the possibility that there is also a lower Force projection to the left of VP in the area that Belletti (2004) calls the ‘clause internal periphery.’ According to Belletti, there is a VP periphery, corresponding to the CP left periphery proposed in Rizzi (1997),

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.22 (1588-1689)

 Manola Salustri and Nina Hyams

which contains Topic and Focus positions. It is possible that RIs are an expression of a Focus position within the VP periphery, and that this is the more economical option than raising to the CP periphery in languages which permit this option, viz. the non-pro-drop languages. Space limitations prevent us from developing this idea further.

References Albright, A. (2002). “Islands of reliability for regular morphology: Evidence from Italian”. Language, 78(4), 684–709. Arbisi-Kelm, T. (2005). “Acquisition of aspect and tense in Spanish”. To appear in UCLA Working Papers in Linguistics. Becker, M. & Hyams, N. (1999). “Modal reference in children’s root infinitives”. In E. Clark (Ed.), Proceedings of the 13th Child Language Research Forum (pp. 113–122). Stanford University, CA: CSLI. Behrens, H. (1993). Temporal Reference in German Child Language. PhD dissertation, University of Amsterdam. Bel, A. (2001). Teoria Lingüística i adquisició del llenguatge. Anàlisi comparada del trets morfològics en català i en castellà. Barcelona: Institut d’Estudis Catalans. Belletti, A. (1990). Generalized Verb Movement. Turin: Rosenberg & Selier. Belletti, A. (2004). “Aspects of the low IP area”. In Rizzi (Ed.), The Structure of CP and IP. The Cartography of Syntactic Structures, Vol. 2 (pp. 16–51). Oxford: Oxford University Press. Berger-Morales, J. & Salustri, M. (2003). “Supporting the separate systems hypothesis: A case study in bilingual acquisition of Italian and German”. In A. T. Pérez-Leroux & Y. Roberge (Eds.), Romance Linguistics: Theory and Acquisition. Amsterdam: John Benjamins. Berger-Morales, J., Salustri, M., & Gilkerson, J. (2003). “Root infinitives in spontaneous speech of two bilingual children: Evidence for separate grammatical system”. To appear in J. Cohen, K. McAlister, K. Rolstad, & J. MacSwan (Eds.), ISB4: Proceedings of the 4th International Symposium on Bilingualism. Somerville, MA: Cascadilla Press. Blom, E. (2003). From RIs to Finite Sentences. PhD dissertation. University of Utrecht. Borer, I. (1984). Parametric Syntax. Dordrecht: Foris. Boser, C. , Lust, B., Santelman, L., & Whitman, J. (1992). “The syntax of CP and V2 in early child German”. NELS, 23, 51–65. Brun, D. & Babyonyshev, M. (2004). “The acquisition of statives in Russian”. ZAS Papers in Linguistics, 29, 59–72. University of Berlin. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Davidson, L. & Goldrick, M. (2003). “Tense, agreement and defaults in child Catalan: An optimality theoretic account”. In S. Montrul & F. Ordoñez (Eds.), Linguistic Theory and Language Development in Hispanic Languages (pp. 193–211). Sommerville, MA: Cascadilla Press. de Houwer, A. (1995). “Bilingual language acquisition”. In P. Fletcher & B. MacWhinney (Eds.), Handbook of Child Language. Oxford: Basil Blackwell. de Mauro, T., Mancini, F., Vedovelli, M., & Voghera, M. (1993). Lessico di frequenza dell’italiano parlato. Milano: Etaslibri. Ferdinand, A. (1996). The Acquisition of the Subject in French. HIL/Leiden University. Guasti, M. T. (1992). “Verb syntax in Italian child grammar”. Geneva Generative Papers, 0, 145– 162.

JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.23 (1689-1799)


Guasti, M. T. (1993/1994). “Verb syntax in Italian child grammar: Finite and nonfinite verbs”. Language Acquisition, 3, 1–40. Grinstead, J. (1998). Subjects, Sentential Negation and Imperatives in Child Spanish and Catalan, Doctoral dissertation. University of California, Los Angeles. Halle, M. & Marantz, A. (1993). “Distributed morphology and the pieces of inflection”. In K. Hale & S. J. Keyser (Eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger (pp. 111–176). Cambridge, MA: MIT Press. Han, C. (2001). “Force, negation and imperatives”. The Linguistic Review, 18, 289–325. Hoekstra, T. & Hyams, N. (1998). “Aspects of root infinitives”. Lingua, 106, 81–112. Hoekstra, T. & Jordens, P. (1994). “From adjunct to head”. In T. Hoekstra & B. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 119–149). Amsterdam: John Benjamins. Hyams, N. (2005a). “Child non-finite clauses and the mood-aspect connection: Evidence from Child Greek”. In P. Kempchinsky & R. Slabakova (Eds.), The Syntax, Semantics and Acquisition of Aspect (pp. 293–316). Dordrecht: Kluwer. Hyams, N. (2005b). “Aspectual effects on interpretation in early grammar”. Unpublished UCLA manuscript. Hyams, N. (2002). “Clausal structure in child Greek: A reply to Varlokosta, Vanikka and Rohrbacher and a reanalysis”. The Linguistic Review, 19, 225–269. Ingram, D. & Thompson, W. (1996). “Early syntactic acquisition in German: Evidence for the modal hypothesis”. Language, 72(1), 97–120. Kramer, I. (1993). “The licensing of subjects in early child language”. MIT Working Papers in Linguistics, 19, 197–212. Lasser, I. (1997). “The interpretation of root infinitive constructions in adult and child German”. In J. Scheaffer (Ed.), Proceedings of the Workshop on the Interpretation of Root Infinitives and Bare Nouns in Child Language, MIT Occasional Papers in Linguistics. MIT. Leonini, C. (2002). “Assessing the acquisition of Italian as L1: A longitudinal study of a child of pre-school age”. Unpublished manuscript, University of Siena. Londe, Z. C. (2004). “The Hungarian Case of Root Infinitives: Too many cases no place for root infinitives”. Unpublished, UCLA Manuscript. López Ornat, S. (1994). La adquisición de la lengua española. Madrid: Siglo XXI. Meisel, J. (1990). “Inflection, subjects and subject verb agreement”. In J. Meisel (Ed.), Two First Languages: Early Grammatical Development in Bilingual Children (pp. 237–298). Dordrecht: Foris. Montrul, S. (2003). The Acquisition of Spanish. Amsterdam: John Benjamins. Newport, E., Gleitman, H., & Gleitman, L. (1977). “Mother, I’d rather do it myself: Some effects and noneffects of maternal speech style”. In C. E. Snow & C. A. Ferguson (Eds.), Talking to Children: Language Input and Acquisition (pp. 109–150). Cambridge: Cambridge University Press. Pertsova, K. & Salustri, M. (2004). “Learning underspecification”. Poster presented at GALANA 2004, University of Hawaii. To appear in UCLA Working Papers in Linguistics. Plunkett, K. & Strömqvist, S. (1990). “The acquisition of Scandinavian languages”. Gothenberg Papers in Theoretical Linguistics, 59. Rhee, J. & Wexler, K. (1995). “Optional infinitives in Hebrew”. In C. Schutze, J. Ganger, & K. Brohier (Eds.), Papers on Language Processing and Acquisition. MIT Working Papers in Linguistics, 26, 383–402. Rizzi, L. (1993/1994). “Some notes on linguistic theory and language development: The case of root infinitives”. Language Acquisition, 3, 371–393.



JB[v.20020404] Prn:5/06/2006; 15:03

F: LALD4108.tex / p.24 (1799-1891)

 Manola Salustri and Nina Hyams

Rizzi, L. (1997). “The fine structure of the left periphery”. In L. Haegeman (Ed.), Elements of Grammar: Handbook in Generative Syntax (pp. 281–337). Dordrecht: Kluwer. Rus, D. (2004). “Embedded imperatives in Slovenian”. Unpublished Georgetown University manuscript. Rus, D. & Chandra, P. (2005). “Bare participles are not root infinitives: Evidence from early child Slovenian”. To appear in Proceedings of the 29th annual BUCLD. Salustri, M. & Hyams, N. (2003). “Is there an analogue to the RI stage in the null subject languages?” Proceedings of the 27th annual BUCLD, 692–703. Sano, T. & Hyams, N. (1994). “Agreement, finiteness and the development of null arguments”. Proceedings of NELS, 24, 543–558. Schaeffer, J. (1990). The Syntax of the Subject in Child Language: Italian Compared to Dutch. MA Thesis, State University of Utrecht. Stephany U. (1997). “The acquisition of Greek”. In D. I. Slobin (Ed.), The Crosslinguistic Study of Language Acquisition (pp. 183–333). Mahwah, NJ: Lawrence Erlbaum. Sigurjónsdóttir, S. (2004). “The different properties of root infinitives and finite verbs in the acquisition of Icelandic”. To appear in Proceedings of the 29th annual BUCLD. Torrens, V. (1995). “The Acquisition of Inflection in Spanish and Catalan”. In C. Schütze, J. Ganger, & K. Broihier (Eds.), Papers on Language Processing and Acquisition. MIT Working papers in Linguistics, 26, 451–472. Torrens, V. (2002). La adquisición del tiempo y la concordancia. La continuidad en el proceso evolutivo. Madrid: Estudios de la Universidad Nacional de Educación a Distancia (UNED). Van Kampen, J. (2004). “Early Operators and late topic-drop/pro-drop”. University of Utrecht unpublished manuscript. Varlokosta, S., Vainikka, A., & Rohrbacher, B. (1998). “Functional projections, markedness, and ‘root infinitives’ in early child Greek”. The Linguistic Review, 15, 187–207. Vila, I. (1985). ICE (Institute of Educational Sciences). University of Barcelona. Wexler, K. (1994). “Optional Infinitives, head movement, and the economy of derivations”. In D. Lightfoot & N. Hornstein (Eds.), Verb Movement (pp. 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. Wexler, K., Schaeffer, J., & Bol, G. (2004). “Verbal syntax and morphology in Dutch normal and SLI children: How developmental data can play an important role in morphological theory”. To appear in Syntax. Wijnen, F. (1994). “Incremental acquisition of phrase structure: A longitudinal analysis of verb placement in Dutch”. Groningen University manuscript. Wijnen, F. (1997). “Temporal reference and eventivity in root infinitives”. In J. Schaeffer (Ed.), Proceedings of the Workshop on the Interpretation of Root Infinitives and Bare Nouns. MIT Occasional Papers in Linguistics, 12, 1–25. Zanuttini, R. (1997). A Comparative Study of Romance Languages. Oxford: Oxford University Press.

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.1 (47-112)

The acquisition of experiencers in Spanish L1 and the external argument requirement hypothesis* Vincent Torrens#, Linda Escobar# and Kenneth Wexler## #

Psycholinguistic Institute / ## Massachusetts Institute of Technology

We present evidence that children have difficulty with the Spanish psych verbs that do not project the subject as the external argument. Our findings support the External Argument Requirement Hypothesis (EARH), according to which children until 5 or more have trouble with base structures that don’t assign a subject/external argument. On the basis of our results, we argue that acquisition of the entire class of psych verbs depends on the acquisition of one linguistic property that allows the L1 grammar to generate structures with no external argument.

.

Introduction

Specific details of linguistic knowledge are biologically determined, since there is an absence of knowledge in the presence of abundant evidence. The biology that supports the relevant knowledge is not available until a comparatively late stage in child development. The proposal that certain properties of language derive from biologically determined properties of the human brain is born out by the fact that research reveals extremely early linguistic knowledge, the existence of linguistic universals and the ‘poverty of stimulus’ argument. Also, the human organism changes state as it matures and the brain structures that instantiate UG are subject to some kind of maturational timetable. There are two possible types of explanation for late learning: input-driven explanations that attribute late knowledge to nurture rather than nature, and non-linguistic maturational explanations attributing late knowledge to non-linguistic developmental factors. One explanation for a child’s lack of some kind of linguistic knowledge at a particular age could be insufficient exposure to certain relevant linguistic facts by that age. Another kind of explanation is non-linguistic, which attributes late knowledge to late maturation of mental capacities other than the human language faculty; in this

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.2 (112-177)

 Vincent Torrens et al.

category one might explore limitations of memory capacity, overall processing speed, or communicative abilities. In contrast to input-driven late knowledge and late knowledge arising from nonlinguistic maturation, purely linguistic maturation proposes that we cannot explain changes in knowledge of specific aspects of grammar at particular ages unless we assume innate structures that also change over time. Borer and Wexler (1987, 1992) proposed that the ability to represent passive constructions in an adultlike manner does not mature until about 4 years of age. Borer and Wexler (1987) explored the possibility that the problem concerns movement. They suggested that children at the relevant ages are unable to represent a passive clause with an A-chain linking object and subject. We will call this proposal A-chain Deficit Hypothesis (ACDH). Another linguistic maturation proposal is the External Argument Requirement Hypothesis (EARH), which localizes the deficit not in the A-chain but in the absence of an external argument (Babyonishev, Ganger, Pesetsky, & Wexler 2001; Borer & Wexler 1992). We postulate these proposals in the following way: (1) The A-Chain Deficit Hypothesis (ACDH) says that children until about 5 or more have trouble with object to subject A-chains. (2) The External Argument Requirement Hypothesis (EARH) says that until 5 or more children have trouble with base structures that don’t assign a subject/external argument. The ACDH and the EARH provide two distinct but similar explanations for the fact that A-chains may be absent in the early grammars. To distinguish between the predictions of these two hypotheses, one would have to examine children’s performance with two types of structures: 1. The first are structures that contain no external argument and no A-chain. The ACDH predicts these to be unproblematic, and the EARH predicts them to be problematic. One example of such structures might be finite complements embedded under raising verbs (e.g. it seems that Mary has left). Another example would be dative experiencers with Spanish psych verbs like gustar (e.g. a María le gustan los globos). 2. The second are the structures that contain an external argument and an A-chain other than (subject-object). The ACDH predicts these to be problematic, and the EARH predicts them to be unproblematic. One example might be raising verbs (e.g. John seems to have left). Another example would be the reflexive clitic construction of Romance languages. In this paper we will mainly test children’s performance with respect to psych verbs.

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.3 (177-239)

The acquisition of experiencers in Spanish L1

. The structure of Experiencers . Belletti and Rizzi (1988) Thematic roles usually have a regular mapping with respect to structure. In an active sentence agents are usually linked to a subject position, and patients are linked to an object position. Baker (1988) has proposed that we can apply general rules for any lexical item: (3) Uniformity of Theta Assignment Hypothesis (UTAH) Identical thematic relationships between items are represented by identical structural relationships between those items at the level of D-structure. However, dative experiencers pose a problem for generating general rules to map lexical representations onto syntactic structures. Belletti and Rizzi (1988) propose that verbs expressing psychological states have a uniform theta-grid, involving an experiencer, the individual experiencing the mental state, and a theme, the content or object of the mental state. This theta grid can be projected onto a variety of syntactic configurations in an apparently arbitrary way. They propose three main primitive lexical classes in Italian: (4) Italian psych verbs a. Gianni teme questo Gianni fears this b. Questo preocupa Gianni This worries Gianni c. 1. A Gianni piace questo to Gianni likes this ‘John likes this’ 2. Questo piace a Gianni This likes to Gianni ‘John likes this’ Following Belletti and Rizzi, (4a) seems to be simple transitive structures where the subject is the experiencer and the object is the theme with temere. In the second class (4b), represented by preoccupare, the subject is the theme and the object is the experiencer. The third class (4c), represented by piacere, involves a dative experiencer and a nominative theme, with both orderings allowed. In Spanish we find the same type of primitive lexical classes, although they have different properties: (5) Spanish psych verbs a. Juan siente frío Juan feels cold



JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.4 (239-300)

 Vincent Torrens et al.

b. 1. Esto (le) asusta a Juan This (to him) frightens to Juan ‘This frightens John’ 2. A Juan *(le) asusta esto To John *(to him) frightens this ‘This frightens John’ c. 1. Esto le gusta a Juan This to him likes to Juan ‘John likes this’ 2. A Juan le gusta eso to Juan to him likes this ‘John likes this’ In Spanish the second and the third type allow both orderings. Also, for the second type doubling of the direct object is optional for non-fronted experiencers whereas doubling of the direct object is obligatory for the fronted experiencers. For the third type doubling is obligatory for fronted and non-fronted experiencers. Following Belletti and Rizzi, the lexical representations of the three classes are identical except for one lexical parameter involving the selection of different inherent cases. They include two components in the theory of the lexicon: (a) lexical representations and (b) a set of principles guiding the mapping of lexical representations onto deep syntactic configurations. Concerning lexical representations, we assume that the lexical entry of each verb contains at least two specifications: a theta-grid and a case-grid. The theta-grid is an unordered list of theta-roles. Theta-grids have a minimal internal structure in that they formally single out, for instance through underscoring, the external theta-role, the theta-role assigned to the subject position. The case-grid is a specification of the inherent cases idiosyncratically selected by a verb. Each inherent case is theta-related in the sense that it is linked to a specific slot in the theta-grid. Verbal entries consisting of theta and case-grids are then mapped to syntactic representations under the constraining effect of a set of mapping principles. These principles guide the shaping of syntactic structures on the basis of the substantive thematic information in lexical entries. These principles can be regarded as the substantive component of the Projection Principle. Belletti and Rizzi propose that the lexical representations that generate the required syntactic configurations for the three classes of psych-verbs are the following: (6) Lexical representations of Italian psych verbs Temere: theta-grid [Experiencer, Theme] case-grid [ – – ] Preocupare: theta-grid [Experiencer, Theme] case-grid [ acc – ] Piacere:

theta-grid [Experiencer, Theme] case-grid [ dat – ]

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.5 (300-367)

The acquisition of experiencers in Spanish L1 

Temere has the Experiencer selected as the external theta-role, and no inherent Case is specified. This gives rise to the simple transitive structure. Preocupare and piacere have no external theta-role, and the Experiencer is linked to an inherent Case – accusative and dative, respectively. The Experiencer must remain VP-internal because it is not marked as the external theta-role, and must be higher than the Theme. The only configuration meeting these requirements is the one with the Theme as sister to V and the Experiencer as sister to V’. Belletti and Rizzi give an unaccusative analysis to the preocupare and piacere class. Pesetsky (1995) has suggested that we need a finer-grained semantics to account for all contrasts in dative experiencers. He has proposed other theta roles for dative experiencers: the subject argument is the role Causer when the experiencer is the Object (second and third primitive lexical classes on the B&R classification), whereas the object argument is the role Target of Emotion/Subject Matter of Emotion when the experiencer is the Subject of the sentence (first primitive lexical class on the B&R classification). Following Pesetsky, the assignment for Experiencer predicates is made on the basis of the hierarchy in (7), where the highest argument is mapped into the highest D-structure position in its clause. (7) Causer > Experiencer > Target/Subject Matter Pesetsky proposes the following syntactic structures for all the kinds of dative experiencers. When Causer and Experiencer are the arguments of the sentence, Causer is in the higher position and Experiencer is in the lower position: (8) [VP [V’ V Experiencer] Causer] with an Experiencer and a Target/Subject Matter, Experiencer is in the higher position and Target/Subject Matter is in the lower position: (9) a. [VP [V’ V Target] Experiencer] b. [VP [V’ V Subject Matter] Experiencer] In the next section we attempt to outline the syntactic structures of the different dative experiencer classes, together with the structure of doubled structures in Spanish.

. Torrego’s (1998) object dependencies in Spanish Torrego (1998) studies verbal objects in Spanish. In her monograph she focuses on a class of verbal objects that displays syntactic and semantic properties that are typical of subjects: the objects of accusative verbs morphologically marked with the dative preposition a. For our purpose, she deals with marked accusatives, clitic doubling and dative experiencers as phenomena that may be analysed as a group and define a particular parameter for Peninsular Spanish. The first important parametrical property to consider is that in Peninsular Spanish clitics are necessarily in the dative when doubling marked accusatives which are analysed categorially as P-DPs:

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.6 (367-413)

 Vincent Torrens et al.

(10) a.

Juan lo visitó al chico Juan to him-acc visited the boy ‘John visited the boy’ (* – Peninsular Spanish, OK – Southern Cone Spanish) b. Juan le visitó al chico Juan to him-dat visited the boy ‘John visited the boy’ (OK – Peninsular Spanish, * Southern Cone Spanish)

She derives the fact that the clitic needs to be in the dative in Peninsular Spanish on internal grounds. In particular, she claims that the dative clitic together with the Double form part of one PP, headed by a functional P as follows: P

(11) le

P’ P

a-DP (DOUBLE)

By adopting this structure for this type of clitic doubling in Peninsular Spanish she may derive the fact that the double may inherently check the Case feature against a functional head with a D-feature, just as is argued to be the case of Case-checking of subjects. This is, in addition, explains why the dative clitic with psych verbs like gustar is compulsory: (12) A los alumnos *(les) gusta el libro to the students *(to them) like the book ‘students like the book’

(Torrego 1998)

Following Belletti and Rizzi (1988), the structure assumed for psych predicates such as temer is a transitive structure, and the structure of predicates such as preocupar and gustar is a VP containing two arguments with the experiencer in a position higher than the theme (with nonthematic subject position), with the only difference that the theme gets accusative case in the first case, and dative case in the second case: (13) a. temer: [S Experiencer [VP V NP]] b. preocupar, gustar: [S ec [VP [V’ V NP] Experiencer]] Torrego suggests that Dative Experiencers in Peninsular Spanish appear in the canonical position of subjects, i.e. SPEC, VP while the theme appears in the canonical position of objects, i.e., V’,COMPL. The resulting partial derivation is as follows:

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.7 (413-465)

The acquisition of experiencers in Spanish L1  T’

(14)

VP

T

V’

PP le

P’ V

P

OBJECT

double DAT

One important fact bearing on her proposal is that in Peninsular Spanish, the experiencer has to be in the dative and the theme has to be in the nominative: (15) a.

le gustas to him-dat like-2s ‘he likes you’ b. *lo gustas to him-acc like-2s ‘he likes you’

tú you tú you

The analysis that Torrego puts forward to deal with this type of psych verbs is within the minimalist framework and she adopts a feature-attract operation in syntax. On the view that clitics are functional categories, she presents strong evidence that accusative clitics are housed in small v, outside the VP, whereas the dative clitic is housed inside the VP, namely in functional p. The presence of T in the partial derivation of (14) is responsible for the attraction (covertly) of the nominative CASE feature of the theme. The dative experiencer checks its Case against the functional preposition p. In this view, she can deal with the general observation that the postulation of v is highly doubtful for unaccusative verbs. Turning to accusative verbs, in the absence of a doubling clitic, Torrego adopts a VP shell à la Larson’s (1988) proposal for English ditransitives, further adapted to Chomsky’s (1995) proposal of including a functional v responsible for the D-feature and the structural CASE checking of internal objects. For grammars like Peninsular Spanish where marked accusatives obligatorily occur with a dative preposition, the analysis is overt raising into Spec of functional v, as follows:

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.8 (465-524)

 Vincent Torrens et al.

(16) a. vp

b. v’

v

v’

P-OBJ VP

V

vp

v P-OBJ

VP V

t

The overt raising operation in (16b) of the Double into Spec, v is motivated by the person and agreement features of D in v, similar to what happens with the raising of subjects into T – also marked with a D feature. The claim that in the previous analysis of clitic doubling functional P acts as small v for the purpose of Case shecking follows the analysis of Kayne (1984) and Pesetsky (1995) concerning datives and other related structures.

. Phases and Wexler’s hypothesis for syntactic development The central proposal of the Minimalist Theory as stated by Chomsky (1995, 2000, 2001) is that language is an optimal solution of the computational system (grammar) to the constraints imposed by the two interfaces: with the articulatory-perceptual system (PF) and the conceptual-intentional system (LF). If this assumption is correct, constraints on syntactic operations follow from the fact that language has to be processed, produced and stored. The conclusion that there are limitations on the quantity of information that can be kept in the workspace/short-term memory gives rise to the proposal that such computational complexity can limit the scope of syntactic operations. Chomsky (2000, 2001) advances the idea of Multiple Spell-Out, which suggests that syntactic computation proceeds in phases. In brief, a phase is a unit of syntactic computation that can be sent to Spell-Out. Phases are defined by non-exhaustive enumeration: vPs (of a particular kind) and CPs are phases, VPs and TPs are not. The reduction in the computational burden is assumed to follow from the (in)accessibility of the content of a phase from outside this phase (strict cycle). According to this view, only the edge of a phase can be accessed from outside this phase. The general idea of this proposal is to limit syntactic computation to smaller chunks, that would be easier to process for the computational system. Chomsky (2000) suggests that phases must be φ-complete, i.e. have all their internal arguments saturated. Thus vPs are φ-complete while VPs, which still lack a subject, are not. In adult language, the accepted assumption is that unaccusative vPs are not phasal. This could be reinterpreted by Torrego’s (1998) observation that a postulation of a vP is doubtful for unaccusatives.

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.9 (524-590)

The acquisition of experiencers in Spanish L1

Wexler (2004) hypothezises that children believe that all vPs are phasal, i.e. project entire phases. Thus (verbal) passives and unaccusative vPs in the child, unlike the adult, will project phases and the resulting structures will be ungrammatical. In this sense, Wexler accounts for the well-known inability of the child to give uniformly correct representations to such structures. This constraint will replace the A-Chain Deficit Hypothesis, doing a better job empirically. In what follows we will fully focus on our experiment and show that Wexler’s constraint will be empirically motivated.

. Psych verbs and left-dislocation structures In Spanish there are three particular clausal structures containing fronted phrases doubled by a clitic. These structures (Clitic-Left Dislocation, Fronted Clitic Doubling, Fronted Dative Experiencers) have been provided with different analyses, ranging from unified to more independent ones in adult grammar (Escobar 1995; Ordoñez 2000; Sportiche 1996). (Clitic Left Dislocation) A la niña la peina su mamá (To) the girl, her combs her mother ‘the girl, her mother combs her’ b. Al niño le asusta una araña (Fronted Dative Experiencer) (To) the child, him frightens a snake ‘the child, the snake frightens her’

(17) a.

Both structures in (17) are homophones. Therefore, children could treat dative experiencers as clitic dislocated structures. To avoid this problem, we will test children with clitic left dislocated structures (not derived from clitic doubling) and with dative experiencer structures with a quantifier. It is well known that in adult language universal quantifiers with a distributive reading cannot appear in a Topic or left-dislocated position but as part of the clause, i.e., as preverbal subjects: (18) a. *A cada niña su madre la peina (To) each girl, her mother combs her ‘each girl, her mother combs her’ b. A su niña cada madre la peina (To) her girl, every mom her combs ‘her girl, each mother combs her’

(Clitic Left Dislocation)

(Clitic Left Dislocation)

In Spanish, dative experiencers can appear with quantifiers because they are not left-dislocated but in clitic doubling configurations in the clause, witness (18a) versus (19a): (19) a.

A cada niña le gusta un globo (To) each girl her likes a balloon ‘each girl likes a balloon’

(Fronted Dative Experiencer)



JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.10 (590-705)

 Vincent Torrens et al.

b. A la niña le gusta un globo (To) the girl her likes a balloon ‘the girl likes the balloon’

(Fronted Dative Experiencer)

If children treat datives in adult language, as in (19) they will also reject CLLDed structures like those in (18a). Clitic left dislocated structures don’t have an shomophone when used with quantifiers. Therefore, these structures are useful to provide evidence for ACDH or EARH. We have chosen the quantifier cada (each) because it’s the universal quantifier, and children don’t make any errors with the kind of items that we will use in the experiments (Escobar & Baauw 1997; Philip 1995). Therefore, if children commit errors, this is not because of the lack of knowledge on quantifiers, but because of the difficulty of interpreting the quantifier in the ungrammatical context of a clitic left dislocation. As for quantifiers, there is a considerable literature on the occurrence of errors in the child interpretation of universal quantifiers in contexts of underexhaustive search (Crain et al. 1996; Drozd 2001; Philip 1995). In other unbiased contexts, however, Chien and Wexler (1990) suggest that in a question like “Is every bear touching her head?” even very young children may obtain the adult distributive reading proper of the wide scope of universally quantified NP “every bear”. These results have been replicated for Spanish universal quantifiers acting as postverbal subjects in contexts of Clitic Left Dislocation (Escobar & Baauw 1997). The implication of this is that if children accept structures in (18a), they may provide a dative experiencer analysis. This would give support to ACDH. If children have trouble interpreting Clitic Left Dislocation with a fronted quantifier as adults do, then EARH would be born out. In order to test these implications we have included verbs that may go with experiencers like: gustar (like), asustar (frighten), divertir (amuse), molestar (bother), faltar (miss).

. Implications of the ACDH and EARH for child language The aim of this paper is to use dative experiencers and clitic left dislocation to see whether we could get evidence for EARH or ACDH. In Tables 1 and 2 we show what the ACD and the EAR hypotheses predict with respect to the acquisition of fronted and non-fronted dative experiencers. On adult analysis Dative Experiencers are missing an external argument/subject and do not have an object to subject A-chain. Thus the structure (19) violates EARH but does not violate ACDH. Thus EARH predicts that children up to age 5 or more will not get the structure right, but ACDH predicts that children will get the structure right. EARH predicts that children give to dative experiencers the clitic-left-dislocation analysis, whereas ACDH predicts that children apply the adult dative experiencer analysis. In this experiment we want to know whether children give the Clitic Left Dislocation structure in (17a) or the adult analysis for dative experiencers. ACDH predicts that they will give the dative experiencer analysis and therefore unlike adults will accept

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.11 (705-744)

The acquisition of experiencers in Spanish L1 

Table 1. Predictions of ACDH and EARH with respect to non fronted experiencers for the three types NONFRONTED [1st type: –A-chains, +External Argument] [2nd, 3rd type: –A-chains, –External Argument]

ACDH EARH

EXPERIENCERS (1st TYPE)

EXPERIENCERS (2nd TYPE)

EXPERIENCERS (3rd TYPE)

No errors No errors

No errors Many errors


Table 2. Predictions of ACDH and EARH with respect to fronted experiencers and clitic left dislocation FRONTED [–A-chains, –External Argument]

ACDH EARH

EXPERIENCERS

LEFT DISLOCATION


Many errors Many errors

(18a), whereas EARH predicts that they will give the clitic-left-dislocation analysis and like adults will reject (18a).

. The experiments In order to test the EARH hypothesis we have designed two experiments. The first experiment tests children’s comprehension of psych verbs in non fronted structures. The second experiment tests children’s comprehension of fronted psych verbs and Clitic Left Dislocated structures. This experiments only test comprehension and not production, since production has already been studied in Torrens and Wexler (2000).

. Experiment 1 The experiment design is based on Chien and Wexler’s (1990). To test children’s comprehension, the experimental items are presented as questions and children are expected to produce “yes” and “no” responses which will be measured according to adult grammaticality judgements. Children answered to four different sentences, containing the first, second (with and without double clitic), and third types of experiencers with a quantifier. This time the experiencers were not fronted. All questions corresponded to a picture containing a subject experiencing a psychological state, and children were asked about what the subject was experiencing. All questions corresponded to a picture containing a number of objects displayed in a distributive fashion in a symmetrical way, where (a) all the actors were displaying the action, or (b) all but one were displaying the action:

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.12 (744-820)

 Vincent Torrens et al.

(20) ¿todos los niños sienten frío? all children feel cold? Child1 feels cold Child2 feels cold Child3 feels cold Answer: Yes Child1 feels cold Child2 feels cold Child3 does not feel cold Answer: No We performed the experiment to 27 Spanish speaking children. We have grouped the children in 3 age groups: 4 year-olds, 5 year-olds and 6 year-olds, on a cross-sectional design. Ages are considered relevant to test both hypotheses. Following are the tables with the number of children for each age: 4-year-olds: 5-year-olds: 6-year-olds:

8 9 10

.. Results Following is the rate of errors for the different kinds of psychological verbs: Table 3. Psychological verb (sentir)

4-year-olds: 5-year-olds: 6-year-olds: total % correct

right

wrong

14 18 20 52 52/54 (96.30%) correct

2 0 0 2 2/54 (3.70%) incorrect

Table 4. Psychological verb (asustar + divertirle)


right

wrong

18 36 20 74 74/88 (84.09%) correct

14 0 0 14 14/88 (15.91%) incorrect

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.13 (820-862)

The acquisition of experiencers in Spanish L1 

Table 5. Psychological verb (gustar)


right

wrong

10 18 20 48 48/54 (88.89%) correct

6 0 0 6 6/54 (11.11%) incorrect

We found that children make a significant number of errors on dative experiencers in non fronted structures with quantifers (P < 0.001); therefore, children have difficulties on interpreting dative experiencers in fronted a non-fronted structures with quantifiers. We found differences on the number of errors between the different kinds of Dative Experiencers in non fronted structures: children make a non significant number of errors for the first type of experiencer (sentir) (P < 0.001). We found that children commit more errors on fronted dative experiencers than in non-fronted dative experiencers (P < 0.10).

. Experiment 2 The experiment design is based on Chien and Wexler’s (1990). To test children’s comprehension, the experimental items are presented as questions and children are expected to produce “yes” and “no” responses which will be measured according to adult grammaticality judgements. Children answered to two different sentences, containing dative experiencers with a quantifier, and clitic left dislocation with a quantifier. We included dative experiencers of the second and third class described by Belletti and Rizzi (1988). All questions corresponded to a picture containing a number of objects displayed in a distributive fashion in a symmetrical way, where (a) all the actors were displaying the action, or (b) all but one were displaying the action: (21) ¿A cada niño le asusta una araña? To each child cl-dat-m-s frightens a spider? ‘a spider frightens each child?’ Spider frightening child1 Spider frightening child2 Spider frightening child3 Answer: Yes Spider frightening child1 Spider frightening child2 Spider not frightening child3 Answer: No

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.14 (862-1008)

 Vincent Torrens et al.

(22) ¿A cada niña la peina su mamá? To each child cl-dat-f-s combs her mom? ‘her mother combs each girl?’ Mom combing child1 Mom combing child2 Mom combing child3 Answer: Yes Mom combing child1 Mom combing child2 Mom not combing child3 Answer: No ACDH predicts that children would make no errors on grammatical dative experiencers, no errors on quantifier dative experiencers, and many errors on left dislocated quantifiers. EARH predicts that children would make many errors on grammatical dative experiencers (because they don’t have an external argument), many errors on quantifier dative experiencers (because they don’t have an external argument), and many errors on left dislocated quantifiers (probably because children analyse them as dative experiencers, since left dislocated structures cannot have a quantifier). We performed the experiment to 35 Spanish speaking children. We have grouped the children in 3 age groups: 4 year-olds, 5 year-olds and 6 year-olds, on a crosssectional design. Ages are considered relevant to test both hypotheses. Following are the tables with the number of children for each age: Subjects 4-year-olds: 5-year-olds: 6-year-olds: 7-year-olds: total: grouped as 4 year-olds: 5-year-olds: 6-7 year-olds:

9 12 12 3 35 9 12 15

age range 4;04,14 to 5;1,17 5;6,17 to 6;2,0 6;1,2 to 7;0,24

.. Results Following is the rate of errors for Clitic Left Dislocation and for Dative Experiencers (Tables 6–9). We have found that children have many difficulties understanding both fronted dative experiencers and fronted left dislocation with quantifiers at ages 4;0 and 5;0: we found that children make a significant number of errors on clitic left dislocation and on dative experiencers (P < 0.001). We propose that children analyse left dislocated and

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.15 (1008-1023)

The acquisition of experiencers in Spanish L1 

Table 6. Clitic left dislocation

4-year-olds: 5-year-olds: 6-year-olds: 7-year-olds: total % correct

right

wrong

33 42 38 15 128 128/160 (80%) correct

12 18 2 0 32 32/160 (20%) incorrect

Table 7. Dative Experiencers (asustar)


right

wrong

18 24 24 9 75 75/96 (78.12%) correct

9 12 0 0 21 21/96 (21.88%) incorrect

Table 8. Dative Experiencers (gustar)


right

wrong

13 18 16 6 53 53/64 (82.81%) correct

5 6 0 0 11 11/64 (17.19%) incorrect

Table 9. Dative Experiencers (collapsed)


right

wrong

31 42 40 15 128 128/160 (80%) correct

14 18 0 0 32 32/160 (20%) incorrect

dative experiencers the same way, as left dislocated. We found the same rate of errors between dative experiencers and left dislocation. Also, we have found that children have the same difficulties with both types of experiencers when we collapse all ages, and for each age. We found that children commit more errors with fronted dative experiencers than with non-fronted dative experiencers (P < 0.10).

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.16 (1023-1071)

 Vincent Torrens et al.

. Discussion The results on the first experiment support EARH, because non fronted experiencers don’t contain A movement and have no external argument. ACDH predicts that children wouldn’t have difficulties in 2nd and 3rd types because these structures don’t have A-chains, whereas EARH predicts that children would have difficulties in 2nd and 3rd types because these structures don’t have an external argument. In addition, we found that children didn’t commit errors with the first type of experiencer, because this type of sentence is a common transitive structure. Furthermore, in order to supply evidence for the ACDH or the EARH, we have tested fronted dative experiencers, because they have no external argument, and have no A-chains. Results on the second experiment suggest that EARH is born out by the data because children make a significant number of errors when interpreting fronted dative experiencers. ACDH predicts that children will have no difficulties with dative experiencers with quantifiers and will have difficulties with left dislocation with quantifiers. EARH predicts that children will have difficulties with dative experiencers with quantifiers and left dislocation with quantifiers. With respect to clitic dislocated sentences, children make a significant number of errors, probably because children are applying the analysis of dative experiencers to clitic left dislocated structures, because the last structure is not possible with quantifiers. Torrens and Wexler (2000) studied the acquisition of clitic left dislocation and dative experiencers in Spanish in a spontaneous speech study. They found that children don’t have any difficulties with either structure. However, children could treat dative experiencers as clitic left dislcation. This could be the reason for the correctness attested in this study. For that reason, we decided to run an experiment where both structures can be differentiated. Studies on the acquisition of unaccusatives support the idea that children don’t have difficulties with the acquisition of A-movement. In the generativist framework, A-movement is the syntactic basis for the grammatical-function changing processes in passive, unaccusative, and raising constructions. These constructions contrast with transitive and unergative constructions in that the latter do not require A-movement or A-chains formation. Auxiliary selection with reflexives has been argued to reflect the unaccusative/unergative distinction. The use of the BE auxiliary with reflexive clitics follows from the fact that reflexive clitic constructions have an unaccusative analysis, in which the surface subject is underlyingly the direct object (e.g. Je me suis/*ai mordu). If children are allowed to violate the UG principles that force A-movement in adult languages, then we expect children to adopt an unergative analysis of reflexive clitic constructions. We thus expect children initially to use the same perfective auxiliary with both reflexive and non-reflexive clitics (e.g. Il m’a/*est mordu). Snyder, Hyams, Crisma (1995) have found that children don’t have any difficulties with reflexive clitics, which have an unaccusative analysis. In French Philippe never overextended BE to non-reflexive clitic constructions in any of 32 relevant utterances, and overextended HAVE to reflexive clitic constructions in only two of 109 cases. In Italian three children consistently used BE with reflexive clitics and HAVE with non-reflexive clitics

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.17 (1071-1139)

The acquisition of experiencers in Spanish L1 

throughout their corpora. The corpora in all cases end well before the fourth birthday. They conclude that all children mastered the adult system of AUX-selection with clitic pronouns well before the ages predicted by the ACDH hypothesis, and without the protracted period of erroneous uses predicted by the ACDH hypothesis. Snyder and Stromswold (1997) present evidence that children acquire English datives, verb-particle constructions, put-locatives, and causative/perceptual constructions all as a group. Their findings favour a parametric model of acquisition in which the acquired knowledge is not construction-specific. They also favour analyses in which all the constructions belong to a single syntactic class. In this sense, they put forward a Principles and Parameters approach to L1 acquisition, strongly claiming that structures related to a particular parameter will be acquired at the same time. If this Principles and Parameters approach is the right way to analyse L1 Acquisition and if Torrego (1998) is right, the three configurations described, being all equal, should be acquired at the same time since the three form part of the same group in Peninsular Spanish. When testing these configurations in CLLD structures, a potential problem may however arise and this should be related to children’s parsing mechanisms, namely, the fact that a clitic is independently compulsory in all CLLD structures (cf. Cinque 1990; Escobar 1995). For our purpose, the CLLD sentences in (23) may in principle be derived either from (24a) or (24b): (23) Al niño *(le) asusta la araña To the child him-dat frightens a spider? ‘the spider frightens the child?’ (24) a.

La the b. La the

araña snake araña snake

asusta al niño frightens the child le asusta al niño him frightens the child

As for ordinary structures, however, the prediction is that children will acquire the three configurations in (25) at the same time. (25) a.

La araña asusta al niño the snake frightens the child b. El payaso le divierte al niño the clown him amuses the child ‘the clown amuses the child’ c. Al niño le gusta el globo to the child him likes the balloon ‘the child likes the balloon’

Based on the data of the second experiment, we can also support the parametric approach to L1 acquisition put forward by Snyder and Stromswold (1997), according to which, structures related with each other under one parametric property are acquired

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.18 (1139-1232)

 Vincent Torrens et al.

at the same time, namely, marked accusatives doubled by a dative clitic and dative experiencers, which supports the unified analysis put forward by Torrego (1998).

Note * We would like to thank Marisé Gil de Gómez for helping us find schools in the Madrid area, and the primary schools of Alkor, Claudio Moyano and Albanta. We are grateful to the audiences of the ‘Lisbon Encounter on the Acquisition of Romance Languages’ and ‘The Romance Turn’ for helpful comments.

References Babyonishev, M., Ganger, J., Pesetsky, D., & Wexler, K. (2001). “The maturation of grammatical principles: Evidence from Russian unaccusatives”. Linguistic Inquiry, 32, 1–44. Baker, M. (1988). Incorporation: A Theory of Grammatical Function Changing. Chicago: University of Chicago Press. Belletti, A. & Rizzi, L. (1988). “Psych-verbs and theta theory”. Natural Language and Linguistic Theory, 6, 291–352. Borer, H. & Wexler, K. (1987). “The maturation of syntax”. In T. Roeper & E. Williams (Eds.), Parameter Setting. Dordrecht: Reidel. Borer, H. & Wexler, K. (1992). “Bi-unique relations and the maturation of grammatical principles”. Natural Language and Linguistic Theory, 10, 47–189. Chien, Y. C. & Wexler, K. (1990). “Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics”. Language Acquisition, 1, 225–295. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. (2000). “Minimalist inquiries: The framework”. In R. Martin, D. Michaels, & J. Uriagereka (Eds.), Step by Step: Essays in Minimalist Syntax in Honor of Howard Lasnik (pp. 89–155). Cambridge, MA: MIT Press. Chomsky, N. (2001). “Derivation by phase”. MIT Occasional Papers in Linguistics, 18. Cambridge, MA. (Reprinted in M. Kenstowicz (Ed.), Ken Hale: A Life in Language (pp. 1–52). Cambridge, MA: MIT Press.) Cinque, G. (1990). Types of A’-Dependencies. Cambridge, MA: MIT Press. Crain, S., Thornton, R., Boster, C., Conway, L., Lillo-Martin, D., & Woodams, E. (1996). “Quantification without qualification”. Language Acquisition, 5, 83–153. Drodz, K. (2001). “Children’s weak interpretation of universally quantified sentences”. In M. Bowerman & S. Levinson (Eds.), Conceptual Development and Language Acquisition. Cambridge: Cambridge University Press. Escobar, M. A. (1995). Lefthand Satellites in Spanish. Utrecht: OTS publications. Escobar, M. A. & Baauw, S. (1997). “The wide scope interpretation of postverbal quantifier subjects: QR in the early grammar of Spanish”. In A. Sorace, C. Heycock, & R. Shillcock (Eds.), Proceedings of the Gala ’97 Conference on Language Acquisition. Edinburgh: The University of Edinburgh. Kayne, R. S. (1984). Connectedness and Binary Branching. Dordrecht: Foris. Larson, R. (1988). “On the double object construction”. Linguistic Inquiry, 19, 335–391.

JB[v.20020404] Prn:5/06/2006; 15:07

F: LALD4109.tex / p.19 (1232-1308)

The acquisition of experiencers in Spanish L1 

Ordoñez, F. (2000). Clausal Structures of Spanish: A comparative Analysis. New York: Garland. Pesetsky, D. (1995). Zero Syntax: Experiencers and Cascades. Cambridge, MA: MIT Press. Philip, W. (1995). Event Quantification in the Acquisition of Universal Quantification. Doctoral dissertation. Amherst, MA: University of Massachusetts. Sportiche, D. (1996). “Clitic constructions”. In J. Rooryck & L. Zaring (Eds.), Phrase Structure and the Lexicon. Dordrecht: Kluwer. Torrego, E. (1998). The Dependencies of Objects. Cambridge, MA: MIT Press. Torrens, V. & Wexler, K. (2000). “The acquisition of clitic doubling in Spanish”. In S. Powers & C. Hamann (Eds.), The Acquisition of Scrambling and Cliticization. Dordrecht: Kluwer. Snyder, W., Hyams, N., & Crisma, P. (1995). “Romance auxiliary system selection with reflexive clitics: Evidence for early knowledge of unaccusativity”. In E. Clark (Ed.), Proceedings of the Twenty Sixth Annual Child Language Research Forum. Somerville, MA: Cascadilla Press. Snyder, W. & Stromswold, K. (1997). “The structure and acquisition of English dative constructions”. Linguistic Inquiry, 28, 281–317. Wexler, K. (2004). Recent theoretical and experimental results in the development of syntax from a minimalist point of view: Clitics, A-chains, Phases. Paper delivered at the workshop on language acquisition ‘The Romance Turn’. Madrid.

Appendix. Items for experiments Experiment 1 Dative Experiencers (with quantifiers) a. b. c. d. e.

¿todos los niños sienten frío? all children feel cold? ¿todos los niños sienten calor? all children feel hot? ¿el payaso divierte a todos los niños? the clown amuses all children? ¿la araña les asusta a todos los niños? the snake frightens all children? ¿les gusta el globo a todos los niños? all children like the balloon?

Experiment 2 Fronted Datives Experiencers a.

b.

¿a cada niño le asusta una araña? to each child him frightens a snake? ‘a snakes frightens each child?’ ¿a cada niño le divierte un payaso? to each child him amuses a clown? ‘a clown amuses each child?’

JB[v.20020404] Prn:5/06/2006; 15:07

 Vincent Torrens et al.

c.

¿a cada niño le gusta su globo? to each child him likes his balloon? ‘each child likes his balloon?’

Clitic left dislocation a.

b.

c.

¿a cada niña la peina su mamá? to each girl her combs her mom? ‘her mom combs each girl?’ ¿a cada niña la abraza su mamá? to each girl her hugs her mom? ‘her mom hugs each girl?’ ¿a cada niña la seca su mamá? to each child her dries her mom? ‘her mom dries each girl?’

F: LALD4109.tex / p.20 (1308-1323)

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.1 (46-124)

Early operators and late topic-drop/pro-drop* Jacqueline van Kampen Universiteit Utrecht This paper compares two reduction principles in child language, Tense/ Agreement Omission (Schütze 1997; Wexler 1998) and Root Peripheral Truncation (Rizzi 1994, 2001). The paper argues that the empirical content of these proposals as well as their partial overlap follows from an inherent hierarchy of acquisition steps of the kind predicted in Jakobson (1942). The empirical support for this reinterpretation follows from the construction of longitudinal graphs. The approach accommodates the lack of optional infinitives in Romance languages.

.

The acquisition of grammatical features

. Jakobson’s order of acquisition steps Children have a special talent to spot grammatical features. Yet, some grammatical features are acquired before others. Jakobson (1942) had a research proposal for it. Phonological features would fit into a natural hierarchy of UG. That hierarchy would predict the earliness of acquisition. Jakobson postulates that such a UG hierarchy of learnablity should exist for syntax as well as it exists for phonology. Jakobson’s research program, as I understand it now, is simple: Find that hierarchy of features and predict the order of acquisition steps given any target grammar. Go

(1)

Initial state {UG set of feature F}

Go

Þ

Gi-1

Þ

Sing le Value acquisition step of a feature F

Gi

…

Þ

Gn

Target grammar {UG features selected by input}

= initial state before the systematic pressure of grammatical features Gi = intermediate grammar Gn = target grammar Gi–1 ⇒ Gi = step towards intermediate grammar Gi by the acquisition of a feature F

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.2 (124-173)

 Jacqueline van Kampen

For each acquisition step one grammatical feature F is singled out, identified and subsequently acquired. Jakobson’s research proposal implies the Single Value Constraint for language acquisition (Clark 1993). In addition, it stresses the predictable hierarchy of feature acquisition c.q. parameter setting (cf. Evers & van Kampen 2001; van Kampen & Evers 2004). The present-day study of language acquisition relates the reductions in child language to the influence of UG principles. I will argue that the child’s order of acquisition steps must become a crucial ingredient of that approach.1 In order to develop that point, I will shortly consider two present-day Reduction Principles that partly predict child language. They are (i) Wexler and Schütze’s Tense/Agreement Omission and (ii) Rizzi’s Root Peripheral Truncation. My question will be how these two Reduction Principles relate. I will answer that question by placing them into a more Jakobson-like linear order of acquisition steps. This approach will reveal why agreement, topic-drop and pro-drop must appear late.

. Wexler’s Optional Infinitives English Child language often uses predicates that are not Inflection-marked, at least not overtly. The remnant predicate may take the form of an infinitive, a gerund, a past participle or a non-verbal head. Wexler and Schütze offer the following explanation. The acquisition procedure predicts a unique checking for each functional feature. Suppose now that the Spec,I subject somehow faces the acquisition procedure with a double checking requirement (Schütze 1997; Wexler 1998). Tense and Agr features both relate to the obligatory presence of the subject.2 (2)

IP Spec Subject case

IP

I0 j-features

Predicate

The double-checking procedure might cause a delay in which the Infl-marking of the predicate is left out. Pro-drop languages like Italian or Spanish require no double checking. They check the subject on Tense only.3 This predicts that pro-drop languages show no optional infinitives to speak of, which is according to the facts we found (Guasti 1994). There is no significant delay for agreement in pro-drop languages, at least not on verbal predicates.

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.3 (173-204)

Early operators and late topic-drop/pro-drop 

. Rizzi’s Truncations Rizzi’s (1994, 2001) abstract truncation principle fits several simplifications in child language. Child language often prefers to drop the peripheral specifier in root clauses. Plausibly, the peripheral specifier truncation may simplify the processing of constituent hierarchy in some general way. CP

(3) Spec

C’ IP

C°

I’

Spec I°

VP predicate

Truncation appears in V-2nd Dutch/German as root topic-drop in Spec,C, and in English/French as root subject-drop in Spec,I. If the truncation includes in addition the functional head Io or Co , one gets the constructions studied by Wexler as Optional Infinitives. Rizzi claims as well that there is no significant delay for agreement in pro-drop languages.

. Conflict of Reduction Principles Now, let’s see how the two Reduction Principles relate. Both Wexler/Schütze’s proposal and Rizzi’s proposal (4) a. target Io /Co projections b. imply Io features and/or Do features (Spec DP, Io /Agr) Besides that common point there is a difference as well, see (5). (5) a. Agr/Tense omission targets heads of the Infl-type b. Truncation targets specifiers (Spec,I) This leads me to the questions in (6).

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.4 (204-257)

 Jacqueline van Kampen

(6) a.

What is reduced first in child language? – an Inflection head? – an Inflection specifier? b. What is the licensing status of its structural partner at that moment? – If a functional head drops, must its specifier have a different licensing status? – If a specifier drops, must its functional head have a different licensing status?

To put it more simple: what happens when? The ‘when’ question will require longitudinal graphs from a single child moving towards a single grammar, as elaborated in van Kampen (1997), Evers and van Kampen (2001). Subsequently, the reductions must be related to the order of acquisition steps (if any).

. The acquisition of I-marking and D-marking The acquisition steps and the intermediate grammars they define follow a linear order. Some steps will precede others. The order of acquisition steps is probably the same for all children, given a target language. The order and the relative speed of acquisition steps can be shown by the construction of acquisition graphs. If we had a clear picture of language acquisition, we should be able to predict the order of acquisition steps given a target language. The acquisition order as pointed out by Jakobson will give a key to the history of grammar and to language typology. The linear order and the relative speed of acquisition steps should be predictable given a target grammar if we knew how learnability is controlled by input and UG. We still have to go some way to reach that goal, but to my mind it can be reached by the careful construction of acquisition graphs. I think that I have succeeded in reaching two important points along that way, see (7) (van Kampen 2004a). (7) a.

The acquisition of I-marking (auxiliaries, copulas, finite verbs) precedes the acquisition of D-marking (articles, possessives, demonstratives). b. The acquisition of D-marking coincides with the acquisition of pronominal reference tracking (the acquisition of anaphoric pronouns/clitics).

The longitudinal graphs for the acquisition order in (7a) are given in (8). They are constructed from the files of Dutch Sarah, between two and three years old.

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.5 (257-312)

Early operators and late topic-drop/pro-drop 

(8) actual graphs 100

percentage

80 A: I-marking 60 40

B: D-marking

20 0 85

95

105

115 125 135 age in weeks

145

155

(van Kampen 2001) Dutch Sarah: The acquisition of finite verbs and determiners Graph A: % of predicates marked by Io (a finite form) + a spelled-out subject Graph B: % of arguments marked by Do ( determiners before nouns) Graph A represents the rising percentage of I(nflection)-marked predicates by {auxiliaries, modals, copulas, inflection}. Within 20 weeks, I-marking rises from some 15% to more than 80% of the predicates. This is close to the level of the adult conversation, which is 90%. Graph B starts shortly after I-marking flattens near the target level. This graph takes some 25 weeks and represents the D(eterminer)marking of arguments by means of {articles, demonstratives, possessives}. It is possible to make a linear idealization of the graphs in (8), see (9). Let the point of irreversible rise be the eureka point and the point within 10–20% of the adult norm the acquisition point (Evers & van Kampen 2001). The time difference between these two points is the acquisition speed and the order between the acquisition point of A and the eureka point of B yields a clear case of acquisition order. Consider the case with graphs A for I-marking and B for D-marking in (9). (9) linear idealization I-marking A

D-marking

B

The two graphs separate early child language (before year 2) from later child language (after year 3). My claim is that only after and due to I- and D-marking, lexical categories, phrasal structure and grammatical relations get a formal grammatical basis, as stated in (10). This option, chosen here and in van Kampen (1997, 2004b, 2005) for language acquisition, was implemented earlier in computational approaches to

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.6 (312-333)

 Jacqueline van Kampen

category assignment (Buszowsky 1987). It has been recently advocated in theoretical approaches (Halle & Marantz 1993; Baker 2005; Borer 2005). (10) input frame

[Y [ Fi

· ·

X ]FP ]FP relation ì selection strict order

í

adjacency

î high frequency

where X = complement Fi Y = specifier Fi a. if Fi = Do , then X = N b. if Fi = Io , then X = V There is another important point. It relates to D-marking and was stated in (7b). It is possible to make a longitudinal graph for the acquisition of free anaphoric pronouns. Consider graph C in (11).4 (11) actual graphs 100

percentage

80 60 40

B

C

20 0 90

100

110

120 130 age in weeks

140

150

160

(van Kampen 2004a) Dutch Sarah: the acquisition of Do articles and free anaphors (graph based on 2 consecutive files) Graph B (again) % of arguments marked by Do ( determiners before nouns Graph C reflects the rising use of free anaphoric pronouns hij/zij/het (‘he/she/it’) The graph for D-marking (B) and the graph for free anaphoric pronouns (C) coincide for Dutch Sarah.5 A linear idealization of the graphs in (11) is given in (12).

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.7 (333-378)

Early operators and late topic-drop/pro-drop 

(12) linear idealization discourse reference tracking

situation bound modes B+C®

articles and free anaphoric pronouns

This is a striking confirmation of Postal (1968) that articles and pronouns represent the same category Do . It can be explained as follows. Both articles and free anaphoric pronouns are D-elements that refer back to previous linguistic discourse. The obsessive need of West European languages to add the article seems somewhat redundant. But it can be understood as a device that supports the speaker in discourse reference tracking. So, the rise of D-marking shows Sarah’s growing ability in reference tracking. She starts to enter real conversations, rather than holding on the situationbound exchanges of early child language.6 Her speech grows situation-free, if I may borrow Chomsky’s (1968) phrasing for this dramatic moment.

. Overlap between the Reduction Principles It is now possible to fit the two Reduction Principles into the acquisition space marked by I-marking and D-marking. See (13) and consider the following re-interpretations. I-marking +EPP(subject)

(13) target grammar

D-marking +free anaphors (pers.pronouns) (Agreement)

% a+b

b

c

acquisition time weeks

The triangular space covers Wexler’s Optional infinitives and copula-less predicates. It fits into the acquisition graph for I-marking and EPP. This space diminishes with the progress of acquisition in time. I-marking and Optional Infinitives are two sides of the same coin. The rectangular space covers Rizzi’s Left Peripheral Truncations. At first sight, it does not fit into the acquisition graph for I-marking and EPP. Left Peripheral Truncations appear before I-marking, after D-marking and in between. The triple cut-

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.8 (378-459)

 Jacqueline van Kampen

ting of the rectangular space does not necessarily weaken the significance of the two fundamental acquisition graphs, although it may seem that way. Rizzi’s Truncation principle is a structural abstraction. It takes different effects in different periods. For that reason, I will distinguish three different types of truncations, exemplified in (14)– (16). Type a (e.g. optional infinitives) Type b (e.g. imperatives) Type c (e.g. topic-drop, pro-drop)

no head Io beertje slapen (bear sleep) no Spec Io head Io kom eens (come to me) o no Spec I head Io /Co (∅) slaapt ook ((he) sleeps too) ∅ Spec Io /Co

Type a, type b, and type c all seem to lack a peripheral specifier at PF, especially if one has the adult system in mind. Type a lacks the peripheral specifier and in addition it lacks the Inflection-marked head. Type a appears mainly left of the I-marking graph, as it coincides with the triangular space. Type c is a predication with unambiguous full Agreement. It appears to the right of D-marking only. Type b, as I will show, exists in adult language productively mainly as the imperative, an agent-implying mood. In my view, child language employs more such agent-implying moods. Most of them give way to EPP predication in the adult language. My main point will be now to defend the relevance of the type a-b-c distinction. The major objective of that attempt will be to get the two Reduction Principles in a UG order of acquisition steps. We will first see some examples of types a, b and c, and then the characteristics.

. Rizzi’s Truncations for the three different predicate types . Examples of the predicate types The predicate types a-b-c are exemplified in (14)–(16). All may be interpreted as Rizzi-type truncations of peripheral material. Note that the part in italics is the child language utterance, the child’s intake. The remaining part of the input sentence is not within the child’s competence. The input reduction leads to preliminary acquisition frames, so-called evidence frames (van Kampen 1997). Type a is typical early child language. It is a bare predicate and represents one of Wexler’s Optional Infinitives. (14) Type a a. [jij]Spec [moet]I [liedje zingen]predicate (Dutch) b. [∅]Spec [vuoi]I [vedere libro?]predicate (Italian) c. [tu]Spec [vas]I [chercher camion]predicate (French)

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.9 (459-495)

Early operators and late topic-drop/pro-drop

Type b is the difficult one. It could be interpreted as a case of an early finite verb with a null subject, but I will argue that its typical restrictions rather fit the picture of a mode-operator like the imperative (see van Kampen 1997: Chapt. 4). (15) Type b a. [ik]Spec b. [∅/io]Spec [∅]Spec c. [je]Spec [∅]Spec

[kan wel]CP/IP [voglio chicco]IP [canta canzone]IP [veux pomme]IP [mets voiture dedans]IP

1st p (speaker) implied (Dutch) 1st p (speaker) implied (Italian) 2nd p (hearer) implied 1st p (speaker) implied (French) 2nd p (hearer) implied

Type c is typical for the second part of a conversation pair, in adult language as well as in later child language. Type c is discourse oriented, whereas type a/b are situationbound. (16) Type c a. [∅ d-topic]Spec [leest een boek]CP/IP (Dutch) b. [∅ pro]Spec [legge uno libro]IP (Italian) Truncations should be universal. So, a legitimate question is why type a (Wexler’s Optional Infinitives) hardly appear in the acquisition of a pro-drop language like Italian? As I will argue below, radically different answers can be given to that question.

. Characteristics of the predicate types The left Peripheral Truncations a, b and c have different relations to the functional categories I or C. CP

(17) Spec

C’ IP

C° Spec

I’ I°

VP predicate

Type a Type c Type b – after D-marking – before D-marking – before I-marking – topic predicate – operator predicate – bare predicate – situation-bound – discourse-oriented – situation-bound – peripheral specifier – mode-implied subject (in peripheral root position)



JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.10 (495-574)

 Jacqueline van Kampen

Type a lacks I/C-marking completely. Type a is early child language. It is a situationbound utterance. Type b is based on a form that reminds of or is , like the imperative. The specifier is absent, but it is not the case that a peripheral specifier has optionally been left out. Rather, the peripheral specifier cannot be present. My contention is that a fixed position for the sentential topic (peripheral specifier) is only present in type c sentences that have a formal discourse relation to previously uttered sentences. Type b, by contrast, marks a situation-bound intention by excluding a formal topic and by excluding the peripheral specifier. Type c is based on a finite form with all I/C oppositions marked. It has a peripheral specifier. The peripheral specifier may remain empty as an option in conversation pairs. The type c construction is typically discourse dependent.

. Type b and type c predicate types . Type b operator predicates Type b predicates were introduced in van Kampen (1997) as ‘mode-implied subjects’. Let us have a closer look at the type b predicates. It may be clear by now that I will not consider the operator predicates with mode-implied agent as examples of early subject-drop or early topic-drop. They rather belong to restricted situation-bound utterances already present in the proto-grammar before I-marking and D-marking. Examples of type b predicates, in Dutch child language (around two years) are given in (18). See also Jordens (2002). (18)

Type b predicates are not examples of early peripheral specifier-drop, because it is the illocutive function of the utterance that relates the predicate to first person (wish/intention), to second person (imperative), or to third person (intention/naming). Topic-drop, by contrast, relates to a topic that has already been expressed, in previous discourse, which is a completely different state of affairs.

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.11 (574-624)


Comparable type b predicates appear in English, French and Italian child language, see the examples in (19) for Dutch Sarah (van Kampen), French Grégoire (Champaud), Italian Diana and Martina (Calambrone), English Adam (Brown). (19)

Dutch French – hoef niet bad (S.2;1) – veux pas lolo (N.2;1) – doe ogen dicht (S.1;11) – mets dedans (G.2;0) – moet liedje aan (S.2;3) – va chercher (G.2;0) – is bal (often) – est ballon (often) Italian English 1st pers. (wish) – voglio chicco (M.1;11) – no wanna open (A.2;6) 2nd pers. (imperative) – do alla mamma (M.1;11) – put glove (A.2;4) 3rd pers. (intention) – va a lavorare (D.2;1) – goes there (A.2;6) 3rd pers. (naming) – è palla (often) – is ball (often)

1st pers. (wish) 2nd pers. (imperative) 3rd pers. (intention)7 3rd pers. (naming)

The examples in (19) all have the characteristics of the type b operator predicates. Firstly, they allow no free variation, they rather have an illocution-fixed person, see (20). (20) a. kwi is typically first person I (agent) want b. doe is typically second person you (agent) do/put c. moet is typically it must be that Secondly, there is no option for a peripheral specifier. The implied subject is never spelled out. The type b verbs are a fairly closed class of modal type qualifiers. Thirdly, like the imperative in the adult language, the type b predicate head in early child language is a root phenomenon.8 (21) a. type b predicate heads are sentence initial in Co /Io b. they typically do not allow the topic-discourse slot in the specifier position (because type b is situation-bound) The modal operator in Dutch child language is often derived from a finite modal verb in the adult language, as the examples in (18) show. But as soon as the finite forms come under the sway of the EPP (the I-graph), the system picks up first and second personal pronouns (I/you). Type b is then changed into an EPP predicate by adding the personal pronoun and it disappears as predicate type b (except for the imperative).

. Type c: Discourse topic-oriented predicates For type c, the peripheral specifier may be substituted by an empty A-bar pronoun in both Italian pro-drop and Dutch topic-drop. A characterization of the discourse properties of the A-bar pronoun in pro-drop and topic-drop languages is given in van Kampen (1997). Note that A-bar pronouns concern third person. First and second person are not discourse-oriented, but situation-bound (speaker/hearer oriented). They appear earlier (see also Pinto 2004).



JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.12 (624-693)

 Jacqueline van Kampen IP/CP Spec empty

(22) Previous Discourse

Spec I/C DPi Æ proi

Io/Co

Æ topici

i

Predicate

The empty A-bar pronoun is 3rd person only, unspecified for phi-features. The A-bar pronouns belong to discourse grammar. The specifier position can, but need not be empty. See respectively (23a) and (23b). (23) a. [∅ d-topic]Spec [∅ pro]Spec b. [die]Spec [mamma]Spec

[leest een boek]CP/IP [legge un libro]IP [leest een boek]CP/IP [legge un libro]IP

Topic-drop and pro-drop 3rd person have the characteristics in (24).9 (24) a. they refer to a previous discourse topic b. they are a 3rd person pro (pro-drop) or d-pronoun (topic-drop) c. they typically appear upon speaker change (Postma 2004) The empty pronoun can only refer to the topic of the preceding sentence. See Grimshaw and Samek-Lodovici (1998) and van Kampen (1992, 1997: 89f.) for an elaboration.

. Rise of type c predicates: Topic-drop (Spec,C ∅) It may appear problematic that type b and c look rather the same. Both seem to display a finite verb and a truncated specifier. Nevertheless, I claim that type b can appear before D-marking, whereas type c can only appear after D-marking. Do I have empirical evidence for that? Sure, I do. It is very well possible to distinguish type b predicates from type c predicates, since (25) a. c.

Type b are a small set of verbs with a fixed person, and there is no 3rd person agent he/she involved. Type c, by contrast, shows full Agreement and there is a 3rd person agent involved.

In (26), the space of finite predicates without peripheral specifier for Dutch Laura, between week 110 (onset of D-marking) and week 220 is given. Within that period, Laura reaches her type c truncation from marginal to over 80%. Her type c subject topic-drop closely follows the rise of D-marking (van Kampen 1997: 107).

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.13 (693-716)


(26) Percentual shift between type b and type c towards the adult norm (taken as 100%) Use of mode-implied versus topic-drop Laura (van Kampen corpus) 100

type b mode-imlied drop

percentage

80

60

type c: topic-drop

40

20

0 100 110 120 130 140 150 160 170 180 190 200 210 220 age in weeks before D-marking

after D-marking

Graph type b: Decline of operator predicates (situation-bound) characterized by (i) fixed truncation (ii) a fixed verb (iii) no 3rd person hij/zij (‘he/she’) involved. Graph type c: Rise of subject topic-drop in Spec,C root (discourse-oriented) characterized by (i) optional truncation (ii) full agreement (iii) 3rd person. The predicate space just before D-marking and just afterwards shows both types. The longitudinal graph of type c truncations, rises as soon as D-marking rises. The use of the mode-implied subject-drop type b becomes marginal in the same period. Type c is really a side effect of D-marking and reference tracking, whereas situation-bound type b is not.

. Conjecture: Type c Pro-Drop/Agreement is late My approach implies that full agreement follows D-marking. If that is true, I am forced to predict that Italian pro-drop is late, after D-marking. This is an empirical issue. Recent literature on the erratic use of person agreement (Grinstead 1998; Soares 2002; Avram & Coene 2004) suggests to me that the acquisition of non-emphatic subject pronouns in the Romance pro-drop languages may be late and not that different from non-emphatic pronouns in the acquisition of Germanic non-pro-drop languages. The



JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.14 (716-772)

 Jacqueline van Kampen

early finite verb may have “agreement” in a fixed default form. Clitic pronouns offer a parallel. Clitic pronouns are a reduced type of free anaphors. Acquisition data in French, Romanian and Italian show how clitics appear right after D-marking (van Kampen 2004a; Avram 2003; Müller & Kupisch 2004). This suggests that little pro/agr in Italian pro-drop is part of discourse reference tracking as well. If this is true, the data should allow a reinterpretation of early prodrop. The potential reanalysis should claim the two points in (27). (27) a.

Early finite verbs without a subject that appear before systematic Dmarking are fixed forms (type b, for example imperatives). b. Like object clitics, the regular use of full agreement and pro-drop will not appear at adult level before systematic D-marking.

Real pro-drop would appear after D-marking. Empirical support for (27a) and (27b) may be found in the direction of (28a) and (28b). (28) The period before D-marking in pro-drop languages is characterized by a. Absence of object clitics and erratic use of subject agreement b. Correct finite forms that are type b: imperatives The fact that object clitic pronouns appear right after D-marking has been attested for French by van Kampen (2004a), for Romanian by Avram (2003), for Spanish by Muñoz (2004), and for German-Italian bilingual children by Müller and Kupish (2004). Since pro-drop and clitics rely on pronominal oppositions (person/number) that function in discourse grammar, this may group them with all reference tracking elements, including pro/agreement. The erratic use of person agreement has been attested for Spanish by Grinstead (1998: Chap. 4.4.2), for Portuguese by Soares (2002: Table 7), and for Romanian by Avram and Coene (2004: Table 3). Avram and Coene (2004) claim as well that early ‘finite’ verbs in child Rumanian are non-adult like and that the pro-drop property appears later. The use of correct imperatives in early Italian child language is reported in Salustri and Hyams (2003). In child Dutch, by contrast, the use of imperatives as type b predicates, i.e. before D-marking, are rather rare. This can be explained as follows. Next to imperatives of a lexical verb, the Dutch imperative equivalents are either a [modal verb + infinitive], or [ga + infinitive]. Both appear in early child Dutch as Optional Infinitives (see Appendix 3).

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.15 (772-828)

Early operators and late topic-drop/pro-drop 

. Conclusions The reduction of functional categories in child language may fit Jakobson’s conjecture of universal feature-hierarchy for syntax. The Io /Co marking of predicates precedes the Do marking of arguments. Wexler’s Optional Infinitives fit into the order of acquisition steps. They are the counterpart of systematic I-/C-marking, but the relation with full Agreement in Io pursued in Schütze (1997) is highly doubtful. Full agreement is late. It follows Dmarking and D-marking follows I-marking (finiteness) (van Kampen 2004b). By consequence, Schütze and Wexler’s conjecture about double checking does not fit the major hierarchy of feature acquisition. The Optional Infinitives disappear well before the Agr-checking is possible at all. Rizzi’s Truncation is a different story. It has no uniform effect, but it does not need to have one. It is an abstract principle and may fit the order of acquisition steps, if one makes the distinction between type a, b and c predicates. (29) a.

Early situation-bound predicates appear before I-marking. They belong to early child language. b. Dutch topic-drop (Spec,C) appears systematically after D-marking. It belongs to discourse grammar and depends on D-features for reference tracking in a context. c. The same must hold for Italian pro-drop (Spec,I). It probably appears and rises after D-marking.

Children seem well aware of the fact that some Tense/Agr systems (such as the Dutch one) allow a benign neglect, whereas others (such as the Italian one) do not. My conjecture (van Kampen 1997) is that Optional Infinitives have after all no direct relation with the richness of the verbal paradigm. They are rather an effect of leaving out grammatical auxiliaries/modals. The decisive circumstance must be that more than 60% of the verbal predicates in the Dutch input are constructions with an auxiliary plus a non-finite lexical verb, or a bare non-finite lexical verb (see for details and percentages Evers & van Kampen 2001). The initial intake of the Dutch child will be a set of lexical content predicates that are infinitival. By contrast, less than 20% of the Italian predicates are of that type. See for these percentages, Appendix 1. The initial conjecture of the Italian child can be that lexical content predicates are based on some not yet decodable Agreement form. They simulate the form by means of a default without real Tense or phi-feature oppositions. It is only after the acquisition of phi-features on Dmarking that the real truncation based on zero pronouns can come in. The brute input percentages (> 60% versus < 60%) are less sophisticated than the Unique Checking Constraint, but they must have some relevance.10 The factual order of acquisition steps has to be established for typological variants. It has to be considered whether and how that order fits the present conjecture about the hierarchy of evidence frames. For instance, the present conjecture predicts that for VSO Irish, language acquisition will heavily depend on Optional Gerunds rather than

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.16 (828-891)

 Jacqueline van Kampen

on forms, although Irish offers a pro-drop system for all predicates. I expect this because the VSO order is a root order (subordinates are SVO) like V2nd Dutch. The root I-to-C types Celtic VSO as well as Germanic V-second can be acquired only due to an input that contains > 60% Aux-marked predicates.

Notes * The research for this paper was supported by NWO (grant 355-70-009). . The issue of acquisition order as such was first raised in Brown (1973: 313ff.). . I use the label Io here for the early grammatical marking of on the predicate. On purpose, I do not yet enter the more technical issue how I-features like Tense, Aspect and Agreement are best represented in the structure. . Pro-drop is of course not obligatory. It remains possible to add a lexical, non-pronominal subject. This subject seems to enter the same complex subject licensing that inspired optional infinitives in non-pro-drop languages, but Wexler (1998) suggests, following Barbosa (1995), that lexical subjects in Romance languages are in a higher position than INFL. . The graph for D-marking in Dutch is based on the opposition . See van Kampen (1997, 2001, 2003, 2004a, 2005), van Kampen and Evers (2004) for the idea that speed and order of acquisition are imposed by the quantity and quality of ± feature oppositions. . The simultaneity between Do -marking by articles and by free anaphors holds especially for Dutch (van Kampen 2004a). The situation in French and English is slightly different. The free anaphors in English may also function in direct connection with the speech situation. They appear before the systematic use of articles. The V-second grammar of Dutch applies for such situation highlighted cases a demonstrative die/dat (‘this/that’). It stresses the topic and is used as an A-bar pronoun in Spec,C (van Kampen 1997). The A-bar property is related to the V-second character of Dutch that has been acquired a few weeks earlier. The A-bar pronouns will soon vary with the A-bar question pronouns wie/wat (‘who/what’) (van Kampen 1997). The French free anaphors are clitics. They can only function within discourse. Their identification follows the full acquisition of article. The typological variations will get quantified and it will be argued that they do not weaken the main conclusion that systematic and intensive use of Co /Io -marking and Do -marking within the sentence is maintained to construct the discourse coherence. The latter makes human language situation-free. . The parallel acquisition of determiners and 3rd person pronouns has also been argued for by Hoekstra and Hyams (1995) and Schaeffer (1997). . See Fleischman (1982) for a tight connection in adult language between modal verbs of obligation like must/devoir/moeten and later aspectual expressions for the future go/aller/gaan. . See van Kampen (2001, 2005) and Evers and van Kampen (2001) for the learning steps that are needed to acquire Io /Co . This point is irrelevant for the present argumentation. . Of course, unlike pro-drop, topic-drop is not restricted to empty subject pronouns, but includes object-drop and adverb-drop. Moreover wh-elements may be dropped in Spec,C. See for the characteristics of topic-drop in Dutch van Kampen (1992, 1997: Chap. 4), and in Swedish Mörnsjö (2002). Van Kampen’s (1989, 1997) analysis generalizes over topic-drop and wh-drop in V-second languages.

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.17 (891-995)

Early operators and late topic-drop/pro-drop  . This does not mean that the acquisition procedure makes decisions on frequency alone. See for the idea that local evidence frames outweigh mere input frequency Evers and van Kampen (2001) and van Kampen and Evers (2004).

References Avram, L. (2003). “From feature selection to feature valuation: The view from early clitics and determiners”. Paper presented at the UiL OTS, Utrecht University. Avram, L. & Coene, M. (2004). “What early clitics can tell us about early subjects”. In J. van Kampen & S. Baauw (Eds.), Proceedings of GALA 2003 (pp. 93–102). Utrecht: LOT Occasional Series. Baker, M. (2005). Lexical Categories: Verbs, Nouns and Adjectives. Cambridge: Cambridge University Press. Barbosa, P. (1995). Null Subjects. Ph.D. Dissertation. Cambridge, MA: MIT. Borer, H. (2005). Structering Sense. Oxford: Oxford University Press. Brown, R. (1973). A First Language: The Early Stages. Cambridge, MA: Harvard University Press. Buszkowski, W. (1987). “Discovery procedures for categorial grammars”. In E. Klein & J. van Benthem (Eds.), Categories, Polymorphism and Unification. Centre for Cognitive Science, University of Edinburgh & Institute for Language, Logic and Information, University of Amsterdam. Chomsky, N. (1968). Language and Mind. New York: Harcourt, Brace & World. Clark, E. (1993). The Lexicon in Acquisition. Cambridge: Cambridge University Press. Evers, A. & Kampen, J. van (2001). “E-language, I-language and the order of parameter setting”. UiL OTS W.P. 00105-S-S, Utrecht. Fleischman, S. (1982). The Future in Thought and Language: Diachronic Evidence from Romance. Cambridge: Cambridge University Press. Grimshaw, J. & Samek-Lodovici, V. (1998). “Optimal subjects and subject universals”. In P. Barbosa et al. (Eds.), Is the Best Good Enough? (pp. 193–219). Cambridge, MA: MIT Press. Grinstead, J. (1998). Subjects, Sentential Negation and Imperatives in Child Spanish and Catalan. Ph.D. dissertation UCLA, Los Angeles. Guasti, M. T. (1994). “Verb syntax in Italian child grammar: Finite and nonfinite verbs”. Language Acquisition, 3, 1–40. Halle, M. & Marantz, A. (1993). “Distributed morphology and the pieces of inflection.” In K. Hale & S. J. Keyser (Eds.), The View from Building 20 (pp. 111–176). Cambridge, MA: MIT Press. Hoekstra, T. & Hyams, N. (1995). “The syntax and pragmatics of early clause structure”. Manuscript. Leiden University and UCLA. Jakobson, R. (1942). Kindersprache, Aphasie und Allgemeine Lautgesätze. Uppsala: Uppsala Universitets årsskrift 9. Jordens, P. (2002). “Finiteness in early child Dutch”. Linguistics, 40(4), 687–765. Kampen, J. van (1989). De verwerving van wh-vraagzinnen (The acquisiton of wh-questions in Dutch). Master thesis Utrecht University. Kampen, J. van (1992). “D-pronouns and topic-drop”. Manuscript. OTS, Utrecht University. Kampen, J. van (1997). First Steps in Wh-movement. Delft: Eburon.

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.18 (995-1083)

 Jacqueline van Kampen

Kampen, J. van (2001). “Pragmatic bootstraps for category assignment: from non-discourse towards discourse oriented language”. In S. Kern (Ed.), Proceedings of ELA (Early Lexicon Acquisition). Lyon. Kampen, J. van (2003). “Rien que des oppositions”. Poster presented at GALA (Generative Approaches to Language Acquisition), September 4–6. Utrecht. Kampen, J. van (2004a). “Learnability order in the French pronominal system”. In R. BokBennema, B. Hollebrandse, B. Kampers-Manhe, & P. Sleeman (Eds.), Romance Languages and Linguistic Theory. Selected Papers from Going Romance 2002 (pp. 163–183). Amsterdam: John Benjamins. Kampen, J. van (2004b). “The rise of predication by EPP”. In A. Brugos, L. Micciulla, & C. E. Smith (Eds.), Proceedings of the BU Conference on Language Development 28 (pp. 267–281). Somerville, MA: Cascadilla Press. Kampen, J. van (2005). “Language specific bootstraps for UG categories”. International Journal of Bilingualism, 253–275. Kampen, J. & Evers, A. (2004). “Locality constraints imposed by acquisition frames”. In O. Crouzet, H. Damirdache, & S. Wauquier-Gravelines (Eds.), Proceedings of JEL (Journées d’Etudes Linguistiques) ‘Domaines’ 2004 (pp. 35–41). Nantes: UFR Lettres et Langues. Mörnsjö, M. (2002). V1 Declaratives in Spoken Swedish. Ph.D. Dissertation. Lund University. Müller, N. & Kupisch, T. (2004). “Wait and see: Determiner omissions and object clitics in the acquisition of Italian by bilingual children”. Paper presented at the Second Lisbon Meeting. Muñoz, L. (2004). “Spanish acquisition of determiners and clitics”. Master thesis. Utrecht University. Pinto, M. (2004). “Cross-linguistic influence in Italian-Dutch bilingual subjects”. Paper presented at The Romance Turn, Madrid 16–18 September. Postma, G. (2004). “On the distinction between AGREE and SHAGR: Forum readings in Dutch imperatives”. Paper presented at the TIN-dag, Utrecht, February 7. Rizzi, L. (1994). “Early null subjects and root null subjects”. In T. Hoekstra & B. D. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 151–176). Amsterdam: John Benjamins. Rizzi, L. (2001). “Relativized minimality effects”. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell. Salustri, M. & Hyams, N. (2003). “Is there an analogue to the RI stage in the null subject languages?”. In B. Beachley, A. Brown, & F. Conlin (Eds.), Proceedings of the BU Conference on Language Development 27 (pp. 692–703). Somerville, MA: Cascadilla Press. Schaeffer, J. (1997). Direct Object Scrambling in Dutch and Italian Child Language. Ph.D. Dissertation UCLA. Los Angeles, CA. Schütze, C. (1997). INFL in Child and Adult Language: Agreement, Case and Licensing, Ph.D. dissertation. Cambridge, MA: MIT. Soares, C. (2002). “The acquisition of D, T and C: Additional evidence”. In I. Lasser (Ed.), Proceedings of GALA 1999 (pp. 135–163). Frankfurt: Peter Lang. Wexler, K. (1998). “Very early parameter setting and the unique checking constraint: A new explanation of the optional infinitive stage”. Lingua, 106, 23–79.

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.19 (1083-1171)

Early operators and late topic-drop/pro-drop 

Appendix 1: Input for types a-b-c (Dutch/Italian) Italian: Diana’s mother (Calambrone corpus) (with thanks to Manuela Pinto) Dutch: Laura’s mother (Kampen corpus)

Table (i). Verbs by types in Italian and Dutch (mother) (100% = all verb types) Type

Italian

Dutch

Vlex Aux/modal + Vlex

1.5% 13.5%

10% 34%

Vlex (imperative) Modal verb

25 % –

4% 10%

Copula (non-formulaic)

20 %

18 %

Vlex

40 %

24%

The potential evidence frames for the verbal configurations in Table (i) are given in Table (ii).

Table (ii). Evidence frames (intake child) for type a, b and c (100% = all verb types) input mother

Italian

Dutch

intake=output child

Vlex

1.5%

10%

Aux/modal + Vlex

13.5%

34%

type a bare V (Infinitive) type a bare V (Infinitive)

Vlex imperative

25%

4%

Modal verb

–

10%

Vlex

40%

24%

type b (?) mode-implied type c Vlex

Copula + N/A/P

20%

18%

type a bare N/A/P type b mode-implied

type b mode-implied type b mode-implied

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.20 (1171-1274)

 Jacqueline van Kampen

Appendix 2: Input for types a–b (Dutch/Italian) The Dutch input presents lexical verbs as infinitives in 62% of the cases, the Italian input in 20% of the cases only.

Table (iii). Lexical verbs and in Italian and Dutch input (mother) (100% = all lexical verbs) Italian (Diana’s mother) Dutch (Laura’s mother) V 20% V 80%

V 62% V 38%

A certain percentage of the finite lexical verbs above constitute imperatives. The Italian input presents some 30% of all finite lexical verbs as imperatives, whereas the Dutch input does so in 5% of the cases only.

Table (iv). Imperatives in Italian and Dutch input (mother) (100% = all lexical finite verbs) Italian (Diana’s mother) Dutch (Laura’s mother) Vimp 31%

Vimp 5%

Appendix 3: Percentages for imperative type a–b (Dutch/Italian) The low 5% finite lexical verbs to express imperative mood in Dutch is due to the fact that Dutch rather uses gaan/modal + V. The use of a Vlex as an imperative is in general considered to be quite impolite. Italian, by contrast, does use finite lexical verbs, see (1). (1) Expressions of imperative mood a. Italian adult: Vlex → type b in Italian child language b. Dutch adult: gaan/modal + Vlex → type a in Dutch child language The figures for the percentages in Tables (iii) and (iv) are given in (2) and (3). (2) Italian (Diana’s mother, Calambrone corpus) Lexical verbs: 590 of which a. Lexical V: 117 (20%) Imperative V: 10 (2%) b. Lexical V: 473 (80%) Imperative V: 182 (31%) (3) Dutch (Laura’s mother, Kampen corpus) Lexical verbs: 657 of which a. Lexical V: 407 (62%) Imperative V: 54+? (?%) b. Lexical V: 250 (38%) Imperative V: 35 (5%) The exact input percentage for the imperative V cannot be calculated because Dutch uses a great number of non-imperative moeten/gaan + infinitive with imperative intention, as in (4c, d).

JB[v.20020404] Prn:5/06/2006; 10:03

F: LALD4110.tex / p.21 (1274-1290)

Early operators and late topic-drop/pro-drop 

(4) Dutch input for imperatives a. [Vlex imp.mode + modal particle] b. [ga (go, mode-implied) + modal particle + Vlex] c. [2nd pers. + gaat (go) + modal particle + Vlex] d. [2nd pers. + moet (must) + modal particle + Vlex] Examples of the three input types are given in (5): (5) Examples of (4) a. was even je handen wash just your hands handen wassen b. ga even je go just your hands wash handen wassen c. je gaat nu even je you go now just your hands wash d. je moet nu even je handen wassen you must now just your hands wash

JB[v.20020404] Prn:31/03/2006; 15:09

F: LALD41P3.tex / p.1 (47-74)

 

Movement and resumptive pronouns

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.1 (46-127)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese* Elaine Grolla University of Connecticut

Previous studies on the acquisition of pronouns have concentrated either on A-bound or on A’-bound (resumptive) pronouns. In both cases, children performed at chance-level. In this work, I argue that the poor performance is due to a single underlying cause. This conclusion receives support from results of tests I conducted with Brazilian Portuguese-speaking children, in which the same children performed poorly with both A-bound and resumptive pronouns. I follow Grodzinsky and Reinhart’s (1993) claim that children’s behavior is due to their limited working memory, but diverge from them in arguing that the problem resides in comparing syntactic computations, not semantic interpretations. The two analyses make different predictions for resumptive pronouns, where the interpretations for a derivation with a gap and a pronoun are the same and children still have problems.

.

Introduction

This study is concerned with the acquisition of pronominal elements in Brazilian Portuguese (henceforth, BP). I investigate the acquisition of pronouns appearing in two distinct environments, namely, pronouns locally A-bound, as shown in (1), and pronouns A’-bound appearing inside relative clauses, as in (2): (1) *O elefantei está lavando elei . The elephant is washing him ‘*The elephant is washing him.’ (2) O sapo que (*ele) está esquiando está contente. The frog that he is skating is happy ‘The frog that (*he) is skating is happy.’ Sentence (1), with the pronoun locally A-bound by the antecedent ‘the elephant,’ is unacceptable. In (2), we have a subject relative clause with a resumptive pronoun (RP) sitting in the relativized position. The sentence with the RP is unacceptable, but its gap counterpart is ok.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.2 (127-188)

 Elaine Grolla

Studies on the acquisition of locally A-bound pronouns in various languages indicate that children behave differently from adults when sentences like (1) are tested. In Chien and Wexler’s (1990) study in English, for example, children allowed the pronoun to corefer with a local DP antecedent around 50% of the time. Adults did not allow such coreference. In the case of RPs, McKee and McDaniel (2001) found out that children acquiring English judged the English counterpart of sentence (2) grammatical around 50% of the time, contrary to adults, who almost never accepted it. The constructions in (1) and (2) have always been treated separately in acquisition studies. This is so because they involve pronominal elements appearing in different environments. In (1), the pronoun is bound by an antecedent sitting in A-position, while in (2) the pronoun is bound by the relative operator, which is sitting in an A’position. Overall, studies dealing with the acquisition of RPs and local coreference have come up with interesting ways to account for children’s problems. However, the crucial observation to be made here is that none of these studies have explored the possibility that children’s behavior in both domains may be correlated. That is, none of these studies have considered the hypothesis that children perform poorly in contexts like (1) and (2) because the source of the problem is the same. This is the hypothesis under investigation in this paper. The language under study is Brazilian Portuguese, but the analysis can be carried over to English as well, as the facts are similar in both adult languages and children’s behavior is also the same in both languages. There are several reasons to pursue such a unifying approach for children’s behavior in tests on A- and A’-bound pronouns. First, both constructions involve pronominal elements. Second, results of studies in various languages indicate that children perform similarly on both tests, that is, they incorrectly accept sentences like the ones in (1) and (2) around 50% of the time, behaving at chance. And third, the age-range when this chance performance is detected is the same in both cases, that is, around 4 and 5 years of age. In this paper, I discuss how such connection is plausible from the point of view of syntactic theory, and how this unification is worth pursuing from the point of view of language acquisition research. I provide acquisition data that corroborate this claim. Using the grammaticality judgment task, I interviewed the same Brazilian Portuguese-speaking children on two experiments, one involving A-bound pronouns and another involving A’-bound pronouns. The data revealed that the majority of the children displayed chance level performance on the tests with A- and A’-bound pronouns. This result constitutes strong evidence in support of the hypothesis that children’s chance performance on both domains has the same cause. The paper is organized as follows. Section 2 is a brief overview of previous studies on the acquisition of pronominal elements. Section 3 outlines an analysis of bound pronouns as elsewhere elements. Section 4 lays out the predictions that such analysis makes for language acquisition. Section 5 discusses the experimental results of a study carried out with Brazilian Portuguese-speaking children. Section 6 is the conclusion.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.3 (188-254)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

. Previous studies on the acquisition of pronouns The vast literature on the acquisition of pronouns has essentially reported two developmental problems in this area, as mentioned above. The first one is that children acquiring languages such as English, Russian, Icelandic and Dutch sometimes accept sentences in which a pronoun is coreferent with a local DP antecedent. The other problem relates to the placement of RPs inside relative clauses. Young children allow RPs to appear in syntactic positions where adult speakers do not allow them. Commenting first on the acquisition of locally A-bound pronouns, Chien and Wexler (1990) is one of the most influential studies in this area. These authors interviewed 177 children on the age range of 2;6 to 7;0 years. They showed children pictures of cartoon characters, such as Mama Bear and Goldilocks, with one character performing a reflexive action and the other watching the scene. In one of the trials, the picture showed Mama Bear touching herself, and Goldilocks next to her. The experimenter then said to the child: (3) This is Mama Bear, this is Goldilocks. Is Mama Bear touching her? Children responded ‘yes’ around 50% of the time, in contrast to adults, who answered ‘no’ close to 100% of the time. When children answered ‘yes,’ they were presumably taking ‘Mama Bear’ as the antecedent for the pronoun. This type of response indicates that sometimes children allow the pronoun to corefer with a local antecedent. Interestingly, a different result emerged when Chien and Wexler tested sentences involving quantified antecedents. Test sentences were like the following: (4) These are the bears, this is Goldilocks. Is every bear touching her? The picture accompanying this question depicted three female bears touching themselves and Goldilocks next to them, watching. Children could pick either the QP ‘every bear’ or the DP ‘Goldilocks’ as the antecedent for the pronoun. If children took the QP ‘every bear’ as the pronoun antecedent, then they should have answered the question affirmatively, as the picture indeed displayed the bears touching themselves. However, if they took ‘Goldilocks’ as the antecedent for the pronoun, they should have answered the question negatively, as the bears were not touching Goldilocks in the picture. Contrary to what happened with sentences like (3), children did not allow the pronoun to be locally A-bound in the case of (4), behaving like adults. That is, they correctly answered question (4) negatively 84% of the time (for the 5 year-olds). Since Chien and Wexler’s experiment, other researchers have replicated these results not only for English, but for other languages as well, generally making use of a similar methodology (see Avrutin 1999; Avrutin & Thornton 1994; Avrutin & Wexler 1992; Boster 1994; Cairns, McDaniel, Hsu, & Konstantyn 1995; Deutsch, Koster, & Koster 1986; Grimshaw & Rosen 1990; McDaniel, Cairns, & Hsu 1990; McDaniel & Maxfield 1992; Philip & Coopmans 1996; Sigurjónsdóttir & Hyams 1992; Thornton & Wexler 1999; Wexler & Chien 1985; among others).

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.4 (254-302)

 Elaine Grolla

Interestingly enough, children acquiring languages such as Italian, French, Catalan and Spanish are adult-like with respect to Principle B (see, among others, Baauw, Escobar, & Philip 1997; Cardinaletti & Starke 1995; Escobar & Gavarró 1999; Hamann 2002; Hamann, Kowalsky, & Philip 1997; Jakubowicz 1984; Lust, Loveland, & Kornet 1980; McKee 1992; McKee, Nicol, & McDaniel 1993; Silva 1989; Solan 1983; and Varela 1988). The main difference between the group of languages where children are adultlike with respect to Principle B and where they are not is the presence or absence of clitics. In languages such as Italian and Spanish, the sentences used in experiments on Principle B contained a clitic instead of a full pronoun and children correctly rejected sentences involving a locally A-bound clitic. The morphological differences between full pronouns and clitics are usually taken as the main factor in explaining children’s different behavior in these two groups of languages. Turning now to the acquisition of RPs, Labelle (1988) discovered that children acquiring French produced RPs in relative clauses where adult speakers did not allow these elements. Since this work, various other languages have been studied and similar results have been found (see Bar-Shalom & Vinnitskaya 2001; Goodluck & Stojanovic 1996; Grolla 2004; Kang 2003; McKee & McDaniel 2001; Pérez-Leroux 1995; and Varlokosta & Armon-Lotem 1998, among others). Mainly, children produce (or judge grammatical) relative clauses with RPs in a manner many times inconsistent with the languages being acquired. Consider the relative clauses below used in McKee and McDaniel’s (2001) study. In (5a), the RP is in the highest subject position and in (5b) it is placed inside an island: (5) a. *This is the man that he’s swimming. b. This is the troll that Ariel doesn’t know what he’s eating. Using a grammaticality judgment task, McKee and McDaniel (2001) interviewed 38 English-speaking children between the ages of 3;5 to 5;11 and 34 adults.1 Children judged sentence (5a) acceptable 47% of the time. English adult speakers judged this sentence acceptable 2% of the time. Sentence (5b) was accepted 80% of the time by adults and 78% of the time by children. McKee and McDaniel’s results demonstrate that children behave like adults when RPs are placed in unextractable positions, but perform poorly when RPs are inserted in extractable positions. These results seem puzzling: if children did not know the rules regulating the use of RPs, they should have performed poorly in both contexts. On the other hand, if they knew these rules, they should have behaved like adults in both cases. As we will see in the next section, the analysis to be proposed here can explain this seemingly puzzling behavior not only for the case of RPs, but for the cases of A-bound pronouns as well.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.5 (302-356)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

. Pronouns as elsewhere elements The proposal to be advocated here assumes Hornstein’s (2001) analysis of bound pronouns. Hornstein proposes that bound pronouns are elsewhere elements that can only be inserted in a derivation if movement is not possible. According to this hypothesis, pronouns are not present in the numeration and their insertion in a derivation is highly costly. The application of movement is considered cheaper than insertion of pronouns. In order to appreciate his proposal in deeper detail, let us consider the relative clauses below involving A’-bound pronouns: (6) a. The man that __ is swimming b. *The man that he is swimming (7) a. *The pirate that Minnie Mouse laughed when __ arrived b. The pirate that Minnie Mouse laughed when he arrived The contrast in (6) shows that, when a gap is possible, the presence of a pronoun is banned. In (7), we see that when the gap is impossible, the pronoun is allowed. Assuming that the constructions displaying gaps involve movement of the relative operator from the relativized position to spec,CP, and that the gap corresponds to a trace left by movement, the contrasts above can be restated as follows. When movement is possible, insertion of the pronoun is prohibited; when movement is impossible, insertion of the pronoun is obligatory. In other words, Hornstein claims that pronouns are parasitic on the impossibility of movement. So, the contrasts above constitute evidence for Hornstein’s proposal. Turning now to A-bound pronouns, anaphors and pronouns are conventionally analyzed as being regulated by Principles A and B of Binding Theory respectively. These Principles are stated below (Chomsky 1981: 188): (8) Principle A: An anaphor is bound in its governing category Principle B: A pronominal is free in its governing category2 That is, Principles A and B state opposite requirements: anaphors must be close to their antecedents, while pronouns must be far from their antecedents. The existence of such principles is challenged in Hornstein (2001), who claims that it is possible to eliminate both Principles A and B from the theory of grammar if we allow movement to occur more generally. In particular, in his system movement into theta-positions is permitted. In the case of anaphors, Hornstein suggests that structures containing these elements involve movement. Several studies have made this claim. Chomsky (1986), for example, proposed that anaphors move to a (non-theta) position close to their antecedents at LF, in a type of LF cliticization. This movement can account for the locality condition of anaphors, which need to be close to their antecedents. Other researchers have also made the claim that anaphor constructions involve movement, although the implementations of this idea vary. Instead of proposing that anaphors involve LF cliti-

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.6 (356-424)

 Elaine Grolla

cization, Hornstein (2001), Lidz and Idsardi (1997) and Zwart (2002) have proposed that this movement takes place in overt syntax and that the anaphor is the residue of movement. This movement generates a structure like the following: (9) Bert1 admires [t1 himself]. Leaving aside the technical implementations of this idea, let us concentrate on the intuition that anaphors involve movement.3 If so, consider the contrast below, intended to have the pronoun and the anaphor locally bound by ‘Bert:’ (10) a. Bert admires himself. b. *Bert admires him. This contrast illustrates the complementarity between anaphors and pronouns. (10a) is a structure involving movement, while (10b) does not involve movement and contains a pronoun in the place of the anaphor. The structure with the pronoun is not acceptable, while the structure involving movement is. If we analyze pronouns as elsewhere elements that can only be inserted when movement has failed to apply, we can rule out cases like (10b) without the need of a stipulation like Principle B. Given that a derivation involving movement is available, the insertion of the pronoun is blocked. So, (10b) is excluded because it violates economy conditions.4 In addition, we can explain the locality requirement on anaphors, without the need of Principle A. Anaphors need to be close to their antecedents due to movement. Thus, this contrast can be taken as corroborating evidence for Hornstein’s proposal. In this system, when movement is impossible, pronouns can be inserted. This is what happens in the examples below (the items in italics are intended as coreferential): (11) a. John likes his mother. b. Peter thinks that he is a genius. In (11a), movement of ‘John’ from inside the DP ‘his mother’ to spec,IP violates the Left Branch Condition. Therefore, the insertion of the pronoun inside the DP is licit. The derivation of this sentence will involve the insertion of the pronoun inside the DP and the merge of ‘John’ in the subject position of ‘likes’. In (11b), movement of ‘Peter’ from the embedded spec,IP to the matrix clause is not possible for Case reasons. In Hornstein’s system, DPs can check Case only once, and after a DP checks Case, it is frozen in place and cannot move further. So, if ‘Peter’ checks nominative Case in the embedded spec,IP, it cannot move up to the matrix clause. Therefore, the alternative is to insert the pronoun in the embedded subject position and merge ‘Peter’ in the matrix sentence, as in (11b).

. Brazilian Portuguese Hornstein’s analysis works in a similar way in Brazilian Portuguese. Starting with RPs, in the examples below it is shown these elements are not possible in local subject position, but are required inside islands:5

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.7 (424-481)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

(12) a.

O menino que (*ele) chegou The boy that he arrived ‘The boy that arrived’ b. O menino que a Maria saiu quando *(ele) chegou The boy that the Maria left when he arrived ‘The boy that Mary left when he arrived’

In (12a), the RP is not possible in subject position because movement of the relative operator from spec,IP to spec,CP is possible. Movement being possible, the pronoun is banned. In (12b), movement out of islands is not licit. Thus, with movement being impossible, the insertion of RP is obligatory. Consider next structures with the oblique position relativized: O menino que a Maria conversou com *(ele) The boy that the Maria talked with him ‘The boy that Mary talked with’ b. O menino que a Maria conversou __ The boy that the Maria talked ‘The boy that Mary talked with’

(13) a.

There are two possibilities for oblique relatives: either the PP is present with a RP as the complement of the preposition, as in (13a), or the whole PP is absent, as in (13b). BP does not exhibit preposition stranding, which means that movement out of PPs is impossible. So, if the preposition is present, there must be a RP as its complement.6 The case in (13b) can be analyzed in a number of ways. The gap in the complement position of ‘talk’ could be analyzed as a null object that, instead of being a DP, is a PP. This would entail that BP has null PPs. Another possibility is that the relative operator moved from oblique position to spec, CP and then deletion of P occurred. As this issue is not relevant to the acquisition study to be discussed below, I will not pursue it here. For a discussion, see Kato (1993) and Tarallo (1983). Turning now to the case of locally A-bound pronouns, observe in (14a) that the pronoun ele ‘him’ cannot be locally A-bound.7 In (14b), we see that the clitic anaphor se is the appropriate form: (14) a. *O Pedroi admira elei . The Pedro admires him b. O Pedroi sei admira. The Pedro himself admires ‘Peter admires himself.’ If we assume that the derivation with the anaphor se involves movement and that the pronoun ele ‘him’ is an elsewhere element only used in case movement cannot be applied, then the facts in (14) are comparable to what happens in English. Note that BP displays a mixed pronominal system. While the anaphor is a clitic pronoun, the object pronoun ‘ele’ is a full pronoun, not a clitic. BP displayed object

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.8 (481-547)

 Elaine Grolla

clitics in the past, but now these forms are only used in formal written registers. The only object pronouns now available for singular third person are the non-clitic forms ele (him) / ela (her). As the presentation above shows, Hornstein’s analysis has the advantage of explaining the complementarity between (local) anaphors and A-bound pronouns: every time the anaphor is possible, the pronoun isn’t. It is also valuable as it is an attempt to eliminate stipulations like the Binding Principles from the theory of grammar. For the acquisition issues under investigation here, this system is relevant as it analyzes both A- and A’-bound pronouns as elsewhere elements.

. Elsewhere elements and reference-set computation As we saw above, it is possible to analyze both A- and A’-bound pronouns in English and BP as elsewhere elements. In both cases, these elements can be inserted in a derivation only if movement cannot occur. This analysis then requires a comparison of derivations in order to decide if pronouns are licit or not. The comparison that takes place in these cases is called ‘reference-set computation.’ The reference-set is comprised of the convergent derivations being compared. Thus, in order to decide if the pronoun is possible in the sentences below, they need to be compared to their movement counterparts. In both cases, given that the movement counterparts (shown in the b sentences) are possible, the insertion of the pronoun is banned: (15) a. *John admires him. b. John admires himself. (16) a. *This is the boy that he likes ice cream. b. This is the boy that __ likes ice cream. In the minimalist framework, only convergent derivations can be compared. This is the case in the derivations above. In the case of (15), both (15a) and (15b) are convergent. Although the derivation with the pronoun does not win the comparison, it is convergent, as it does not violate any other constraint besides economy, which is precisely the issue being evaluated. The same considerations hold for the pair in (16). Both (16a) and (16b) are convergent and so can be compared. Things are different when pronouns appear inside islands, as the derivation involving movement does not converge: (17) a. This is the boy that Mary left when he arrived. b. *This is the boy that Mary left when __ arrived. Derivation (17b) involves illicit movement out of an island and is not convergent. Therefore, this derivation is not included in the reference-set. Thus, in order to check if the pronoun is licit in derivation (17a), the reference-set contains only one derivation: the structure with the pronoun is derived without reference-set computation.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.9 (547-610)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

Thus, in some environments like the one in (17), reference-set computation is not needed to check if pronouns are licit or not.8 In other cases, as in (15) and (16), reference-set computation is required in order to exclude a derivation with a pronoun.

. The acquisition of pronouns As discussed in Section 2, English-speaking children are not adult-like in cases like (15a) and (16a), where pronouns are inserted in extractable positions. Children are adult-like when the pronoun is inserted in unextractable positions, like in (17a). It is interesting that in the cases where children have problems, they behave similarly. Children accept RPs in extractable positions at roughly the same rate that they accept locally A-bound pronouns, that is, around 50% of the time. Another significant fact is that children of the same age range exhibit chance level performance on these tests, roughly from 4 to 5 years of age. The question to be addressed then is this: is it a coincidence that both constructions with which children have problems involve pronouns in extractable positions and that children at around the same age range behave at chance level in both cases? My claim is that this is not a coincidence; it is precisely the fact that both cases involve pronouns in extractable positions that is critical. It is also relevant that children accept both of these constructions at chance level. For if they did not know the rules regulating when pronouns can be present in such constructions, we would expect them to accept these constructions close to 100% of the time, not 50%. According to what was discussed in the previous section, we could characterize children’s problems in the following way. Children exhibit chance level performance in cases that require reference-set computation. In the cases where reference-set computation is not required, as in island contexts, children behave like adults. In a (1999) paper, Reinhart discussed another case where reference-set computation is required and showed that in that case, children behaved at chance.9 She argued that reference-set computation requires greater load on working memory than local computation and proposed that whenever reference-set computation is involved, there should be some evidence of processing complexity. Her hypothesis is that if referenceset computation exceeds children’s processing ability, one should expect to find a guess pattern in areas where we assume this computation is involved. Reinhart claimed that, although children know what they have to do in these computations, their working memory is not big enough to hold the materials needed to complete the task. The computation required is beyond young children’s abilities, which makes them give up and guess, leading to the 50% pattern of response. In the cases under scrutiny here, the computation required to exclude a derivation with a pronoun involves the following steps. While holding the sentence under processing in memory, the child has to build a set with two possible derivations, one involving movement, the other containing a pronoun. Then the child has to compare these derivations to check which one is more economical. The hypothesis entertained here is that children cannot perform all these steps.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.10 (610-664)

 Elaine Grolla

It is important to note that this analysis does not raise learnability problems. The claim that children’s working memory is more limited than adults’ has been demonstrated in various studies (see, for example, Gathercole & Baddeley 1993). As children grow older, their working memory develops, which explains how children reach the adult state. That is, children will not have processing problems anymore when their working memory is fully developed. This proposal, though clearly inspired by Grodzinsky and Reinhart’s (1993) analysis for children’s problems with coreference, differs from that theory in crucial respects. Grodzinsky and Reinhart (1993) claim that children’s processing problem in sentences involving local coreference arises when they have to compare the bound and coreferential interpretations of a pronoun. My claim is that the problem is in comparing syntactic derivations and deciding which one is more economical. The two analyses make different predictions for RPs, where the interpretations for a derivation with a gap and a pronoun are the same and children still behave at chance. Summarizing, adopting the proposal discussed in Reinhart (1999) regarding young children’s processing limitations, the hypotheses considered here are that (a) constructions involving pronouns in extractable positions require reference-set computation to be excluded and (b) children are not capable of handling this kind of computation. The predictions we make are the following. First, children should perform around chance level (that is, around 50% correct responses) when pronouns are placed in extractable positions. In these cases, both the derivation with a pronoun and the one without it converge, requiring reference-set computation. Second, children should perform like adults (that is, close to 100% correct responses) when pronouns are inserted in non-extractable positions. In these cases, the derivation without the pronoun does not converge. If so, no computation is needed: the derivation with the pronoun is the only one in the reference-set. If no reference-set computation is needed, children should not have processing problems. And third, given that the same kind of computation is required in tests with both A- and A’-bound pronouns, children who have chance performance on the test with locally A-bound pronouns should also have chance performance on the test with RPs in extractable positions.10 Before moving on to the experimental section, I would like to discuss the difference between clitics and full pronouns. Hornstein’s theory of bound pronouns is centered on data from English, which does not display clitics, and in his (2001) book there is not a discussion about these elements. The most straightforward analysis for them would be to assume that they are also elsewhere elements, requiring reference-set computation. However, as mentioned in Section 2 above, children acquiring languages such as French and Spanish do not exhibit chance level performance when locally Abound clitics are tested. This fact from language acquisition requires us to analyze object clitics in a manner distinct from full pronouns in adult languages. Given that children do not over-accept locally A-bound clitics, it must be the case that these elements do not require reference-set computation to be excluded. A detailed analysis of these elements is beyond the scope of this study. What is relevant for our purposes is that in the case of BP, since the sentences presented to children displayed full object

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.11 (664-727)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

pronouns, and not clitics, children should have processing problems in these cases and resort to guessing.11 In order to test the predictions above, I interviewed the same children on tests with A- and A’-bound pronouns. As far as I know, this kind of data has never been collected. In the next section, I turn to these experiments.

. Method Subjects. I tested 40 children acquiring BP as their native language. They ranged in age from 3;4 to 6;6 (mean age = 4;7). The children came from a daycare center in Franca, a town in the state of São Paulo. Procedure. The experiment was a grammaticality judgment task (Hiramatsu & LilloMartin 1998 and McDaniel et al. 1990). Children were introduced to a puppet which came from the moon and spoke moon-talk. The puppet was presented as a creature that was willing to learn BP, but got confused sometimes. The child was then invited to help the puppet to learn BP. In this task, only one experimenter was present. The experimenter showed to the child and the puppet pictures of cartoon characters. After a brief presentation of the character(s) in the picture, the experimenter manipulated the puppet, uttering the target sentence. Children had to say whether the sentence uttered by the puppet was right or wrong.12 The whole study was comprised of 4 sessions. The first session was devoted to teaching children the task and to applying a pretest in order to check if children had learned it. In the training part of this first session, children were given feedback on their answers.13 When children started to give only correct responses, the pretest was applied. Only children who answered 5 out of 6 sentences correctly were included in the study. The subsequent sessions, done at least two days apart from each other, tested the target sentences. In testing sentences with RPs, the pictures depicted two identical animals engaged in different activities, and then a short story was told. For example, in one of the trials, the picture depicted two frogs; one was skating and smiling, and the other had his skating shoes on, but was being carried by a swan. A short story along the following lines was then told to the child: (18) In this picture, we have two frogs. This one is very happy skating! This other one was skating too, but he fell! When the frog fell, the swan laughed, but then he was sorry and went to help the poor frog. After the experimenter told the story, she pointed to the frog being carried by the swan and asked the puppet who was that frog. The target sentence in this case was “this is the frog that the swan laughed when he fell.” When locally A-bound pronouns were tested, there was no need to tell a story. The experimenter simply introduced the character in the picture and the verb that was going to be used in the target sentence. In one of the trials, the picture showed a female

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.12 (727-790)

 Elaine Grolla

spider inside a bathtub soaping up its body. The experimenter presented the picture in the following way: (19) In this picture we have a spider and she is soaping. The target sentence could be one of the following: “the spider is soaping her,” “the spider is soaping herself.” A clarification is in order here regarding the choice of the methodology. Usually, studies investigating Principle B have made use of the truth-value judgment task to elicit children’s responses.14 In studies on RPs, the grammaticality judgment task is more frequently used. In the present study, it was desirable to gather data from the same children using the same type of methodology. This is the reason why a grammaticality judgment task was used in both cases.

Materials. There were six conditions investigated with four items for each type, giving a total of 24 sentences tested. The types of sentences were: (a) simple sentences with DP and QP antecedents locally binding pronouns and anaphors and (b) relative clauses with RPs in the highest subject position and inside islands. Examples of target sentences are presented below: (20) A-bound (pronouns and anaphors): a. DP – him: *A aranha está ensaboando ela. The spider is soaping her b. QP – him: *Todo urso está escovando ele. Every bear is brushing him c. DP – self: O elefante está se ensaboando. The elephant is himself soaping. d. QP – self: Todo bichinho está se vestindo. Every animal is himself dressing. (21) A’-bound (RPs in subject and island positions) a. Subject: *O sapo que ele está esquiando está contente. The frog that he is skating is happy. b. Island: Esse é o sapo que o cisne riu quando ele caiu This is the frog that the swan laughed when he fell. In addition to these target sentences, 8 filler sentences were included. These were simple sentences that could be grammatical or ungrammatical, depending on the pattern of the child’s responses.

. Results Children had a high rate of acceptance of the grammatical cases, namely, sentences involving anaphors (20c/d) and sentences with RPs in island contexts (21b). The rate of acceptance of these structures, as shown in the table below, is 95%. Children judged

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.13 (790-863)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

Table 1. Frequencies of acceptance of A- and A’-bound pronouns Structure

Percentage of acceptance

DP – him QP – him DP – self QP – self Subject Island

71/160 (44.4%) 79/160 (49.3%) 152/160 (95%) 152/160 (95%) 93/160 (58.1%) 152/160 (95%)

sentences with RPs in the highest subject position (21a) acceptable 58% of the time. Sentences with pronouns locally A-bound by DP antecedents, such as (20a), were judged acceptable 44% of the time. 49% of the pronouns locally A-bound by QP antecedents (20b) were judged acceptable. No child was adult-like (i.e., rejecting all of the ungrammatical cases) in all tests and no child accepted all of the sentences. That is, chance performance is observed for children individually, in both domains. Thus, in analyzing individual children, the majority of them had chance performance on both A- and A’-bound pronouns in extractable positions.15 The results of these experiments show that the majority of children displayed chance level performance with A- and A’-bound pronouns in extractable positions. Also relevant is that these children did not exhibit problems with A’-bound pronouns in unextractable positions. This last fact is critical, as it shows that it is not the case that children have problems with RPs generally. It is only when they are inserted in extractable positions that their performance gets poor. The fact that the majority of children had problems in both domains (that is, A- and A’-bound pronouns placed in extractable positions) can be either taken as a coincidence or analyzed as one sole problem. In this paper, I claim that the latter holds. That is, the fact that the same children had problems in both cases is taken as evidence that the problems they face have the same source. As mentioned earlier, I propose that children’s problems reside in performing the reference-set computations that are needed to exclude pronouns placed in extractable positions, whether it is in A or A’ environments. A final issue to be discussed regards pronouns A-bound by Quantified antecedents. This is done in the next section.

. Discussion As the results in Table 1 show, children acquiring BP accepted pronouns locally bound by QP antecedents at a considerable rate. This contrasts with the results in Chien and Wexler (1990) and Thornton and Wexler (1999), which showed a clear difference in the acceptance rate of pronouns locally bound by DP and QP antecedents. In Chien

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.14 (863-923)

 Elaine Grolla

and Wexler’s study, children accepted pronouns A-bound by DPs around 50% of the time and pronouns locally bound by QPs around 16% of the time. In Thornton and Wexler’s study, the difference is even more significant: pronouns bound by DPs were accepted 58% of the time and pronouns bound by QPs were accepted 8% of the time, for the simple sentences.16 This pattern of behavior is not the one we expect in the framework adopted here. In order to see this, consider the sentences below: (22) a. *Every elephant is washing him. b. Every elephant is washing himself. Given that the derivation with the anaphor is possible, the insertion of the pronoun is banned. This shows that sentences like (22a) require reference-set computation to be excluded. If so, we expect children to behave at chance in these cases. The results of my experiment, unlike those of Chien and Wexler (1990) and Thornton and Wexler (1999), conform to this prediction. I believe that the difference between my results and the results obtained in Chien and Wexler’s and Thornton and Wexler’s studies for QP antecedents has a principled explanation, which is related to the methodology employed in the studies. I will start commenting first on Chien and Wexler’s experiments. In these cases, children saw a picture with three identical characters performing a reflexive action and a fourth distinct character, which was only watching the scene. Children then were asked the question: “these are the bears, this is Goldilocks. Is every bear touching her?” As discussed in example (4) above, 5-year-olds answered this question negatively 84% of the time. In this case, they probably were picking Goldilocks as the antecedent for the pronoun. Chien and Wexler claimed that because Principle B blocks an interpretation where ‘her’ has ‘every bear’ as antecedent, children picked Goldilocks as the antecedent for the pronoun. The observation I want to make here is that Chien and Wexler did not consider another possibility in interpreting children’s answers. Children could have taken Goldilocks as the antecedent for the pronoun not because of Principle B, but because Goldilocks was highly salient in the context. That is, if children’s attention was drawn to Goldilocks in the picture because of her saliency, then their answers do not bear on their knowledge of Principle B. I believe this possibility is highly likely, as the following factors indicate. First, in the picture shown in Chien and Wexler’s paper, Goldilocks was much bigger than each of the three bears, which obviously made her stand out. Second, all three bears were identical and Goldilocks was physically different from them. We know from studies on the acquisition of universal quantifiers like ‘every’ that children tend to concentrate their attention in the different character present in the pictures in those experiments (see Crain et al. 1996; Drozd & van Loosbroek 1998; Philip 1995; Sugisaki & Isobe 2001; among others). For example, in Philip’s (1995) study, children were shown a picture with four elephants – three of them were being ridden by a boy and the fourth one was not. When asked the question Is every boy riding an elephant?, 97 out of

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.15 (923-994)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

216 preschoolers responded ‘no,’ pointing to the elephant that was not being ridden. The same phenomenon could have happened in Chien and Wexler’s experiment, as their pictures were similar to the ones in Philip’s experiment, in that they also had three identical characters and a fourth individual which was somehow different and more salient. Besides these problems, Boster (1994) discusses the possibility that children in Chien and Wexler’s study might have had difficulty in recognizing the bears as female, as the experimenter did not identify them as such or name them. The picture of the bears did not make their gender clear also; the only hint about it was the bows in the bears’ heads. So, although the bears might or might not be female, Goldilocks was clearly a girl and so doubtlessly an appropriate antecedent for the pronoun. Boster (1994), taking this gender issue under consideration, tested 24 children acquiring English (ages 3;3 to 6;2). She used a yes/no question task, modeled after Chien and Wexler’s experiment. Boster’s trials were of three types. The first two types were similar to Chien and Wexler’s, where sentences with DP and QP antecedents were tested. In the case of DPs, children accepted local coreference around 37% of the time. In the case of QPs, children accepted local coreference around 34% of the time. One difference between Boster’s and Chien and Wexler’s experiment in these cases is that Boster made the gender of the animals clear to children. Also, the fourth character in the QP case was roughly of the same size as the other animals. The third type of pictures depicted three identical animals performing a reflexive action and two other animals of a different type watching the scene. For example, one of the trials had three monkeys patting themselves and two zebras watching them. All animals were of the same size. Because the pictures had 3 animals of one type and 2 of another type, there was no character more prominent than the others. The sentences in these cases were like the following: (23) Is every monkey patting them? Children accepted local coreference in this case around 42% of the time. Boster reports that there was no real difference between children’s rates of acceptance of Principle B violations in sentences such as Every bear is washing her and in sentences such as Mama Bear is washing her. This contrasts with Chien and Wexler’s results where children accepted the former 16% of the time and the later 50% of the time. The increase in acceptance rate observed in the case of (23) has different possible explanations. One possibility, entertained by Boster, is that children analyze QPs of the type ‘every NP’ as binding plural pronouns. If so, children might reject sentences where the QP binds a singular pronoun and accept it more often when the QP binds a plural one. Another possibility, mentioned above, is that the saliency of the character outside the sentence in Chien and Wexler’s study drew children’s attention, which made them pick that character as the antecedent for the pronoun more often. Given that in the case of (23) there was not a more salient character to be considered, children picked the sentence internal QP as its antecedent more often.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.16 (994-1045)

 Elaine Grolla

This discussion entertains the possibility that children chose Goldilocks as the antecedent for the pronoun in sentences involving QPs not because of Principle B, but because the experiment pictures were flawed, as they had the extra sentential character stand out. If this was the case, then Chien and Wexler’s experiments had a confounding factor, and the authors’ conclusion about children’s knowledge of Principle B does not necessarily go through. Note that these problems do not arise in the pictures showing only Mama Bear and Goldilocks, for sentences of the type “Is Mama Bear touching her?”. In these cases, the pictures displayed two equal sized characters, which were clearly identified as female. The context did not make one character more salient than the other. Therefore, there is no clear salient antecedent for the pronoun in this case. Thus, in sentences involving a possible DP antecedent for the pronoun, the confounding factors mentioned above did not arise and children’s answers could not have been guided by the saliency of one of the characters. In the case of Thornton and Wexler’s study, similar problems can be detected. Their study was a truth-value judgment task and stories were acted out in front of children using toys and props. Let us take a look at the general guidelines of a story leading up to a sentence with a potential QP antecedent for the pronoun (story taken from Thornton and Wexler’s book, page 142):17 (24) “Bert and three reindeer friends have a snowball fight, and they all get covered in snow. When they go inside, Bert is shivering, so he asks the reindeer to brush the snow off him. Two of the reindeer (separately) refuse, saying they have too much snow to deal with, and they brush themselves. The third reindeer helps Bert a little bit, but then brushes the snow off himself. Bert thanks the helpful reindeer for starting to brush him. He says he’s sorry he can’t reciprocate by helping brush the reindeer; he needs to finish brushing all the snow off himself because he’s still very cold. Puppet: Every reindeer brushed him.” In this story, Bert is more salient than the other characters. This is acknowledged by Thornton and Wexler themselves, when they claim that the background behind this story is that Bert has a problem: he wants the reindeer to brush the snow off him. The way the story is told, we have Bert as the protagonist, and three other animals involved in it. The three animals are not as prominent in the story as Bert is. They do not have names (they are referred to as ‘the reindeer,’ ‘the third reindeer,’ ‘the helpful reindeer’). They do not have the urgency that Bert has, as they do not ask each other to brush the snow off them. Only Bert is depicted as someone having this need. Also, I suspect that the three reindeer are identical looking. Bert, on the other hand, is physically different, which makes him stand out. So, the fact that Bert is more prominent in the story might be the reason why children took him as the antecedent for the pronoun. One piece of evidence corroborating this possibility comes from a study with adult speakers of English. Morrow (1985) investigated the influence of protagonist status on referent assignment. In his experiment, subjects read a story with a protagonist and a non-protagonist. At the

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.17 (1045-1112)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

end of the story, a sentence containing an ambiguous pronoun was presented. Subjects were asked what the ‘he’ referred to in that last sentence. The results are that subjects are more likely to choose the protagonist when the protagonist was thematically prominent or most recently mentioned. Subjects preferred the non-protagonist only when the non-protagonist was thematically prominent and most recently mentioned. Oppy and Long (1996) also found that adults are more likely to pick the protagonist of a story as the referent of an ambiguous pronoun. Therefore, if children are like adults in this matter, they will pick Bert in the story above as the antecedent for the pronoun, given that Bert is more thematically prominent and was most recently mentioned. If this is so, children’s behavior in those tests might be telling us nothing about their knowledge of Principle B.18 This discussion leads me to conclude that these studies investigating children’s knowledge of Principle B ended up having a confounding factor when the possible local antecedent for the pronoun was a QP. In these cases, the character not mentioned in the target sentence (e.g., Goldilocks or Bert) is necessarily made more salient than the characters that make up the QP (e.g., three bears or three reindeer). This might be the reason why children took that DP as the antecedent for the pronoun, and if that is the case, then the results of these experiments do not bear on children’s knowledge of Principle B. In the experiments I conducted with Brazilian Portuguese-speaking children, reported above, the context did not provide other potential antecedents for the pronoun. In this grammaticality judgment task, children were presented with a context (the picture) and were asked to judge if the sentence they heard was acceptable or not for that context. For example, in one of the trials, children saw a picture with three elephants of equal size doing the same action: washing themselves. The puppet then said: every elephant is washing him. Here, ‘every elephant’ was the only possible antecedent for ‘him,’ as there were no other individuals in the picture. When the confounding factor pointed out above does not exist, as in my grammaticality judgment task, children behaved differently from the children in Chien and Wexler’s and Thornton and Wexler’s experiments, accepting pronouns locally bound by QPs and DPs at chance level. So, the fact that the stories or pictures in these other studies had one of the characters more salient than the others cannot be ignored when we analyze children’s answers. If their attention was drawn to that salient character, then their answers cannot be taken as evidence that they were obeying Principle B. In this case, their answers only indicate that they are sensitive to the saliency of the characters in a story. This makes the results of these other experiments unreliable. Therefore, the difference in the results of my experiments on the one hand and in Chien and Wexler’s and Thornton and Wexler’s studies on the other, can be explained in terms of the experimental differences of these studies, indicating that the methodology employed here has advantages over the methods they used for investigating locally A-bound pronouns.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.18 (1112-1171)

 Elaine Grolla

. Conclusion The results of this study with Brazilian Portuguese-speaking children are relevant in two respects. First, they suggest that the search for a unified way to explain children’s problems with A- and A’-bound pronouns in extractable positions is well motivated. This paper showed how such unified explanation can be implemented. Secondly, these results suggest that the theories proposing that bound pronouns are elsewhere elements provide the most straightforward tools to account for children’s behavior in both domains. Therefore, our results can be viewed as bringing evidence for such theories.

Notes * I would like to thank Marcelo Ferreira, Howard Lasnik, Diane Lillo-Martin, Jairo Nunes and William Snyder for discussing the ideas presented here with me. I would also like to thank Diane Lillo-Martin and an anonymous reviewer for written comments and criticisms. All errors remain mine. Thanks also to the children in Brazil and their teachers for their warm welcome. This research was partially funded by Capes. . McKee and McDaniel (2001) also interviewed 44 English-speaking children between the ages of 6;0 and 8;11. I will not discuss the results for this group here because the children’s ages in this group differ from the ages of children in the other studies being reported in the text. . ‘Free’ and ‘bound’ are defined as follows (Chomsky 1981: 184–185): (i) α is bound by β iff α and β are coindexed, and β c-commands α. (ii) α is free iff it is not bound. ‘Governing category’ is defined as follows (Chomsky 1981: 188): (iii) β is a governing category for α if and only if β is the minimal category containing α, a governor of α, and a SUBJECT (accessible to α). ‘SUBJECT’ is defined as follows: (iv) The SUBJECT of a category is its most prominent nominal element (including the agreement features on the verb in finite clauses). . The technical implementations of this idea are discussed in detail in my (2005) UConn Ph.D. dissertation. . Hornstein’s system is an attempt to derive the complementarity between pronouns and anaphors by allowing movement to occur more freely. However, there are cases where pronouns and anaphors are not in complementary distribution. For example, in some environments, both anaphors and pronouns are possible, as shown in (i) below; in others, neither is possible, as shown in (ii): (i) John pulled the blanket over him/himself. (ii) We like (*me)/(*myself).

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.19 (1171-1237)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

Neither Hornstein’s theory nor Principle B account for these cases. I will abstract away from them in the discussion to follow and will concentrate instead on the cases where complementarity between anaphors and pronouns occur both in BP and in English. For a discussion of cases like (i), see Lees and Klima (1963), Lakoff (1968), Chomsky (1981), among others. For a discussion of cases like (ii), see Lasnik (1981). For an overview on these issues, see Reuland and Everaert (2001). . As for the direct object position, since this is an extractable position, RPs should be banned from it. However, there seems to be a dialectal difference in this case, as some BP speakers consistently accept RPs in this position while others don’t. In a grammaticality judgment task conducted by me with adult native speakers of BP, I found that relative clauses with a RP in direct object position as shown below are judged grammatical 20% of the time, with a group of the speakers tested always accepting these constructions and the other, more numerous group, always rejecting them: (i)

Esse é o menino que a Maria viu ele. This is the boy that the Maria saw him ‘This is the boy that Mary saw.’

In order to account for such level of acceptability, we could follow Shlonsky (1992), who reports a similar phenomenon for RPs in object position in Hebrew. Shlonsky’s proposal is the following. Hebrew has two homophonous complementizers, one identifies its Spec as an A-position, and the other identifies its Spec as an A’-position. The choice between these complementizers is free. If the A-complementizer is chosen, movement from object position to Spec,CP is blocked, as it crosses the subject position, also an A-position. This constitutes a Relativized Minimality violation. If the A’-complementizer is chosen, movement from object position to Spec,CP can occur. So, Shlonsky proposes that for the case of direct object position in Hebrew relative clauses, movement will be allowed or not depending on the complementizer chosen, hence the alternation gap/RP. The same analysis could be proposed for BP. The speakers who accept RPs in direct object position choose the A-complementizer, while the speakers who reject RPs in this position choose the A’-complementizer. . In Grolla (2004), I tested children on constructions like (13a), with RPs in oblique position (among other positions). 11 BP-speaking children between 3;0 and 5;5 years of age were interviewed in a grammaticality judgment task. The results are that children accepted RPs in oblique position 90% of the time. Adult speakers also accepted this construction 90% of the time. . An alternative to the derivations shown in (14) would be to have a null element in object position, as these elements are possible in BP, as shown below: (i)

Quando o João comprou o livro, ele não mostrou __ pra ninguém. When the João bought the book, he not showed __ to nobody ‘When John bought the book, he didn’t show it to anyone.’

However, as the example below illustrates, null and overt objects are not in competition: (ii) Quando o João comprou o livro, ele não mostrou ele pra ninguém. When the João bought the book, he not showed it to nobody Although BP displays null objects, it is not the case that these elements are possible everywhere. Example (14) above, for example, cannot have a null object, as shown in (iiia). There seems to be a restriction on the animacy of the object, as also illustrated in (iiib), which must have an overt object pronoun:

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.20 (1237-1298)

 Elaine Grolla

(iii) a. *O Pedro admira __ . The Peter admires __ b. *Quando a Maria encontra o Pedro, ela abraça __ . When the Maria meets the Peter, she hugs __ ‘* When Mary meets Peter, she hugs.’ Given these complications, which are not relevant to the acquisition study reported below, I will not consider these cases. For analyses of these elements in BP, see Cyrino (1997) and Ferreira (2000). . Other examples of structures where reference-set computation is not used are presented in the text, in examples (11), repeated here: (i)

a. b.

John likes his mother. Peter thinks that he is a genius.

As was discussed in the text, the movement counterparts of these sentences do not converge. So, no reference-set computation is involved in order to check if pronouns are licit in these cases. . Reinhart’s (1999) paper discussed cases of stress shift, which she claims require reference-set computation. . Note that we predict children to behave in such a way in grammaticality judgment tasks, but not in production, especially in spontaneous production. In the GJ task, children are prepared to give a judgment. Therefore, when they hear a sentence with a pronoun, knowing of its elsewhere character, they know that they have to check whether the pronoun is licit or not in the derivation. However, the same should not occur in production, as the pronoun will only be inserted if needed. . One possibility is to assume that clitics are agreement markers, rather than pronominal elements. If so, they would not be elsewhere elements requiring reference-set computation. Note that children have not been shown to display chance level performance on tests with anaphors/reflexives in any language tested. In this case, the prediction is the same for languages with non-clitic anaphors, like in English, and for languages with clitic anaphors, like BP and the other Romance languages. . Ideally, this experiment would involve two experimenters, one to manipulate the puppet and another to present the pictures to children and to teach them the task. Because I could not find someone able to help me in carrying out the task in Brazil, I developed a way to play both roles. When talking to the child as the experimenter, I used my normal voice and left the puppet sideways, not intervening between the child and me. When pretending to be the puppet, I’d put it in front of my face and spoke with a high-pitched voice. The younger children did not seem to mind that I was the one talking for the puppet. The older children seemed more curious about it, but found it fun to talk to the puppet anyway. So, I do not think that the results of the study were in any way compromised by the presence of just one experimenter. . The sentences used in the training session were formulated in such a way as to teach children to pay attention to subtleties on them. For example, the ungrammatical sentences were not ‘word salad’, but almost grammatical sentences with just one word missing or one word extra. . An exception is McDaniel et al. (1990), who have made use of the grammaticality judgment task to test children’s knowledge of locally A-bound pronouns. Their results are comparable to what has been reported in other studies using the truth-value judgment task.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.21 (1298-1393)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese  . Five children were not considered in this analysis, as they accepted all the sentences, including fillers. This makes their answers unreliable and this is the reason why they were excluded from the analysis. . Thornton and Wexler (1999) tested both simple sentences like the ones in the text and complex sentences involving VP ellipsis. . Thornton and Wexler mention that these are just general guidelines of the story told to children. In narrating the stories, the experimenter was careful not to use reflexive pronouns, for example. . Note that, exactly like in Chien and Wexler’s case, this problem does not arise when the possible antecedent for the pronoun is a DP. In the story with only two characters, both are salient, have names and thus the context does not provide an obvious antecedent for the pronoun, as it did in the case of QPs.

References Avrutin, S. (1999). Development of the Syntax-Discourse Interface. Dordrecht: Kluwer. Avrutin, S. & Thornton, R. (1994). Distributivity and Binding in Child Grammar. Linguistic Inquiry, 25, 265–271. Avrutin, S. & Wexler, K. (1992). Development of Principle B in Russian: Coindexation at LF and Coreference. Language Acquisition, 2, 259–306. Baauw, S., Escobar, L., & Philip, W. (1997). A Delay of Principle B Effect in Spanish Speaking Children: The Role of Lexical Feature Acquisition. In A. Sorace, C. Heycock, & R. Shillock (Eds.), Proceedings of Generative Approaches to Language Acquisition (pp. 16–21). Edinburgh. Bar-Shalom, E. & Vinnitskaya, I. (2001). Relative Clauses in Child Russian: A Comparison with Serbo-Croatian. Manuscript. Storrs, University of Connecticut. Boster, C. (1994). Children’s Failure to Obey Principle B: Syntactic Problem or Lexical Error? In J. Abe, L. Ferro, L. Laporte-Grimes, D. Takahashi, & M. Yamashina (Eds.), UConn Working Papers in Linguistics 4. Storrs, University of Connecticut. Cairns, H., McDaniel, D., Hsu, J., & Konstantyn, D. (1995). Grammatical and Discourse Principles in Children’s Grammars: The Pronoun Coreference Requirement. In E. Fernandez (Ed.), Proceedings of CUNYForum 19. City University of New York. Cardinaletti, A. & Starke, M. (1995). The Tripartition of Pronouns and Its Acquisition: Principle B Puzzles are Ambiguity Problems. In J. Beckman (Ed.), Proceedings of NELS 25. University of Massachusetts, GLSA, Amherst. Chien, Y.-C. & Wexler, K. (1990). Children’s Knowledge of Locality Conditions in Binding as Evidence for the Modularity of Syntax and Pragmatics. Language Acquisition, 1, 225–295. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1986). Barriers. Cambridge, MA: MIT Press. Crain, S., Thornton, R., Boster, C., Conway, L., Lillo-Martin, D., & Woodams, E. (1996). Quantification without Qualification. Language Acquisition, 3, 83–153. Cyrino, S. (1997). Objeto Nulo no Português do Brasil, Ph.D. Dissertation, Universidade de Londrina. Deutsch, W., Koster, C., & Koster, J. (1986). Children’s Errors in Understanding Anaphora. Linguistics, 24, 203–225.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.22 (1393-1500)

 Elaine Grolla

Drozd, K. & van Loosbroek, E. (1998). The Effect of Context on Children’s Interpretation of Universally Quantified Sentences. Manuscript, Max Planck Institute for Psycholinguistics and Nijmegen University. Nijmegen. Escobar, L. & Gavarró, A. (1999). The Acquisition of Catalan Clitics and Its Implications for Complex Verb Structure. In I. Lasser (Ed.), Proceedings of Generative Approaches to Language Acquisition. Frankfurt/Berlin: Peter Lang Verlag. Ferreira, M. (2000). Argumentos Nulos no Português Brasileiro, Masters Thesis, IEL, UNICAMP. Gathercole, S. & Baddeley, A. (1993). Working Memory and Language. Mahwah, MA: Lawrence Erlbaum. Goodluck, H. & Stojanovic, D. (1996). The Structure and Acquisition of Relative Clauses in Serbo-Croatian. Language Acquisition, 5, 285–315. Grimshaw, J. & Rosen, S. (1990). Knowledge and Obedience: The Developmental Status of the Binding Theory. Linguistic Inquiry, 21, 187–222. Grodzinsky, Y. & Reinhart, T. (1993). The Innateness of Binding and Coreference. Linguistic Inquiry, 24, 69–101. Grolla, E. (2004). Resumptive Pronouns as Last Resort: Implications for Language Acquisition. In S. Arunachalam, T. Scheffler, S. Sundaresan, & J. Tauberer (Eds.), Proceedings of the 28th Annual Penn Linguistics Colloquium. Philadelphia, PA. Grolla, E. (2005). Pronouns as Elsewhere Elements: Implications for Language Acquisition. PhD Dissertation, University of Connecticut. Hamann, C. (2002). From Syntax to Discourse. Dordrecht: Kluwer. Hamann, C., Kowalsky, O., & Philip, W. (1997). The French “Delay of Principle B” Effect. In E. Hughes, M. Hughes, & A. Greenhill (Eds.), Proceedings of the 21st Annual Boston University Conference on Language Development. Somerville, MA: Cascadilla Press. Hiramatsu, K. & Lillo-Martin, D. (1998). Children Who Judge What They Produce Ungrammatical. In A. Greenhill, M. Hughes, H. Littlefield, & H. Walsh (Eds.), Proceedings of the 22nd Boston University Conference on Language Development (pp. 337–347). Somerville, MA: Cascadilla Press. Hornstein, N. (2001). Move! A Minimalist Theory of Construal. Oxford: Blackwell. Jakubowicz, C. (1984). On Markedness and Binding Principles. In C. Jones & P. Sells (Eds.), Proceedings of NELS 14. University of Massachusetts, GLSA, Amherst, MA. Kang, B. (2003). The Nature of Children’s Resumptives. Manuscript. General Examination Paper, University of Connecticut. Kato, M. (1993). Recontando a História das Relativas. In I. Roberts & M. Kato (Eds.), Português Brasileiro: Uma Viagem Diacrônica. Campinas: Unicamp. Labelle, M. (1988). Predication et Mouvement: Le Developpement de la Relative Chez les Enfants Francophones, PhD Dissertation, University of Ottawa. Lakoff, G. (1968). Pronouns and Reference. In J. D. McCawley (Ed.), Syntax and Semantics. Vol 7: Notes from the Linguistic Underground (pp. 275–336). New York: Academic Press. Lasnik, H. (1981). On Two Recent Treatments of Disjoint Reference. Journal of Linguistic Research, 1, 48–58. Lees, R. B. & Klima, E. S. (1963). Rules for English Pronominalization. Language, 39, 17–28. Lidz, J. & Idsardi, W. (1997). Chains and Phono-Logical Form. UPenn Working Papers in Linguistics, 8, 109–125. Lust, B., Loveland, K., & Kornet, R. (1980). The Development of Anaphora in First Language. Linguistic Analysis, 6, 217–249.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.23 (1500-1598)

The acquisition of A- and A’-bound pronouns in Brazilian Portuguese 

McDaniel, D., Cairns, H. S., & Hsu, J. R. (1990). Binding Principles in the Grammars of Young Children. Language Acquisition, 1, 121–138. McDaniel, D. & Maxfield, T. (1992). Principle B and Contrastive Stress. Language Acquisition, 2, 337–358. McKee, C. (1992). A Comparison of Pronouns and Anaphors in Italian and English Acquisition. Language Acquisition, 1, 21–55. McKee, C. & McDaniel, D. (2001). Resumptive Pronouns in English Relative Clauses. Language Acquisition, 9, 113–156. McKee, C., Nicol, J., & McDaniel, D. (1993). Children’s Application of Binding Principles during Sentence Processing. Language and Cognitive Processes, 8, 265–290. Morrow, D. (1985). Prominent Characters and Events Organize Narrative Understanding. Journal of Memory and Language, 24, 304–319. Oppy, B. & Long, D. (1996). The Influence of Working Memory Constraints on Pronominal Reference. In R. J. Kreuz & M. S. MacNealy (Eds.), Empirical Approaches to Literature and Aesthetics (pp. 69–82). Norwood, NJ: Ablex. Pérez-Leroux, A. T. (1995). Resumptives in the Acquisition of Relative Clauses. Language Acquisition, 4, 105–138. Philip, W. (1995). Event Quantification in the Acquisition of Universal Quantification, PhD Dissertation, University of Massachusetts. Philip, W. & Coopmans, P. (1996). The Double Dutch Delay of Principle B Effect. In A. Stringfellow, D. Cahana-Amitay, E. Highes, & A. Zukowski (Eds.), Proceedings of the 20th Annual Boston University Conference on Language Development. Somerville, MA: Cascadilla Press. Reinhart, T. (1999). The Processing Cost of Reference-Set Computation: Guess Patterns in Acquisition: Lectures given in Rutgers Cognitive Colloquium. Reuland, E. & Everaert, M. (2001). Deconstructing Binding. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory. Malden, MA: Blackwell. Shlonsky, U. (1992). Resumptive Pronouns as a Last Resort. Linguistic Inquiry, 23, 443–468. Sigurjónsdóttir, S. & Hyams, N. (1992). Reflexivization and Logophoricity: Evidence from the Acquisition of Icelandic. Language Acquisition, 2, 359–413. Silva, C. (1989). Language Acquisition of Portuguese: Some Aspects of Binding Theory, PhD Dissertation, University of California. Solan, L. (1983). Pronominal Reference: Child Language and the Theory of Grammar. Dordrecht: Reidel. Sugisaki, K. & Isobe, M. (2001). Quantification without Qualification without Plausible Dissent. Paper presented at Semantics of Under-Represented Languages in the Americas, University of Massachusetts, Amherst. Tarallo, F. (1983). Relativization Strategies in Brazilian Portuguese, PhD Dissertation, University of Pennsylvania. Thornton, R. & Wexler, K. (1999). Principle B, VP Ellipsis, and Interpretation in Child Grammar. Cambridge, MA: MIT Press. Varela, A. (1988). Binding in Spanish: A Theoretical and Experimental Study. PhD Dissertation, University of Connecticut. Varlokosta, S. & Armon-Lotem, S. (1998). Resumptives and Wh-Movement in the Acquisition of Relative Clauses in Modern Greek and Hebrew. In A. Greenhill, M. Hughes, H. Littlefield, & H. Walsh (Eds.), Proceedings of the 22th Annual Boston University Conference on Language Development (pp. 737–746). Sommerville, MA: Cascadilla Press.

JB[v.20020404] Prn:5/06/2006; 10:27

F: LALD4111.tex / p.24 (1598-1608)

 Elaine Grolla

Wexler, K. & Chien, Y.-C. (1985). The Development of Lexical Anaphors and Pronouns. Papers and Reports on Child Language Development, 24, 138–149. Zwart, J.-W. (2002). Issues Relating to a Derivational Theory of Binding. In S. Epstein & T. D. Seely (Eds.), Derivation and Explanation in the Minimalist Program (pp. 269–304). Oxford: Blackwell.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.1 (46-105)

Acquiring long-distance wh-questions in L1 Spanish A longitudinal investigation María Junkal Gutiérrez Mangado* Euskal Herriko Unibertsitatea

In this article we present data from an oral experiment eliciting LD (Long Distance) Wh-questions from a L1 Spanish child aged 4;9 at the beginning of the experiment. The purpose of the experiment was to observe the development of this structure and more specifically to compare the results with those reported in the L1 acquisition of English by Thornton (1990). The results show that the acquisition of LD wh-questions in Spanish, like in L1 English, is characterized by a stage where non-adult structures are produced systematically. These non-adult structures, known as Partial wh-movement and wh-copying, are concluded to reflect UG options since they obey the same characteristics as those in the natural languages which allow them such as Bavarian and Frisian.

.

Introduction

In this article we present the results of a longitudinal experimental study investigating the acquisition of L1 Spanish long-distance (LD) wh-questions by a Spanish speaking child named Maider, aged 4;9 at the beginning of the experiment. A Spanish LD whquestion is shown in (1) involving object extraction: (1) A quién crees que ama Miguel? to whom think-2sg that love Michael ‘Who do you think Michael loves?’ The main goal of the study was to determine the development of Maider’s LD whquestions in order to find out whether she produced non-adult LD wh-questions involving two related wh-phrases of the type reported in the L1 acquisition of other languages such as English (Thornton 1990), Dutch (van Kampen 1997), French (Oiry 2002) or Basque (Gutiérrez 2003). Examples of the non-adult questions in the acquisition of these languages are given below:

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.2 (105-187)

 María Junkal Gutiérrez Mangado

(2) L1 English (examples from Thornton 1990: 213, 212) a. What do you think who jumped over the can? b. Who do you think who is in the box? (3) L1 Dutch (examples from van Kampen 1997: 140, 155) (L. 7;11.8) a. Wat denk je bij de hoeveelste ik ben? what think you at which number I am TARGET: ‘At which number do you think I am?’ b. Waarom denk je waarom ik op swemles zit? (L. 8;3.8) why think you why I on swimming.lesson am TARGET: ‘Why do you think I take swimming lessons?’ (4) L1 French (examples from Oiry 2002) Tu crois quoi qui est caché dans le sac? you think what that/which is hidden in the bag TARGET: ‘What do you think is hidden in the bag?’ (5) L1 Basque (examples from Gutiérrez 2003) (A. 5;1) a. Zer uste duzu zein bizi dela etxean? what think aux-2sg who live aux-comp house-the-in TARGET: ‘Who do you think lives in the house?’ b. Nor uste duzu nor bizi dela etxe horretan? (A. 5;10) who think aux-2sg who live aux-comp house that-in TARGET: ‘Who do you think lives in that house?’ As will be shown below these non-adult questions in child grammar are fully UG constrained since there are natural languages which use these strategies for forming LD dependencies and exhibit the same restrictions as those observed above. It is also the case that these types of questions are absent from the adult grammars of English, Dutch, French, Spanish and Basque. The results from the L1 Spanish elicitation study showed that Maider also produced LD wh-questions involving two related wh-phrases. Some sample questions are featured below: (6) a.

¿Tú qué crees cómo ha hecho el castillo? you what think-2sg how has made the castle TARGET: ‘How do you think he made the castle?’ b. ¿Dónde crees dónde ha ido el niño? where think-2sg where has gone the child TARGET: ‘Where do you think the child has gone?’

(M. 5;5)

(M. 5;7)

In the first type of non-adult question, (6a) above, the questioned wh-phrase is spelledout in the intermediate CP domain, while a second wh-phrase appears in the matrix clause marking the scope of the question. In the literature, this type of questions are known as ‘Partial movement’ or ‘Wh-scope marking’. In the second type of non-adult question, (6b), two copies of the questioned wh-phrase are spelled-out, one in the

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.3 (187-244)

Acquiring long-distance wh-questions in L1 Spanish 

embedded CP domain and the other one in the matrix. This strategy is known as ‘whcopying’ in the literature (see the collection of papers in Lutz et al. 2000 for more details on partial movement and wh-copying questions). The finding of these non-adult structures in the development of Spanish LD whquestions adds to the body of evidence pointing that in the course of acquiring LD whquestions children produce Partial wh-movement and wh-copying structures before converging on the adult grammar of their respective languages. On the other hand, the Spanish longitudinal data reported in this paper offers new insights into the phenomenon. First, the data show that in Maider’s production, Partial movement questions precede wh-copying and adult-like questions. This finding is also supported by the longitudinal data obtained from an independent study investigating the acquisition of L1 Basque LD wh-questions (Gutiérrez 2004). Second, the production of new non-adult variants of the patterns already reported in the literature leads to an alternative analysis of these non-adult questions. Finally, the contribution from the Spanish data shows how options allowed by UG can vary as a result of language particular characteristics. The paper is organized as follows. In the first section, we present wh-movement options crosslinguistically concentrating on LD wh-movement options and present several analyses proposed in the literature for Partial movement and wh-copying questions. Next we briefly present some characteristics of Spanish and previous findings relating to the acquisition of LD wh-questions in L1 English and Dutch. The rest of the paper describes the experiment on L1 Spanish LD wh-questions and discusses the main findings.

. Wh-movement options cross-linguistically Basically there are two ways in which a wh-question can be formed in natural languages: in the first, questions are formed without displaying wh-movement overtly. This option is found in Chinese, Japanese or Hindi: (7) Example from Mandarin Chinese (Cheng 1991: 11) a. hufei mai-le yi-ben-shu. Hufei buy-asp one-cl-book ‘Hufei bought a book’ b. hufei mai-le sheme Hufei buy-asp what ‘What did Hufei buy?’ Example (7a) shows a declarative question while (7b) shows that when the object is questioned the wh-phrase sheme ‘what’ remains in the same position it occupies in the declarative counterpart.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.4 (244-310)

 María Junkal Gutiérrez Mangado

On the other hand, there are also languages where the questioned wh-element undergoes displacement to the left periphery of the clause. English (8) and German (9) serve as examples: (8) Whoi did he kiss ti ? (9) Was hat Hans gesagt? what has Hans said ‘What has Hans said?’ Among the wh-movement languages, a further distinction has to be made. On the one hand, when two wh-phrases are questioned, there are languages which move only one of the wh-phrases to the matrix while the second one remains in situ. This is the case of English, as in (10) below: (10) Who did what? On the other hand, there are languages in which all the questioned wh-phrases move to the front of the clause. These are known as ‘Multiple fronting languages’ and include Hungarian and Bulgarian, among others: (11) Example from Bulgarian (taken from Rudin 1988 by Cheng 1991: 77) Koj kogo e vidjal. who whom saw-3sg ‘Who saw who?’ The examination of the movement options these languages display in the formation of LD wh-questions shows that the pattern is not uniform. It is the case that in each language subgroup presented above there is a different alternative for forming LD wh-questions. In some cases a wh-phrase in the matrix marks matrix scope, while the questioned wh-phrase remains in the embedded clause: either in situ (in wh-insitu languages), or in the left periphery of the embedded clause (in wh-movement languages). Among the wh-in-situ languages, an SOV language like Hindi exhibits this strategy: (12) Example from Hindi (Lutz et al. 2000: 7) Raam-ne kyaa kahaa thaa ki kis-ne kis-ko maaraa? Raam-erg WH said who whom hit ‘Who did Ram say hit whom?’ Among the languages where only one wh-word moves in matrix questions - alongside English type LD wh-movement shown in (13) below, (13) Whoi do you think [CP ti he kissed ti ?] we find languages, like German and Frisian where Partial movement questions are allowed:1

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.5 (310-372)

Acquiring long-distance wh-questions in L1 Spanish 

(14) Example from Frisian (Hiemstra 1986: 97) Wat tinke jo wa’t ik sjoen haw? what think you who-cl I seen have ‘Who do you think I have seen?’ Finally, among the languages where all wh-phrases have to be fronted in matrix questions we also find some languages like Hungarian which allow Partial movement in LD wh-questions as in (15a) in addition to English type LD wh-movement as in (15b) – cf. Marácz (1990: 297, 325): (15) a.

Mit gondolsz hogy Jánost kinek mit adott? wh-acc you think that János whom-dat what-acc gave ‘What do you think that János gave to whom?’ b. Kit gondolsz hogy Jánost látta? Who-acc you think that János saw ‘Who do you think saw János?’

In the literature, the relationship between the wh-scope marker and the true wh-phase in the embedded clause has received much attention. Basically, two types of analyses have been proposed: the Direct Dependency and the Indirect Dependency Approaches. According to the Direct Dependency Approach, there is a direct relationship between the wh-scope marker and the questioned wh-phrase in the embedded clause. Such a view is proposed for German by Van Riemsdijk (1982), McDaniel (1989), Rizzi (1992) or Müller (1997) among others, for Hungarian by Marácz (1990) and Brody (1995) and for Hindi by Mahajan (1990). Most of these proposals take the wh-scope marker to be an expletive wh-phrase which is merged in the matrix and forms a chain with the wh-phrase in the embedded clause at some level of representation: (16) Under the Direct Dependency Approach also fall analyses which postulate that the wh-phrase in the matrix arises from overt movement of the wh-feature of the whphrase to the matrix where it is spelled out as ‘what’. This view has been proposed by Hiemstra (1986) and Cheng (1997, 2000). According to these accounts, a Partial movement question would have the following derivation: (17) Under the Indirect Dependency Approach, the wh-phrase marking the scope is not a wh-expletive but rather a wh-phrase argument that quantifies over propositions rather than individuals. This wh-phrase moves to the matrix in wh-movement languages such as German or Hungarian and stays in situ in wh-in-situ languages such as Hindi. So in this view, the relationship between the wh-phrase in the matrix and the one in the embedded clause is indirect in the sense that the wh-phrase in the matrix establishes

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.6 (372-445)

 María Junkal Gutiérrez Mangado

a direct relationship with the whole embedded clause rather than with the wh-phrase in the embedded clause.2 An Indirect Dependency Approach would represent Partial movement in wh-movement languages in the following way: (18) English and Spanish pattern with German and Frisian in that in multiple questions only one wh-phrase can and must undergo movement while the second remains in situ. However, like Frisian, German has the alternative Partial movement strategy discussed above, as in example (19), an alternative which lack languages like English and Spanish, as illustrated by the ungrammatical examples in (20) and (21) below: (19) Example from German (McDaniel 1989: 574) Was glausbt du wer heute ankommt? what thinks you who today comes ‘Who do you think is coming today?’ (20) *What do you think who is coming today? (21) *¿Qué crees quién va a venir hoy? what think-2sg who go to come today ‘Who do you think is coming today?’ Alongside English-type LD wh-movement and Partial movement, German and Frisian also allow a third strategy for forming LD wh-questions, known as wh-copying: (22) Example from Frisian (Hiemstra 1986: 97) Wa tinke jo wa’t ik sjoen haw? who think you who-cl I seen have ‘Who do you think I have seen?’ One of the characteristics of wh-copying questions is that it is impossible with Dlinked wh-phrases. Thus, when the questioned wh-phrase is a D-linked wh-phrase such as which boy only Partial movement or an English type LD wh-question are possible.3 That is, when a D-linked wh-phrase moves to the embedded CP, the wh-phrase in the matrix has to be a bare wh-phrase such as ‘what’ in German as illustrated by the following examples adapted from Thornton (1990: 229): (23) a. *Wessen Buch glaubst du wessen Buch Hans liest? which book think you which book Hans is reading ‘Which book do you think which book Hans is reading?’ b. Was glaubst du welchen Mantel Jakob heute angesogen hat? what think you which coat Jakob today put on has ‘What do you think which coat Jakob put on today?’ A third alternative for D-linked wh-phrases has been reported to be possible in some ´ dialects of German (Fanselow & Cavar 2001). In these, the D-linked wh-phrase is in

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.7 (445-509)

Acquiring long-distance wh-questions in L1 Spanish 

the matrix CP and the lower wh-phrase is not an exact copy but rather ‘the minimal spellout of the wh-features that should be present in the lower Spec, C position. . .’ ´ (Fanselow & Cavar 2001: 18). This is illustrated by the example in (24) below, taken ´ from Fanselow and Cavar (2001: 18): (24) Welchen Mann denkst du wen er kennt? which man think you who he knows ‘Which man do you think he knows?’ Besides the restriction of not allowing wh-copying with D-linked wh-phrases discussed above, it appears to be the case that both Partial movement and wh-copying are restricted to a certain subclass of bridge verbs such as ‘think’. In addition, extraction from tense-dependent or non-finite embedded clauses is not allowed in German Partial movement questions as illustrated by the example (25) below taken from McDaniel (1989: 573):4 (25) *Was will Hans wen Jakob besticht? what want Hans whom Jakob bribe ‘Who do you want Jakob to bribe?’ The relevance of these two restrictions observed in the languages which allow whcopying and Partial movement presented above is crucial when non-adult questions in child language are examined: in L1 English and Dutch Partial movement and whcopying questions observe both restrictions, which has been taken as evidence that these non-adult questions are not mere performance errors, but stem from UG since they are subject to the same type of restrictions observed in the adult languages.

. A few remarks about Spanish Spanish is an SVO null subject language where the verb shows agreement with the subject: (26) (Los niños) hicieron un enorme muñeco de nieve. (the children) made-3pl an enormous doll of snow ‘The children made an enormous snowman’ As (26) above shows, the sentence is grammatical with and without the overt subject ‘the children’. The verb hicieron ‘made’ agrees with the subject being specified as third person plural. One of the characteristics of Spanish is that it is a free subject inversion language. That is, both SVO and VOS word orders are allowed. Compare (26) with (27) below: (27) Hicieron un enorme muñeco de nieve (los niños). made-3pl an enormous doll of snow the children ‘The children made an enormous snowman’

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.8 (509-576)

 María Junkal Gutiérrez Mangado

Traditional accounts of the postverbal subject analyse it as being derived by adjunction to the right edge of VP (Torrego 1984 and Suñer 1994). More recently, Ordóñez (1999) has proposed that the postverbal subject is the result of movement (scrambling) of the object to the left of the subject. Another characteristic of Spanish is that the complementizer que ‘that’ is obligatory in the embedded clause in subject, object and adjunct LD wh-questions: (28) a.

b.

c.

d.

e.

f.

¿Quién crees que ha hecho el muñeco de nieve? who think-2sg that has made the doll of snow ‘Who do you think has made the snowman?’ ¿Qué crees que han hecho los niños? what think-2sg that have made the children ‘What do you think the children made?’ ¿Dónde crees que han hecho el muñeco de nieve? where think-2sg that have made the doll of snow ‘Where do you think they have made the snowman?’ *¿Quién crees ha hecho el muñeco de nieve? who think-2sg has made the doll of snow ‘Who do you think has made the snowman?’ *¿Qué crees han hecho los niños? what think-2sg have made the children ‘What do you think the children made?’ *¿Dónde crees han hecho el muñeco de nieve? where think-2sg have made the doll of snow ‘Where do you think they have made the snowman?’

Torrego (1982) points out, however, that there are some dialects of Spanish which allow complementizerless embedded clauses when the embedded verb is in the subjunctive mood: (29) Lamento no esté contenta. regret-1sg not be-3sg-subj happy-fem ‘I regret she is not happy’ Note also that sequences where the complementizer is followed by a wh-word are possible in certain dialects of Spanish (including the dialect spoken by Maider): (30) Example taken from Suñer (1994: 335): Preguntaron que qué le compró Bri a su amiga asked-3pl that what to-her bought-3sg Bri to her friend ‘They asked what Bri bought for her friend’ It has been argued that in these cases Spanish can exhibit a double CP (Plann 1982; Suñer 1986, 1991). Thus, the (simplified) structure for questions such (30) would be (31):

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.9 (576-610)

Acquiring long-distance wh-questions in L1 Spanish 

(31) Double CP structure for Spanish: IP CP preguntaron C’ CP

que qué

C’ C

IP pro

I’ VP

le comprói V’

Bri V ti

a su amiga

Finally note that in matrix wh-questions the verbal complex has to precede the overtly realized subject which results in a sequence where the verb immediately follows the fronted wh-phrase:5 (32) a.

¿Qué han hecho los niños? what have made the children ‘What have the children made?’ b. *¿Qué han los niños hecho? what have the children made ‘What have the children made?’ c. *¿Qué los niños han hecho? what the children have made ‘What have the children made?’

The fronting of the verbal complex is also observed in embedded environments: (33) a.

¿Qué crees que han hecho los niños? what think-2sg that have made the children ‘What do you think the children have made?’

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.10 (610-664)

 María Junkal Gutiérrez Mangado

b. *¿Qué crees que han los niños hecho? what think-2sg that have the children made ‘What do you think that the children have made?’ However, as Torrego (1984) already noted, the VS order observed in matrix questions is obligatory only in the clauses the wh-phrase moves through. Thus, in long distance extraction the subject may precede the verbal complex in the embedded clause: (34) ¿Qué crees que los niños han hecho? what think-2sg that the children have made ‘What do you think the children have made?’ Different analyses have been proposed for the observed obligatory adjacency between the wh-phrase and the verb. Torrego (1984) analysed it in terms of V to C movement, an instance of Germanic-like V2 effect. On the other hand, Suñer (1994) adopting the VP-Subject-Internal Hypothesis (Zagona 1982; Koopman & Sportiche 1986, 1988; among others) proposed that the VSO order is derived by movement of the verb to the highest inflectional projection while the subject remains in situ in the VP internal position. Ordóñez (1999) on the other hand proposes that in Spanish preverbal subjects are dislocated to a more peripheral position (see also Contreras 1991; Zubizarreta 1994 and Uribe Etxebarria 1995). He proposes that such position is the Spec of a Top phrase above IP and which is part of the CP field (following Rizzi 1995). The head of this TopP is filled by a null complementizer in matrix questions which is specified as [–wh] via agreement with the subject in its specifier. Ordóñez proposes that in matrix clauses in Spanish there is movement of the null complementizer creating CP through head movement. When a wh-phrase moves to the matrix Spec CP there is a mismatch between the features of the head which continue being [–wh] and the wh-phrase in its specifier, which is [+wh]. According to Ordóñez, then, it is this mismatch of features between the wh-phrase in the Spec of CP and its head (the null complementizer) which renders the question ungrammatical. So, in his view, it is not the lexical subject in Spec TopP which creates the ungrammaticality in matrix questions but rather the feature mismatch between the elements in Spec CP and its head.

. Previous language acquisition findings: Thornton (1990) and van Kampen (1997) Pioneering work by Thornton (1990) on the acquisition of LD wh-questions showed that some children acquiring L1 English produced non-adult LD wh-questions featuring two related wh-phrases as in the German dialects discussed above. In an oral experimental task Thornton (1990) and Thornton and Crain (1994) found that some of the children participating in the study produced these non-adult LD wh-questions consistently while yet others did so sporadically. The subjects in Thornton (1990) ranged in age from 2;11 to 5;5 (mean 4;3). Non-adult LD wh-questions featuring extra

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.11 (664-729)

Acquiring long-distance wh-questions in L1 Spanish 

medial wh-phrases in the embedded CP from Thornton’s data are shown below. In (35a) we can see a Partial movement question and in (35b) a wh-copying question, cf. Thornton (1990: 213, 212):6 (35) Examples from child English a. What do you think who jumped over the can? b. Who do you think who is in the box? The elicitation experiment used by Thornton engaged the children in a guessing game. The children had to ask questions to a puppet called ‘Ratty’ manipulated by a second investigator. By using this technique, Thornton elicited subject, object and adjunct questions. (36) Protocol for eliciting subject extraction questions (Crain & Thornton 1998: 190–191) Experimenter in low voice to child, so that Ratty can’t hear We know where all the things are hidden, right? We know that there’s a marble in the box, a bear under the blanket and we know that Grover is under the yogurt carton. Let’s see if Ratty can guess where we hid them. Let’s do the box first, OK? We know that there’s a marble in the box, but ask the rat what he thinks. Child What do you think is in the box? Ratty Can you rattle the box for me? Hmm, I think there’s a marble in the box. Child You’re right! Experimenter Hey, he made a good guess. Now let’s do the blanket . . . game continues in the same way Note that the lead-in the investigator uses for eliciting the question is an embedded question where the LD question has been elided (ask the rat what he thinks. . .). In this way the child has no clues as to the structure of the LD wh-question he is asked to produce.

. Non-adult questions are not performance errors A careful examination of these non-adult questions led Thornton to conclude that they bore none of the hallmarks of performance errors. One of the characteristics of performance errors, not found in the data in Thornton, is that they are more likely to occur either at the beginning of the session, due to ‘warm-up effects’ or at the end, induced by tiredness. Thornton also tested whether length effects affected the production of these non-adult questions. She elicited questions such as (37) below, where the adverb really intervened between the matrix verb and the embedded CP position. She found that despite the distance, an extra wh-phrase still appeared, cf. Thornton (1990: 332): (37) What do you think really really really really really what’s in there?

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.12 (729-792)

 María Junkal Gutiérrez Mangado

As Thornton pointed out, performance errors are typical of younger children. But Thornton found that a subset of the older children also produced these non-adult questions. Also, very often performance errors stemming from memory overload result in deletion of material: the non-adult questions produced by these children feature the insertion rather than deletion of material. Had these non-adult questions been performance errors they would have appeared with more frequency in object extraction questions. In object extraction questions the empty gap is further away than in subject extraction questions. Thornton found that her subjects persisted in employing these non-adult patterns in questions involving subject extraction long after having ceased to use them in object extraction questions. In a follow-up study of some of the children who produced these non-adult questions the following developmental pattern emerged: 1. Children who produce wh-copying (and ‘that’ trace questions) were found to produce an extra wh-phrase in both subject and object extraction questions. 2. The extra wh-phrases disappear from object extraction questions and only appear in subject extraction questions. 3. The non-adult questions disappear altogether from both subject and object extraction questions. The non-adult questions produced by Thornton’s subjects had the following characteristics:

1. They never appeared in questions extracting from infinitival embedded clauses. Thornton’s subjects were never found to produce questions such as (38) below: (38) Example of non-attested structure (Thornton 1990: 213) *Who do you want who to eat the pizza?

2. A copy of a D-linked wh-phrase cannot appear in the embedded CP. That is, when a D-linked wh-phrase moves to the embedded CP, the wh-phrase in the matrix has to be a bare wh-phrase such as ‘what’ in German. Questions (39a–c) below show the strategies used by the children to avoid producing wh-copying questions with D-linked wh-phrases, cf. Thornton (1990: 245, 246): (39) a. Which Smurf do you think is holding a toothbrush? b. What do you think which Smurf really has roller skates? c. Which Smurf do you think who has roller skates on? Question (39a) is an adult-like LD wh-question where the D-linked wh-phrase has moved to the matrix. Questions (39b) and (39c), on the other hand, are non-adult but possible question formation strategies in the German dialects as shown above. Question (39b) is a Partial movement question with the D-linked wh-phrase in the embedded CP and ‘what’ as a scope marker in the matrix. Question (39c), on the

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.13 (792-863)

Acquiring long-distance wh-questions in L1 Spanish 

other hand features the D-linked wh-phrase in the matrix CP and a reduced wh-phrase agreeing with the matrix one in person features is overtly realized in the embedded CP position. Following McDaniel’s (1989) analysis of Partial movement questions, Thornton (1990) concluded that the Partial movement questions produced by the English children did not involve a LD dependency. Rather, the wh-phrase ‘what’ in the matrix was taken to be an expletive-like element merged in the matrix to mark matrix scope. Thus, questions as in (39b) above have the following (simplified) structure in (40): (40) In the structure (40) above, the wh-phrase in the embedded clause moves from its base position to the specifier of the embedded CP and does not undergo any further movement. As for wh-copying questions, which Thornton (1990) termed ‘medial wh-’questions, she proposed that these were a reflection of spec-head agreement. Thornton observed a parallelism between the production of questions with the complementizer ‘that’ and questions with a medial wh-phrase. Both types of questions were produced with higher frequency in subject extraction questions, where in adult English the presence of the overt complementizer ‘that’ renders the question ungrammatical. Thornton adopted Rizzi’s (1990) proposal according to which subject traces must be properly governed. In English the trace left in subject extraction questions is properly governed by an agreeing complementizer, which is null: (41) [CP Whoi do you think [CP ti’ [C Øagri [IP ti came]]] When the subject in the embedded IP position moves to the specifier of the embedded C, the trace establishes an agreement relationship with the null complementizer, turning it into an agreeing complementizer, which can properly govern the original trace. In adult English, only the null complementizer can be an agreeing complementizer. Looking at the parallelisms between questions with an overt complementizer and a medial wh- in her child data, Thornton concluded that the overt complementizer ‘that’ and the copy of the wh-phrase in the intermediate CP in wh-copying questions were both instances of agreeing complementizers in child grammar. In her account, children who produced these non-adult questions had wrongly concluded that in English the agreeing complementizer could be realized as ‘that’ instead of the null complementizer required in the adult grammar. As for the medial wh-copy, Thornton proposed that this was the result of spec-head agreement. As the questioned wh-phrase passed through the intermediate spec CP it established an agreement relationship with the complementizer which is realized as a complementizer identical to the moved whphrase. Instances of non-adult questions with an overt complementizer and a medial wh- were concluded to be attempts by children to abide by the ECP. Thornton provided several pieces of evidence for analysing the medial wh- as a complementizer rather than the spelling out of the trace of the moved wh-phrase in the

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.14 (863-923)

 María Junkal Gutiérrez Mangado

intermediate CP. On the one hand, Thornton found no instances where the medial whphrase appeared in the embedded clause also featuring embedded inversion. Movement of the auxiliary in the embedded clause lands it in the complementizer position in English. If the medial wh-phrase had been in the specifier position, then, in principle this structure would have been allowed. Thornton did not find any instances where the embedded clause showed subject auxiliary inversion and a medial wh-question. This was taken as evidence that the medial wh- was in the complementizer position. On the other hand, the type of wh-phrase in the embedded CP also provided evidence for analysing it as an agreeing complementizer. The medial wh- produced by the children in Thornton’s data usually involved a bare wh-phrase rather than a D-linked wh-phrase as in (39c) above. Similarly, van Kampen (1997) reported that L1 Dutch children also produced nontarget-like LD wh-questions in the course of acquisition. van Kampen found that most of the LD wh-questions produced by the two children in her experiment consisted mainly of wh-copying questions, cf. van Kampen (1997: 188): (42) a.

Op welke school denk je waar Laura zit? at which school think you where Laura is TARGET: ‘At which school do you think Laura is?’ b. Hoe denk je hoe hij praat? how think you how he talks TARGET: ‘How do you think he talks?’

(L. 8;3)

(L. 7;10)

In addition, van Kampen (1997) reported that the non-adult questions produced by Dutch children obeyed the two restrictions found in the L1 English non-adult LD whquestions, namely, avoidance of D-linked wh-phrases in wh-copying questions and absence of wh-copying in questions extracting from infinitival embedded clauses. Following Thornton (1990), van Kampen (1997) analysed the L1 Dutch medial wh-phrase as an agreeing complementizer. Van Kampen further suggested that the property of child grammars that allows for these non-adult questions resides in a preference by children for structures with a reduced PF-LF discrepancy. Van Kampen (1997: 158) proposes that child grammars where the intermediate wh-phrase is spelled-out “. . . is closer to, and more explicit about the LF representation than the PF spell-out in the adult language”. She concludes that “the overtly marked chain reflects the LF wh-chain in overt PF elements.” To sum up, Thornton found that some of the children in her study produced non-adult wh-questions involving Partial movement and wh-copying structures. She concluded that these questions could only arise from UG directly as questions involving two related wh-phrases, one in the matrix and the other in the embedded CP, are found in languages such as German, Romani, Russian and Hindi among others. What is more, these types of questions are absent in adult English, hence in the input these L1 English children are exposed to. Similarly, van Kampen (1997) reported that L1 Dutch children also make use of Partial movement and wh-copying questions in a stage prior to converging on

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.15 (923-973)

Acquiring long-distance wh-questions in L1 Spanish 

adult type LD wh-questions. The analysis van Kampen proposed for the non-adult questions in L1 Dutch follows Thornton’s proposal that the medial wh-phrase in child wh-copying questions is an agreeing complementizer and proposes that these structures are a reflection of children’s preference for derivations with reduced PF-LF discrepancy.

. The experiment In this section we present the experiment which was used to elicit LD wh-questions from a Spanish speaking child, Maider, aged 4;9 at the beginning of the experiment.7 Maider is a Spanish speaking child growing up in a Spanish speaking environment and has been attending a state school from age 3, where she is instructed in Basque. However, the main language of communication for Maider is Spanish, which she speaks with her parents (who don’t speak any Basque) and her friends. The subject was tape recorded every four weeks to five weeks and the recordings were later transcribed by the investigator. The recordings cover a time span of one year and 5 months at the last recording to date.

. Design and procedure The experiment used for eliciting LD wh-questions from Maider was an adaptation of Thornton’s (1990) elicitation task. However, a few changes were made (see Note 8). An elicitation task was needed due to the fact that LD wh-questions are very hard to find in spontaneous speech.8 First, we eliminated the toys and puppet. Then, we devised a situation which forced the subject to produce LD wh-questions, meeting the requirement that the target structure be uniquely felicitous under the circumstances. We kept the original idea of involving the subject in a guessing game as in Thornton (1990). In our guessing game the subject had to choose a photograph from two alternatives. If the researcher guessed which photograph had been chosen by the subject she received one point. If the investigator did not guess correctly, the subject obtained the point. The one with more points at the end of the session would win the guessing game. See Appendix I for a sample of the materials used. First the experimenter explained to the subject how the game worked. The materials for all the questions were placed in front of the subject and the questions were chosen by the subject in random order. The subject was asked to choose one of the see-through envelopes containing a color cardboard with a photograph on it. Below the photograph a square was drawn with a question mark in the middle. The main picture in the cardboard represented the scenario for a particular question. Then, two additional smaller pictures were presented to the subject. The investigator described the scenario and the small photographs to the subject. The subject had to choose which small picture she thought would fit the scenario. The subject placed the small photo-

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.16 (973-1042)

 María Junkal Gutiérrez Mangado

graph face down on the square with the question mark. While the subject chose the photograph, the investigator kept her eyes closed so as not to see which photograph the subject had chosen. When the subject was ready, the investigator, with her eyes still closed, asked the subject to produce the question. This technique was used for eliciting subject, object and adjunct LD wh-questions with bare wh-phrases such as qué ‘what’ or a dónde ‘where’ and D-linked wh-phrases such as qué chica ‘which girl’. Questions were elicited featuring extraction out of tensed as well as infinitival complements. The types of target questions elicited are shown in (43) below: (43) Target samples: a. Quién crees que vive en la casa? (tensed, subject, bare wh-) ‘Who do you think lived in this house?’ b. Qué crees que vieron? (tensed, object, bare wh-) ‘What do you think they watched?’ c. Cómo crees que comió la comida? (tensed, adjunct, bare wh-) ‘How do you think she ate her food?’ d. Qué chica crees que compró esta planta? (tensed, subject, D-linked wh-) ‘Which girl do you think bought this plant?’ e. Qué coche crees que compró? (tensed, object, D-linked wh-) ‘Which car do you think he bought?’ f. A qué hora crees que se fueron a hacer surf? (tensed, adjunct, D-linked wh-) ‘What time do you think they went surfing?’ g. Quién quieres que coma el helado? (infinitive, subject) 9 ‘Who do you want to eat the ice-cream?’ h. Qué quiere comer? (infinitive, object, bare wh-) ‘What does he want to eat?’ i. A dónde quiere conducir? (infinitive, adjunct, bare wh-) ‘Where does he want to drive?’ A sample protocol used for eliciting these questions is presented below (see Appendix I for the corresponding set of material): (44) Protocol for eliciting subject extraction questions:10 Experimenter Here we have a man (showing the main photograph to the subject) and two cars (showing the two small photographs to the subject). OK? You choose which car the man bought. (The experimenter places folder between herself and the subject so as not to see what the subject chooses. The child chooses which of the two small photographs to place face down and which to hide (in this case Maider has to choose between a picture of a blue car and a picture of a yellow car).) Experimenter Ready? Subject Yes

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.17 (1042-1122)

Acquiring long-distance wh-questions in L1 Spanish 

Experimenter

Pregúntame qué coche creo que compró ask-me which car think-1sg that bought-3sg ‘Ask me which car I think he bought’ Subject Qué coche crees que compró? which car think-2sg that bought ‘Which car do you think he bought?’ Experimenter (with her eyes still closed) the blue one. Subject No. Experimenter OK well done, you beat me so you get the point. Now let’s do the next one. Besides eliminating the toys and puppet, another change was introduced in the elicitation task. In Thornton (1990) the lead-in included an elided wh-question. We decided to run the risk of including the elided part of the question introduced by the overt complementizer que ‘that’. We were aware that by doing so we were giving the subject an important clue as to how to form LD wh-questions. However, we reasoned that if Maider ignored the lead in and nevertheless produced the non-adult questions, then, her productions would strongly show to be part of UG rather than a performance error. The complete lead-in also helped her construct the LD wh-question.

. Results The results showed that Maider did not produce adult-like LD wh-questions at the beginning of the experiment. In the first two sessions, she produced mainly matrix questions in response to the prompts above or alternatively yes/no questions. It was not until her third session when she started producing questions where the questioned whphrase underwent movement. However, these questions were not target-like. Rather, they involved first Partial movement questions and in subsequent sessions wh-copying and adult-like questions, as we will show below. Maider produced a total of 160 questions distributed in the following way: 42 (26,2%) were subject extraction questions, 66 (41,2%) were object extraction questions and finally, 52 (32,5%) were adjunct extraction questions. The types of questions Maider produced can be seen in the graph below (Graphic 1). Sample questions of each type of non-adult question are given below: 1. yes/no questions: ¿Tú crees que habrá comprado la señora? you think that aux-fut bought the lady ‘Do you think the lady will have bought?’ 2. what . . . wh-: ¿Tú qué crees cómo ha hecho el castillo? you what think how has made the castle ‘What do you think how he made the castle?’

(M. 4;10)

(M. 5;5)

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.18 (1122-1159)

 María Junkal Gutiérrez Mangado 100 80 60 40 20 0 4;9

5;2

yes/no

what... Wh-

5;7

5;9;29

wh-copy

adult

6;2 other

Graphic 1. Development of Maider’s LD wh-question types

3. wh-copy:

4. other:

(M. 5;7) ¿Dónde crees dónde ha ido el niño? where think-2sg where has gone the child ‘Where do you think where the child has gone?’ (M. 5;10) ¿Tú qué crees que ha leído los libros? you what think-2sg that has read the books TARGET: ‘Who do you think has read the books?’ ¿Tú por qué crees que que qué máquina compró you why think-2sg that that which machine bought la chica? the girl TARGET: ‘Which machine do you think she bought?’

From these 160 questions, 68 (42,5%) involved Partial movement questions, 14 (8,7%) were wh-copying questions and 19 (11,8%) were yes/no questions. Questions classified as ‘other’ were of two types: on the one hand, she produced questions involving Partial movement where the wh-phrase marking the matrix scope was other than ‘what’ 11 (6,8%). On the other hand, Maider produced subject extraction questions where instead of the corresponding quién she used the [–animate] wh-phrase qué ‘what’, typically used for objects 23/160 (14,3%). Finally, from the 160 questions produced, 25 were adult-like (15,6%). The Partial movement questions produced by Maider were of two different types. Although the majority of Partial movement questions involved the overt wh-phrase qué ‘what’ in the matrix, she also produced questions where the wh-phrase marking matrix scope was null. From the total of 68 Partial movement questions produced by Maider, 4 (5,8%) were of the type in (45) involving a null wh-scope marker: (45) Question with a null wh-scope marker: ¿Tú crees dónde fue el niño? you think where went the child TARGET: ‘Where do you think the child went?’

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.19 (1159-1226)

Acquiring long-distance wh-questions in L1 Spanish 

Maider also produced Partial movement and wh-copying questions unattested in previous experiments. These involved a wh-phrase in the embedded CP together with a complementizer. In fact, Maider produced questions with the following sequence in the embedded clause: (1) compl. + wh, (2) wh- + compl. and (3) compl. + wh- + compl. These are illustrated in (46), (47) and (48) respectively: ¿Tú crees que que porqué le ha regalado una flor? (5;2) you think-2sg that that why to.him has given a flower TARGET: ‘Why do you think he has given her a flower?’ (5;7) b. ¿Dónde crees que dónde ha ido el señor? where think-2sg that where has gone the man TARGET: ‘Where do you think the man went?’ (5;9) c. ¿Tú qué crees que a quién le besó? you what think-2sg that to whom to.him kissed TARGET: ‘Who do you think he kissed?’

(46) a.

(47) ¿Tú qué crees quién que ha hecho el pastel? you what think-2sg who that has made the cake TARGET: ‘Who do you think has made the cake?’

(5;7)

(48) ¿Tú qué crees que que que porqué que se quedaron en you what think-2sg that that that why that refl.stayed in (5;5) casa? house TARGET: ‘Why do you think they stayed at home?’ From this type of questions involving an overt complementizer in the embedded clause, the most frequent one was sequence (1) exemplified in (46c) (11/68, 16,1%). The distribution of Maider’s non-adult questions across the different question types is shown below (Graphic 2).11

70 60 50 40 30 20 10 0 subject partial movement

object wh-copy

Graphic 2. Distribution across question types

adjuct adult

other

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.20 (1226-1293)

 María Junkal Gutiérrez Mangado 100 80 60 40 20 0 4;9

5;2

5;7

PM with ‘quién’

5;9;29

6;2

type (46)

Graphic 3. Subject extraction questions

Graphic 2, containing Maider’s non-adult questions revealed that subject extraction questions pose an additional difficulty for her. First, it is noteworthy that no adult questions were produced involving subject extraction questions, unlike object and adjunct questions where 29,4% and 25,6% of the questions respectively were adultlike. Similarly, no subject wh-copying was produced by Maider either, unlike the case in object and adjunct wh-questions. Secondly, the developmental pattern of her LD subject extraction questions shows that the first subject extraction questions with the wh-phrase quién in Maider’s data appear in the fifth recording at age 5;7 in Partial movement questions. A sample question of Maider’s subject extraction questions prior to the emergence of quién questions is given below in (49). The full range of her qué subject questions is given in Appendix II. (49) ¿Tú qué crees que ha leído los libros? you what think-2sg that has read the books TARGET: ‘Who do you think has read the books?’

(M. 5;10)

From age 5;7 to 5;10 both types of subject questions (type (49) and Partial movement questions) coexist until type (49) completely disappears in the last recording at age 6;2, when all her subject extraction questions involve Partial movement structures. The distribution of Maider’s subject extraction questions is shown in Graphic 3 above. Note that Maider is able to produce subject matrix questions (50a) and questions involving an embedded subject in exclamatives (50b) from the very first session, where she produced the following questions spontaneously: (50) a.

¿Quién ha hecho las galletas? who has made the cookies ‘Who has made the cookies?’ b. ¡A ver quién ha leído los libros! to see who has read the books ‘Let’s see who has read the books!’

(M. 4;9)

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.21 (1293-1346)

Acquiring long-distance wh-questions in L1 Spanish 

As for the type of questions in (49) it could be hypothesized that Maider is turning her subject extraction questions into object extraction questions as she invariably uses the wh-phrase typically introducing direct objects qué ‘what’ instead of quién ‘who’. However, this seems unlikely since the object of the question is lexically realized in the postverbal position in the embedded clause (los libros ‘the books’ in (49) above). The special status of subject extraction questions is well-known both in the syntactic and acquisitional literature. Recall that one of the characteristics of the non-adult questions reported by Thornton (1990) was precisely that they persisted longer in subject extraction questions due to the special licensing requirements of subject traces in English. As we saw above, Spanish is characterized by being a null subject language and allowing free subject inversion, characteristics which are absent from English. The exact nature and relationship of Maider’s special subject extraction questions with the adult grammar is left for further research.

. Discussion The finding that non-adult questions involving Partial movement and wh-copying are produced in the L1 acquisition of Spanish adds weight to the conclusion reached by Thornton (1990) and van Kampen (1997) that these questions are not performance errors but are fully UG constrained and that they are part of the process of acquiring LD wh-questions. One may wonder whether the relationship between the matrix clause and the embedded clause is actually a subordinating dependency or whether they could involve a paratactic relationship of the type illustrated in (51) below: (51) Tú qué crees? Quién vive en esa casa? you what think-2sg who live in that house ‘What do you think? Who lives in that house?’ However, two observations from Maider’s data lead us to reject this analysis. First, if these questions involved a paratactic dependency no complementizer que ‘that’ would be expected to appear in the embedded clause. However, Maider produces questions of the following type: (52) Tú qué crees que quién vive en la casa? you what think-2sg that who lives in the house TARGET: ‘Who do you think lives in the house?’ The presence of the complementizer clearly establishes that the relationship between the two clauses is a subordinating relationship. Similar questions were elicited from the L1 Basque data, as we saw in (5a) above and repeated here as (53): (53) Zer uste duzu zein bizi dela etxean? what think aux-2sg who live aux-comp house-the-in TARGET: ‘Who do you think lives in the house?’

(A. 5;1)

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.22 (1346-1429)

 María Junkal Gutiérrez Mangado

As the glosses show, in the non-adult Basque question in (53) the subordinating suffix -ela is attached to the embedded auxiliary as in adult Basque. Secondly, a paratactic analysis of these non-adult questions would not be able to account for wh-copying questions, since wh-copying questions do not involve a paratactic dependency: (54) *Tú quién crees? Quién vive en esa casa? you who think-2sg who lives in that house ‘Who do you think? Who lives in that house?’ (55) *Dónde crees? Dónde ha ido el niño? where think-2sg where has gone the boy ‘Where do you think? Where has the child gone?’ The non-adult questions produced by Maider share the two characteristics reported by Thornton in child English and by van Kampen in child Dutch, which are also observed in the adult grammars which allow Partial movement and wh-copying. These are presented below.

.. Extraction from infinitival complements First, the non-adult questions in L1 English and L1 Dutch are produced exclusively in extractions involving embedded finite clauses and are absent from infinitival embedded extraction questions. Maider’s non-adult questions were also restricted to finite embedded questions. This is shown in the exchange below: (56) INV: MAIDER:

INV: MAIDER:

Maider pregúntale a tu madre quién quiere que haga el pastel. Maider ask your mother who she wants to make the cake ama, ¿quién quieres que haga el Mummy, who want-2sg that make-subjun.2sg the pastel? cake ‘Mummy who do you want to make the cake?’ ahora pregúntame a mí quién creo que ha hecho el pastel. now ask me who I think made the cake ¿tú qué crees quién que ha hecho el pastel? you what think-2sg who that has made the cake TARGET: ‘Who do you think has made the cake?’

None of the questions involving an infinitival embedded complement featured Partial movement or wh-copying in Maider’s data. All her questions extracting from an infinitival complement were adult-like. The fact that Maider avoids Partial movement and wh-copying in questions extracting from infinitival complements completely supports previous observations. It is also interesting to observe that unlike in English, the embedded verb in Spanish ‘want’type subject extraction questions is in the subjunctive mood and that the embedded

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.23 (1429-1483)

Acquiring long-distance wh-questions in L1 Spanish 

clause is usually headed by an overt complementizer. This shows that in Spanish, the embedded clause in ‘want’-type structures is clearly headed by a CP projection. So in this respect, both finite and subjunctive clauses in Spanish provide a potential landing site for where the medial wh-phrase can be overtly realized. The fact that the medial wh-phrase is never spelled out in ‘want’-type questions in none of the languages examined so far thus, poses an interesting question. It has been suggested that the contrast allowing partial movement and wh-copying in finite complements but not in infinitival complements may be related to the nature of indicative complements which have been argued to constitute islands for extraction (Uriagereka 1988). Similarly, Etxepare (p.c.) suggests that the presence of a [+wh] feature is restricted to a certain type of C-T systems, namely, indicative C-T systems allowing the presence of a [+wh] feature to the exclusion of subjunctive C-T systems. The precise formulation of this restriction, however, needs to be further explored and is left for future research.

.. D-linked wh-phrases and wh-copying The second characteristic reported by Thornton (1990) and also observed in the L1 Dutch data, was that wh-copying was avoided when the questioned wh-phrase was Dlinked. Maider’s questions also observe this restriction. None of Maider’s wh-copying questions involved D-linked wh-phrases. The types of questions produced by Maider were Partial movement questions as the following example shows: (57) Maider (5;7): ¿Tú qué crees qué chicha ha comprau la planta?12 you what think-2sg which girl has bought the plant TARGET: ‘Which girl do you think has bought the plant?’ It is noteworthy that although Maider produces sporadic adult-like questions from her third recording session at age 5;7 involving bare wh-phrases, in all the questions she produces involving a D-linked wh-phrase, the D-linked wh-phrase remains in the embedded clause. In other words, Maider is not able to move the D-linked wh-phrase successive cyclically long-distance. So, the non-adult questions produced by Maider obey the same restrictions as those in child English and child Dutch. However, several characteristics of Maider’s non-adult questions challenge the analyses proposed by Thornton (1990) and van Kampen (1997) according to which the wh-phrase in the embedded CP is a complementizer bearing spec-head agreement. We propose an alternative analysis by which the intermediate wh-phrase is in the specifier of the embedded CP. That is, the medial wh-phrase is the spelling out of the questioned wh-phrase in the specifier of the embedded CP. In adult Spanish, this position is filled by the trace of the moving whphrase. Several characteristics of Maider’s non-adult questions support this analysis: (1) the developmental sequence of Maider’s LD wh-questions; (2) the production of certain types of non-adult questions involving a D-linked wh-phrase in the periphery of the embedded clause; (3) the presence of the complementizer(s) in the embedded clause. We will present each of the arguments below.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.24 (1483-1560)

 María Junkal Gutiérrez Mangado

.. Developmental sequence of Maider’s LD wh-questions The development of the LD wh-questions produced by Maider so far has shown the following stages: 1. production of yes/no questions instead of LD wh-questions (4;9–4;10) 2. appearance of her first LD wh-questions involving Partial movement structures with a null wh-scope marker (second session). However, the majority of her questions at this stage are still yes/no questions (4;10–5;5) 3. the next type of question involves PM with an overt wh-scope marker. Questions of this type are not produced until the fourth session, where we also find the first adult-like questions (5;5–5;7). Her first subject extraction questions with quién appear at age 5;7 4. in the fifth recording wh-copying questions are produced for the first time, although, the production of these is marginal compared to PM 5. up to her last recording session, Partial movement, wh-copying and adult-like questions coexist, as Graphic 1 above shows. Note that the first attempts at producing LD wh-questions by Maider resulted in what look like yes/no questions, that is, no wh-phrase is present in her questions. However, when Maider starts producing questions where the questioned wh-phrase is spelled out, the wh-phrase appears displaced from its base position in the embedded clause to the periphery of the embedded clause. Her first LD wh-question involving a wh-phrase is repeated below in (58): (58) Tú crees dónde fue el niño? you think-2sg where went the boy TARGET: ‘Where do you think the boy went?’

(M. 4;10)

The developmental pattern observed in the acquisition of L1 Basque parallels the stages reported in Maider’s development of LD wh-questions in Spanish. Longitudinal data from the acquisition of L1 Basque (Gutiérrez 2004) were elicited using the same materials and technique used with Maider. The results have shown that the first LD wh-questions produced by a child, named Axel, acquiring Basque as an L1 consisted solely of Partial movement question types. These were followed by the appearance of wh-copying questions, a stage in which these coexisted with Partial movement structures. Eighteen months after the first recording session this Basque child still produces mainly Partial movement and wh-copying questions, adult-like questions being produced sporadically after the 4th session at age 5;6. His developmental pattern is presented in Graphic 4. As Graphic 4 shows, Axel’s first attempts at producing Basque LD wh-questions all involve Partial movement questions. It is in his third session when he starts producing alternative structures such as wh-copying questions. The questions labelled as ‘other’ involve LD wh-questions where the wh-phrase in the matrix is other than ‘what’. As the graphic shows, Axel’s LD wh-questions in his last recording are mainly Partial movement and wh-copying questions.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.25 (1560-1602)

Acquiring long-distance wh-questions in L1 Spanish  100 80 60 40 20 0

5;1

5;2

5;4

what-wh

5;6 wh-cp

5;9

5;10

adult

other

6;0

6;1

Graphic 4. Development of Axel’s LD wh-questions (Basque)

The question that arises in view of this developmental pattern found in the L1 acquisition of Spanish and Basque LD wh-questions is why Partial movement questions precede wh-copying and adult-like LD wh-questions. Chomsky (1995) proposed a general economy principle of the computational system by which movement seeks to raise only the feature in need of checking. This is reflected in Chomsky (1995: 262) in the following way in (59) below (his (26)): (59) F carries along just enough material for convergence. Pied-piping of any other elements is a reflection of PF, where scattered features may not be interpreted. As way of example Chomsky (1995: 263) considers wh-movement. Whphrases are made up of at least three components: the wh-feature, an indefinite part and the feature [±human]. In adult English the wh-feature of the wh-word cannot rise alone as this would cause the derivation to crash at PF. So it would appear that the less marked option allowed by UG is movement of the wh-feature, an option that seems to be realized in languages such as Japanese (Watanabe 1991). Supposing that child grammars favour less marked options and that additional movement of other nonwh-features is the result of PF conditions which have to be learned from the input, then questions where the wh-feature alone is moved are predicted to be available to children. We propose that Maider’s first questions involving partial movement of the questioned wh-phrase to the embedded CP are instances of the default application of the economy condition above. Her first attempts at producing LD wh-questions involve movement of the wh-feature of the wh-phrase alone, which is realized as a null element in the matrix clause.13 LD wh-questions where the wh-phrase moves partially to the embedded CP and the matrix scope is marked by a phonologically empty wh-scope marker of the type first produced by Maider, are found in languages such as Malay (Cole & Hermon 2000) or Kikuyu (Sabel 2000). In the acquisition literature such structures have been reported in L1 French (Oiry 2002) and L2 English by L1 Japanese speakers (Wakabayashi & Okawara 2003; Yamane 2003):

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.26 (1602-1671)

 María Junkal Gutiérrez Mangado

(60) Example from L1 French (Oiry 2002) (repeated from (4) above): Tu crois quoi qui est caché dans le sac? you think what that/which is hidden in the bag ‘What do you think is in the bag?’ (61) Example from L2 English (Wakabayashi & Okawara 2003: 231) Do you think what is in the bag? LD wh-questions where the matrix scope is realized by a phonologically null element in Maider’s production give way to Partial movement questions where the scope marker is overtly realized as qué. The choice of this wh-word as the overt realization of the wh-scope marker seems to be general in all the child grammars which allow for these structures, as the examples in (2), (3) and (5) above show.

.. The nature of the intermediate wh-phrase One of the arguments for analysing the medial wh- in child English and Dutch as an agreeing complementizer came from the absence of wh-copying questions involving D-linked wh-phrases. In the majority of the cases the D-linked wh-phrase occupied the matrix Spec CP position while the medial wh- in those cases was a bare wh-phrase showing some type of agreement with the D-linked wh-phrase in the matrix: (62) Example from child English (repeated from (35c) above): Which Smurf do you think who has roller skates on? (63) Examples from child Dutch (van Kampen 1997: 187): a. In welk huis denk je waar Femke woont? in which house think you where Femke lives TARGET: ‘In which house do you think Femke lives?’ b. Over wie denk je wie wij praten? about whom think you who we talk TARGET: ‘Who do you think we are talking about?’ c. Op welke dag denk je wanneer ik jarig bent? on which day think you when I birthday have TARGET: ‘On which day do you think I have my birthday?’

(L. 7;10.1)

(L. 7;10.1)

(L. 7;10.1)

In particular, van Kampen (1997: 156) predicts that sequences where the D-linked whphrase remains in the embedded clause and an agreeing bare wh-phrase marks matrix scope are not possible derivations in child Dutch: (64) *Wo glaubst du in welchem House Lotte wohnt? where think you in which house Lotte lives TARGET: ‘In which house do you think Lotte lives?’ The grammaticality of such examples in child English and Dutch would constitute evidence against analysing the medial wh- as a complementizer occupying the embedded C head, since the C head could serve as landing site for a head not a wh-phrase such as ‘in which house’. A question which arises in the light of structures such as (64) is how

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.27 (1671-1737)

Acquiring long-distance wh-questions in L1 Spanish 

the wh-phrase in the matrix gets to agree with the embedded D-linked wh-phrase: where . . . in which house. So evidence in child grammars where this type of questions were produced would constitute evidence against an analysis of the medial wh- as an agreeing complementizer. Crucially, questions of this type were produced by Maider at age (6;2): (65) ¿Quién te ha dicho la ama qué señor ha comprado el who cl has said the mother which man has bought the tractor? tractor TARGET: ‘Which man did Mummy say has bought the tractor?’ This question was produced in the last recording session to date. Note that quién in the matrix cannot be interpreted as the subject of the matrix clause since the subject la ama (Mummy) appears postverbally. In view of this, it could be argued that this type of questions result from the following movement steps: the D-linked wh-phrase qué señor ‘which man’ moves to the embedded Spec CP. From this position, the wh-phrase quién is subextracted. The bare wh-phrase in the matrix would be the spelling out of the [+person] feature together with the [+wh] feature: the fact that the bare wh-phrase agrees with the embedded D-linked supports this analysis. Questions of this type are not exclusive of Maider’s grammar. Similar questions have been found in the data from the acquisition of L1 Basque and also in the L2 English of Basque/Spanish learners: (66) Examples from L1 Basque (Gutiérrez 2004): (A. 6;9) a. Nora uste dezu ze tokitara jun ziela? where.to think aux.2sg what place-to go aux.past.comp TARGET: ‘Which place do you think they went to?’ (A. 6;9) b. Nun uste dezu zein tokitan busti zirela? Where.in think aux.2sg which place.in wet aux.past.comp TARGET: ‘Where do you think they got wet?’ (67) Examples from L2 English (Gutiérrez 2005): a. Who do you think which baby eats the ice cream? (subject 254)14 b. Who do you think which girl did buy the plant? (subject 254) If this view is correct, then, this type of questions would involve subextraction of the features [+wh] and additional features such as [+human] from the embedded Dlinked wh-phrase in the embedded Spec CP to the matrix. If we allow for the possibility of selected feature subextraction for these cases and adopting Chomsky’s (1995) economy principle, then Partial movement questions could also be analysed as involving a LD relationship. Under our account then, children would start with the default option of marking the matrix scope in LD wh-questions by moving only the wh- feature from the wh-phrase in the embedded Spec CP. This would result in Partial movement questions. In the case of Maider, the subextracted wh-feature in the matrix is null. This type of questions in child grammar then would be derived as in (68) below:

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.28 (1737-1799)

 María Junkal Gutiérrez Mangado

(68)

Note that analyses for German and Frisian Partial movement questions have been proposed involving this type of feature movement. While Hiemstra’s (1986) analysis is set in a pre-minimalist account, Cheng (2000) offers an analysis based on the view of movement outlined by Chomsky (1995) sketched above. However, both analyses of PM questions involve movement of the questioned wh-phrase to the embedded CP. Then only the wh-feature of the wh-phrase would be moved further to the matrix CP where it is spelled out as the default que ‘what’. Under this view, PM would be an instance of long-distance wh-movement, albeit only movement of the wh-feature leaving the rest of the features stranded in the intermediate position. On the other hand, Felser (2004) proposed a similar account adopting a copy theory of movement and allowing selective feature deletion for wh-copying questions. When children realize that movement of the wh-feature alone is not sufficient then they produce wh-copying structures, which under our account would involve the subextraction of additional features besides the [+wh] feature moved in Partial movement questions: (69)

Note also that besides accounting for the presence of a bare wh-phrase agreeing in features with the embedded D-linked wh-phrase, the analysis we propose can also account for the presence of a D-linked in the embedded position, which under our account occupies the embedded Spec CP position. Adopting an analysis along these lines for child language implies that, in our case, Maider possesses the mechanism for long-distance extraction from the beginning and that as a default she opts for the UG allowed option of moving only the wh-feature to the matrix. Convergence on the adult grammar takes place gradually, allowing for subextraction of additional features as she progresses towards the adult grammar.

.. Complementizers and intermediate wh-phrases A further piece of evidence arguing in favour of analysing the medial wh- as occupying the embedded Spec CP position comes from the possibilities Maider shows of overtly realizing the embedded complementizer: (70) ¿Tú qué crees que a quién le besó? you what think-2sg that to whom to.him kissed TARGET: ‘Who do you think he kissed?’

(5;9)

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.29 (1799-1877)

Acquiring long-distance wh-questions in L1 Spanish 

(71) ¿Tú qué crees quién que ha hecho el pastel? you what think-2sg who that has made the cake TARGET: ‘Who do you think has made the cake?’

(5;7)

(72) ¿Tú qué crees que que que porqué que se quedaron en you what think-2sg that that that why that refl.stayed in (5;5) casa? house TARGET: ‘Why do you think they stayed at home?’ Question (70) is expected given that the dialect Maider speaks allows for this type of structures as shown in (30) above. Question (71) on the other hand shows that the whphrase quién occupies the embedded Spec CP and that the complementizer following it can be overtly realized. Structures where the wh-phrase in adult Spanish is followed by an overt complementizer can be found in certain expressions: (73) Cómo que no has hecho los deberes! how that not have made the homework ‘What do you mean you haven’t done your homework!’ Finally, that both complementizers project their own heads is shown by the cooccurrence of both complementizers before and after the medial wh-phrase. Investigations of the structure of the left periphery, most notably by Rizzi and his colleagues (dating back to 1997 to the most recent proposals in the edited 2004 volumes), have shown that the left periphery is made up of several complex layers. Rizzi (1997) proposes that the CP layer is made up of at least a Force Phrase and a Finiteness Phrase where the force and finiteness of the following clause are specified respectively. Other CP internal positions have also been identified sandwiched between the top ForceP and the bottom FinP: Focus and Topic phrases which host the focussed and topicalized elements. Adopting this complex CP structure, we propose that questions of the type in (72) produced by Maider can be given an analysis along these lines: (74) [TOP Tú [CP qué [IP crees [FORCEP que [FOCP por qué. . . [C que [IP se you what think that why that rfl quedaron en casa]]]]]]] stayed in house The ForcePhrase part of the embedded CP field in the derivation above hosts the subordinating complementizer que necessary in Maider’s dialect. Although Maider does occasionally overtly realize both complementizers, questions where the complementizer precedes the medial wh-phrase are the most common. However, the production of questions where the medial wh-phrase is followed by a complementizer do indeed provide evidence that the medial wh-phrase is in the specifier position rather than the complementizer position. To sum up, we have presented evidence that in the process of acquiring Spanish LD wh-questions non-adult questions involving Partial movement and wh-copying

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.30 (1877-1954)

 María Junkal Gutiérrez Mangado

Table 1. Developmental stages Stage

Sample question

1. Tú crees que habrá comprado la señora? unattested # Tú crees fue el niño dónde? 2.

Tú crees dónde fue el niño?

3.

Tú qué crees dónde fue el niño?

4.

Tú dónde dices dónde fue el niño?

Description Yes/no question null wh-feature in matrix + in situ wh-phrase in embedded clause null wh-feature in matrix + wh-phrase in embedded Spec CP overt wh-feature in matrix + wh-phrase in embedded Spec CP overt wh-feature plus some additional feature in matrix + wh-phrase in embedded Spec CP

questions are produced. More precisely, we have shown that Partial movement questions precede wh-copying and adult-like LD wh-questions. The developmental path followed by Maider in the process of acquiring LD wh-questions in L1 Spanish is shown schematically above in Table 1. The non-adult questions found in child Spanish share the same restrictions previously reported in L1 English and L1 Dutch: avoiding D-linked wh-phrases in whcopying questions and absence of both Partial movement and wh-copying questions in extractions from infinitival complements. However, in view of some of the characteristics of Maider’s non-adult questions we have concluded that the wh-phrase in the periphery of the embedded clause moves to the Specifier of the embedded Focus Phrase, rather than being analysed as an agreement bearing complementizer as proposed by Thornton (1990) and van Kampen (1997). In order to account for the fact that Partial movement questions precede wh-copying and adult-like questions and also for the presence of D-linked wh-phrases in the embedded clause, we have adopted Chomsky’s (1995) view of movement. According to this proposal the computational system follows a general economy principle which favours the movement of as little material as possible for convergence. Thus, we have concluded that Partial movement questions in child Spanish involve movement of the wh-feature of the wh-phrase in the Spec of the embedded clause. Wh-copying questions would involve movement of additional features such as the person features of the wh-phrase in addition to the wh-feature. The present analysis has several advantages over previous ones. To start with, analysing the medial wh-phrase in child non-adult LD wh-questions as occupying the specifier of the embedded clause allows us to extend the account for wh-copying questions to Partial movement questions too. This implies that Maider possesses the mechanism for forming LD dependencies from the time she starts producing Partial movement questions. This analysis also accounts for the presence of a D-linked whphrase in the embedded Specifier in questions where the matrix wh-phrase shows some type of agreement with it. It also accounts for the non-adult questions where the medial wh-phrase coexists with an overt complementizer.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.31 (1954-1999)

Acquiring long-distance wh-questions in L1 Spanish 

. Conclusion The findings reported in this paper support the conclusion that the non-adult questions reported in the acquisition of LD wh-questions in languages such as English and Dutch are not mere performance errors. In all cases the non-adult questions involve Partial movement and wh-copying questions, which are absent in the adult grammars of English, Dutch and Spanish. Furthermore, they follow the same restrictions this type of questions obey in natural languages. These observations lend support to the hypothesis that they are a reflection of UG. However, the non-adult questions reported in this paper shed new light on the derivation of these non-adult questions. It shows that the Partial movement and whquestions may feature slight variations depending on the characteristics of the adult language being acquired. Thus, we found that the embedded clause can host both a whphrase and an overt complementizer in child Spanish, unlike in child English or Dutch, where coexistence of an overt wh-phrase and a complementizer are not reported in the adult languages. The data from L1 Spanish have led us to propose an alternative analysis for the medial wh-phrase in both Partial movement and wh-copying child questions. According to our analysis the medial wh-phrase is the spelling out of the questioned wh-phrase in the embedded Spec CP. From this position, different features can be subextracted. Thus, the wh-phrase in the matrix clause in Partial movement questions would be the spelling out of the subextracted wh-feature from the intermediate wh-phrase. The wh-phrase in the matrix clause in wh-copying questions is analysed as the spelling out of additional features (e.g.: the wh-feature + person features) subextracted from the medial wh-phrase. This analysis can account for the full range of non-adult questions discussed in this paper. Finally, the developmental data elicited from Maider have shown that Partial movement questions precede wh-copying questions. Our analysis of Maider’s non-adult questions, which adopts Chomsky’s (1995) general economy principle, also accounts for the observed developmental pattern, where Partial movement questions are predicted to precede wh-copying structures since the latter involve movement of additional features besides the default option of subextracting the wh-feature only.

Notes * This research was supported by grants from the Spanish Ministry of Science and Technology (BFF 2000-0101) and the Basque Government/ University of the Basque Country (00103.13013578/2001). This support is hereby gratefully acknowledged. Special thanks are due to Maider for the inthusiasm with which she takes part in each of the experiments and to her parents for patiently facilitating and kindly keeping the appointments for the elicitation sessions. I also want to thank Ricardo Etxepare, Juan Uriagereka and Vidal Valmala for comments on previous versions of this paper and the audience at the Romance Turn Workshop for their comments and discussion.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.32 (1999-2070)

 María Junkal Gutiérrez Mangado . Note that in most wh-movement languages where Partial movement is allowed, English-type LD wh-questions are also possible. . The reader is referred to the volume edited by Lutz et al. (2000) for detailed analyses of partial movement and wh-copying questions in the languages discussed. . The term D-linked (Pesetsky 1987) is used instead of complex wh-phrases for ‘which boy’ type of wh-phrases. The reason behind this lies in the observation that complex wh-phrases such as ‘with who’ are allowed in wh-copying in certain languages such as Bavarian or Afrikaans (du Plessis 1977: 725) illustrate below: (i)

Met wie het jy nou weer gesê met wie het Sarie gedog met wie gaan with who did you now again said with who did S. thought Jan trou? with who go J. marry ‘Whom did you say (again) did Sarie think Jan is going to marry?’

L1 Dutch data also showed that children allow complex-wh-phrases in the embedded position: (i)

Hoe duur denk je hoe duur dit is? how expensive think you how expensive this is ‘How expensive do you think this is?’ (ii) Op wie denk je op wie Sarah verliefd is? with whom think you with whom Sarah in-love is ‘With whom do you think Sarah is in love?’

(L. 8;3.24)

(L. 7;10.25)

It will be shown that the L1 Spanish LD wh-questions produced by Maider also obey this restriction on avoiding wh-copying when the questioned wh-phrase is D-linked, but allows it when it involves a PP wh-phrase. . However, Simpson (2000: 163) gives the following wh-copying questions with a tensedependent complement clause as grammatical, in contrast with the Partial movement question (ii), which is ungrammatical: (i)

Wen willst du wen Jakob bestich? who want you who Jakob bribe ‘Who do you want Jakob to bribe?’ (ii) *Was willst du wen Jakob besticht? what want you who Jakob bribe ‘Who do you want Jakob to bribe?’ . This restriction is not observed in adjunct extraction questions where the subject can intervene between the fronted wh-phrase and the verb: (i)

¿Por qué Miguel llegó tarde?

. Thornton labelled wh-copying questions in her child data medial wh-questions. This term is not used to refer to the Partial movement questions she found. In the present article the discussion of non-adult questions refers both to partial movement and wh-copying questions. . The experiment reported here for L1 Spanish was originally designed and used in an experiment eliciting LD wh-questions from Basque/Spanish bilingual learners learning English in a school context (Gutiérrez 2005). The age range of the learners in this experiment ranged from 8 to 18. In order to adapt Thornton’s task to all the age ranges, a few changes were made, which

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.33 (2070-2133)

Acquiring long-distance wh-questions in L1 Spanish 

will be described below. The experiment was subsequently used with Maider and also in an experiment eliciting LD wh-questions from a child acquiring Basque as an L1 (Gutiérrez 2003, 2004). . See, however, van Kampen’s (1997) investigation where her subjects produced this type of questions in spontaneous production. . Note that in the subject infinitival question the embedded verb is in the subjunctive while in the object and adjunct questions the embedded verb bears the infinitival ending. This stems from one of the characteristics of Spanish infinitival complementation namely the possibility of having infinitival and subjunctive complements: (i)

Maria quiere PRO comer Maria wants eat.to ‘Mary wants to eat’ (ii) Maria quiere que pro coma Mariai wants that pro *i/j eat-subj ‘Maria wants him/her/it to eat’ As the examples show, when the embedded verb is in the infinitive, the embedded subject is PRO, which is obligatorily controlled by the subject in the main clause. When the embedded verb is in the subjunctive form, the embedded subject is pro and cannot be controlled by the main subject, as indicated by the indices in (ii). . The protocol was entirely presented in Spanish. However, in the article only the actual prompt and the garget answer are presented in Spanish and glossed for ease of exposition. It has to be said that the game proved to be rather easy to set up from the experimental point of view and very successful with the children in the different experiments. Maider’s attention was invariably focused on beating the experimenter and making sure she didn’t cheat and kept her eyes shut during the game. . Questions with wh-scope markers other than ‘what’ and yes/no questions have been excluded from the graph. . The verb form comprau Maider uses corresponds to the adult form comprado ‘bought’. The form used by Maider is characteristic of child Spanish and is also found in colloquial speech in the Spanish peninsula. . Note however, that a strict application of this general economy principle would predict that Maider’s first productions would leave the wh-phrase in situ as exemplified below: (i)

Tú crees fue el niño dónde?

However, Maider never produces a LD wh-question where the questioned wh-phrase is spelled out in its base position as in (i) above. We assume that evidence from matrix clauses has led Maider to establish that wh-phrases cannot stay in situ. This is evidenced by her matrix questions, all of which show displacement of the wh-phrase to the matrix CP. Also note that no LD wh-questions have been reported in the literature where the wh-phrase remains in-situ. That is, no questions such as (ii) below have been reported in the child data of languages where wh-questions are formed by movement: (ii) Do you think she saw what? Furthermore, data from the L2 acquisition of English by L1 Japanese speakers has shown that subjects acquiring English produce questions of the following type:

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.34 (2133-2238)

 María Junkal Gutiérrez Mangado

(iii) Example from L2 English (Wakabayashi & Okawara 2003: 231) Do you think what is in the bag? As the example shows, the wh-phrase in the embedded clause has undergone movement. Being Japanese a wh-in-situ language, it would not have been surprising to find type (ii) questions. However, none have been reported. The question that arises now is why the general economy principle by which only the wh-feature of the wh-phrase needs to be moved doesn’t apply to the first movement step, resulting in a derivation such as (i) above. This question shall be explored in further research. . This subject in particular was a teenage girl aged 18 at the time of the experiment who had been learning English in a school context for 9 years (around 810 hours). Her score in a general knowledge of English test (Oxford Placement test, 1992) was of 6 over 10.

References Brody, M. (1995). Lexico-Logical Form. Dordrecht: Kluwer. Cheng, L. L.-S. (1991). On the Typology of Wh-Questions. Doctoral dissertation, MIT. Cheng, L. L.-S. (1997). “Partial wh-movement”. UCI Working Papers in Linguistics, 3, 27–50. Cheng, L. L.-S. (2000). “Moving just the feature”. In U. Lutz, G. Müller, & A. von Stechow (Eds.), Wh-Scope Marking (pp. 77–99). Amsterdam: John Benjamins. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Cole, P. & Hermon, G. (2000). “Partial wh-movement: Evidence from Malay”. In U. Lutz, G. Müller, & A. von Stechow (Eds.), Wh-Scope Marking (pp. 101–130). Amsterdam: John Benjamins. Contreras, H. (1991). “On the position of subjects”. In S. Rothstein (Ed.), Perspectives on Phrase Structure: Heads and Licensing, Syntax and Semantics 25 (pp. 61–79). San Diego, CA: Academic Press. Crain, S. & Thornton, R. (1998). Investigations in Universal Grammar: A Guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge, MA: MIT Press. Du Plessis, H. (1977). “Wh-movement in Afrikaans”. Linguistic Inquiry, 8, 723–726. ´ Fanselow, G. & Cavar, D. (2001). “Remarks on the economy of pronunciation”. In G. Müller & W. Sternefeld (Eds.), Competition in Syntax (pp. 107–150). Berlin: de Gruyter. Felser, C. (2004). “Wh-copying, phases and successive cyclicity”. Lingua, 114, 543–574. Gutiérrez, M. J. (2003). “Extra medial wh-questions in long distance wh-questions in L1 Basque acquisition”. Paper presented at the 1st EHU-U.Nantes-CSL Workshop on Syntax and Semantics. Vitoria-Gasteiz, Spain. Gutiérrez, M. J. (2004). “The acquisition of L1 Basque Long Distance wh-questions: a longitudinal investigation”. Paper presented at BIDE04, Bilbo, Spain. Gutiérrez, M. J. (2005). The Acquisition of English Long-Distance Wh-questions by Basque/Spanish Bilingual Subjects in a School Context. Doctoral dissertation. University of the Basque Country. Hiemstra, I. (1986). “Some aspects of wh-questions in Frisian”. Nowele, 8, 97–110. Koopman, H. & Sportiche, D. (1986). “A note on long extraction in Vata and the ECP”. Proceedings of the North East Linguistic society, 4, 357–374. Amherts, MA: Graduate Linguistic Student Association, University of Massachusetts. Koopman, H. & Sportiche, D. (1988). Subjects. Unpublished manuscript. UCLA.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.35 (2238-2344)

Acquiring long-distance wh-questions in L1 Spanish 

Lutz, U., Müller, G., & von Stechow, A. (Eds.). (2000). Wh-Scope Marking. Amsterdam: John Benjamins. Mahajan, A. (1990). The A/A-bar Distinction and Movement Theory. Doctoral dissertation, Cambridge, MA, MIT. Marácz, L. (1990). Asymmetries in Hungarian. Doctoral dissertation, University of Groningen. McDaniel, D. (1989). “Partial and multiple wh-movement”. Natural Language and Linguistic Theory, 7, 565–604. Müller, G. (1997). “Partial wh-movement and Optimality Theory”. The Linguistic Review, 14, 249–306. Ordóñez, F. (1999). Word Order and Clause Structure in Spanish and Other Romance Languages. Doctoral dissertation, City University of New York. Oiry, M. (2002). Acquisition des questions á longue distance. Mémoire de Maîtrise, Université de Nantes, France. Pesetsky, D. (1987). “Wh-in situ: Movement and unselective binding”. In E. Reuland & A. ter Meulen (Eds.), The Representation of (In)definiteness (pp. 98–129). Cambridge, MA: MIT Press. Plann, S. (1982). “Indirect questions in Spanish”. Linguistic Inquiry, 13, 297–312. Rizzi, L. (1990). Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, L. (1992). “Argument/adjunct (a)ssymetries”. Proceedings of the North East Linguistic society, 22, 365–381. Amherts, MA: Graduate Linguistic Student Association, University of Massachusetts. Rizzi, L. (1995). The Fine Structure of the Left Periphery. Manuscript, Université de Genève. Rizzi, L. (1997). “The fine structure of the left periphery”. In L. Haegeman (Ed.), Elements of Grammar: A Handbook of Generative Syntax (pp. 281–337). Dordrecht: Kluwer. Rizzi, L. (Ed.). (2004). The Structure of CP and IP: The Cartography of Syntactic Structures, Volume 2. Oxford: Oxford University Press. Rudin, C. (1988). “On multiple questions and multiple wh constructions. Proceedings of the North East Linguistic society, 6, 445–502. Amherts, MA: Graduate Linguistic Student Association, University of Massachusetts. Sabel, J. (2000). “Partial wh-movement, and the typology of wh-questions”. In U. Lutz, G. Müller, & A. von Stechow (Eds.), Wh-Scope Marking (pp. 409–446). Amsterdam: John Benjamins. Simpson, A. (2000). Wh-movement and the Theory of Feature Checking. Amsterdam: John Benjamins. Suñer, M. (1986). On the Structure of the Spanish CP. Unpublished manuscript, Cornell University. Suñer, M. (1991). “Indirect questions and the structure of CP: Some consequences”. In H. Campos & F. Martínez-Gil (Eds.), Current Studies in Spanish Linguistics (pp. 238–312). Washington, DC: Georgetown University Press. Suñer, M. (1994). “V-movement and the licensing of argumental wh-phrases in Spanish”. Natural Language and Linguistic Theory, 12, 335–372. Thornton, R. (1990). Adventures in Long-distance Moving: The Acquisition of Complex Whquestions. Doctoral dissertation, University of Conecticut, Storrs. Thornton, R. & Crain, S. (1994). “Successful cyclic movement”. In T. Hoekstra & B. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 215–252). Amsterdam: John Benjamins. Torrego, E. (1982). “More effects of successive cyclic movement”. Linguistic Inquiry, 13, 561–565.

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.36 (2344-2406)

 María Junkal Gutiérrez Mangado

Torrego, E. (1984). “On the inversion in Spanish and some of its effects”. Linguistic Inquiry, 15, 103–129. Uriagereka, J. (1988). On Government. Doctoral dissertation, University of Connecticut. Uribe Etxebarria, M. (1995). “On the structure of spec IP and its relevance for scope asymmetries in Spanish and English”. In J. Amastae et al. (Eds.), Contemporary Research in Romance Linguistics (pp. 335–367). Amsterdam: John Benjamins. van Kampen, J. (1997). First Steps in Wh-movement. Ponsen & Looijen: Wageningen. Van Riemsdijk, H. (1982). “Correspondence effects and the Empty Category Principle”. Tilburg Papers in Language and Literature, 12. Wakabayashi, S. & Okawara, I. (2003). “Japanese learners’ errors on long distance wh-questions”. In S. Wakabayashi (Ed.), Generative Approaches to the Acquisition of English by Native Speakers of Japanese (pp. 215–245). Berlin: Mouton de Gruyter. Watanabe, A. (1991). Wh-in-situ, Subjacency and chain formation. Manuscript, MIT. Yamane, M. (2003). On the interaction of first-language transfer and universal grammar in adult second language acquisition: Wh-movement in L1-Japanese and L2-English interlanguage. Doctoral dissertation: University of Connecticut. Zagona, K. (1982). Government and Proper Government of Verbal projections. Unpublished doctoral dissertation, University of Washington. Zubizarreta, M. L. (1994). “El orden de palabras en español y el caso nominativo”. In V. Demonte (Ed.), Gramática del Español. Mexico, DF: Publicaciones de la Nueva Revista de Filología Hispánica.

Appendix I: Sample corresponding to the question Qué coche crees que compró? ‘Which car do you think he bought?’

JB[v.20020404] Prn:27/06/2006; 10:59

F: LALD4112.tex / p.37 (2406-2499)

Acquiring long-distance wh-questions in L1 Spanish 

Appendix II: Maider’s subject extraction questions 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

¿Tú qué crees que habrá comido el helau? TARGET: ¿Qué niño crees que habrá comido el helado? (5;2) ¿Tu qué crees que habrá hecho las galletas? TARGET: ¿quién crees que hizo las galletas? ¿Tú qué crees que ha que ha hecho el pastel? TARGET: ¿Quién crees que ha hecho el pastel? (5;5) ¿Tú qué crees que le ha dado un beso? TARGET: ¿Quién crees que le ha dado un beso? ¿Tú qué crees que que ha hecho las galletas? TARGET: ¿Quién crees que ha hecho las galletas? ¿Tú qué crees que um que que vive en la casa? TARGET: ¿Quién crees que vive en la casa? ¿Tú qué crees que ha leído los libros? TARGET: ¿quién crees que ha leído los libros? (5;7) ¿Tú qué crees que que que ha comido el helau? TARGET : ¿Qué bebé crees que comió el helado? ¿Tú dónde crees tú qué crees tú qué crees que que ha hecho las galletas? TARGET: ¿Qué chica crees que ha hecho las galletas? ¿Tú qué crees que que ha comprau el tractor? TARGET: ¿Qué señor crees que ha comprado el tractor? (5;9) ¿Tú qué crees que que que qué ha comprau el señor el tractor? TARGET: ¿Qué señor crees que ha comprado el tractor? ¿Tú qué crees que ha hecho las galletas? TARGET: ¿Quién crees que ha hecho las galletas? ¿tú qué crees que ha hecho las galletas? TARGET: ¿Quién crees que ha hecho las galletas? ¿Tú qué crees que ha leído los libros? TARGET: ¿Quién crees que ha leído los libros? ¿Tú qué crees que vive en la casa? TARGET: ¿Quién crees que vive en la casa? ¿Tú qué crees que ha comido el helau? ¿Tú qué crees que ha comido el helau? TARGET: ¿Qué bebé crees que comió el helado? ¿Tú qué crees que que cocinó el el pastel? TARGET: ¿Quién crees que cocinó el pastel? ¿Tú qué crees que ha comprado el tractor? TARGET: ¿Qué señor crees que ha comprado el tractor? ¿Tú qué crees que la ha reñido a Masao? TARGET: ¿Quién crees que ha reñido a Masao? (5;10) ¿Tú qué crees que ha leído los libros? TARGET: ¿Quién crees que ha leído los libros? ¿Tú qué crees que ha comido el helau? TARGET: ¿Qué niño crees que ha comido el helado? ¿Tú qué crees que vive en la casa? TARGET: ¿Quién crees que vive en la casa? ¿Tú qué crees a qué a qué está persiguiendo? TARGET: ¿A quién crees que está persiguiendo?

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.1 (46-123)

Evidence from L1 acquisition for the syntax of wh-scope marking in French* Magda Oiry and Hamida Demirdache Université de Nantes

We argue for alternative wh-scope marking strategies in the acquisition of LD-questions. Direct Dependency involves a matrix non-referential wh-scope marker licensing a medial referential wh-phrase. L1 Partial wh-movement is well-documented cross-linguistically. We take the claim that children go through a stage involving wh-scope marking strategies, ungrammatical in the target language but reflecting parametric settings of other languages, a step further by arguing for the existence of L1 Indirect wh-Dependencies: the matrix scope marker is a referential wh-phrase quantifying over propositions, restricted by a subordinate wh-question over individuals. In L1 French, the matrix and subordinate wh-phrases appear in-situ or locally fronted, transparently reflecting the covert/overt syntax of Hindi wh-questions (Dayal 2000). We conclude that the syntax of L1 scope marking should be correlated with the acquisition of complementation.

.

Introduction

The goal of this paper is twofold. We first provide empirical arguments from L1 acquisition of French questions for the syntax of wh-in-situ in the adult grammar. In particular, we provide arguments for the existence of a non-lexical Q morpheme in French (Cheng & Rooryck 2000; Mathieu 1999). The central claim we make is that this Q morpheme licenses both wh-in-situ – be it in the child or the adult grammar – and partial wh-movement in French L1 acquisition (Oiry 2002). The second goal of this paper is to provide empirical arguments for alternative wh-scope marking structures in L1 acquisition of French. We argue that the seemingly surprising syntax of certain Long Distance (henceforth, LD) questions in L1 French reflects the existence in the child grammar of alternative non-adult strategies for forming wh-questions. These strategies, which involve either an argumental/referential or a non argumental/referential scope marker in the matrix, are ungrammatical in the target language but reflect parameters settings for other languages. In particular, we identify two classes of scope marking strategies: indirect vs. direct dependency strategies.

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.2 (123-176)

 Magda Oiry and Hamida Demirdache

Direct dependency yields both partial wh-movement and wh-in-situ: a non-lexical Q morpheme merged in the matrix Spec CP ensures that the scope of a lower wh-phrase, either in-situ or partially fronted at Spell-out, is extended over the matrix clause. Indirect dependency (Dayal 1996, 2000) involves an argument wh-phrase quantifying over propositions in the matrix clause and a subordinate wh-clause serving as a restriction on the matrix wh-quantifier. The latter appears either in-situ or fronted to the matrix Spec CP in the overt syntax. We then turn to the question of how our findings bear on the issue of the acquisition stages involved in the language development process. The experimental results from the production task carried out to elicit root questions does not validate the claim that wh-in-situ is the default option in French child grammar (see Hulk & Zuckerman 2000 or Zuckerman 2001, for instance). Our findings suggest that, if indeed there is a preference for forming root questions, the preference is for overt wh-movement; see Deprez (1995) and Soares (2003) for similar conclusions regarding the status of wh-in-situ in L1 French and L1 European Portuguese, respectively. This preliminary generalization requires further careful and systematic investigation. The syntax of long distance dependencies in French child grammar suggests, however, that the child goes through acquisition stages where long distance dependencies are not established via long movement but rather always involve local movement in the subordinate clause. The dependency between the matrix and the subordinate clause is then established via either (i): coindexation of an argument wh-phrase in the matrix clause quantifying over propositions with the subordinate wh-clause itself (indirect dependency strategy); or (ii) coindexation of a Q morpheme in the matrix clause with a lower wh-phrase itself fronted to the subordinate [–wh] Spec CP (direct dependency strategy). We then suggest three stages in the acquisition of long-distance questions in French: (1) an indirect dependency stage which involves simultaneous local covert or overt movement of an argument wh-phrase in both the matrix and the subordinate clause; (2) a direct dependency stage involving local overt wh-movement to the subordinate Spec CP – licensed by a Q morpheme, generated in a non-argument/operator position in the matrix; and (3) a long movement stage involving overt movement of a subordinate wh-phrase to the matrix Spec CP. This acquisition sequence reflects a semantic shift from scope marking structures where the scope marker is an argument of the matrix predicate and whose restrictor is a questioned proposition (indirect dependency), to scope marking structures where the scope marker is no longer an argument but a Q-morpheme merged directly in an operator/A’-position and whose restrictor is a wh-phrase, and, finally, to LD questions without a scope marker: overt long movement. Our proposal supports the Intermediate State Default Grammars Hypothesis (Abdulkarim & Roeper 2003), according to which, the child goes through various acquisition stages involving default grammars that are gradually selected to match the adult grammar. Building on proposals in Abdulkarim and Roeper, we conclude by

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.3 (176-252)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

suggesting that the above acquisition sequence of LD questions could be correlated with the acquisition of subordination.

. Wh-in-situ in French To account for cross-linguistic strategies in wh-movement, Cheng (1997: 22) proposes the Clausal Typing Hypothesis, which requires every interrogative clause to be overtly typed. Cheng identifies two strategies for typing a clause as interrogative: either a whparticle is merged in C◦ or else the wh-word is fronted to Spec of CP – as illustrated in (1) and (2). (1) Insertion of a Q-particle (e.g. Japanese) (anata-wa) [kare-ga dare-o aisiteita to] omoimasu ka (you-top) he-nom who-acc loved comp think q ‘Who do you think he loved?’ (2) Wh-movement (e.g. English or French) a. Which booki do you want to read ti ? b. Quel livrei veux-tu lire ti ? As is well known, both in-situ and wh-movement strategies are attested in French. Since French lacks an overt scope marker, the question of how wh-in-situ is licensed in French has been the subject of much debate in the literature. (3) Jean a acheté quoi ? John has bought what As shown in (3), no lexical Q morpheme appears in the clause licensing wh-in-situ. How then is the Clausal Typing Hypothesis satisfied in French? Cheng and Rooryck (2000: 5) provide an attractive answer to this question. They argue that wh-in-situ structures do in fact exhibit a phonological overt Q morpheme licensing both wh-in-situ and intonational yes/no questions (that is, yes/no questions without raising of I◦ to C◦ ). In particular, they note that in-situ questions reveal a special intonation that is absent in sentences with wh-movement. They conclude that wh-in-situ is licensed by an intonation morpheme inserted in the syntax as a Qmorpheme and spelled out at PF as a rising intonation. The Q-morpheme is a root morpheme merged in the syntax in a null C◦ position. It serves to check the [+wh] features carried by this null root C◦ . (4) a.

Jean a acheté un livre ? John has bought a book ‘Did John buy a book?’ b. [CP Qi [Jean a acheté quoii ?]] John has bought what ‘What did John buy?’

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.4 (252-299)

 Magda Oiry and Hamida Demirdache

Mathieu (1999: 444) argues on independent grounds for the existence of a non-lexical Q morpheme in French. Under his analysis, wh-phrases in French consist of a variable and a null wh-operator, as illustrated in (5). (5) [CP OPi [IP Jean aime [DP ti quoii ]]] John likes what ‘What does John like?’ The question operator moves to Spec CP full-filling three functions: it serves to indicate the scope of the stranded wh-phrase, to provide a binder for the wh-phrase, and to check the strong Q features of C◦ . In conclusion, under both Cheng and Rooryck (2000) and Mathieu (1999) analysis, the syntax of wh-in-situ in French involves a non-lexical Q morpheme. For Cheng and Rooryck, this Q morpheme is merged in the syntax in the matrix C◦ and is overtly realized via intonation. For Mathieu, the Q morpheme is phonologically null and moves in the syntax to Spec CP. At this stage, we leave open the question of whether the Q morpheme is null or intonational, as it cannot be answered without a thorough comparative study of the intonational properties of both children and adult wh-questions.1

. Partial wh-movement in first and second language acquisition of English LD questions We now turn to the syntax of non-adult strategies for forming wh-questions in first and second language acquisition of English.

. Thornton (1990): L1 English acquisition of LD wh-questions Thornton (1990) carried out an experimental task designed to elicit long-distance questions from English children. The results of this experiment revealed that some children either consistently or sporadically produce non-adult questions with an extra medial wh-phrase in the intermediate CP, as illustrated in (6). (6) L1 English (Thornton 1990: 246) a. What do you think which animal says “woof woof ”? b. What do you think which Smurf really has roller skates? Thornton draws a parallel between the syntax of exceptional questions in the child grammar of English and the syntax of so-called partial wh-movement in the adult grammar of languages such as German. The syntax of partial wh-movement in German is illustrated in (7) quoted from Mc Daniel (1989: 569).

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.5 (299-365)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

(7) *(Wasi ) glaubt Hans mit wemi Jakob jetzt ti spricht? What believes Hans with who Jacob now talk to ‘With whom does Hans believe that Jacob is now talking?’ The matrix verb believe in (7) selects a [–wh] complement. Partial movement of the embedded wh-phrase to the intermediate [–wh] Spec CP position thus violates the Wh-criterion (Rizzi 1996, among others) which requires every wh-phrase to show up in the specifier of a [+wh] C◦ . Now, (7) would be ungrammatical if the wh-phrase (was) in the matrix [+wh] Spec CP were omitted. The latter is analyzed as a basegenerated dummy wh-phrase acting as a scope marker – that is, signaling where the medial wh-phrase is to be interpreted. Since the syntax of the exceptional questions in (6) parallels the syntax of partial wh-movement in (7), Thornton concludes that children produce questions involving partial wh-movement. What in (6) is thus analyzed as a base-generated scope marker indicating the (matrix) scope of the intermediate wh-phrase. Thornton (1990) and Crain and Thornton (1998) conclude that the non-adult LD questions illustrated in (6), are determined by principles of Universal Grammar. That is, children produce questions that, although not well-formed in the target language (English), are nonetheless well-formed in other languages (e.g. German). This is expected under the Continuity Hypothesis according to which children’s developing grammars can differ only in the way adult grammars can differ from each other. In Section 6 below, we will take the proposal that children go through a stage involving wh-scope marking strategies, ungrammatical in the target language but reflecting the parametric setting for other languages, a step further by arguing for the existence of alternative wh-scope marking strategies in L1 French acquisition – that is, scope marking strategies involving either a direct or an indirect dependency between the matrix and the subordinate clause. The classic reference for direct dependency is German partial wh-movement whereas the classic reference for indirect dependency is Hindi (Dayal 2000).

. Gutierrez (forthcoming): L2/L3 English acquisition of LD wh-questions We now turn to the acquisition of LD questions by subjects learning English as a second or third language. Using an adaptation of Crain and Thorntons’ (1998) protocol for inducing oral production of LD questions, Gutierrez (forthcoming) elicited LD questions from a total of 260 bilingual Basque/Spanish children and teenagers ranging in age from 8 to 18. Note that the parameter setting for both Basque and Spanish wh-questions is the same as in English, i.e. wh-movement (in the overt syntax). Out of the 260 subjects tested, 32 (12,3%) produced the same type of non-adult questions as those reported in Thornton (1990). The paradigm in (8) illustrates the partial wh-movement questions produced by these L3 learners of English:

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.6 (365-433)

 Magda Oiry and Hamida Demirdache

(8) L3 English (Gutierrez forthcoming: 25) a. What do you think which baby had eaten the cake? b. What do you think who lived in that house? Gutierrez further reports that according to Wakabayashi and Okawara (2003: 231– 232), Japanese university students learning English as an L2 also produce partial wh-movement questions. Partial wh-movement structures produced by these adult Japanese learners of English are illustrated in (9). (9) L2 Japanese (Gutierrez forthcoming: 17) a. What do you think who loved Mr. Yellow? b. What do you think who did he loved? In both (8) and (9), a wh-phrase appears in the intermediate Spec CP position and a dummy wh-phrase is inserted in the higher CP to signal that the medial wh-phrase has matrix scope. Gutierrez concludes that since partial wh-movement structures are ungrammatical in both the subject’s L1 (Basque/Spanish or Japanese) and the target language (English), they must be determined by UG – and, hence, that L2 learners have access to UG.

. The experiment: Long-distance wh-questions in L1 acquisition of French The goal of this section is to first provide empirical evidence for a partial whmovement strategy in L1 acquisition of French. We then draw out the implications that the overt syntax of partial movement in child French has for the syntax of wh-in-situ in the adult grammar.

. Participants, method and results The results reported here are based on an elicited production protocol originally designed by Celia Jakubowicz (see Jakubowicz 2004) to induce both root and LD whquestions in French. The elicitation technique engages the child in a guessing game which leads the child to ask a question to a teddy bear called ‘Tommy’ in relation to a situation acted out with toys and props in front of the child – as in Crain and Thornton (1998). We tested 20 monolingual children consisting of 9 boys and 11 girls in a kindergarden setting. The children ranged in age from 3;02 to 5;11. The subjects were divided into 3 chronological age groups: 3 children between 3;02 & 3;05 (mean: 3;03.10) – 8 children between 3;07 & 4;02 (mean: 3;11.08) – and 9 children between 4;06 & 5;11 (mean: 5;05.11). 12 adult controls were also tested, ranged in age from 22 to 54 (mean 26;4). Table 1 below presents the general results reported in Oiry (2002) for the 20 children tested – in raw numbers and in percentages. The subjects’ responses are divided

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.7 (433-514)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

Table 1. Typology of questions produced Fronted-wh LD Medial-wh LD Wh-in-situ LD Root / yes-no questions Total

87 11 3 43 144

(60%) (8%) (2%) (30%)

into 4 categories. Fronted-wh groups together all LD questions involving long extraction of a wh-phrase. The syntax of these questions is illustrated in (10) below. Medialwh includes all LD questions with a wh-phrase occurring at the left-periphery of the [–wh] complement clause; see examples in (12)–(14) and Section 6 below. Wh-in-situ refers to LD questions involving a wh-phrase in situ. The last category characterizes unfelicitous adult like questions – that is, root or yes/no questions produced when a LD wh-question is expected. The results show that most subjects, 19/20 (95%), produced adult-like LD whquestions. In contrast, only 2 children (10%) volunteered in-situ LD questions. 8 out of the 20 subjects tested (40%) produced non-adult questions with a medial wh-phrase occurring in the intermediate CP domain of the embedded clause. Strik (2003) conducted an experimental study of the acquisition of wh-questions by L1 learners of French based on the protocol designed by Jakubowicz to induce oral production of root and LD wh-questions. 32 children ranged in age from 3 to 6 were tested. The general results reported in Strik (2003) are consistent with those presented in Table 1. In particular, although the percentage of both wh-in-situ and medial-wh LDs questions is higher in Strik (2003), the distribution of responses across question types is comparable. Wh-in-situ represents 8% of the total number of LD questions elicited (11 items out of 130), and medial-wh questions represent 25% of the total number of responses (33 items out of 130). Both studies thus converge on the same pattern of comparative preferences for establishing LD dependencies: long movement > partial movement > wh-in-situ. It should be noted that the subjects tested in Strik (2003) and Oiry (2002) are from different geographical areas (Paris and Loire Atlantique, respectively). (10) Fronted-wh (Oiry 2002) a. Qui tu crois qui saute ? Who you believe C◦ jumps ‘Who do you believe/think is jumping?’ b. Quoi tu penses qui saute ? What you think C◦ jumps ‘What do you think is jumping?’ c. Quel animal tu penses qui saute ? Which animal you think C◦ jumps ‘Which animal do you think is jumping?’

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.8 (514-575)

 Magda Oiry and Hamida Demirdache

WH + ESK / KESK d. Qui est-ce que tu crois qui saute ? Who-is-it C◦ you believe C◦ jumps ‘Who do you believe/think is jumping?’ e. Qu’est-ce que tu crois qui saute ? What-is-it C◦ you believe C◦ jumps ‘What do you believe/think is jumping?’ Clefts f. C’est qui que tu penses qui saute ? It’s who C◦ you think who jumps ‘Who do you think is jumping?’ (11) Wh-in-situ (Oiry 2002) a. Tu crois que lequel saute ? ◦ You believe C which one jump ‘Which one do you believe/think is jumping?’ b. Tu penses que Lala aime quoi, le sac, le chapeau ou ballon ? or ball You think C◦ Lala likes what the bag the hat ‘What do you think Lala likes, the bag, the hat or the ball?’ c. Tu crois c’ Tinky Winky, il aime quoi ? You believe dem Tinky Winky he likes what ‘What do you believe/think this Tinky Winky, he likes ?’ The strategies for forming LD questions illustrated in (10) and (11) reflect parameter settings of the target grammar. Note, however, that the status of wh-in-situ in tensed subordinate clauses is highly controversial in the literature. For authors such as Boeckx (1999a, b), Cheng (1997), Cheng and Rooryck (2000) or Mathieu (2002), LD wh-in-situ is restricted to infinitive or subjunctive subordinate clauses. In contrast, for authors such as Aoun and Li (1981), Baunaz (2004), Tellier (1991) or Starke (2001), wh-in-situ is allowed in finite complement clause. Judgments elicited from our own informants confirm that there are two dialects when it comes to wh-in-situ in French. The paucity of in-situ LD questions elicited could thus be correlated with dialectal variation in the status of LD wh-in-situ in adult French. We return to LD wh-in-situ in Section 7.1 below (see also Notes 6 and 10).

. Partial wh-movement questions in L1 acquisition of French We now turn to the syntax of LD questions where a medial-wh appears at the leftperiphery of the [–wh] complement clause, illustrated in (12) through (14). (12) a.

tu crois quoi qui est caché dans l’sac ? you believe what C◦ is hidden in the-bag ‘What do you believe/think is hidden in the bag?’

(Oiry 2002)

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.9 (575-632)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

b. tu penses quoi c’qui est caché dans le sac ? ◦ You think what dem-C is hidden in the bag ‘What do you believe/think is hidden in the bag?’ c. tu penses quoi dans l’sac qu’ il-y- a ? you think what in the-bag C◦ there is ‘What do you think there is in the bag? d. Tu penses quoi # que # Tinky Winky l’adore ?2 you think what C◦ Tinky Winky cl-loves ‘What do you think that Tinky Winky likes?’ (13) Tu veux lequel caresser ? you want which one pet ‘Which one do you want to pet?’

(Chaussy 2002)

(Strik 2003) quoi que je lis ? Tu penses que c’est you think C◦ dem-is what C◦ I read ‘What do you think I’m reading?’ b. Tu penses quoi que je lis ? you think what C◦ I read ‘What do you think I’m reading?’ c. Tu penses que c’est qui qui me lit des histoires ? you think C◦ dem-is who C◦ me read det stories ‘Who do you think reads me stories?’ d. Tu penses qui qui me lit des histoires ? you think who C◦ me read det stories ‘Who do you think reads me stories?’ e. Tommy, tu penses quoi que Laa Laa préfère ? Tommy, you think what C◦ Laa Laa prefers ‘Tommy, what do you think Laa Laa prefers?’

(14) a.

The above data show that children acquiring L1 French produce the same type of nonadult questions first reported by Thornton (1990) for the acquisition of L1 English. All the non-adult questions in (12) to (14) involve partial movement of a wh-phrase to the left periphery of the CP/IP domain of the complement clause. We conclude that French children, just like English children, produce questions that are not part of the target grammar, involving partial movement of the wh-phrase to the intermediate Spec CP/IP. There is, however, a crucial difference in the syntax of partial wh-movement questions in L1 French vs. English acquisition. Recall that in both L1/L3 English and adult German, a scope marker appears in the matrix clause. In contrast, in (12) through (14), no overt scope marker appears in the matrix clause to license the medial wh-phrase and indicate that the latter has wide (matrix) scope. The hypothesis of a non-lexical Q morpheme in French provides a straightforward explanation for the syntax of French children’s exceptional LD questions. We assume

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.10 (632-717)

 Magda Oiry and Hamida Demirdache

that this non-lexical Q morpheme licenses both wh-in-situ – whether it is in the child or the adult grammar – and partial wh-movement in the child grammar. Under this proposal, the syntax of partial wh-movement in L1 French parallels the syntax of wh-in-situ. In both (15a) and (15b), a non-lexical Q morpheme is merged in the matrix [+wh] Spec CP in the syntax. The Q-morpheme serves three functions: it types the clause as interrogative; checks the [+wh] features of the null matrix C◦ and acts as a scope marker signaling that the medial or in-situ wh is to be interpreted as having wide (matrix) scope. (15) a.

Partial wh-movement [Qi ] Tu penses [CP quoii [que [je lis ti ]]] you think what C◦ I read b. Wh-in-situ [Qi ] Tu penses lire quoii you think read what

In sum, once we adopt the proposal that French has a non-lexical Q morpheme, then the syntax of partial wh-movement in L1 acquisition of French is no longer surprising.3 This proposal is further supported by the fact that partial wh-movement structures without an overt scope marker are attested cross-linguistically in languages such as Quechua (16), Bahasa Indonesia (17) or Kitharaka (18). We analyze these partial movement structures as involving a null Q morpheme signaling where the medial-wh is to be interpreted at LF. (16) Ancash Quechua (Cole & Hermon 1994: 240) Ø Jose munan may-man Maria away-na-n-ta ? Jose wants where-to Maria go-nom-3-acc ‘Where does Jose want Maria to go?’ (17) Bahasa Indonesia (Saddy 1991: 189) Ø Bill tahu siapa yang Tom cintai ? Bill knows who foc Tom loves ‘Who does Bill know that Tom loves?’ (18) Kitharaka (Muriungi 2004: 10) Ø U - ri-thugania ati n-uu John a-ring-ir-e-t ? ◦ 2sg-T -think that foc-who John subj-beat-T◦ -finalvowel ‘Who do you think that John beat?’ Finally, partial wh-movement without an overt scope marker is also attested in L2 acquisition of English by Japanese adult learners, as illustrated in (19) from Wakabayashi and Okawara (2003: 231–232), quoted from Gutierrez (forthcoming: 17). (19) Ø Do you think what is in the bag? Under the analysis outlined here, the grammar of child French (and presumably of Japanese L2 learners of English) differs from the target adult grammar exactly in the

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.11 (717-765)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

same way as the grammar of adult French (or adult Japanese) differs from Bahasa Indonesia or Ancash Quechua. We thus conclude that the syntax of the non-adult long-distance questions produced by English and French children reflects a parameter setting which, although not part of the target grammar, is part of Universal Grammar: partial wh-movement licensed by a lexical vs. non-lexical Q morpheme. The existence of UG constrained child language variation from the target language is expected under the Continuity Hypothesis (see Crain & Thornton 1998 for discussion of this issue with respect to medial-wh questions in the L1 grammar of English) or the Intermediate State Default Grammars Hypothesis (see Abdulkarim & Roeper 2003), according to which language development is a process involving stages of default multi-grammars that get gradually selected to match the adult input.4 We close this section with a comment on the variation in the syntax of partial wh-movement in child language shown above: an overt scope marker appears in L1 English (see (6)) whereas no overt scope marker appears in L1 French (see (12)–(14)). Fanselow (to appear) observes that “Simple Partial Movement” – that is, in our terms, partial movement without an overt scope marker – always coexists with the wh-in-situ strategy, as well as full wh-movement. Variation in the syntax of partial movement in L1 French vs. L1 English very nicely fits this generalization. That is, in French, which allows wh-in-situ, as well as full wh-movement, no lexical scope marker appears in L1 partial movement questions. In contrast, in English, which does not allow wh-in-situ, a lexical scope marker appears in L1 partial movement questions. Fanselow’s generalization thus allows us to nicely correlate the lexical vs. non lexical status of the scope marker in L1 LD questions with the availability of the in-situ strategy in the target language.5

. Direct dependency scope marking strategies: Wh-in-situ in French and partial wh-movement in L1 French We now argue that both partial wh-movement in the child grammar and wh-in-situ (whether it is in the child or adult grammar) are wh-scope marking strategies involving a direct dependency between the scope marker and the wh-phrase – itself either in-situ or fronted to the intermediate Spec CP, by Spell-out. Direct dependency analyses have been proposed for partial wh-movement in German (illustrated in (7) above), by McDaniel (1989), Beck and Berman (2000), Cheng (1997) or Riemsdijk (1982) among others. The basic tenet underlying direct dependency is that the wh-element (Was in (7)) appearing in the matrix clause is merely a scope marker – directly merged in the matrix Spec CP – whose function is to mark the syntactic scope of a lower wh-phrase stranded in the [–wh] Spec CP position immediately subordinate to the scope marker (mit wem in (7)). The scope marker presumably also serves to overtly type the matrix clause in which it occurs as interrogative. In contrast, under Dayal’s (1996, 2000) indirect independency analysis of Hindi scope marking structures (discussed in Section 6 below), the scope marker is not merged di-

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.12 (765-834)

 Magda Oiry and Hamida Demirdache

rectly in Spec CP – rather, it is an argument wh-phrase quantifying over propositions and, as such, based generated directly in an A-position within the matrix VP. McDaniel (1989) suggests that the scope marker (in scope marking structures involving partial movement) is directly coindexed with the contentful medial wh-phrase with which it is associated. It is kind a of expletive wh-phrase forming a wh-chain with the wh-phrase whose scope marks in the overt syntax and is replaced by the latter at LF. We do not assume here that the scope marker licensing both partial wh-movement in the child grammar of French and wh-in-situ (whether it is in the child or adult grammar) is a semantically vacuous element subject to expletive replacement at LF, for at least three reasons. First, LF-raising of the associate to the expletive position is argued not to be conceptually motivated in Chomsky (1998) – that is, expletives do not attract and do not need to be replaced. The associate simply does not move. Moreover, as Fanselow and Mahajan (2000) point out, merging an expletive into Spec CP is in fact no longer even an available option in the framework. Second, we do not take the scope marker base-generated in the matrix Spec CP and licensing partial whmovement/wh-in-situ in child/adult French to be semantically vacuous – but rather to be a full-fledged Q morpheme serving three functions. It types the matrix clause as interrogative, binds the medial/in-situ wh-phrase and checks the later’s wh/Q feature via Agree (see discussion below). Thirdly, this proposal allows us to draw a principled distinction between the grammar of overt long movement in French on the one hand, and that of wh-in-situ and partial movement on the other, as shown in (20). (20) a.

Partial wh-movement (at Spell-out & LF) [CP [Qi ] tu veux [cp lequeli [PRO caresser ti ]]] ‘Which one do you want to pet?’ b. Wh-in-situ [CP [Qi ] il mange quoii ] ‘What does he eat?’

(21) Long wh-movement [CP1 whi [IP1 . . . [CP2 t’i [IP2 . . . ti ]]]] Long movement in (21) involves overt phrasal movement to the matrix Spec CP to check the [+wh] feature of C◦ . In contrast, partial wh-movement and wh-in-situ are scope marking strategies which do not involve movement of the medial/in-situ whphrase to the matrix Spec CP – whether it is in the covert or overt syntax. Rather a non-lexical Q-morpheme is base generated directly in the matrix Spec CP. The question then is how to compositionally assign matrix scope to the medial/in-situ wh-phrases in (20a–b) – without further (covert) movement. There are at least two well-defined semantic mechanisms available in the literature for encoding scope without movement. (i) Unselective Binding as in Pesetsky (1987) and Nishigauchi (1990): the lower wh-phrase is analyzed as an indefinite introducing an individual variable subject to existentially closure, and the matrix Q provides the existential binder – see Fanselow and Mahajan (2000) for an analysis of partial movement in German along

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.13 (834-897)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

these lines. (ii) A choice function analysis (Reinhart 1997): the lower wh-phrase is analyzed as an indefinite introducing a variable over choice functions, and the matrix Q provides the existential quantifier binding this variable – see Brandner (2000) for an analysis of partial movement in German along these lines. We leave open the question here of what exactly is the appropriate mechanism for capturing matrix scope in (20) without movement. In sum, under the proposal in (20), wh-in-situ in child/adult French and partial movement in child French are scope marking strategies which assign the same syntax: the medial/ in-situ wh-phrase in (20a/b) is bound by the matrix Q at LF. We now provide an indirect argument for the parallel we draw between wh-in-situ and partial movement in French. As is well known, partial wh-movement is blocked by negation. The incompatibility of negation with partial movement is illustrated below with examples from German (Rizzi 1991; in Beck 1996: 3). The ungrammaticality of (22a) contrasts with the grammaticality of the corresponding question involving overt long movement (22b). (22) a.

Partial wh-movement *Was glaubst du nicht mit wem Maria gesprochen hat? what believe you not with who Maria spoken has b. Long wh-movement √ Mit wem glaubst du nicht dass Maria gesprochen hat? has with whom believe you not that Maria spoken ‘Who don’t you believe that Maria talked to?’

Negation is likewise incompatible with wh-in-situ in French, as the paradigm in (23) illustrates. The ungrammaticality of (23a) contrasts with the grammaticality of the corresponding question involving overt long movement (23b). (23) a.

Wh-in-situ *Il ne mange pas quoi ? He neg eat neg what b. Overt wh-movement Qu’est-ce qu’il ne mange pas ? What-is-it that-he neg eat neg ‘What doesn’t Jean eat?’

Summarizing, there are no negative intervention effects with either long overt movement in German or long overt movement in French. In contrast, negative intervention effects show up with both partial wh-movement crosslinguistically and wh-in-situ in French. We take the incompatibility of negation (and other scopal elements) with both wh-in-situ in adult French and partial movement cross-linguistically to indirectly provide support for the parallel we have drawn in (20) between the syntax of partial movement and wh-in-situ in French. We refer the reader to Mathieu (1999) for further arguments for the parallel drawn. In particular, Mathieu argues that constraints on wh-in-situ in French are similar to those governing partial movement in German:

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.14 (897-959)

 Magda Oiry and Hamida Demirdache

scopal elements (e.g. negation), as well as an overt C◦ , count as interveners for the licensing of a lower wh-phrase. We follow Fanselow and Mahajan (2000) in assuming that interventions effects and, more generally, the locality constraints governing partial wh-movement – and, by extension, wh-in-situ in French – follow from the wh-agreement relation established between the matrix Q and the lower wh-phrase in order to ensure that the latter checks its wh/Q feature. That is, since there is no covert movement of the lower wh-phrase in (20), the latter must agree directly with the matrix Q for the feature [+wh]. Locality effects then follow from the locality of this agreement relation. For an alternative account of the locality constraints governing both wh-in-situ in French and partial wh-movement cross-linguistically, see Mathieu (1999).6 To conclude this section, under the proposal illustrated in (20), both partial whmovement in French L1 and wh-in-situ in child/adult French are wh-scope marking constructions involving a scope marker generated in a non-argument/operator position in the matrix clause (Spec, CP) and directly associated (via binding) with a lower whphrase, itself either in-situ or stranded in the specifier of [–wh] C◦ . We now provide evidence from L1 French for scope marking strategies where the scope marker is, this time, merged in an argument-position and where the lower wh-phrase is not directly associated with the scope marker – rather, it is the CP containing the lower wh-phrase that is itself associated with the scope marker.

. Indirect dependency wh-scope marking strategies in L1 French In the preceding sections, we have discussed two wh-scope marking strategies: wh-insitu and partial wh-movement. These options do not exhaust the wh-scope marking strategies attested cross-linguistically for forming long distance dependencies without having recourse to overt long movement. Two proposals have been made to account for the typology of wh-scope marking strategies employed by natural languages: direct vs. indirect dependency. We now argue for the existence of wh-scope marking strategies in L1 acquisition of French involving an indirect dependency between the matrix and the subordinate clause. The classic reference for indirect dependency is Dayal’s (1996, 2000) analysis of Hindi.

. Indirect dependency in Hindi The wh-scope marking construction in Hindi is illustrated in (24), from Dayal (2000: 160–162). Notice that two wh-phrases appear in (24): kyaa (‘what’) appears in the object position of the main clause, and kisse (‘who’) in the object position of the embedded verb talk.

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.15 (959-1009)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

(24) Jaun kyaa soctaa hai ki merii kis-se baat karegii? John what think-pr that Mary who-ins talk do-fut ‘Who does John think Mary will talk to?’ (25) Jaun kyaa soctaa hai? John what think-pr ‘What does John think? Dayal argues that the wh-scope marker kyaa occurring in the matrix clause is crucially not a non referential (expletive) scope marker. Rather, it is an ordinary wh-phrase appearing in its base argument position. Dayal thus draws a parallel between the matrix clause in (24) and the independent clause in (25). In both (24) and (25), the object whphrase kyaa occurs in the internal argument position of the verb ‘think’ and is used to question over the set of propositions that John stands in the think relation to. In (24), we thus have two clauses, each containing a contentful wh-phrase and interpreted as a wh-question in its own right. The matrix (CP1) is a question over propositions, and the subordinate clause (CP2), syntactically analyzed as an appositive clause adjoined to the matrix, is a question over individuals. The LF for (24) is given in (26). The in-situ wh-phrases each move to the specifier position of the CP dominating them, yielding two local wh-dependencies. The connection between the two clauses is established indirectly by coindexing the matrix wh-phrase and the subordinate wh-question, as shown in (26). CP1

(26) CP1

Spec Whati

C P2i

IP

DP John

IP

Spec whoj VP

DP Mary

VP

DP

V

DP

V

ti

think

tj

will talk

Semantically, the subordinate wh-question forms the restriction of the wh-quantifier in the matrix clause: it restricts the set of propositions that are possible answers to the matrix question to all and only those propositions that are also possible answers to the subordinate question. e.g., in (26), CP2 restricts the matrix wh-question (What does John think?) to propositions concerning the possible people that Mary will talk to. The interpretive procedure given for indirect wh-dependencies thus creates the effect of long distance wh-extraction. Dayal further argues that languages can differ with respect to the syntactic realization of indirect dependency. She identifies three syntactic options: (i) juxtaposition of

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.16 (1009-1084)

 Magda Oiry and Hamida Demirdache

two clauses, CP1 and CP2 are adjoined; (ii) indirect syntactic subordination, CP2 is embedded within IP1; and (iii) direct subordination, CP2 is generated as the internal argument of the matrix verb. Dayal illustrates the first option with sequential questions in English and parenthetical was constructions in German. She takes sequential questions to have properties characteristic of scope marking. In (27a), the wh occurring in CP2 is construed as taking scope outside its syntactic domain as the possible answers to (27a) show: the answer in (27b) embeds the proposition corresponding to CP2 as a complement to the verb in CP1, supplying a value for the variable in CP2. (27a) is assigned the representation in (27c). Syntactically, the two independent clauses are adjoined. Semantically, the wh-phrase in CP1 is a quantifier over propositions restricted by the wh-question with which it is coindexed (CP2).7 (27) a. What do you think? Who will Mary see? b. I think Mary will see Tom. c. LF: [[CP1 whati do you think ti ] [[CP2i whoj tj will Mary see] (28) a.

Indirect dependency Was glaubst du wohin ist er gegangen? what think you where has he gone b. Partial wh-movement Was glaubst du wohin er gegangen ist? what think you where he gone has c. ‘Where do you think he has gone?’

(28a) illustrates the syntax of so-called parenthetical was constructions in German. The syntax of this construction differs from the corresponding partial movement question (28b), in that V2 occurs in CP2 – signaling that the clauses in (28a) are independent / root questions, which in turn entails that was in (28a) is not a scope marker but an ordinary wh-phrase. Conversely, the absence of V2 in the partial movement structure (28b) signals that CP2 is a subordinate clause. Indirect dependency straightforwardly explains how the two root questions in (28a) combine together to yield the meaning of the LD question in (28c).

. Indirect dependency in L1 French We now argue that L1 acquisition of French provides empirical evidence for indirect dependency wh-scope marking strategies as a means of forming LD dependencies. The relevant data are given below: (29) a.

Qu’est-ce que tu crois qu’est-ce caché dans what-is-it-that you believe what-is-it hidden in le sac ? the bag ‘What do you think that is hidden in the bag?’

(Oiry 2002)

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.17 (1084-1137)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

b. Qu’est-ce que tu penses qu’est-ce que j’aime lire ? what-is-it-that you think what-is-it-that I like read ‘What do you think that I like to read?’ c. Ce qu’il pense # c’est qui qui est caché dem-what-cl thinks dem-is who C◦ is hidden dans le sac ? in the bag ‘What do you believe is hidden in the bag?’ (30) a.

Tu crois quoi # lala elle aime bien quoi ? you believe what Lala she likes well what ‘What do you believe Lala likes?’ b. Tu crois quoi que je bois quoi ? you believe what C◦ I drink what ‘What do you believe I’m drinking?’

(31) tu crois quoi qui est caché dans l’sac ? you believe what who is hidden in the-bag ‘What do you believe is hidden in the bag?’

(Strik 2003)

(Oiry 2002)

(Oiry 2002)

(Strik 2003)

(Oiry 2002)

Consider first the paradigm in (29). Questions introduced by qu’est-ce que are matrix questions in standard French. We take the occurrence of qu’est-ce que in CP2 to signal that the clauses in (20) are independent/root questions. Note that there are alternative analyses of these root questions in the literature. The first option is that their derivation involves two movements: wh-raising of the interrogative object pronoun que, and V2 – that is, raising of the verb est over the subject ce. The second option is fronting of the object wh-phrase que into the specifier of the complex interrogative C◦ ESK. The third option would be to front the object wh-phrase KESK analyzed as a single word. Notice, however, that in (29a) repeated below as (32), the complementizer que (‘that’) in CP2 has been elided. Elision of que argues against the last two alternatives which take estce que and qu’est-ce que to be unanalyzable, undecomposable words (standing for the complex interrogative C◦ ESK and the wh-word KESK, respectively). We thus conclude that qu’est-ce que root questions involve V2 with subsequent raising of the wh-phrase que. The crucial point for us here is that the syntax of the questions in (29a–b) is exceptional in that it involves two root/independent questions. Under Dayal’s proposal, however, the syntax and the semantics of these surprising non-adult questions become transparent. Syntactically, we have two juxtaposed/adjoined matrix questions, as shown in (32). Semantically, the wh-phrase que/KESK in CP1 is a quantifier over propositions restricted by the wh-question with which it is coindexed (CP2).

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.18 (1137-1221)

 Magda Oiry and Hamida Demirdache

(32) Overt syntax of wh-scope marking in L1 French [CP1 Qui ’est-ce que [tu crois ti ] [CP2i quj ’est-ce [t’j caché tj dans what-is-it-that you believe what-is-it hidden in le sac]] the bag ‘What do you think that is hidden in the bag?’ We conclude that the overt syntax of the exceptional questions (29)/(32) in L1 French transparently reflects the covert syntax of wh-scope marking structures in Hindi, illustrated in (33). (33) Covert syntax of wh-scope marking in Hindi (24)/(26) [CP1 kyaai [Jaun ti soctaa hai] [CP2i kis-sej ki Merii tj baat karegii] think-pr talk do-f what John who-ins that Mary ‘Who does John think Mary will talk to?’ The same analysis can be extended to the exceptional question in (29c), to which we assign the representation in (34): (34) [CP1 Ce qui ’il pense ti ] # [CP2i c’est quij qui est t’j caché tj dans dem-what-cl thinks dem-is who C◦ is hidden in le sac] the bag ‘Who does he think is hidden in the bag?’ Note that we have analyzed que in CP1 as the neutral interrogative pronoun ‘what’, and not as the complementizer ‘that’. The evidence for this analysis is diachronic: ce que, in contemporary adult French, introduces indirect questions involving extraction of an object, as in (35). However, according to Grévisse (1980: 1282), bare interrogative que (derived from latin quid) was used as an interrogative direct object pronoun in both direct and indirect wh-questions. Bare que in indirect interrogatives disappeared in the XXVII century. (35) Je me demande ce que Jean pense I me ask dem what John thinks ‘I wonder what John thinks.’ Notice finally the phonological pause after the matrix ‘think’ in (34). The occurrence of this pause supports our analysis of (34) as involving two independent clauses, neither of which is subordinated to the other, each containing a contentful wh-phrase and interpreted as a wh-question in its own right. We have established a parallel between the overt syntax of wh-scope marking in non-adult French questions and the covert syntax of wh-scope marking in languages such as Hindi. We now take this syntactic parallel a step further. Consider (30) above. Notice that the overt syntax of the French non-adult question in (30) and that of the Hindi adult question in (24)/(26) are identical in all relevant

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.19 (1221-1287)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

respects. That is, both (24)/(26) and (30) exhibit two in-situ wh-phrases: the first whphrase appears in the object position of the matrix verb (‘believe’/‘think’), and the second wh in the object position of the subordinate verb. We assign (30) the representation in (36). The higher wh-phrase quantifies over the set of propositions that the subject stands in a belief relation to. CP2 restricts the denotation of the matrix wh-question to propositions concerning the possible things that Lala likes. Indirect dependency thus creates the effect of LD wh-extraction: (36) [CP1 Tu crois quoii ] # [CP2i Lala elle aime bien quoij ] you believe what Lala she likes well what ‘What do you believe Lala likes? The occurrence of a phonological pause in (36) after the wh-phrase occupying the internal argument position of ‘believe’ signals once again a wh-scope marking strategy involving juxtaposition of two syntactically independent questions. The proposal that the L1 French questions illustrated in (29) through (31) are wh-scope marking structures instantiating indirect dependency explains the seemingly ungrammatical syntax of these non-adult questions. In particular, the syntax of these questions involves two root questions with a wh-phrase occurring in both the matrix and the subordinate clause. Both whs can either remain situ or be fronted in the overt syntax. The wh occurring in the first clause can be any of the wh-phrases used to quantify over propositions in French – that is, either quoi, que/KESK or (ce) que. Indirect dependency straightforwardly explains how these two root questions combine semantically together to yield the meaning of a long distance question.

. Direct or indirect dependency? Finally, consider (31) repeated below. (31) Tu crois quoi qui est caché dans l’sac ? you believe what who is hidden in the-bag ‘What do you think that is hidden in the bag?’ We now have two possible analyses for this exceptional question, depending on the status of qui. If we analyze the latter as the subject wh-pronoun qui, then (31) instantiates an indirect dependency scope marking strategy, as in (37a). That is, we have two semantically contentful wh-phrases appearing in argument positions: quoi appears insitu in the object position of croire and qui appears in a derived argument position, the subject position of the embedded passive verb. (37) a.

Indirect dependency [CP1 Tu crois quoii ] [CP2i [IP2 quij est caché tj dans l’sac]] you believe what who is hidden in the-bag

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.20 (1287-1355)

 Magda Oiry and Hamida Demirdache

b. Direct dependency [CP1 Qi [Tu crois [CP2 quoii [C◦ qui] [t’i est caché ti dans l’sac]]]] you believe what that is hidden in the-bag Alternatively, qui could be the complementizer que that becomes qui (que/qui alternation)8 when a subject is extracted, as in (37b). (31) would then instantiate a direct dependency scope marking strategy. That is, the wh-phrase quoi has undergone partial movement to the intermediate Spec CP, and is bound by the non-lexical Q morpheme in the matrix. No further movement ensues at LF. To conclude, we have argued that the syntax of exceptional questions in child French reveals the existence of alternative wh-scope marking strategies for forming long distance dependencies. We have identified two classes of wh-scope marking strategies: indirect vs. direct dependency. Direct dependency yields both partial whmovement and wh-in-situ: the matrix non-lexical Q morpheme is merged in the matrix in an operator/A’ position and directly associated (via binding) with a lower wh-phrase, itself either in-situ or stranded in the specifier of [–wh] C◦ at Spell-out. The indirect dependency strategy (in the sense of Dayal 2000) involves two clauses, each containing a contentful wh-phrase, and interpreted as a wh-question in its own right. Both whs can simultaneously appear at Spell-out either in-situ or else fronted to the specifier position of the CP in which they occur. The wh-phrase in the matrix is not directly associated with the wh-phrase in the subordinate clause – rather, it is associated (coindexed) with the CP containing the latter. This proposal is recapitulated in (38). (38) L1 French wh-scope marking Direct dependency Partial wh-movement (14b) Qi Tu penses quoii que je lis you think what that I read LD wh-in-situ (11a) Qi Tu penses que lequeli saute you think that which one jumps Indirect dependency Overt wh-movement + V2 in both CP1 & CP2 (29a–b) Qui ’est-ce que tu crois ti [CPi quj ’est-ce (que) j’aime lire tj ] what-is-it-that you believe what-is-it-(that) I-like to read Covert wh-movement in both CP1 & CP2 (30a) Tu crois quoii # [CPi lala elle aime bien quoij ] you believe what Lala she likes well what

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.21 (1355-1431)

Evidence from L1 acquisition for the syntax of wh-scope marking in French 

Table 2. Typology of root questions produced Fronted-wh Wh-in-situ In-situ/Fronted Total

81 28 15 124

(65%) (22,5%) (12,5%)

. Acquisition stages We conclude with a brief discussion of how our findings bear on the question of the stages involved in language acquisition.

. Wh-in-situ as the least marked strategy? The experimental results from the production task carried out to elicit root questions (Chaussy 2002) are presented in Table 2 in raw numbers and in percentages. Note that the same 20 children were tested for both root and LD questions. The category In-situ/Fronted in table 2 refers to subject wh-questions analyzable as either wh-in-situ or wh-fronting questions (e.g. Qui saute ? ‘Who jumps?’). Note that the category wh-in-situ receives only 22,5% of the children’s responses (or maximally 34% if we take into consideration the ambiguous In-situ/Fronted responses). Moreover, we did not find any correlation between age and in-situ responses (younger children did not produce more in-situ responses than older children9 ). These findings do not validate the claim that wh-in-situ is the default option in French child grammar – see Hulk and Zuckerman (2000) or Zuckerman (2001), where wh-in-situ is ranked as the most economical option. Recall further that only 3 wh-in-situ LD questions where produced out of a total of 144 items (by two subjects). These results are surprising under the view that overt movement is more costly than either non-movement or covert movement.10 We conclude that our findings do not reflect a preference for wh-in-situ as the least marked strategy for forming (non) local wh-dependencies in L1 French. This conclusion supports the claim in Deprez (1995), based on Deprez and Pierce (1993), according to which wh-movement is enforced very early in child French and there is no stage reflecting a preference for wh-in-situ. Note finally that a similar conclusion is put forth in Soares (2003) for L1 European Portuguese (EP). According to Soares, the in-situ strategy only becomes available in child EP much later that wh-movement to the left periphery.

. Long-distance dependencies We have argued that the seemingly surprising syntax of certain LD questions in L1 French reflects alternative non-adult scope marking strategies for forming longdistance dependencies in the child grammar. The existence of these scope marking strategies suggests that the child goes through acquisition stages where long distance

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.22 (1431-1503)

 Magda Oiry and Hamida Demirdache

dependencies are not established via long movement – whether it is covert or overt – but rather always involve local movement in the subordinate clause. The dependency between the matrix and the subordinate clause is then established via coindexation of either (i) an argument wh-phrase in the matrix clause quantifying over propositions with the subordinate wh-clause itself (indirect dependency strategy), or (ii) a Q morpheme in the matrix clause with a wh-phrase itself fronted to the subordinate [–wh] Spec CP (direct dependency strategy). This leads us to suggest the following sequence in the acquisition of LD questions, which would reflect a semantic shift from scope marking structures where the scope marker is an argument of the matrix predicate and whose restrictor is a questioned proposition, to scope marking structures where the scope marker is not an argument but a Q-morpheme merged directly into an operator/A’-position and whose restrictor is a wh-phrase, yielding LD extraction at the last stage – that is, LD questions without a scope marker. (39) a.

indirect dependency stage Local covert movement in both CP1 & CP2 Overt syntax [Tu penses quoii ] [CPi lala elle aime bien quoij ] you think what Lala she likes well what LF [quoii tu penses ti ] [CPi quoij lala elle aime bien tj ] ‘What do you think that Lala likes?’ Local overt movement in both CP1 & CP2 Overt syntax & LF [qui ’est-ce que tu crois ti ] [CPi quj ’est-ce j’aime lire tj ] what-is-it-that you believe what-is-it I-like read ‘What do you believe that I like to read?’

b. direct dependency stage Partial wh-movement [CP Qi [IP Tu penses [CP quoii [C◦ que [IP je lis ti ]]]]] you think what C◦ I read ‘What do you think (that) I am reading?’ c.

overt long movement stage [CP Quii [IP tu penses [CP t”i [C◦ qui [IP t’i est caché ti dans le who you think c◦ is hidden in the sac]]]]] bag ‘Who do you think is hidden in the bag?’

Now, recall that for Dayal (2000), languages differ with respect to the syntax of indirect dependency, which can involve either juxtaposition of two clauses, indirect syntactic

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.23 (1503-1558)

Evidence from L1 acquisition for the syntax of wh-scope marking in French

subordination, or direct subordination (Section 6.1). She further argues that variation in the syntax of scope marking, from juxtaposition to genuine subordination, reflects diachronic stages in the process of language change. Thus, for instance, Reis (2000) argues that the diachronic evolution of partial movement scope marking structures in German involves a syntactic shift from juxtaposition to genuine subordination that can be correlated with a semantic shift from indirect to direct dependency. Abdulkarim and Roeper (2003) argue for a sequence in the acquisition of subordination moving from adjunction, to VP-complement, to subcategorized V◦ complement. Their proposal raises the question of whether the acquisition sequence in (43), which reflects a semantic shift from scope marking structures where the scope marker is referential/an argument of the matrix verb (indirect dependency) to scope marking structures where the scope marker is a non-referential/a Q morpheme (direct dependency), can be correlated (and, if so, to what extent) with the acquisition of subordination. The fact that variation in the syntax of scope marking in L1 French correlates with the acquisition of complementation would be the null hypothesis. This question, which we hope to answer in the future, is a matter of empirical investigation requiring tests to determine the syntactic status (juxtaposition/adjunction vs. true subordination) of the “complement” clause in LD questions in the child grammar of French.

Notes * The work reported here is part of larger collaboration with Celia Jakubowicz (CNRS, UMR8581) on the syntax of wh-questions in L1 French. We are very grateful to Celia for invaluable and generous feedback concerning every aspect of this research, theoretical, methodological or experimental. We thank Lamya Abdulkarim, Tom Roeper, Nelleke Strik and an anonymous reviewer for very helpful comments and questions. Thanks also to Valérie Gautier, Nicolas Guilliot and Frédéric Martin. This research was funded by the Laboratoire de Linguistique de Nantes (LLING, EA 3827) and is part of the Program #4 (Clausal Architecture) of the CNRS “Fédération Typologie et Universaux Linguistiques”. . The results of the preliminary acoustic studies we have carried out do not confirm the strict correlation established by Cheng and Rooryck (2000) between in-situ and yes-no questions in the adult grammar. Indeed, our results suggest that there are different intonational patterns associated with wh-in-situ, as argued by Baunaz (2004) who distinguishes, alongside the raising intonation (also characteristic of yes-no questions, see (4)), two other intonational patterns for wh-in-situ. Whether Cheng and Rooryck’s correlation proves to be correct or not (at least as stated), the proposal that intonation licenses wh-in-situ in French remains an intuitively correct and attractive idea. . The symbol # indicates a phonological pause. . An anonymous reviewer asks if, assuming Cheng’s and Rooryck analysis, one might not expect there to be a non lexical Q-morpheme in English, since this language has yes no questions without an overt Q◦ morpheme. This is not the case since, for Cheng and Rooryck, English yes no questions further differ from French yes no questions in that the former instantiate Aux to



JB[v.20020404] Prn:5/06/2006; 13:01



F: LALD4113.tex / p.24 (1558-1640)

Magda Oiry and Hamida Demirdache Comp – which serves precisely to overtly type the matrix C◦ as interrogative. In contrast, French has yes no questions without either raising of I◦ to C◦ or an overt particle, to type the matrix as interrogative (see (4a)). Note that none of the (roughly) 300 – root and LD – questions elicited instantiated Aux to Comp. . An anonymous reviewer points out that UG constrained child language variation from the target language in the syntax of wh-questions (be it in English or French) is at odds with the idea that parameters are set at a very early stage and asks what would distinguish the parameter for wh-movement from other parameters. We offer two comments in reply to this question. First, under multiple grammar models of language development (e.g. the Intermediate State Default Grammars Hypothesis, see Roeper 1999; Abdulkarim & Roeper 2003), we would expect UG constrained variation from the target language to extend well beyond the acquisition of whquestions. Second, as far as French is concerned, it is not clear that the two wh-scope marking strategies that we identify in French L1 acquisition in this paper – that is, direct dependency (partial wh-movement; Section 5) and indirect dependency (Section 6) are actually confined to the child grammar. First, take the indirect dependency strategy. Dayal (2000) argues that this strategy is in fact universally available in all languages for forming LD wh-questions – since all languages have sequential questions (see Section 7 for discussion). Turning next to the direct dependency strategy. To our surprise, we elicited 3 LD questions involving a (clefted) medial wh-phrase from 3 out of 12 adults in the control group. Note that syntax of partial movement in (i)–(iii) parallels in all relevant respects (i.e. null scope marker, medial focused/clefted wh-phrase) the syntax of partial movement in Bahasa Indonesia and Kitharaka, illustrated in (17)–(18) in the text. qui qui joue du tambour ? Tu penses que c’est you think C◦ dem-is who C◦ play drums ‘Who do you think is playing drums?’ (ii) Tommy, tu crois que c’est qui qui saute ? Tommy you think C◦ dem-is who C◦ jumps ‘Tommy, who do you think is jumping?’ quel animal qui saute ? (iii) Tommy, tu crois qu’c’est Tommy you think C◦ -dem-is which animal C◦ jumps ‘Tommy, which animal do you think is jumping?’

(i)

Now, although clefted medial wh-questions have never been reported (to our knowledge) in the literature on French wh-questions and are certainly not part of the grammar of standard French, the question in (i)–(iii) sound perfectly fine to our ears and to those of others speakers we have consulted. These findings suggest that partial wh-movement is in fact a licit (albeit marginal?) strategy in colloquial French. See Demirdache and Oiry (in progress) for a discussion of the implications of these data for both the adult and the child grammar of LD questions in French. . This correlation can be extended to variation in the syntax of scope marking in L2 acquisition of LD questions. The appearance of an overt scope marker in partial movement questions (see (8)) in L3 acquisition of English by bilingual Basque/Spanish learners correlates with the unavailability of the in-situ strategy in either the L1 or the L2 grammar of these learners. In contrast, the occurrence of a null scope marker in L2 acquisition of English by Japanese learners (see (19)) correlates with the availability of the in-situ strategy in the L1 grammar of these learners. . Note that both Mathieu’s account of the locality constraints governing partial movement and French wh-in-situ, and Fanselow and Mahajan’s account, which as we have suggested here can be extended to cover wh-in-situ in French, would have to be parametrized in order to explain why

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.25 (1640-1704)


locality constraints can be relaxed for those speakers of French who allow single wh-questions with wh-in-situ in finite complement clauses (see Section 4.1). Note, however, that even for these speakers, the distribution of wh-in-situ is not as free as it is in ‘true’ wh-in-situ languages (e.g. Chinese). For instance, wh-in-situ is in French is not allowed in an indirect single wh-question or a strong island. Moreover, further embedding the finite complement clause containing whin-situ yields a question whose status ranges from degraded to uninterpretable as a non-echo question. Note further that, if Baunaz (2004) is correct, then in dialects of French, which allows wh-in-situ in non-root finite contexts, there are different locality effects correlated with the interpretation of wh-in-situ (whether it is +/–specific and +/–presuppositional). . Dayal nicely supports the claim that sequential questions are wh-scope marking structures by arguing that they are subjects to constraints characteristic of scope marking. For instance, negation cannot occur in the first question of a sequence of questions, as the contrast in (i)–(ii) illustrates. (i) What do you think? Who is coming? (ii) *What don’t you think? Who is coming? Recall that negation in the matrix clause is illicit with scope marking strategies such as partial wh-movement in German or wh-in-situ in French (see Section 5 for discussion). . Note that we found no errors regarding the que/qui alternation in all elicited LD questions involving overt long extraction of a wh-phrase (illustrated in (10)): the complementizer qui appeared systematically and only when a subject was LD extracted. . In particular, in the youngest age group (3;2,3 to 3,5), the category object in-situ received only 20% of the children’s responses, as compared with 70% for fronted questions (including 30% for KESK questions). . Recall (from Section 4.1) that there are two dialects when it comes to wh-in-situ in French: for some speakers LD wh-in-situ is restricted to root and infinitive complement clauses, for others, wh-in-situ is allowed in finite complement clauses. In this context, it is not clear how to interpret the paucity of LD questions elicited. We can take this finding either, to be all the more surprising since LD wh-in-situ is indeed an option available in certain grammars of French, or we can merely take it to reflect dialectal variation in the status of LD wh-in-situ. Interestingly, the percentage of LD in-situ responses is higher in the adult control group: 38% (20/53). 5 out of 10 adults were responsible for these in-situ responses.

References Abdulkarim, L. (2001). Complex WH Questions and Universal Grammars: New evidence From the Acquisition of Negative Barriers. Unpublished Ph.D. Thesis. University of Massachussets, Amherst. Abdulkarim, L. & Roeper, T. (2003). “From universal to language-specific grammars: How do children acquire embedded Yes/No questions in English?” Manuscript. University of Massachusets, Amherst. Aoun, J. & Li, Y. A. (1981). “The interaction of operators.” In R. Freiden (Ed.), Principles and Parameters in Comparative Grammar (pp. 163–181). MIT Press. Cambridge. Baunaz, L. (2004). “Un NPs and Wh in situ: An argument for an indefinite analysis”. In Generative Grammar in Geneva, 4.



JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.26 (1704-1818)

 Magda Oiry and Hamida Demirdache

Beck, S. (1996). “Quantified structures as barriers for LF movement”. Natural Language Semantics, 4, 1–56. Beck, S. & Berman, S. (2000). “Wh-scope marking: Direct vs. indirect dependency”. In U. Lutz, G. Müller, & A. von Stechow (Eds.), WH-Scope Marking (pp. 17–44). Amsterdam: John Benjamins. Boeckx, C. (1999a). “Decomposing French questions.” In J. Alexander, W. R. Han, & M. Minnick Fox (Eds.), University of Pennsylvania Working Papers in Linguistics 6.1, Proceedings of the 23rd Annual Penn Linguistics Colloquium (pp. 69–80). Boeckx, C. (1999b). “Decomposing French questions”. Manuscript. University of Pennsylvania. Boskovic, Z. (2000). “Sometimes in [Spec CP], sometimes in-situ”. In R. Martin, D. Michaels, & J. Uriagereka (Eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik (pp. 53–88). Cambridge, MA: MIT Press. Brandner, E. (2000). “Scope marking and clausal typing”. In U. Lutz, G. Müller, & A. Von Stechow (Eds.), WH-Scope Marking (pp. 45–76). Amsterdam: John Benjamins. Chaussy, F. (2002). Acquisition des Questions Racines. Master Thesis. University of Nantes. Cheng, L. L.-S. (1997. On The Typology of Wh-Questions. New York: Garland. Cheng, L. L.-S. & Rooryck, J. (2000). “Licensing Wh-in-situ”. Syntax, 3, 1–19. Chomsky, N. (1998). “Minimalist inquiries: The framework”. Manuscript. MIT. Cole, P. & Hermon, G. (1994). “Is there LF movement?” Linguistic Inquiry, 25(2), 239–262. MIT. Crain, S. & Thornton, R. (1998). Investigations in Universal Grammar. A Guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge, MA: MIT Press. Dayal, V. (1996). Locality in Wh-quantification: Questions and Relative Clauses in Hindi. Dordrecht: Kluwer. Dayal, V. (2000). “Scope marking: Cross-linguistic variation in indirect dependency”. In U. Lutz, G. Müller, & A. Von Stechow (Eds.), WH-Scope Marking (pp. 157–193). Amsterdam: John Benjamins. Demirdache, H. & Oiry, M. (in progress). “Wh-scope marking in child and adult French”. University of Nantes. Deprez, V. (1995). “Sous-spécification, projections fonctionnelles et fixation des paramètres”. Recherches Linguistiques de Vincennes, 24, 177–207. Presses Universitaires de Vincennes. Deprez, V. & Pierce, A. (1995). “Negation and functional projections in early grammar”. Linguistic Inquiry, 24, 26–67. Fanselow, G. (to appear). “Partial movement”. In M. Everaert & H. Van Riemsdijk (in prep.), SYNCOM (the Syntax Companion): An Electronic Encyclopedia of Syntactic Case Studies. The LingComp Foundation. Fanselow, G. & Mahajan, A. (2000). “Towards a minimalist theory of wh-expletives, wh-copying and successive cyclicity.” In U. Lutz, G. Müller, & A. Von Stechow (Eds.), WH-Scope Marking (pp. 195–230). Amsterdam: John Benjamins. Grévisse, M. (1980). Le bon usage. Gembloux: Duculot. Gutierrez, M. J. (in preparation). The Acquisition of English LD Wh-questions by Basque/Spanish Bilingual Subjects in a School Context. Ph.D. dissertation. University of the Basque Country. Hulk, A. & Zuckerman, S. (2000). “The interaction between input and economy: Acquiring optionality in French wh-questions”. Proceedings of the Boston University Child Language Development Conference, 24, 438–449. Jakubowicz, C. (2004). “Is movement costly?”. Proceedings of the Journées d’études linguistiques. University of Nantes. Mathieu, E. (1999). “French wh-in-situ and the intervention effect”. University College London Working Papers in Linguistics, 11, 441–472.

JB[v.20020404] Prn:5/06/2006; 13:01

F: LALD4113.tex / p.27 (1818-1909)


Mathieu, E. (2002). The Syntax of Non-Canonical Quantification: A Comparative Study. Ph.D. Dissertation. University College London. McDaniel, D. (1989). “Partial wh-movement”. Natural Language Theory, 7, 565–604. Muriungi, P. K. (2004). “Wh-movement in Kitharaka as focus movement”. Proceedings of the Workshop on the Syntax, Semantics and Pragmatics of Questions. European Summer School in Logic, Language and Information 16. Nishigauchi, T. (1990). Quantification in the Theory of Grammar. Dordrecht: Kluwer. Oiry, M. (2002). Acquisition des Questions à Longue Distance. Master Thesis. University of Nantes. Pesetsky, D. (1987). “Wh-in-situ: Movement and unselective binding.” In E. Reuland & A. ter Meulen (Eds.), The Representation of (In)definiteness. Cambridge, MA: MIT Press. Reinhardt, T. (1997). “Quantifier scope: How labor is divided between QR and choice functions”. Linguistics and Philosophy, 20, 335–397. Reis, M. (2000). “On the parenthetical features of German Was. . . W-constructions and how to account for them”. In U. Lutz, G. Müller, & A. Von Stechow (Eds.), WH-Scope Marking (pp. 359–407). Amsterdam: John Benjamins. Riemsdijk, H. van (1982). “Correspondence effects and the Empty Category Principle”. Tilburg Papers in Language and Litterature. Rizzi, L. (1991). “Argument/adjunct (a)symmetries”. Manuscript. University of Genève. Rizzi, L. (1996). Parameters and Functional Heads: Essays on Comparative Syntax. Oxford: Oxford University Press. Roeper, T. (1999). “Universal bilingualism”. Bilingualism: Language and Cognition, 2, 169–186. Saddy, D. (1991). “Wh scope mechanisms in Bahasa Indonesia”. MIT Working Papers in Linguistics, 15, 183–218. Soares, C. (2003). “The C-domain and the acquisition of European Portuguese: The case of wh-questions”. Probus, 15, 147–176. Starke, M. (2001). Move Dissolves into Merge: A Theory of Locality. Ph.D. Dissertation. University of Geneva. Strik, N. (2003). Où tu as caché ton sac? Qu’est-ce que tu penses que je lis? Acquisition des Questions Wh- chez les Enfants Francophones de 3 à 6 Ans. DEA Thesis. University of Paris 8. Tellier, C. (1991). Licensing Theory and French Parasitic Gaps. Dordrecht: Kluwer. Thornton, R. (1990). Adventures in Long-Distance Moving: The Acquisition of Complex WhQuestions. Ph.D. dissertation. University of Connecticut. Wakabayashi, S. & Okawara, I. (2003). “Japanese learners errors on long-distance wh-questions”. In S. Wakabayashi (Ed.), Generative Approaches to the Acquisition of English by Native Speakers of Japanese. Berlin: Mouton de Gruyter. Zuckerman, S. (2001). “The acquisition of “optional” movement”. Manuscript. University of Groningen.



JB[v.20020404] Prn:3/04/2006; 14:25

 

Syntax/discourse interface

F: LALD41P4.tex / p.1 (47-74)

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.1 (46-122)

Acquisition of focus marking in European Portuguese Evidence for a unified approach to focus João Costa* and Kriszta Szendröi** *Universidade Nova de Lisboa / ** Universiteit Utrecht

This paper investigates whether children make the distinction between syntactic marking and prosodic marking of focus. The acquisition of focus marking provides a good testing ground to compare parametric and interface views on focus. If focus marking is parametric, children acquiring a language with syntactic focus marking should fail to comprehend marked word orders. If a language marks focus only prosodically, they should initially fail to comprehend prosodic markedness. Crucially, in Portuguese, in which both strategies co-exist, children are expected to fail on both. The results show that children have problems with the prosodic strategy only, favoring an interface approach to focus.

.

Introduction

Recent literature on the syntax-discourse interface in Romance languages indicates that the word order variation found in these languages correlates with the position of nuclear stress. For instance, in Romance languages allowing subject-verb inversion, it has been claimed that the subject is clause-final, because this is the position where nuclear stress falls (Zubizarreta 1998; Costa 1998; Ordóñez 1997; among others). The nature of focus marking across languages is subject to a lively debate. According to some authors, focus-marking is subject to parametric variation, in the sense that some languages mark focus syntactically, while others do it prosodically (Horvath 1986; Rizzi 1997). Under this view, if a language marks focus syntactically, as many Romance languages do, additional prosodic effects are coincidental. In other words, syntactic marking is sufficient for encoding focus information, and prosody plays no role. According to other authors, focus is universally marked by prosodic prominence. In case syntactic effects are associated with focus, they are a direct consequence of the prosodic organization of the language (Reinhart 1995; Szendröi 2001).

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.2 (122-164)

 João Costa and Kriszta Szendröi

The goal of this paper is to investigate whether children make the distinction between syntactic marking and prosodic marking of focus. Since there is variation in focus-marking, children have to find out which strategy is used in their language, and, therefore, in the course of development, it is expected that they will make mistakes both in the production and comprehension of focus. In fact, initial results confirm the problematic status of focus-marking in different languages (Crain et al. 1994). There is reason to believe that the characteristic properties of the acquisition of focus may shed light on the theoretical debate with respect to the parametric nature of focusmarking. If focus marking was truly parametric, we would expect that children who acquire language with syntactic focus marking fail to comprehend marked word orders. In languages where focus is marked only prosodically, the expectation is that they initially fail to comprehend cases of prosodic markedness. Crucially, in a language such as European Portuguese, in which both strategies co-exist in some constructions, children are expected to fail on both. Therefore, children’s mastery of word order and stress in focus contexts provides a good testing ground for the comparison between the two approaches to focus. The paper is organized as follows: in Section 2, we present the basic differences between two views on focus marking, the parametric view (e.g. Horvath 1986)1 and the interface theory (Reinhart 1995, to appear), spelling out some of the predictions made by each of them for a language like European Portuguese, in which, in some specific constructions, stress-shift and word-order changes co-exist as strategies for marking focus. In Section 3, we explain how research on language acquisition may contribute to shed some light on the debate on focus. In Section 4, the results of an experiment on the comprehension of focus marking strategies by Portuguese children are presented. Section 5 presents the main conclusions of the experiment.

. The debate on the nature of focus There is no consensus in the literature regarding the nature of focus marking in languages. According to some authors, as in Horvath (1986) and Rizzi (1997), focus is to be understood as a syntactic feature that may or may not correspond to a functional category, depending on the analysis. According to this view, languages in which there is syntactic movement for focus reasons, this type of movement occurs for checking or identification of a focus feature. If a language does not exhibit this type of movement, this is due to a parametric split between languages related to the grammar component in which focus is marked. This type of parametric approach is made explicit in Horvath (1986), who formulates it as follows: (1) Focus parameter: 1. nature of the feature: (i) freely occurring, i.e. not transferred from another category (English) (ii) assigned by a specific X0 category (Hungarian)

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.3 (164-219)

Acquisition of focus marking in European Portuguese 

2. which X0 functional category of the clause is the assigner, i.e. the source of the feature 3. whether the feature-assigning category needs to be lexicalized 4. the mode/nature of the process of feature-assignment: (i) feature transfer (ii) Spec-head agreement According to this type of formulation, a difference between Hungarian and English is whether focus is marked syntactically or prosodically. In what follows, we will refer to this view on focus as the parametric view. The assumption that languages split in the nature of focus marking (syntactic vs. prosodic) leads the parametric view to make some problematic predictions. If a language marks focus syntactically, additional prosodic effects are coincidental. This is problematic, since it is known that languages may resort to both prosodic and syntactic strategies or to prosodic strategies alone for marking focus; but, typologically, there appears to be no language that resorts to syntax alone, as mentioned in Givón (1990). In order to illustrate this problem, we can mention the case of Hungarian. According to Szendröi (2001), the target of focus movement is the place where nuclear sentence stress falls in Hungarian. According to the parametric view, this behaviour is coincidental and, therefore, left unexplained. It is also known that some languages resort to both prosodic and syntactic strategies in different syntactic environments. Consider the case of focalized possessives in European Portuguese (Castro & Costa 2003). Cardinaletti (1998) shows that postnominal possessives in Italian are used in definite contexts, only if they are focussed. In Portuguese, unlike in Italian, a focalized possessive is not post-nominal. Rather, the focussed possessive in pre-nominal position receives a marked stress. This difference between the two languages is illustrated in (2) and (3), where capital letters signal prosodic prominence: (2) Italian a. il mio libro the my book b. il libro MIO the book my “my book” (3) European Portuguese a. o meu livro the my book b. *o livro MEU the book my b’. o MEU livro the my book “my book”

(non-focused possessive) (focused possessive)

(non-focused possessive)

(focused possessive)

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.4 (219-281)

 João Costa and Kriszta Szendröi

The case of possessives in European Portuguese is an instance of stress-shift in a language that typically marks focus syntactically, as most Romance languages do. If the type of focus marking were a consequence of a parametric split between languages, it should not vary with the type of construction at stake within the same language.2 The parametric view contrasts with the Interface Theory, as proposed in Reinhart (1995, to appear). According to this author, languages always mark focus prosodically. Under this view, syntactic movement does not operate for the sake of licensing a syntactic focus feature, but rather in order to place a focused constituent in the position where nuclear sentence stress falls. If, for some reason, the syntax is unable to perform this, stress shift is required. Accordingly, stress shift is a costly operation requiring reference set computation at the interface (Reinhart 1995, to appear). The basic idea of this approach is that stress-shift is an uneconomical operation requiring that one determines whether there is no syntactic derivation satisfying the same purposes without involving the superfluous operation of stress-shift. Consider the following example, taken from Reinhart (1999). Let us start by defining focus set of IP as consisting of the constituents containing the main stress of IP. Thus, for a sentence like (4) uttered with neutral stress (assigned independently of focus), the focus set is the one indicated below: (4) [IP My neighbour is building a DESK] Focus set: {IP, VP, Object} Now, suppose that, for some reason, one wants to focus the verb alone. In that case, (4) is not good enough, since the verb alone is not part of the focus set. Stress-shift must apply, as shown in (5). (5) [IP My neighbour is BUILDING a desk] Focus set: {IP, VP, V} Since part of the focus set of (5) coincides with the focus set of (4), it must be computed whether there is no alternative without stress shift for the specific focus purposes. It is clear that reference-set computation is involved, since otherwise (4) and (5) would be equally good in IP-focus and VP-focus contexts, contrary to fact. Costa (2004) and Szendröi (2004) independently argue that the apparent stressdriven movement in the syntax is an operation available for free, hence not imposing computational cost. In order to illustrate this claim, consider the case of subject-verb inversion in European Portuguese. As in other Romance languages, the subject may occur inverted if it is focused. Its position will then coincide with the position where sentence nuclear stress falls. Costa (2004) argues, based on locality effects, that this position for the subject is made available in the syntax whenever Case can be licensed under Agree. In other words, its position is not dependent on interpretation matters, but rather on different strategies for syntactic licensing. The interface theory of focus crucially predicts that both syntactic and prosodic strategies may co-exist in the same language. In fact, this is the case of European Portuguese in certain constructions. Apart from the case of possessives illustrated above,

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.5 (281-339)

Acquisition of focus marking in European Portuguese 

it can be shown that in ditransitive constructions narrow focus on the object may be obtained either by stressing it in-situ, shifting the stress from its normal clause-final position (6b), or by placing the direct object in clause-final position (6c): (6) a.

O João só atirou a cadeira the João only threw the chair (Focus on VP or on Indirect Object) ao Pedro. to the Pedro “João only threw the chair to Pedro” b. O João só atirou A CADEIRA the João only threw the chair (Focus on Direct Object) ao Pedro. to the Pedro “It was only the chair that João threw to Pedro” c. O João só atirou ao Pedro the João only threw to the Pedro (Focus on Direct Object, neutral) a cadeira. the chair “It was only the chair that João threw to Pedro”

According to the Interface Theory, only (6b) involves reference set computation, since its focus set includes the VP, and requires comparison with (6c). Therefore, (6b) is ruled out at the interface with a VP-focus interpretation, and is computationally more costly. This difference will become relevant shortly for testing children’s behaviour to the two different types of focus.3

. Contribution from acquisition for the debate on focus In what follows, we intend to investigate whether children make the distinction between syntactic marking and prosodic marking of focus. Since there is crosslinguistic variation in focus-marking, children have to find out which strategy is used in their language, and, therefore, in the course of development, it is expected that they will make mistakes both in the production and comprehension of focus. In fact, initial results confirm the problematic status of focus-marking in different languages (Crain et al. 1994). There is reason to believe that the characteristic properties of the acquisition of focus may shed light on the theoretical debate with respect to the parametric nature of focus-marking. If focus marking was truly parametric, we would expect that children who acquire language with syntactic focus marking fail to comprehend marked word orders. In languages where focus is marked only prosodically, the expectation is that they initially fail to comprehend cases of prosodic markedness. Crucially, in a language such as Portuguese, in which both strategies co-exist, children are expected to fail on both. Therefore, children’s mastery of word order and stress in

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.6 (339-400)

 João Costa and Kriszta Szendröi

focus contexts provides a good testing ground for the comparison between the two approaches to focus. An experiment run for Dutch by Szendröi (2003) reveals that children fail to interpret stress-shift in ditransitive constructions as a strategy for assigning narrow focus to a direct object. Instead, to a Dutch sentence patterning like (6b), they assign a VPfocus reading. This shows that stress shift is problematic in a language that uses it for marking focus. This behaviour is expected under the interface view on focus, since stress-shift requires reference set computation at the interface and additional working capacities. The assignment of VP-focus interpretation straightforwardly follows from the Subset Principle (Crain et al. 1994) and reveals that children fail to assign the relevant interpretation to stress-shift. Elaborating on these results, it becomes relevant to ask whether children also have problems in comprehending the effects of word order variation related to focus. The case of European Portuguese ditransitives is of particular interest, since it may contribute to compare the parametric view and the interface theory discussed in the previous section. Let us then consider the predictions made by each approach for acquisition. Under the interface theory, children are expected to exhibit different behaviours when exposed to syntactic or marked prosodic focus marking. Since, according to what we have demonstrated above, only stress shift requires reference set computation at the interface, stress-shift is the only strategy expected to be problematic. On the contrary, under the parametric view, since stress-shift is a strategy for marking focus parallel to the syntactic strategy, just reflecting different settings of the same parameter, children are expected to have problems with syntactic focus marking, since it is known from previous studies that they do have problems with interpreting stress-shift (cf. Szendröi 2003 for Dutch, and Crain et al. 1994 for English). Since, to our knowledge, children’s sensitivity to stress-shift as a strategy for focus-marking has not been tested for European Portuguese, it is not clear whether the results will match those obtained for English or Dutch, but it is clear that the parametric view predicts no contrast in behaviour, since the two strategies follow from one single setting for the focus-parameter. Bearing these predictions in mind, we will present, in the next section, the results of an experiment conducted with Portuguese children in order to test their behaviour under the two relevant conditions: stress-shift and word order change.

. Experiment on the comprehension of focus marking strategies . Methodology We conducted a truth-value judgement task with 21 normally developing children aged between 3 years 11 months and 5 years 9 months (average age 4 years 11 months) and 6 adult controls. Two subjects were excluded, since one did not understand the task

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.7 (400-459)

Acquisition of focus marking in European Portuguese 

and another one had a yes-bias. The task was modelled after Gualmini et al. (2002). It consisted of a story acted out with props, followed by a statement by a puppet. The child’s task is to determine whether the puppet’s statement was true or false. The conditions tested were: a.

Condition 1: Word order change – the aim of this condition was to test whether children have problems comprehending focus-related word order change. b. Condition 2: Stress-shift – the aim of this condition was to determine whether Szendröi’s (2003) results for Dutch are reproduced in European Portuguese, that is, detect whether children fail to comprehend focus-related stress-shift. c. Condition 3: Control – the control condition was included in order to show that children do not have problems with unmarked stress and unmarked word order. The tested sentences all included the focus-sensitive operator ‘só’ (only) in a preverbal position. The reason for this is that otherwise changes in focus do not give rise to truth-conditional differences. Examples of the test items are given in (7): (7) Condition 1: a. O Tigre só deu ao Piglet o jogo. Tigger only gave to Piglet the game. “It was only the game that Tigger gave to Piglet” Condition 2: b. O Tigre só deu O JOGO ao Piglet. Tigger only gave the game to Piglet “It was only the game that Tigger gave to Piglet” Condition 3: c. Só caiu o Pooh. Only fell Pooh “Only Pooh fell” or “Pooh only fell” The context stories were designed in such a way that adult-like competence was reflected by a yes-answer. Six stories were used: one per condition and three fillers.

. Expected results Crain et al. (1994) show that, in the case of focus ambiguity, children, unlike adults, opt for the widest possible interpretation. According to the authors, this is due to the semantic subset principle. Since the widest interpretation entails the other possible interpretation, it would be impossible for the child to find out that the widest reading is possible on the basis of positive evidence, in case the narrowest is adopted. For this argumentation to become clearer, consider the following example:

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.8 (459-603)

 João Costa and Kriszta Szendröi

(8) O Tigre só deu o jogo ao Piglet. the Tigger only gave the game to Piglet. “Tigger only gave the game to Piglet” The focus set of (8) contains VP and the Indirect Object, as illustrated in the possible paraphrases in (9): (9) a. Giving the game to Piglet was the only thing that Tigger did. b. It was only to Piglet that Tigger gave the game. Crucially, (9a) entails (9b). Hence, the child can discover the (9b) interpretation by assuming (9a) as a possible interpretation, but she will not be able to do the reverse, without facing a learnability problem. The results obtained in Szendröi (2003) confirm this. Assuming this background and the predictions made by the interface theory, the expected results for each of the test items at stake are the following: –

S only V PPIO DPDO

Focus set: {DPDO } Word-order condition

The story presented makes the Direct Object reading true and the VP reading false, therefore the expected response, both for children and for adults is yes. In other words, since there is no ambiguity, and the interface theory predicts that syntactic focusmarking is unproblematic, it is expected that children interpret this sentence in an adult-like manner. –

S only V DPDO PPIO

Focus set: {DPDO , VP}

Stress-shift condition

The story makes the Direct Object and Indirect Object readings true and the VP reading false. If children, unlike adults, fail to comprehend the role played by stress-shift in marking focus, and if the Semantic Subset Principle is operative, it is expected that children’s and adults’ behaviours are different. Children, unlike adults, should have problems interpreting this sentence, assigning it a VP-focus interpretation, and interpreting it as false. Adults, on the contrary, should judge the sentence as true. –

Only VUnacc S Focus set: {Subj, IP}

Control condition

Since the story makes the subject reading true and the IP reading false, it is expected that children, unlike adults, fail to interpret it correctly, assigning it the IP reading. Accordingly, the expected responses are “no” for children and “yes” for adults.

. Results In the following Tables 1 and 2, the results of the experiment are presented. The results presented above enable us to draw the following general conclusions: a. Syntactic focus marking is unproblematic for children; b. Stress-shift is problematic for children.

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.9 (603-629)

Acquisition of focus marking in European Portuguese 

Table 1. Results for children

Word order condition (expected response: YES) Stress-shift condition (expected response: NO) Control condition (expected response: NO)

YES

NO

78,95% (15/19) 5,26% (1/19) 84,21% (16/19)

21,05% (4/19) 89,47% (17/19) 15,79% (3/19)

Table 2. Results for adults

Word order condition (expected response: YES) Stress-shift condition (expected response: YES) Control condition (expected response: YES)

YES

NO

100% (6/6) 100% (6/6) 66,6% (4/6)

0% (0/6) 0% (0/6) 33,3% (2/6)

. Discussion Focusing on a comparison between the results obtained for the word order condition and the stress shift condition, it is clear that only the latter is problematic for children. This result supports the interface theory over the parametric view. Recall that the parametric view predicts no contrast between the two conditions, while the interface approach predicts that children should have problems with stress shift only, since this is the only strategy requiring reference set computation. One problematic result is the one obtained for the control condition. The problem is that, since VS is ambiguous, children should assign it the IP-reading, contrary to fact, because of the Semantic Subset Principle. A possible way of explaining this unexpected behavior is to link it to the fact that, in unaccusative contexts, both VS and SV orders are grammatical with wide focus reading. If children analyse VS as a non-canonical order, there would be no ambiguity for them. The evidence for this potential explanation is mixed. On the one hand, adults response was only 66% YES in this condition, indicating the presence of the ambiguity, while children performed more uniformly (84%). However, production data indicate that a younger child treats unaccusative VS as a canonical order (Adragão 2001). An alternative explanation stems from the fact that although both an IP- and a S-focus reading is potentially available in the case of unaccusatives, nevertheless intervening factors block the ‘blind’ application of the Semantic Subset Principle (e.g. subject is often not part of the focus). In such a case, an alternative strategy could be invoked to resolve the ambiguity. One such strategy that has been proposed to apply in a similar ambiguity situation is that children try to be ‘charitable’: i.e. facing two possible interpretations, they adopt the one that makes the utterance true. This would also explain why children adopt the S-focus interpretation in the case of unaccusatives. Further research is needed in order to clarify this issue.

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.10 (629-701)

 João Costa and Kriszta Szendröi

. Conclusions The experiment reported in this paper allowed for showing that Portuguese children do not treat different strategies for focus marking in the same way. It was shown that children have problems with comprehending stress shift, a result already achieved for other languages. Conversely, syntactic focus-marking is not problematic for children. These results permit comparing two approaches to focus: the interface theory, as outlined in Reinhart (1995), and parametric views, such as the one advocated in Horvath (1986). It was argued that the results obtained favour the former approach, since a parametric view predicts that children should not exhibit a contrast in responding to prosodic and syntactic focus marking. Therefore, children’s performance favours a unified approach to focus, according to which syntax and prosody are not independent, and work together for the purpose of focus marking. Moreover, the results obtained cast doubt on the prosodic bootstrapping hypothesis for the setting of parameters related to word order. This hypothesis crucially relies on the assumption that sensitivity to the location of prosodic prominence precedes sensitivity to word order variation and stress shift. In particular, it is fundamental to this hypothesis that children ‘ignore’ data with non-canonical stress placement. However, as we demonstrated, children are able to identify the interpretative consequence of word order variation, while they fail to do the same in the case of marked prosody.

Notes . Cf. Horvath (2000, 2004) for a different approach. . Cf. Costa and Figueiredo Silva (to appear) for further discussion. . Actually, under the interface theory (6b) should be ruled out, given the less costly alternative (6c). In Costa (2004), it is argued that the two co-exist for independent reasons due to binding effects in ditransitive constructions. This matter must be subject to further research.

References Adragão, M. M. (2001). Aquisição da inversão numa criança entre os dois e os três anos. Manuscript, Universidade Nova de Lisboa. Cardinaletti, A. (1998). “On the deficient/strong opposition in possessive systems”. In A. Alexiadou & C. Wider (Eds.), Possessors, Predicates, and Movement in the Determiner Phrase (pp. 17–53). Amsterdam: John Benjamins. Castro, A. & Costa, J. (2003). “Weak forms as X0 : Prenominal possessives and preverbal adverbs in Portuguese”. In A. T. Pérez-Leroux & Y. Roberge (Eds.), Romance Linguistics: Theory and Acquisition (pp. 95–110). Amsterdam: John Benjamins. Costa, J. (1998). Word Order Variation. A Constraint-based Approach. PhD dissertation, HIL/Leiden University.

JB[v.20020404] Prn:5/06/2006; 13:13

F: LALD4114.tex / p.11 (701-783)

Acquisition of focus marking in European Portuguese 

Costa, J. (2004). Subject Positions and Interfaces: The Case of European Portuguese. Berlin: Mouton de Gruyter. Costa, J. & Figueiredo Silva, M. C. (to appear). “On the (in)dependence relations between syntax and pragmatics”. In V. Molnar & S. Winkler (Eds.), Architecture of Focus. Berlin: Mouton de Gruyter. Crain, S., Ni, W., & Conway, L. (1994). “Learning, parsing and modularity”. In C. Clifton, L. Frazier, & K. Rayner (Eds.), Perspectives on Sentence Processing (pp. 443–467). Mahwah, NJ: Lawrence Erlbaum. Givón, T. (1990). Syntax. A Functional Typological Introduction. Vol. 2. Amsterdam: John Benjamins. Gualmini, A., Maciukaite, S., & Crain, S. (1992). “Children’s insensitivity to contrastive stress in sentences with only”. Proceedings of the 25th PLC, 87–110. University of Pennsylvania. Horvath, J. (1986). Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Horvath, J. (2000). “Interfaces vs. the computational system in the syntax of focus”. In H. Bennis, M. Everaert, & E. Reuland (Eds.), Interface Strategies (pp. 183–207). Amsterdam: Royal Netherlands Academy of Arts and Sciences. Horvath, J. (2004). Is “focus movement” driven by stress? Manuscript. Tel Aviv University. Ordóñez, F. (1997). Word Order and Clause Structure in Spanish and Other Romance Languages. PhD Dissertation, City University of New York. Reinhart, T. (1995). Interface Strategies. OTS Working Papers in Linguistics. Utrecht University. Reinhart, T. (1999). The Processing Cost of Reference-set Computation: Guess Patterns in Acquisition. OTS Working Papers in Linguistics. Utrecht University. Reinhart, T. (to appear). Interface Strategies. Cambridge, MA: MIT Press. Rizzi, L. (1997). “On the fine structure of the left-periphery”. In Haegeman L. (Ed.), Elements of Grammar (pp. 281–337). Dordrecht: Kluwer. Szendröi, K. (2001). Focus and the Syntax-Phonology Interface. PhD dissertation, University College London. Szendröi, K. (2003). “Narrow and wide focus interpretation in the acquisition of only sentences”. Paper presented at GALA 2003, Utrecht University. Szendröi, K. (2004). A Prosodic Approach to the Syntax of Focus – From a Theoretical and Experimental Perspective. Manuscript. Utrecht University. Zubizarreta, M. L. (1998). Focus, Prosody and Word Order. Cambridge, MA: MIT Press.

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.1 (46-115)

Subject pronouns in bilinguals Interference or maturation? Manuela Pinto Universiteit Utrecht

Longitudinal data from two Dutch-Italian bilinguals show that third person subject pronouns are not acquired before the age of 3;1 and that the usage of these pronouns does not always converge with the monolingual target. This paper accounts for this acquisitional delay by integrating Müller and Hulk (2001) and Serratrice et al.’s (2004) core insights on crosslinguistic influence in bilingual language acquisition with a global maturational view as proposed in Van Kampen (2004). As for the observed bilinguals’ difficulty in anaphora resolution, we speculate on the existence of an economy principle that might be part of the Language Acquisition Device and guide children’s choices in cases of ambiguity in language.

.

Introduction*

In adult language subjects can be realized in different forms varying from proper names and definite DPs to various sorts of pronominals. The set of possibilities is determined by the grammar of the relevant language. The felicity, that is, the correct usage of these nominal forms in specific contexts, is regulated by discourse-pragmatic rules. The acquisition of subjects involves thus development of competence on both grammar and pragmatics. Studies on the acquisition of subjects in child language have traditionally focused either on grammatical competence (Hyams 1986; Rizzi 1994) or on cognitive maturation and performance limitations (see among others Bloom 1990; Valian 1991). Recent work by Serratrice (in press), Serratrice and Sorace (2003) and Serratrice, Sorace and Paoli (2004) show that both approaches fail in giving a clear picture of this phenomenon. Besides grammar development and general physical and cognitive maturation, children have to acquire (discourse) pragmatic competence. The combination of these factors gives as a result the language in its adult form. The present study examines how bilingual children acquire subjects pronouns, more specifically, third person singular pronouns. Longitudinal data from two DutchItalian bilinguals show that the acquisition of pronominal subjects presents two pecu-

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.2 (115-173)

 Manuela Pinto

liar traits: (a) pronominals, particularly third person pronouns, are acquired late (not before the age of 3;1) and (b) the usage of pronouns does not always converge with the monolingual target. Similar findings are reported in Serratrice and Sorace (2003), Serratrice et al. (2004) and Serratrice (in press) for English-Italian bilinguals. This paper speculates on the possible reasons for this divergence and aims at proposing a principled explanation for the observed facts. Sections 2 and 3 contain a presentation of the relevant data found in the corpora of two Dutch-Italian bilinguals. These data partly confirm Serratrice et al.’s (2004) findings and call for an explanation. The observed divergence in bilingual pronominal subjects is analyzed in Section 4.1 and Section 4.2 according to two hypotheses that have been recently proposed: Müller and Hulk (2001) and Serratrice et al. (2004). In Section 4.3 the data are examined from a developmental perspective (van Kampen 2004) and compared to data from monolingual Dutch and monolingual Italian children. Finally in Section 4.4 we speculate on the existence of a learnability principle that might be responsible for the peculiar data found in bilingual corpora. The relevance of this research topic resides in the fact that the two language pairs involved, Dutch-Italian and English-Italian, are typologically quite different. Italian is a pro-drop language, allowing null subjects and postverbal subjects. According to Grimshaw and Samek-Lodovici (1998) the distribution of null subjects is determined by the information status of the topic: when the topic is known, the subject is omitted, when the topic shifts (due either to contrast or selection or reintroduction of a previously introduced topic) an overt pronoun is required. Dutch is a topic-drop and a verb second language: in finite sentences the verb raises to C0 , the second position and the topic of the sentence (subject, object, predicate, etc.) moves to the first position, the specifier of CP. From this position the topic can be dropped. Dutch has another peculiarity: although this language does not have a null pronoun, subject pronouns are distinguished in strong and weak. The weak forms appear to show up in those contexts where the topic is given. In addition, Dutch has the demonstrative die that alternates with third person weak pronouns. The third language involved in this study is English. English is neither pro drop nor topic drop and pronouns have no weaker counterpart. Morphosyntactic encoding of discourse-pragmatic aspects in this language is limited to a few specific constructions. The three languages involved in this study show clear typological differences both at the syntactic and at the discourse-pragmatic level. The analysis of the simultaneous acquisition of two of such languages may detect possible crosslinguistic influence.

. Method The empirical data discussed here come from two Dutch-Italian bilingual longitudinal corpora, Heleen’s corpus (Hulk & van der Linden) and Lapo’s corpus, collected by the author. Both children were born in the Netherlands to an Italian mother and a Dutch father and are brought up as bilinguals according to the one-parent-one-

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.3 (173-247)

Subject pronouns in bilinguals 

Table 1. Descriptive statistics for Heleen’s Italian data Age

MLU

N of child utterances

N of finite verbs

N of NS

%NS

1;9 1;10 2;0.1 2;0.16 2;1 2;2 2;4 2;6 2;7 2;9 2;11 3;1 3;2 3;3 4;0 4;1

1.79 1.57 2.04 2.51 1.81 3.04 3.03 2.45 3.77 2.25 5.39 4.04 3.04 4.28 3.20 3.65

206 179 183 180 204 242 199 188 436 307 169 159 262 321 304 303

15 20 12 10 19 61 25 18 159 52 76 60 100 211 114 132

7 17 3 5 11 13 17 12 86 35 31 24 52 72 56 67

46.6 85.0 25.0 50.0 57.8 21.3 68.0 66.6 54.0 67.3 40.7 40.0 52.0 34.1 49.1 50.7

Table 2. Descriptive statistics for Lapo’s Italian data Age

MLU

N of child utterances

N of finite verbs

N of NS

%NS

2;9 2;10 2;11 3;2 3;4 3;8 3;9

2.48 3.64 3.78 3.88 4.38 5.02 3.88

109 171 93 113 135 123 235

37 72 35 43 53 38 120

10 23 8 14 15 13 30

27.0 31.9 22.8 32.5 28.3 34.2 25.0

language model. Heleen’s available files cover the age period 1;9–4;1, although there is a gap between 3;3 and 4;0. Lapo’s available files cover the age period 2;9–3;9. Table 1 and 2 provide descriptive statistics of the age and MLU of the two children. Also included are the number of child utterances, the number of finite verbs, the number of null subjects (NS) and the percentage of null subjects for each age stage. The null subjects rate refers to all finite sentences without an overt subject, independently from the pragmatic felicity of the missing pronoun. All overt referential subject arguments in the corpora were counted and coded for morphosyntactic form: noun phrase, proper name, personal pronoun, demonstrative pronoun. Subjects of meteorological verbs and impersonal si constructions were excluded. Imperatives, repetitions, unintelligible utterances and songs were also excluded from the analysis. All overt subject pronouns were coded for person and for pragmatic felicity. Pragmatic felicity was determined on the base of the presence or absence of what Grimshaw and Samek-Lodovici (1998) call the topic-shift feature. As no video recording was available to us, we had to rely exclusively on the transcripts.

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.4 (247-316)

 Manuela Pinto

Table 3. Distribution of subject pronouns in Heleen’s Italian files Age

1S

2S

3S

1P

2P

3P

1;9 1;10 2;0.1 2;0.16 2;1 2;2 2;4 2;6 2;7 2;9 2;11 3;1 3;2 3;3 4;0 4;1

– – 2 – 1 18 3 – 18 5 21 1 9 19 13 11

– – – – – 3 3 1 8 – 3 4 3 – 4 2

1 – – – – – – 1 – 1 – 6 2 6 0 –

– – – – – – – – – – – – – 3 1 2

– – – – – – – – – – – – 2 1 – –

– – – – – – – – – – – – – 1 – 2

. Results The central issue of the present study is the observation that bilinguals appear to have problems with subject pronouns, particularly with the first and the third person. In Heleen’s files (age 1;9–4;1) the average rate of null subjects in Italian in finite sentences is around 55%. As observed for Italian monolinguals (Serratrice 2005), her early subjects are mainly demonstratives, proper names and DPs and NPs. The number of her first and second person pronouns increases around the age of 2;2 (Table 3). The first person subject pronoun io clearly outnumbers the second person, a common phenomenon in early language acquisition (Serratrice 2005). As long as the interpretation of these pronominals is contrastive (narrow focus), the child seems to use them correctly (but see observations below). However, in all the other cases io is used in contexts where no topic shift has occurred and is therefore pragmatically inadequate. The abundant use of io diminishes around the three years, when Heleen starts using more null subjects. Heleen’s first third subject pronoun is found at the age of 1;9 and is reported in (1). This is the very first pronominal the child uses. (1) Lei suona ‘She rings/plays’

(Heleen 1;9)

Utterance (1) follows the mother’s question quello che cos’è? ‘what’s that?’. From an adult point of view, (1) is not a felicitous answer. Apparently, the child is following her own line of thoughts and, indeed, there seems to be a little piano among the toys the child is playing with. Be that as it may, (1) seems to be just an accidental occurrence of

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.5 (316-406)

Subject pronouns in bilinguals 

a third person pronoun. Before the age of 3;1 only two more instances of third person pronouns are found: (2) Dottore, lui è malato ‘Doctor, he is ill’

(Heleen 2;6)

(3) E’ lei! ‘It’s she!’

(Heleen 2;9)

Interestingly, (2) and (3) have a narrow focus interpretation and are perfectly felicitous in the context of utterance. It seems that the pragmatic notion of narrow focus is available to children at an early stage. Still, they appear to encounter some difficulty in mapping this notion into a language specific structure. Consider the examples below: (4) Io fa tutto! Io so! I do everything I know ‘I do everything! I know!’ (5) *MOT: Una cartolina per me! E chi mi scrive? ‘A postcard for me! Who writes me?’ *HEL: Io ti scrivo! I to-you write ‘I write to you!’

(Heleen 2;1)

(Heleen 2;11)

In all these cases the child wants to put special emphasis on the subject (narrow focus). In Italian, narrow focus is encoded in syntax and it corresponds to the most embedded position in the sentence (Pinto 1997). It might thus be the case that this language specific construction is still too difficult for this age. In fact, postverbal subjects with wide focus are also hardly available before the age of 3;3. An alternative explanation for this word order problem might be in terms of crosslinguistic influence, by adducing the fact that narrow focus in Dutch has the SV word order, as in the examples above.1 Serratrice (2002) reports a similar phenomenon in the production of an English-Italian bilingual child and the same is found by Antelmi (1997) for monolinguals. However, rather than crosslinguistic influence, these word order errors might reveal a general problem of children in moving from universal discourse-pragmatic strategies to language specific rules. Heleen’s production of third person pronominals appears to increase around the age of 3;1. An examination of the pragmatic felicity of overt pronominals between 3;1 and 4;1 gives the following results in Table 4.2 Example (6) shows an instance of a pragmatic felicitous pronoun: (6) *MOT: Senti, mi aiuti a cercare i mattoncini verdi, Listen, pro me help-2ps to look for the bricks-dim green allora, per fare gli alberi? then to make the trees ‘Listen, will you then help me to find the little green bricks to make the trees?’

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.6 (406-454)

 Manuela Pinto

Table 4. Pragmatic felicity of Heleen’s overt pronominals (age 3;1–4;1). The number of tokens for each pronominal form is reported into brackets. Age

1st [+FEL]

2nd [+FEL]

3rd [+FEL]

3;1 3;2 3;3 4;0 4;1 Totals %

1 (1) 7 (8) 16 (19) 8 (13) 9 (11) 41 (52) 78.8

4 (4) 2 (2) 1 (1) 4 (4) 1 (1) 12 (13) 92.3

6 (6) 2 (2) 5 (6) 0 (0) 2 (2) 15 (16) 93.7

*HEL: Sì, sì, ti aiuto. Io ne ho già due! Yes yes pro you help-1ps I of.it have already two ‘Yes, yes, (I) will help you. I already have two of them!’ (Heleen 4;1) In (6) the pronouns io is perfect with a contrastive interpretation. Conversely, the pronoun in (7) is not pragmatically felicitous: allora conosci qualcuno che vorrebbe fare (7) *MOT: E And then pro know-2ps someone that would make compagnia al porcellino? company to.the pig-dim ‘And then, do you know someone that would like to stay with the little pig?’ *HEL: Eh certo, io conosco la pecora. Oh sure I know the sheep ‘Sure, I know the sheep’ (Heleen 3;3) In this latter sentence the pronoun io is not necessary as the topic has not changed. In sum, Table 4 shows that Heleen makes a few errors in the use of overt pronominals. Interestingly, Serratrice and Sorace (2003) and Serratrice et al. (2004) report similar findings in the corpus of an English-Italian bilingual. These authors observe that the child they studied uses a larger number of overt subjects than his Italian monolingual peers and that 9% of his overtly realized third person subjects are infelicitous. Note, however, that the errors found in Heleen’s corpus involve mainly the first subject pronoun. In Lapo’s corpus (2;9–3;9) the average rate of null subjects is lower than in Heleen’s corpus: 28,8%. Null subjects are mainly found with verbs in the first or in the third person singular form. Interestingly, the few cases in which the verb has the second singular form are almost all questions. (8) Viene dalla nonna pro come-3ps to.the grandma ‘He comes to see grandma’

(Lapo 2;10)

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.7 (454-547)

Subject pronouns in bilinguals 

(9) Voglio Pinocchio pro want-1ps Pinocchio ‘I want Pinocchio’ (10) Fai anche al Lapo? pro do-2ps also to.the Lapo ‘Are you going to do it to Lapo as well?’

(Lapo 2;10)

(Lapo 3;2)

The child makes a massive use of nouns (proper nouns and quasi proper nouns) and he uses his own name and a correctly inflected third person singular verb when referring to himself: this form outnumbers all other overt subjects found in the transcriptions and it is not pragmatically adequate as the topic does not change. (11) Biancaneve va ill’acqua Snowwhite goes to-the water ‘Snowwhite goes to the water’

(Lapo 2;9)

(Lapo 2;10) (12) *MOT: E così lo hai buttato indietro. And so it pro have-2ps thrown back ‘And in this way you have thrown it back’ *LAP: Sì e Lapo arriva e Lapo va dietro, Lapo lo cerca Yes and Lapo arrives and Lapo goes back, Lapo it looks for o no, ecco l’ho trovato. or no, here it pro have-1ps found ‘Yes and Lapo arrives and Lapo goes back, Lapo looks for it or not, here it is, (I) have found it’ (13) *MAN: Ascolta, non hai un altro giochino che giochiamo Listen, not pro have-2ps an other toy that pro play-1pp (Lapo 3;9) lì sul tappeto? there on the carpet ‘Listen, don’t (you) have another toy so that (we) can play there on the carpet?’ *LAP: No, Lapo vuole la televisione. No, Lapo wants the television ‘No, Lapo wants tv’ In the transcripts collected in the period the child was 2;9–3;9 we counted 398 finite clauses. 285 have an overt subject, yet only four contain a subject pronoun. In all these four cases the pronoun is the first person singular io. This is illustrated in Table 5. Table 5 clearly shows that Lapo’s use of overt pronominals is extremely low, even at the age of 3;9. The examples below report this child’s sentences containing an overt pronoun. (14) Io sono cattivo ‘I am naughty’

(Lapo 3;2)

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.8 (547-599)

 Manuela Pinto

Table 5. Distribution of subject pronouns in Lapo’s Italian files Age

1S

2S

3S

1P

2P

3P

2;9 2;10 2;11 3;2 3;4 3;8 3;9

– – – 1 2 – 1

– – – – – – –

– – – – – – –

– – – – – – –

– – – – – – –

– – – – – – –

(15) Anch’io voglio xxx con le rosentje also I want xxx with the roses-dim (Dutch word) ‘I want xxx with the small roses, too’

(Lapo 3;4)

(16) Io ne voglio ancora una I of it want again one ‘I want another one of it’

(Lapo 3;4)

(17) Io faccio così I do so ‘I do it in this way’

(Lapo 3;9)

The discourse-pragmatic felicity of these pronominals in the given context was also inspected. (15), (16) and (17) contain a shift in the topic of the sentence. The child uses an overt pronoun according to the Italian discourse-pragmatic rule. Conversely, in (14) no particular pragmatic reason justifies the use of an explicit pronoun. Unfortunately, Lapo’s data are not sufficient to confirm the tendency that emerged from the examination of Heleen’s data. Yet, we would like to explore further these interesting facts and speculate on a possible explanation. To which extent are these observations peculiar to bilinguals? Serratrice (2005) examined subject development in Italian monolingual children aged 1;7–3;3 and found that the bilinguals’ pattern for subject acquisition appears to be partly shared by monolinguals. So are monolinguals’ early overt third person subjects mainly demonstrative, definite DP’s and proper nouns and the control of the io subject is often problematic.3 Conversely, the pragmatic felicity of overt third person subjects appears to meet the adult norm. The acquisition of subject pronouns, although in different respects, seems hence to be a problem for both monolinguals and bilinguals. Third person subject pronouns are late in the acquisition schedule of both groups. Empirical evidence that is so far available indicates that in the case of bilinguals this may be even later. Summarizing, the amount of empirical data available for this research is too limited for a solid generalization. Nevertheless, the analysis of the files of two ItalianDutch children reveals two interesting facts: (a) subject pronouns, particularly third person singular, are acquired late and (b) the use of certain subject pronouns, particularly the first and the third person singular, is not always pragmatically adequate, that is, the occurrence of overt pronouns in contexts where the subject should be omitted

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.9 (599-675)

Subject pronouns in bilinguals 

seems to be above the monolinguals’ norm. These differences between monolinguals and bilinguals might be the result of crosslinguistic influence effects from one language onto the other. Alternatively, they might be seen as necessary steps of a maturation process that, in the case of bilinguals, requires additional learning devices. This issue will be discussed in the next sections.

. Discussion . Separated systems and yet influence The claim that bilinguals strictly separate the languages they learn (Meisel 1989 and others) encounters nowadays a large consensus. Yet, recent studies on bilingual first language acquisition have focused on crosslinguistic influence effects, particularly in those areas where the formal system interfaces with the system of use. In a study of the phenomenon of object omission in French-Dutch and Italian-German bilinguals, Müller and Hulk (2001) observed omission errors that they claimed reveal the influence of the Germanic language on the Romance one. Interestingly, this phenomenon seems to be localized in a specific area of grammar and it seems to appear at a specific stage of language development. The area of grammar involved is the syntax-discourse pragmatic interface, which is the locus where universal pragmatic strategies are encoded into language specific forms. Müller and Hulk claim that this mapping procedure takes place only when the C-domain becomes well established, that is when the child has completed the acquisition of syntactic structure. Before the C-system is in place, children will have a preference for general pragmatic strategies. Mapping induced influence effects show up, according to Müller and Hulk, when the following three conditions are complied with: (a) language A allows a potentially ambiguous structural analysis of a syntactic construction: the analysis can be in terms of formal structure or it can bear a universal pragmatic strategy; (b) the input in language B offers substantial positive evidence for one of these two analyses and (c) the structural analysis that both languages share is compatible with a universal pragmatic strategy. The bilingual child, who is naturally inclined to choose for the pragmatic option, will indeed linger in this phase. What we see is the influence of one language onto the other. This must not be seen as a failure of children in distinguishing the two language systems. It is rather the natural consequence of a general principle of economy that guides the child in choosing the option that works for both languages and that costs less computational effort. Müller and Hulk do not make explicit predictions about the direction of this influence; however, given their conditions for the occurrence of mapping induced influence and assuming the existence of a general economy principle, it logically follows that the child will select the analysis that comes closer to a universal pragmatic strategy. Let us return to Dutch-Italian bilinguals. Subject omission in Italian is regulated by discourse-pragmatics: subject pronouns are overtly realized only in case of topic-

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.10 (675-721)

 Manuela Pinto

shift. Dutch is a topic-drop language and, as such, it does show a few instances of subject drop.4 Yet, null subjects are not included in the array of possible nominals of Dutch grammar. Müller and Hulk’s analysis predicts that the phenomenon of subject omission with respect to the Dutch-Italian language pair might give rise to mapping induced influence. From a child viewpoint, language A (Dutch) allows a potentially ambiguous analysis of subjects as optionally null elements (topic drop, diary style), and the input in language B (Italian) contains substantial positive evidence for the analysis of subjects as null elements. In addition, subject drop is compatible with a universal pragmatic licensing strategy. The effect of mapping induced influence should result in a syntactic error: an overgeneralization of the Italian null subject in the Dutch grammar. In other words, the rate of subject omissions in Dutch-Italian bilinguals should be higher than in Dutch monolinguals, and bilinguals would rely for a longer time on a general pragmatic device that works for both languages. This prediction does not seem to be confirmed by the data that have been found so far. At 1;11 Heleen’s rate of null subjects in her Dutch files is around 40%. After the age of 2;3.22 the null subject rate remains below 4%. As for Lapo, his null subjects rate in Dutch reaches the highest rate of 10% at the age of 2;11. In the last transcript (age 3;5) no null subject is found. The strongest evidence for the claim that these children do not make syntactic errors comes from questions. It’s well known that subject pronouns are never overt in questions in Italian. Crucially, no subject omission was found in the questions in the Dutch data. A similar conclusion seems to hold for the English-Italian language pair. Serratrice et al. (2004) found no target deviant null subject rate in the English utterances of the bilingual child they studied. Therefore, the bilinguals examined here appear to know the syntactic differences between Dutch and Italian pronominal subjects. What has been observed, however, is that bilingual children seem to encounter problems in a later stage of language development. The errors they make in the Italian files involve syntax but only to the extent that certain pronominal forms are not available to them yet. In addition, they appear to make errors at a discourse pragmatic level, that is they use overt subjects in a target deviant way.

. Features opposition and markedness Serratrice et al.’s (2004) study provides a detailed analysis of the discourse-pragmatic felicity of preverbal subjects in the production of an Italian-English bilingual child aged 1;10–4;6. Recall that Müller and Hulk predicted that crosslinguistic influence occurs before the C-domain is in place, giving thus rise to errors of a syntactic nature. Serratrice et al. found errors also after the C-domain was settled. At this stage the language specific syntax is in place, as expected. The bilingual child knows which forms are grammatical in each of the two languages she is learning. Yet, the child seems to have some difficulties in assessing the pragmatic appropriateness of the (pro)nominals in Italian. Interestingly, this phenomenon clearly emerges when the child’s MLUw is greater than 4 (at about 4 years of age), that is when his monolingual peers show a target-like discourse pragmatic use of overt pronominals. Serratrice et al. suggest

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.11 (721-770)

Subject pronouns in bilinguals 

that these data might give evidence for the existence of another form of crosslinguistic influence, one occurring after syntax has been acquired and involving discoursepragmatic felicity. Like in the case of Müller and Hulk’s mapping induced influence, crosslinguistic influence at a discourse-pragmatic level might be the result of a general principle of economy. In bilingual acquisition the child is faced with the difficult task of keeping apart different language specific procedures for the mapping of pragmatic knowledge onto syntactic structure. It sounds plausible that the child might optionally select the cheapest solution, i.e. the pragmatically unconstrained option. As suggested by Serratrice et al., regular and simultaneous exposure to two languages might lead to the fading out of certain specific interface features. This would be the logical consequence of a choice based on economy considerations. In the case of the distribution of overt pronominal subjects, English has a default rule: always use overt pronominals. Conversely, Italian has a marked rule: use overt pronouns only in case of topic shift. On the ground of economy considerations, the bilingual child would select the default option, in this case the English rule. Repetitive application of this strategy would cause underspecification of distinctive pragmatic features in Italian. Serratrice et al. make an important prediction: no underspecification effects should emerge in the case of two typologically similar languages as, in this case, there is no cheaper option. In other words, pragmatic crosslinguistic influence is to be expected with language pairs that differ at the level of discourse pragmatics. In the case of the English-Italian language pair the features opposition hypothesis works well because of the [–marked] versus [+marked] opposition; the child selects the most economical option, the unmarked one. The other side of the prediction is that where this opposition does not exist, no crosslinguistic influence will occur. Accordingly, it would be reasonable to expect that Spanish-Italian bilinguals do not show any features opposition effects as overt subjects in Spanish and in Italian obey a similar discourse pragmatic rule. Interestingly, in the case of the Dutch-Italian language pair the opposition does not exist either, as both languages have a marked rule that requires computation of the same interface feature: [+/–topic shift]. Van Kampen (1997) observes that the Dutch pronoun die shows a striking parallel with the Italian null subject.5 If this analysis is correct, less or no discourse-pragmatic errors should be expected for Italian third person pronouns in Dutch-Italian bilingual corpora. Only Heleen’s corpus provided instances of third person pronouns and, as Table 4 shows, the use of these pronouns involves a very low percentage of pragmatic errors (less than 7%). On the contrary, first person pronouns provided almost 11% of pragmatic errors. Although these data seem to support our prediction for the Dutch-Italian language pair, more empirical evidence is needed to give a sound basis to this hypothesis. Summarizing, drawing on Serratrice et al.’s features opposition hypothesis, DutchItalian bilinguals are predicted to make less errors in the use of subjects than their English-Italian peers as both Dutch and Italian have the interface feature [+/–topic shift].

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.12 (770-819)

 Manuela Pinto

. Syntax, discourse and cognition In the two previous sections crosslinguistic influence in bilingual subject acquisition was extensively discussed. The predictions we made on the ground of Müller and Hulk (2001)’s analysis do not seem to be corroborated by the data collected so far. The data seem however to provide support for Serratrice et al. (2004)’s hypothesis of crosslinguistic influence at the discourse-pragmatic level. In the introduction we already advocated the explanatory advantage of a global, developmental-oriented approach to this topic. In the present section we take up that discussion again and, from this perspective, we try to explain the first fact we observed, i.e. why third person pronouns are acquired late by all children and, in particular, by bilinguals.

.. Syntax Null subjects are ubiquitous in early child language, independently of the pro-drop nature of the target language (see Hyams 1986 and many others). Adult null subjects have to be licensed and interpreted (Rizzi 1986). For child null subject sentences Rizzi (1994) proposes an analysis in terms of truncated structures with a null constant as a subject. As the null constant occupies the specifier of a truncated structure, it is exempted from identification by an antecedent. This is claimed to be a particular case of licensing. The interpretation of the null constant takes place by means of a null discourse operator. By the time the child acquires subject pronouns and verbal morphology, truncated structures disappear and null subjects must satisfy adult-like licensing and interpreting conditions. At this stage children set the pro-drop parameter according to the value of their target language. As observed by Serratrice et al. (2004), Rizzi’s analysis does not consider the fact that the acquisition of personal subject pronouns is phased (third person pronouns are acquired last) and that their interpretation involves different degrees of complexity (third person pronouns are more difficult to interpret). Van Kampen (1997, 2004) proposes a different analysis for Dutch and French early null subjects. She observes that the verbs children use in this first stage of language acquisition carry an implicit mode that appears to correlate with a specific personal form: the intentional mode for the first person, the imperative mode for the second person and the constative mode for the third person. These lexically fixed verbal forms enable the identification of the missing subject with a fixed person. The interpretation of these mode-implied subjects is situation-bound and gesture sustainable. The appealing trait of this analysis is that besides grammar development it also considers cognitive development and the acquisition of pragmatic competence. The world of a child at such an early age is limited to the relevant situation in which speech time and event time coincide and in which participants share the same knowledge with respect to physically present persons and objects. This analysis also explains why children use at first only the first and the second person subject pronoun and why the acquisition of the third person takes place only later. The first and second person pronouns are deictic elements that can be interpreted in the context of a situation. Third

JB[v.20020404] Prn:5/06/2006; 13:22

F: LALD4115.tex / p.13 (819-882)

Subject pronouns in bilinguals 

person pronouns, instead, involve reference to an antecedent that can be outside the current situation. Identification of the correct antecedent presupposes then sophisticated discourse pragmatic competence and cognitive and processing skills that become available to the child only at a later stage. Van Kampen’s theory reflects the gradual cognitive and psycholinguistic development of children. The early language development stage with mode-implied subjects and verbs functioning as predicate operators is followed by an I-marking stage. For Dutch children this stage is around the age of 2;6. I-marking involves the acquisition of copula, auxiliary, modal and finite morphology in predications. However, subjects are not yet marked for reference. The nominal forms that are available at this stage, mainly quasi proper names and demonstratives, must be interpreted in the context of the current situation. Van Kampen reports that before this stage Sarah uses demonstrative pronouns, but no third person (strong or weak) pronouns. These data confirm Haegeman’s (1995) observation that Dutch children at first make an almost unexceptional use of the situation-bound demonstrative die. Personal pronouns in Dutch appear when D-marking takes place, that is when the feature D (article, demonstrative, possessor and quantifier) appears before nouns. Feature D is the grammatical device of anaphoric D-marking and it reflects the anchoring of reference in discourse. Van Kampen’s analysis offers a theoretical framework, which is compatible with the global view of language acquisition advocated in the introduction. Several observations concerning monolingual and bilingual acquisition can be nicely accommodated into this model. Van Kampen’s analysis was initially based on data from Dutch and from French. A thorough study of Italian still has to be done, but there is some evidence that this might be the right way to follow. The point we want to make in this paper is that, in order to understand how subject acquisition takes place, it is necessary to examine the issue in its different aspects. A full mastering of subjects involves not only knowledge of the forms but also of their rules of usage. In this section we concentrated on formal properties and we showed that van Kampen’s approach might explain why third person pronouns are acquired late. In the next section we turn to discourse-pragmatics.

.. Discourse Grammar provides an array of different forms of denotation for subjects. This can be best represented as a continuum containing intermediate options depending on the nature of the relevant language (Prince 1981): nominals pronominals proper name . . .< . . . definite . . .< . . . demonstrative . . . .50 for all comparisons), which shows an increasing trend towards the native norm.

. Presentationally focused-subject contexts with unergative verbs In presentationally focused-subject contexts with unergative verbs (Figure 4), the main effect of word order is not significant (F(1,81) = .03, p = .86), the main effect of group was highly significant (F(3,81) = 4.78, p < .01) and the interaction between word order and group was also highly significant (F(3,81) = 6.59, p < .01). A visual inspection of Figure 4 crucially reveals that all learner groups prefer the pragmatically anomalous unergative #SV to the pragmatically correct unergative SV, though the difference is intuitively small. Within-group comparisons indeed confirm the intuition that none of the learner groups is able to distinguish between VS and #SV with unergatives in presentational contexts (p > .05 for each comparison), which entails that learners are optionally allowing both SV and #VS simultaneously, hence the non-significant main effect of word order shown previously. This result is in sharp contrast with the natives’ performance, as they show the opposite behaviour, i.e. they significantly (p < .01) prefer the pragmatically acceptable VS (mean = 1.25) to the pragmatically anomalous #SV (mean = .17). Once again we find a familiar pattern in the pragmatically infelicitous condition between the learners and the natives, i.e. while there is no sharp contrast between the

Mean acceptability rate

2 1,5

1,43

1,26

1 0,97

1,32 1,12

1,25 Unerg Foc VS Unerg Foc SV

0,93

0,5

0,17

0 –0,5 Gk Upp Int Gk Low adv Gk Upp adv

Natives

Figure 4. Unergative verbs in focused subject contexts (word order by group)

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.18 (1010-1080)

 Cristóbal Lozano

learner groups and the native groups in the pragmatically correct condition (VS), there is a visually marked contrast between the learners and the natives in the pragmatically odd condition (#VS), which is highly significant (p < .01 for each comparison).

. Discussion Recall from earlier sections that, while our main interest rests on focused contexts, we need to analyse neutral contexts for comparative purposes. In neutral contexts with unaccusatives (Figure 1), learners clearly preferred VS to #SV, as Spanish natives did, which could imply that learners are sensitive to the constraints operating at the lexicon-syntax interface from (at least) intermediate stages of development. Such sensitivity to the syntactic effects of the Unaccusative Hypothesis (UH) is also operative from early stages in the L2 Spanish of English-speaking natives (Hertel 2000, 2003). In the Hertel studies it is safely claimed that learners’ sensitivity to unaccusativity cannot be a direct consequence of their L1 since unaccusativity is not manifested overtly in English, hence it must be the consequence of universal constraints (i.e. UH) operating at the lexicon-syntax interface. It is difficult to decide the source of such sensitivity in our study as it could be due to two factors: either (i) our learners are sensitive to the universal constraints of the UH, or (ii) they are simply transferring from their L1 Greek into their L2 Spanish, since the surface syntactic manifestation of unaccusativity is identical in Greek and Spanish. While it has been argued that such sensitivity derives from universal constraints in the case of Greek learners of Spanish (at least in advanced stages, Lozano 2006), the debate is not crucial for the current purposes of this study, since the rationale is to use word order in neutral contexts simply as a comparative tool, as will be clear later. The results for unergatives in neutral contexts (Figure 2) revealed that Greek learners of Spanish show a clear preference for SV to #VS, similar to the Spanish natives’ patterns. Once again, this sensitivity from intermediate stages can be a direct consequence of the universal constraints operating at the lexicon-syntax interface which derive from UH, or they could simply be surface transfer from L1 Greek into L2 Spanish, since SV is preferred to #VS with unergatives in native Greek. In presentationally focused-subject contexts with unaccusatives (Figure 3), the upper intermediate group allows both #SV and VS orders to the same extent, which results in clear optionality. While the low advanced group shows a mild distinction by preferring VS to #SV, the difference is not significant, which technically implies that they are allowing #SV and VS to the same extent, i.e. they show optionality. Apparently, discursive VS appears to be acquired late, since only the upper advanced group significantly prefers VS to #SV. This pattern is in line with the native pattern, though natives dramatically and significantly prefer VS to #SV. It may be also the case that this apparent native-like behaviour is simply a reflection of the learners’ overgeneralization of unaccusative VS order from neutral to focused contexts, as found in previous research (Hertel 2003; Lozano 2006).

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.19 (1080-1133)

The development of the syntax-discourse interface 

The crucial results come from presentationally focused-subject contexts with unergatives (Figure 4). Clearly, discursive VS has not been acquired, since learners show optionality at all levels of development, i.e. they equally prefer VS and #SV (see Sorace 2000 on optionality). Interestingly, a visual inspection of Figure 4 reveals that their acceptability pattern is the opposite of the Spanish natives’ pattern: #SV is slightly but not significantly preferred to VS, yet natives strongly and significantly prefer VS to #SV. Optionality then pervades learners’ intuitions at all levels of proficiency in focused contexts with unergatives (Figure 4) and in the two initial levels with unaccusatives (Figure 3). This implies that the most advanced learners’ apparent native-like attainment in focused contexts with unaccusative verbs (Figure 3) is probably due to an overgeneralization, as suggested above. To summarize, results from focused contexts with unaccusatives and unergatives suggest that Greek learners of Spanish show deficits with word order distribution at the syntax-discourse interface, while their intuitions converge with natives’ when word order is constrained by universal properties at the lexicon-syntax interface. Consider the possibility of surface transfer from L1 Greek as a possible explanation of the results. If learners were simply transferring, then we would expect the following. First, while learners’ distinctions are admittedly significant in neutral contexts (VS > #SV with unaccusatives but SV > #VS with unergatives), we would expect such distinctions to be ‘sharper’ if they were transferring directly from their L1, i.e. their acceptability rates in the ungrammatical condition (unaccusative #SV and unergative #VS) would be expected to be lower, around the Spanish mean, as Greek and Spanish word order is constrained by identical mechanisms deriving from UH. This is contrary to fact, as shown in the results section. Secondly, even if we ignore the learners’ lack of ‘sharp’ intuitions in neutral contexts and still maintain that transfer is taking place, we would expect learners to significantly prefer SV to VS with both verb types in focused contexts, since presentationally focused subjects appear in sentence-initial position in Greek with both verb types (SV order). This is contrary to fact, as reported in the results. In short, these results clearly indicate that learners’ behaviour is not the result of surface L1 transfer in either neutral or focused contexts. This suggests that (i) universal properties deriving from UH (lexicon-syntax interface) constrain word order in neutral contexts at all levels of proficiency, as found in previous research (de Miguel 1993; Hertel 2000, 2003; Hertel & Pérez-Leroux 1999; Lozano 2003, 2006) and, most importantly (ii) that the source of learners’ deficits in focused contexts (syntax-discourse interface) is other than L1 transfer. Consider now random behaviour as a possible explanation of the data. The optional behaviour shown in focused contexts with unaccusatives (upper intermediate and low advanced groups) and with unergatives (all groups) clearly indicates that focused contexts are persistently problematic and result in learners’ divergent knowledge, since neutral contexts typically result in (near) native-like behaviour. This cannot be attributed to learners’ random behaviour, since in neutral contexts they systematically prefer SV > VS with unaccusatives but SV > VS with unergatives at all levels of proficiency, while their intuitions are optional in focused contexts: SV = #VS with both

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.20 (1133-1198)

 Cristóbal Lozano

unaccusatives and unergatives (except for the most advanced group that preferred VS > SV with unaccusatives). This suggests again that learners are sensitive to the universal constraints operating at the lexicon-syntax interface, while their intuitions are not fine-tuned in focused contexts that are constrained at the syntax-discourse interface. In other words, it appears that learners are sensitive to the fact that word order operates differently in neutral vs. focused contexts. Finally, consider overgeneralization as possible explanation of the results. A comparison between neutral vs. focused contexts (Figure 1 and Figure 2 vs. Figure 3 and Figure 4) might initially give the impression that learners slightly overgeneralize the word order distribution from neutral contexts to focused contexts. Consider unaccusatives first. While in neutral contexts they clearly and significantly prefer VS > #SV, in focused contexts they show a similar trend starting at low advanced level (a mild but non significant VS > #SV preference) and reaches its peak at upper advanced level (a significant VS > #SV preference). Consider now unergatives. In neutral contexts learners significantly prefer SV > #VS. A slightly similar word order pattern can be observed in focused contexts, though, importantly, the difference never reaches statistical significance. In other words, learners are not simply overgeneralizing from neutral to focused contexts since, otherwise, we would expect the same patterns of behaviour in both neutral and focused contexts and not slight overgeneralization patterns. This suggests again that they may be aware that focused contexts are different from neutral contexts, but are unsure about how focus is syntactically marked. In short, once we have discarded several factors as the possible source of learners’ behaviour (namely, transfer, random behaviour and overgeneralization), results on optionality clearly indicate that the discursive properties constraining word order at the syntax-discourse interface are not as readily acquired as the properties at the lexicon-syntax interface, even at very advanced levels of proficiency. This confirms our hypothesis in (17) and entails that learners’ representation for word order alternations in neutral contexts in Spanish must be similar to representations in (7) and (8), while their representations for focused contexts must be somewhat different from Spanish natives representations in (12) and (13), which will be discussed later. What remains to be explained now is the precise etiology or source of learners’ non native-like, optional behaviour at the syntax-discourse interface, which is persistently problematic. The crucial question is to determine the exact locus of the observed deficits in the current study. There are three possibilities, which we will examine: either the deficits are located outside the faculty for human language (i.e. in the conceptualintentional module), or inside the computational system (i.e. at narrow syntax) or perhaps at the very interface between these two systems. Let us analyze each in turn. Current acquisitional studies are debating the precise etiology of deficits at the syntax-discourse interface. It has been proposed that interpretable discursive features (like topic and focus) are responsible for the observed deficits in language attrition, since interfaces are more ‘vulnerable’ (Montrul 2004a; Serratrice 2004; Sorace 2004; Tsimpli et al. in press). In particular, interpretable features become ‘unspecified’, which results in optionality. The implication is that this deficit must be lo-

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.21 (1198-1258)

The development of the syntax-discourse interface 

cated somewhere in the interface between the computational system (syntax) and the discourse/information-structure module (i.e. the Conceptual-Intentional System, CIS, an external module of thought). The consequence for the current study is that the interpretable [Foc] feature cannot be interpreted by CIS as a result of its being unspecified, i.e. unspecified [Foc] cannot be interpreted by CIS as new information. It follows that our acquirers are insensitive to the informational specification of discursive features, i.e. they do not know that a constituent specified as [Foc] must be interpreted as new information. There is no principled reason to believe that this is the case for the following reasons: i.

It has been shown that L2 learners tend to draw upon information organization principles (topic/focus, theme/rheme) that are already present in their L1 (Murcia-Serra 2003). This entails that our learners must somehow know from their L1 Greek that certain elements in the discourse are specified as new information, since they contain an interpretable [Foc] feature that is interpreted by CIS as new information. ii. It is well known that information packaging into topic and focus is universal, though different languages may select different mechanisms to express it (syntactically, morphologically, prosodically or by a combination of these) (Vallduví 1995; Vallduví & Engdahl 1996). Then, there is no reason to assume that our learners’ interpretable [Foc] feature is somehow unspecified, as they must know from their L1 Greek that certain elements are interpreted as new in the discourse, irrespective of whether such element is marked prosodically or syntactically in their L1. iii. Unlike children acquiring their L1 (Grinstead 2004), both the interfaces and the external systems of adults are in their final, fully-fledged state (Chomsky 1998). This implies that our learners’ CIS module and its interface must be ‘intact’, i.e. they are well-formed and developed enough to interpret interpretable discursive features like [Foc]. Hence, there is no reason to believe that the deficits observed in our learners derives from the external module or from its interface. iv. A few Greek adult native speakers (who had participated in the test) were informally tested via email about their intuitions on the Greek sentences shown in (4), (6), (14) and (15) several weeks before the test. They correctly preferred VS to #SV with Greek unaccusatives and SV to #VS with Greek unergatives in neutral contexts, as well as SV > #VS with unaccusatives and unergatives in presentationally focused-subject contexts, as expected. This entails that, in their L1, adult Greek natives are sensitive to the pragmatically adequate word order that is required by the eliciting question (general out-of-the-blue questions and wh- questions). This is additional evidence supporting the idea that there is no reason to believe that the syntax-discourse interface of adult Greek natives is somehow defective in their non-native Spanish. The arguments presented above indicate that (i) learners do not have any problems with the interpretable [Foc] feature and (ii) there is no reason to believe that the mechanisms responsible for the interpretation of such a feature (i.e. the syntax-discourse

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.22 (1258-1333)

 Cristóbal Lozano

interface and the discourse or information-structure module itself) are defective. The only possibility left is that the source of the observed deficits resides in the computational system. This has been claimed (though not fully developed) in a recent study where it was observed that learners show deficits with focused structures (Belletti & Leonini 2004). The authors point out that learners do not have problems in identifying the informational value of presentationally-focused subjects in L2 Italian, which leads the authors to believe that the difficulty must be grammatical in nature. Given the above arguments, the source of optionality at the syntax-discourse interface must be located somewhere in the syntax (computational system). The source of the deficit cannot be the unvalued [uFoc] feature of the focus head, whose only purpose is to probe for the matching goal (the interpretable [Foc] feature of the subject) and agree with it in order to be deleted. Given the above arguments, it is clear that this operation must have taken place since the CIS is able to interpret the [Foc] feature of the subject as interpretable, which implies that [uFoc] must have been previously deleted in the course of the derivation. The only remaining possibility is that learners are not sensitive to the uninterpretable [EPP] feature of the TP-internal focus head (Foc0 ), which is responsible for displacing focused elements to its specifier. Indeed, if such a feature is defective, presentationally focused subjects cannot be displaced to [Spec,FocP]. Note that in more recent terminology, the [EPP] feature corresponds to what has been termed ‘internal’ merge, whereby an object that was previously created via merge, is newly merged with a greater object. This is what was traditionally termed ‘movement’, which is seen now as a type of merge operation, and has to be distinguished from ‘pure’ merge, whereby an element from the lexicon is selected and is merged with a syntactic object. In particular, our learners may have problems with the internal merge operation responsible for displacing the focused element to a designated syntactic configuration. This amounts to saying that for our learners’ internal merge (i.e. ‘move’) is more costly than pure merge when the former is motivated by [EPP] features that have a clear outcome on the discursive interpretation of the sentence. This is not surprising since it is well known in the theoretical literature that move is more costly than merge in terms of computation. In simpler terms, our learners are aware that the focused subject is interpreted as new information, but are unable to grammaticalize it syntactically via word order. These findings are in line with Lozano (2006, in press) who postulated that the strength of the focus head (or, in more recent terminology, the EPP feature of the focus head) is the locus of deficits at the syntax-discourse interface. In the context of recent L2 acquisition theories, there have appeared several proposals to account for learners’ non native-like patterns of acquisition. Given that our learners show an impairment, we will discuss only those theories subsumed under what is generally termed Impairment Hypotheses (see Hawkins 2001; White 2003 and Montrul 2004b for general overviews). One of the versions, the Failed Functional Features Hypothesis, FFFH (Hawkins & Chan 1997; Hawkins 2000), postulates that learners are unable to acquire L2 parametrizable uninterpretable functional features that are not present in their L1 beyond the Critical Period. It could be argued that our

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.23 (1333-1374)

The development of the syntax-discourse interface 

learners are indeed unable to acquire the parametrizable and uninterpretable [EPP] feature of the focus head, hence the observed deficits at the syntax-discourse interface with the pragmatically odd construction in focused contexts. However, our learners do not show any deficits with the pragmatically felicitous construction in focused contexts, hence it could also be argued that learners have acquired the [EPP] feature. In other words, what our learners show is optionality, which is not predicted by the FFFH, since it simply predicts that the functional feature in question cannot be acquired. A different version of the Impairment Hypothesis is the Local Impairment Hypothesis, LIH (Beck 1998), which proposes that the strength value of functional features cannot be acquired, which leads to optionality. It could be argued that the LIH can straightforwardly account for the data, since our learners show optionality with the [EPP] feature of the focus head. This entails (in Beck’s terminology) that they are having problems with the [strong] value of the focus head, which is ultimately responsible for the raising of the focused subject and, hence, leads to optionality. This is indeed what we find in the data. However, LIH would overpredict since, if this were the case, we would also expect verb raising to become optional (since, by hypothesis, the strong [V] feature of T would become impaired). This would lead to a chaotic behaviour by the learner, as both verb-to-tense raising and subject-to-focus raising would be expected to occur. This is not certainly the case, as we have shown earlier that results are not due to random behaviour. To summarize, the developmental data shown here reveal that Greek learners of Spanish are unable to grammaticalize discursive focus syntactically, probably as a result of an impaired functional feature operating at the syntax-discourse interface. Future L2 research will need to provide more fine-grained accounts of these deficits at the discourse interface, since what current theories can provide is just a partial explanation of the data.

. Conclusion This study has contributed to the current debate of whether discursive properties operating at the syntax-discourse interface (like presentational focus) are persistently problematic in second language acquisition. It has been shown that this is the case for Greek learners of Spanish at three levels of proficiency. A possible source of these deficits has been attributed to learners’ inability to grammaticalize discursive focus via word order. Further L2 evidence is needed to (dis)confirm this proposal.

Notes * This research is part of a larger project which was funded by grants no. R00429934281 (ESRC, UK) and no. EX2004-0364 (MEC, Spain) to the author. . Only intransitive alternations (SV and VS) will be investigated in this paper.

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.24 (1374-1427)

 Cristóbal Lozano . While it is an indisputable fact that the subject of Spanish and Greek unaccusatives remains in postverbal position in neutral contexts, there is some debate as to the precise syntactic analysis. It is standardly assumed that the subject can remain postverbally in [Comp,VP] due to the presence of expletive pro in [Spec,TP] (e.g. Rizzi 1997a; Zagona 2002; amongst others). Alternative analyses argue that, either expletive pro cannot enter the derivation due to economy principles, since EPP is weak in pro-drop languages and does not need to be satisfied (Picallo 1998 for Spanish) or the EPP feature of T can be checked via the [D], Case and Phi features of V that raise to T (Alexiadou & Anagnostopoulou 1998 for Greek). I will follow the standard analysis since, whether an overt expletive is required in [Spec,TP] or not, it is irrelevant for the interpretation of the results in this study. . Importantly, presentational focus has to be distinguished from contrastive focus, as they show different syntactic and interpretive effects, which have been often overlooked in the literature. Interpretively, contrastive focus picks up an entity out of a very limited set of entities (which are known to both speaker and hearer) for contrastive purposes, while presentational focus picks up an entity out of an unlimited set of entities (which are unknown to the hearer) to express new information (Gundel 1998; Kiss 1998). Syntactically, contrastive focus is a leftperipheral, CP-domain phenomenon that has been widely investigated cross-linguistically, e.g. Spanish (Zubizarreta 1998; Domínguez 2004), Greek (Georgiafentis 2004), Italian (Rizzi 1997b), Hungarian (Puskas 1997), etc., while presentational focus is a less well-known TP-internal phenomenon in Spanish (Lozano 2003), Italian (Belletti 2001), Hungarian and Basque (Horvath 1986), Malayalam (Jayaseelan 2001), etc. For semantic, syntactic and prosodic differences between presentational and contrastive focus, see Breul (2004), Gundel (1998) and Kiss (1998), and particularly Domínguez (2004) for Spanish and Georgiafentis (2004) for Greek. Further note that presentational focus is also known as information focus in the literature (Kiss 1998) and that both contrastive and presentational focus have been termed narrow focus. I will use the terms contrastive and presentational throughout, yet only presentational focus is relevant in the current study. . Note that this operation would not violate the Phase Impenetrability Condition (PIC) since FocP is located at the very edge of the VP phase, which means that any computation below FocP still ‘belongs’ to the first phase. . Feature deletion is marked with double strikeout font, while displacement (copying and deletion) is marked with single strikeout font. . Crucially, there has been confusion in the Greek literature as to the syntactic analysis of focus since ‘classic’ studies (e.g. Tsimpli 1995) did not distinguish between the two types of focus (presentational vs. contrastive). While Kiss (1998) first observed this problem, recent work (Georgiafentis 2004) clearly shows that presentational focus differs from contrastive focus in Greek in several respects (syntactically, semantically and intonationally). In this paper, I will be concerned with presentational focus only (see Note 3 for a distinction between presentational vs. contrastive focus). . Note that when performing a mixed ANOVA, SPSS allows post-hoc comparisons to betweensubject factors only, and not to within-subject factors. Given our hypothesis in (17), the most suitable test to perform a within-subject factor post-hoc test is a repeated measures t-test (see Brace et al. 2000: 195). . Post-hoc tests with a mixed ANOVA in SPSS only allow comparing the overall difference between, say, natives’ #VS and SV against Greek upper advanced learners’ #VS and SV. By contrast, a one-way ANOVA permits to compare the difference between, say, natives’ #VS and Greek up-

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.25 (1427-1529)

The development of the syntax-discourse interface 

per advanced #VS. Hence the need to use one-way ANOVAS for more fine-grained comparisons (see Brace et al. 2000).

References Alexiadou, A. & Anagnostopoulou, A. (1998). “Parametrizing Agr: Word order, verb movement and EPP checking”. Natural Language and Linguistic Theory, 16, 491–539. Alexiadou, A. & Anagnostopoulou, A. (1999). “Tests for unaccusativity in a language without tests for unaccusativity”. In G. Bampiniotis & A. Mozer (Eds.), Proceedings of the Third International Conference on Greek Linguistics. Athens: University of Athens. Alexiadou, A., Anagnostopoulou, A., & Everaert, M. (Eds.). (2004). The Unaccusativity Puzzle: Explorations of the Syntax-lexicon Interface. Oxford: Oxford University Press. Al-Kasey, T. & Pérez-Leroux, A.-T. (1998). “Second language acquisition of Spanish null subjects”. In S. Flynn, G. Matohardjono, & W. O’Neil (Eds.), The Generative Study of Second Language Acquisition (pp. 161–185). Mahwah, NJ: Lawrence Erlbaum. Antonopoulou, E. & Sifianou, M. (2003). “Conversational dynamics of humour: the telephone game in Greek”. Journal of Pragmatics, 35, 741–769. Beck, M.-L. (1998). “L2 acquisition and obligatory head movement: English-speaking learners of German and the Local Impairment Hypothesis”. Studies in Second Language Acquisition, 20, 311–348. Belletti, A. (2001). ““Inversion” as focalization”. In A. Hulk & J.-Y. Pollock (Eds.), Subject Inversion in Romance and the Theory of Universal Grammar (pp. 60–90). Oxford: Oxford University Press. Belletti, A. (2003). “Aspects of the lower IP area”. In L. Rizzi (Ed.), The Structure of CP and IP. The Cartography of Syntactic Structures, Vol. 2. Oxford: Oxford University Press. Belletti, A. & Leonini, C. (2004). “Subject inversion in L2 Italian”. In S. Foster-Cohen, M. Sharwood Smith, A. Sorace, & O. Mitsuhiko (Eds.), EUROSLA Yearbook: Volume 4 (pp. 95–118). Amsterdam: John Benjamins. Belletti, A. & Shlonsky, V. (1995). “The order of verbal complements: A comparative study”. Natural Language and Linguistic Theory, 13, 489–526. Bolinger, D. (1954). “English prosodic stress and Spanish sentence order”. Hispania, 37, 152–156. Brace, N., Kemp, R., & Snelgar, R. (2000). SPSS for Psychologists. Basingstoke: Palgrave. Breul, C. (2004). Focus Structure in Generative Grammar: An Integrated Syntactic, Semantic and Intonational Approach. Amsterdam: John Benjamins. Brody, M. (1995). “Focus and checking theory”. In I. Kenesei (Ed.), Approaches to Hungarian, V: Levels and Structures. Szeged: Jate. Casielles-Suarez, E. (2004). The Syntax-Information Structure Interface: Evidence from Spanish and English. Oxford: Routledge. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N. (1998). “Minimalist inquiries: the framework”. MIT Occasional Papers in Linguistics, 15, 1–56. Chomsky, N. (2004). “Beyond explanatory adequacy”. In A. Belletti (Ed.), Structures and Beyond: The Cartography of Syntactic Structures. Oxford: Oxford University Press. Cowart, W. (1997). Experimental syntax: Applying Objective Methods to Sentence Judgments. Thousand Oaks, CA: Sage.

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.26 (1529-1636)

 Cristóbal Lozano

Domínguez, L. (2004). Mapping Focus: The Syntax and Prosody of Focus in Spanish. Boston University, Unpublished PhD dissertation. Eguren, L. & Fernández Soriano, O. (2004). Introducción a una sintaxis minimista. Madrid: Gredos. Georgiafentis, M. (2004). Focus and Word Order Variation in Greek. University of Reading, Unpublished PhD dissertation. Georgiafentis, M. & Sfakianaki, A. (2004). “Syntax interacts with prosody: The VOS order in Greek”. Lingua, 114, 935–961. Grinstead, J. (2004). “Subjects and interface delay in child Spanish and Catalan”. Language, 80, 40–72. Gundel, J. K. (1998). “On different kinds of focus”. In P. Bosch (Ed.), Focus: Linguistic, Cognitive and Computational Perspectives (pp. 293–205). Cambridge: Cambridge University Press. Hawkins, R. (2000). “Persistent selective fossilisation in second language acquisition and the optimal design of the language faculty”. Essex Research Reports in Linguistics, 34, 75–90. Hawkins, R. (2001). Second Language Syntax: A Generative Introduction. Oxford: Blackwell. Hawkins, R. & 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. Hertel, T. J. (2000). The Second Language Acquisition of Spanish Word Order: Lexical and Discourse Factors. Pennsylvania State University, Unpublished PhD dissertation. Hertel, T. J. (2003). “Lexical and discourse factors in the second language acquisition of Spanish word order”. Second Language Research, 19, 273–304. Hertel, T. J. & Pérez-Leroux, A. T. (1999). “The second language acquisition of Spanish word order”. In A. Greenhill, H. Littlefield, & C. Tano (Eds.), Proceedings of the 23rd Boston University Conference on Language Development (pp. 228–239). Somerville, MA: Cascadilla Press. Horvath, J. (1986). Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Jayaseelan, K. A. (2001). “IP-internal topic and focus phrases”. Studia Linguistica, 55, 39–75. Keller, F. & Alexopoulou, T. (2001). “Phonology competes with syntax: Experimental evidence for the interaction of word order and accent placement in the realization of Information Structure”. Cognition, 79, 301–372. Kiss, K. E. (Ed.). (1995). Discourse Configurational Languages. Oxford: Oxford University Press. Kiss, K. E. (1998). “Identificational focus versus information focus”. Language, 74, 245–273. Levin, B. & Rappaport-Hovav, M. (1995). “Unaccusativity. At the syntax-lexical semantics interface”. Cambridge, MA: MIT Press. Liceras, J. M. (1989). “On some properties of the “pro-drop” parameter: Looking for missing subjects in non-native Spanish”. In S. M. Gass & J. Schachter (Eds.), Linguistic Perspectives on Second Language Acquisition (pp. 109–133). Cambridge: Cambridge University Press. Liceras, J. M. & Díaz, L. (1999). “Topic drop versus pro-drop: Null subjects and pronominal subjects in the Spanish L2 of Chinese, English, French, German and Japanese speakers”. Second Language Research, 15, 1–40. Lozano, C. (2003). Universal Grammar and Focus Constraints: The Acquisition of Pronouns and Word Order in Non-native Spanish. University of Essex, Unpublished PhD dissertation. Lozano, C. (2006). “Focus and split intransitivity: The acquisition of word order alternations in non-native Spanish”. Second Language Research, 22(2), 1–43.

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.27 (1636-1730)

The development of the syntax-discourse interface 

Lozano, C. (in press). “La adquisición del español como L2: la interfaz sintaxis-discurso”. In M. Amengual Pizarro, M. Juan Garau, & J. Salazar Noguera (Eds.), Actas del XXIII Congreso Internacional de la Asociación Española de Lingüística Aplicada. Universitat de les Illes Balears. Mendikoetxea, A. (2000). “Relaciones de interficie: Los verbos de cambio de estado”. Estudios de Lingüística, 7, 125–144. Miguel, E. de (1993). “Construcciones ergativas e inversión en la lengua y la interlengua española”. In J. M. Liceras (Ed.), La lingüística y el análisis de los sistemas no nativos (pp. 178–195). Ottawa: Dovehouse. Miguel, E. de (2003). “Construcciones ergativas e inversión en la lengua y la interlengua española”. In J. L. Liceras (Ed.), La lingüística y el análisis de los sistemas no nativos. Ottawa: Devehouse. Montrul, S. (1999). “Causative errors with unaccusative verbs in L2 Spanish”. Second Language Research, 15, 191–219. Montrul, S. (2004a). “Subject and object expression in Spanish heritage speakers: A case of morphosyntactic convergence”. Bilingualism: Language and Cognition, 7, 125–142. Montrul, S. (2004b). The Acquisition of Spanish: Morphosyntactic Development in Monolingual and Bilingual L1 Acquisition and Adult L2 Acquisition. Amsterdam: John Benjamins. Montrul, S. & Rodríguez-Louro, C. (2004). “Beyond the syntax of the Null Subject Parameter: A look at the discourse-pragmatic distribution of null and overt subjects by L2 learners of Spanish”. Paper presented at The Romance Turn, Universidad Nacional de Educación a Distancia (Madrid). Murcia-Serra, J. (2003). “Acquiring the linkage between syntactic, semantic and informational roles in narratives by Spanish learners of German”. In C. Dimroth & M. Starren (Eds.), Information Structure and the Dynamics of Language Acquisition (pp. 289–309). Amsterdam: John Benjamins. Pérez-Leroux, A. T. & Glass, W. R. (1997). “OPC effects on the L2 acquisition of Spanish”. In A. T. Pérez-Leroux & W. R. Glass (Eds.), Contemporary Perspectives on the Acquisition of Spanish, Volume I: Developing Grammars (pp. 149–165). Sommerville, MA: Cascadilla Press. Pérez-Leroux, A. T. & Glass, W. R. (1999). “Null anaphora in Spanish second language acquisition: Probabilistic versus generative approaches”. Second Language Research, 15, 220– 249. Pérez-Leroux, A. T., Scott, A. M., Hertel, T., Kellar, V., & Glass, W. R. (1999). “Sources of knowledge in second language acquisition”. Spanish Applied Linguistics, 3, 33–63. Perlmutter, D. (1978). “Impersonal passives and the Unaccusative Hypothesis”. Proceedings of the Fourth Annual Meeting of the Berkeley Linguistics Society. Berkeley: University of California. Phinney, M. (1987). “The pro-drop parameter in second language acquisition”. In T. Roeper & E. Williams (Eds.), Parameter Setting (pp. 221–238). Dordrecht: Reidel. Picallo, M. C. (1998). “On the Extended Projection Principle and null expletive subjects”. Probus, 10, 219–241. Polio, C. (1995). “Acquiring nothing? The use of zero pronouns by nonnative speakers of Chinese and the implications for the acquisition of nominal reference”. Studies in Second Language Acquisition, 17, 353–377. Puskas, G. (1997). “Focus and the CP domain”. In L. Haegeman (Ed.), The New Comparative Syntax. London: Longman. Rizzi, L. (1997a). “A parametric approach to comparative syntax: Properties of the pronominal system”. In L. Haegeman (Ed.), The New Comparative Syntax (pp. 268–285). London: Longman.

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.28 (1730-1806)

 Cristóbal Lozano

Rizzi, L. (1997b). “The fine structure of the left periphery”. In L. Haegeman (Ed.), Elements of Grammar. Dordrecht: Kluwer. Roussou, A. & Tsimpli, I.-M. (2002). “Postverbal subjects and microparametric variation”. Paper presented at GLOW 2002, University of Amsterdam/Utrecht. Satterfield, T. (2003). “Economy of interpretation: Patterns of pronoun selection in transitional bilinguals”. In V. Cook (Ed.), Effects of the Second Language on the First (pp. 214–233). Clevedon: Multilingual Matters. Serratrice, L. (2004). “Anaphoric interpretation of null and overt pronominal subjects in Italian”. Paper presented at The Romance Turn. Universidad Nacional de Educación a Distancia, Madrid, September. Serratrice, L., Sorace, A., & Paoli, S. (2004). “Crosslinguistic influence at the syntax-pragmatics interface: Subjects and objects in English-Italian bilingual and monolingual acquisition”. Bilingualism: Language and Cognition, 7, 183–205. Sorace, A. (2000). “Gradients in auxiliary selection with intransitive verbs”. Language, 76, 859– 890. Sorace, A. (2004). “Native language attrition and developmental instability at the syntaxdiscourse interface: Data, interpretations and methods”. Bilingualism: Language and Cognition, 7, 143–145. Szendröi, K. (2004). “Focus and the interaction between syntax and pragmatics”. Lingua, 114, 229–254. Tsimpli, I.-M. (1995). “Focusing in modern Greek”. In K. E. Kiss (Ed.), Discourse Configurational Languages. Oxford: Oxford University Press. Tsimpli, I.-M., Sorace, A., Heycock, C., & Filiaci, F. (in press). “First language attrition and syntactic subjects: A suty of Greek and Italian near-native speakers of English”. International Journal of Bilingualism. Vallduví, E. (1995). “Aspects of information structure in a dynamic setting”. In C. Martín (Ed.), Proceedings of the 11 Congreso de Lenguajes Naturales y Lenguajes Formales. Barcelona: Publicaciones de la Universidad de Barcelona. Vallduví, E. & Engdahl, E. (1996). “The linguistic realization of information packaging”. Linguistics, 34, 459–519. White, L. (2003). Second Language Acquisition and Universal Grammar. Cambridge: Cambridge University Press. Yuan, B. (1999). “Acquiring the unaccusative/unergative distinction in a second language: Evidence from English-speaking learners of L2 Chinese”. Linguistics, 37, 275–296. Zagona, K. (2002). The Syntax of Spanish. Cambridge: Cambridge University Press. Zubizarreta, M. L. (1998). Prosody, Focus, and Word Order. Cambridge, MA: MIT Press.

JB[v.20020404] Prn:6/06/2006; 9:52

F: LALD4117.tex / p.29 (1806-1884)

The development of the syntax-discourse interface 

Appendix

Learners’ bio-data Greek upper advanced Greek lower advanced Greek upper intermediate Subject ID Age (yrs) Prof (%) Subject ID Age (yrs) Prof (%) Subject ID Age (yrs) Prof (%) MB DF AP ZG ET IM KC MV PF FP SJ CP AD KS FAK AC ED FT C

25 32 31 23 30 22 20 19 21 37 21 22 23 25 21 25 19 21 22

100 100 98 98 98 98 95 95 95 95 95 95 95 95 95 95 95 95 95

MEAN:

24 yrs

96%

DA FP FK HB GA EL AMS EG MP VR KP MK KK JA ASM TK CL AK NL CT CQ KZ KT NM MEAN:

30 22 23 22 21 24 26 30 25 25 28 24 28 33 20 22 21 24 20 25 30 25 23 21 25 yrs

93 93 93 93 93 93 93 91 91 91 91 91 91 88 88 88 88 88 88 88 88 88 88 88 90%

AG GK DP1 DP2 SE GM LV EA JK KT AA XZ FA AP1 AD AP2 AFK DK1 EJ DK2 ND SK PA

19 30 22 23 19 23 21 20 25 24 24 20 21 36 20 23 24 29 21 21 21 30 22

86 86 86 86 84 84 84 84 84 81 81 79 79 79 77 77 77 72 70 67 67 60 60

MEAN:

24 yrs

78%

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.1 (46-124)

Beyond the syntax of the Null Subject Parameter A look at the discourse-pragmatic distribution of null and overt subjects by L2 learners of Spanish Silvina Montrul and Celeste Rodríguez Louro University of Illinois at Urbana-Champaign

This study investigates whether and how English-speaking L2 learners at various proficiency levels acquire the morphosyntactic and discourse-pragmatic properties of subject expression in Spanish. The crucial question is whether discourse-pragmatic properties are acquired together with syntactic features or later in the process of interlanguage development. According to the Full Transfer/Full Access Hypothesis, transfer effects should be observed at earliest stages of development and full parameter resetting at later stages. An oral production task was administered to 15 intermediate, 15 advanced and 15 near-native learners and 20 native speakers. Results indicated that the near-native speakers performed like the native speakers with the morphosyntactic and the discourse-pragmatic properties of the Null Subject Parameter, while in intermediate and advanced learners the acquisition of these two properties were dissociated: the learners were accurate with morphosyntax and inaccurate with discourse-pragmatics.

.

Introduction

Most recently there has been increased interest in the potential instability and vulnerability of the syntax-pragmatics interface in different types of developing grammars, such as in adult L2 acquisition, child and adult bilingualism, and language change. Much of this research has focused on the expression and the position of subjects, traditionally subsumed under the Null Subject Parameter (see, for example, Duarte 2000; Hertel 2003; Montrul 2004; Paradis & Navarro 2003; Serratrice, Sorace, & Paoli 2004; Sorace 2004; Toribio 2004; Tsimpli et al. 2004; among others). “Vulnerability” as used in recent theorizing appears to refer to grammatical domains that are acquired late, are difficult to acquire, or are never acquired completely, or are subject to change in attrition (see Müller 2003).

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.2 (124-161)

 Silvina Montrul and Celeste Rodríguez Louro

In pro-drop languages like Spanish and Italian, overt and null subject expressions are regulated by morphosyntactic and discourse pragmatic factors. Early studies of the Null Subject Parameter in L2 Spanish carried out in the 80s investigated primarily the acquisition of the syntactic properties associated with the parameter, with particular emphasis on learners whose L1 had the opposite value of the parameter. The issues investigated were (1) whether parameter resetting was in principle possible (i.e, UG access), and (2) whether initially L2 learners adopted the parameter value of their L1 (L1 transfer) or another “unmarked” parameter value different from that of the L1. Other studies conducted more recently have focused instead on the discourse-pragmatic aspects of subjects. Although discursive properties are learned, the level of mastery of these properties is variable and advanced learners and near native speakers appear to show some degree of deficit (Sorace 1999). That is why it has been characterized as a vulnerable area. In looking at a variety of bilingual situations (L2 acquisition, L1 attrition in near-natives, bilingual acquisition by children, etc.), Sorace (2004) has stressed the unidirectionality of pragmatic deficits which, according to her, affects the overproduction of overt subjects but crucially not the overproduction of null subjects. This study adds an important contribution to the field by investigating both morphosyntactic and discourse-pragmatic aspects of subject expression in the same L2 learners across different proficiency levels. It also adds to the growing interest on the acquisition of interfaces and the potential vulnerability of these domains during the language acquisition process. In particular, we ask whether and when English-speaking L2 learners at various proficiency levels, including near-native speakers, acquire the morphosyntactic and the discourse-pragmatic properties of subject expression in Spanish. The crucial question guiding this work is whether discourse-pragmatic properties are acquired together with the morphosyntactic aspects of the parameter, or later, in the process of interlanguage development. We will see that morphosyntactic aspects of the Null Subject Parameter in Spanish appear to be learned before discoursepragmatic properties of the distribution of null/overt subjects. Mastery of these two related aspects of the parameter appear to catch up at the advanced/near-native level of proficiency. However, in this study (and see also Montrul 2004), and in contrast to what Sorace (2004) claimed, we show that both null subjects and overt subjects are affected in advanced developing grammars. Indeed, a 5% overproduction of null subjects persists at the near-native level.

. Theoretical background . Some history The acquisition of the Null Subject Parameter in L2 Spanish and its associated morphosyntactic agreement properties was widely investigated in the early days of generative approaches to L2 acquisition, most notably in the work of Liceras (Liceras 1988, 1989) and White (see White 1989 and references therein). Table 1 summarizes the mor-

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.3 (161-238)

Beyond the syntax of the Null Subject Parameter 

Table 1. The Null Subject Parameter (Jaeggli 1982; Rizzi 1982) setting

+pro-drop: no that-t effect

– pro-drop: that-t effect

language properties

Standard Spanish rich verbal agreement null and overt subjects null expletives preverbal and postverbal subjects that-t effect

English poor verbal agreement overt subjects overt expletives preverbal subject *that-t effect

phosyntactic properties associated with the plus and minus value of the parameter, as formulated by Jaeggli (1982) and Rizzi (1982). Spanish, which has the [+pro-drop] setting, has rich person and number verbal agreement that licenses and identifies null subjects. Spanish allows both null and overt subjects, has no overt expletives, and allows preverbal and postverbal subjects. Spanish also allows sentences like ¿Quién crees que t vino? *“Who do you think that came”?, which illustrate the possibility of having a trace after the complementizer que. These syntactic properties, which were subsumed under a cluster and can be seen in the column illustrating the [+pro-drop] value, were not allowed in English, a [–pro-drop] language. The possibility of parameter resetting due to access to UG and the role of language transfer are two issues that have received prominent attention in generative L2 acquisition research since then. For example, early research on the L2 acquisition of the Null Subject Parameter showed that L2 learners of a non-pro-drop language like English initially start with the parameter value of their L1, but soon reset to the Spanish value. For extensive discussion of these issues see Liceras (1988, 1989), Phinney (1987), and Al-Kasey and Pérez Leroux (1998). Because not all structures related to the parameters illustrated in Table 1 have been shown to be acquired simultaneously in the L2, many researchers questioned the original formulation of the parameter and the cluster of properties associated with it, especially as related to the status of postverbal subjects and the that-t phenomenon (Liceras 1988, 1989; Chao 1981; and Safir 1982). More recently, Alexiadou and Anagnostopoulou (1998) have proposed an analysis of pro-drop within Minimalism (Chomsky 1995). Their analysis captures the null subject phenomenon, the word order facts, and the lack of overt expletives in languages like Spanish as opposed to languages like English. Let us now turn to the discourse-pragmatic properties of subjects as discussed in Section 3.

. Discourse pragmatic properties of subjects Pro-drop languages like Spanish and Italian exhibit another layer of complexity, as opposed to non-pro-drop languages. In addition to being regulated by morphosyntactic factors, the null vs. overt expression of subjects in Spanish is governed by discourse

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.4 (238-317)

 Silvina Montrul and Celeste Rodríguez Louro

rules. This happens in referential contexts, where the distribution of null or overt pronouns depends on discourse-pragmatic factors such as whether the subject is a topic, as in (1) and (3), introduces a switch reference, as in (2), or establishes focus or contrast, as in (4). (1) Pepe no vino hoy a trabajar. *Pepe/?él/Ø estará Pepe no came today to work Pepe/?él/Ø will be enfermo. sick ‘Pepe did not come to work today. He must be sick.’ (2) Hoy no fui a trabajar. Pepe/él/*Ø pensó que today I no went to work Pepe/él/*Ø thought that estaba enferma. I was sick ‘Today I did not go to work. Pepe/he thought I was sick.’

same referent

different referent

(3) ¿Quién vino? El/Mario*Ø vino. ‘Who came? He/Mario/*Ø came.’ (4) El periodistai dijo que éli no había escrito ese the journalist said that he (himself) not had written that reporte. report ‘The journalist said that he had not written that report.’

topic

focus

According to Sorace (2000), overt subjects in null subject grammars are specified for the feature [+topic shift]. In English, a non-pro drop language, overt subjects do not have the [topic shift] feature. Overt pronouns cannot have variable expressions like nadie “nobody” or quien “who”, as antecedents, as in (5) and (6). (5) Nadiei cree que *éli /Øi es inteligente. ‘Nobody believes that he is intelligent.’ (6) ¿Quiéni cree que *éli /Øi es inteligente? ‘Who believes that he is intelligent?’ The examples in (7) and (8) show that when the overt pronoun is not in subject position, coindexing with the variable expressions like nadie in (7), and quién in (8), is grammatical. (7) Nadiei quiere que María hable de éli . nobody wants that María talk about him ‘Nobody wants María to talk about him.’ (8) ¿Quiéni quiere que María hable de éli ? who wants that María talk about him ‘Who wants María to talk about him?’

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.5 (317-392)

Beyond the syntax of the Null Subject Parameter 

The distribution of null and overt subjects with variable expressions has been captured by a constraint formulated by Montalbetti (1984) as the Overt Pronoun Constraint. The Overt Pronoun Constraint states that “Overt pronouns cannot link to formal variables iff the alternation overt/empty obtains” (Montalbetti 1984: 94). Finally, in non-referential contexts, null subjects are obligatory, as in existential expressions with haber in (9), weather verbs like llover “rain“ in (10), and impersonal expressions, as in (11): existential

(9) Ø Hay poco trabajo. There is little work.’

weather verb

(10) Ø Llueve mucho en primavera. rains a lot in spring ‘It rains a lot in spring.’ (11) Ø Es obvio que Ø va a nevar. is obvious that is going to snow ‘It is obvious that it is going to snow.’

impersonal expression

. Previous L2 studies of the Null Subject Parameter in Spanish Early studies of L2 acquisition focused primarily on the morphosyntactic aspects of the Null Subject Parameter. However, there was also an awareness at that time that discourse pragmatic properties were similarly involved, particularly when data came from production of connected discourse (as in compositions). Although the early studies mentioned in passing discourse-pragmatic or stylistic aspects of the distribution of null and overt subjects, the road was not paved to carry out systematic investigation of the phenomenon. For example, Phinney (1987) tested the acquisition of the prodrop parameter in Spanish and in English using written compositions. She found that English-speaking learners of Spanish correctly omitted both referential and pleonastic pronouns in Spanish, and showed evidence of parameter resetting. By contrast, the Spanish-speaking learners of English made errors of subject omission in English, but these errors followed Spanish discourse rules. Another influential study on Spanish L2 acquisition was Liceras (1988) who tested both grammatical and stylistic properties of the pro-drop parameter by French and English-speaking learners. In this study, 2 English- and 2 French-speaking learners were administered a Grammaticality Judgment Task (GJT) and were asked to correct ungrammatical sentences and to provide other renditions for sentences they considered stylistically odd. Overall, the 4 learners were accurate with agreement but not with all constructions related to the parameter, which led Liceras to question the original formulation of the parameter. Except for one learner, all others knew the stylistic conventions that govern the use of null subjects in Spanish. An awareness that stylistic or discourse rules were at play as well was also voiced by White (1989) on her review of the studies conducted at that time. She said:

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.6 (392-442)

 Silvina Montrul and Celeste Rodríguez Louro

. . . If native speakers of [–pro-drop] languages tend to use pronouns in their production, this is not a guarantee that they have failed to reset the parameter. They may not have yet worked out the precise discourse considerations that govern use of null versus lexical pronouns. In other words, there are two different things that an L2 learner of a [+pro-drop] language has to acquire: (i) the fact that null subjects are permitted, and (ii) the circumstances in which the language actually makes use of the fact that null subjects are permitted. (pp. 86–87)

Assuming that morphosyntactic properties of pro-drop in Spanish were easily acquired, later work addressed discourse-pragmatic properties of subject expression. Another innovative study in this respect was that of Pérez Leroux and Glass (1999), who tested in two experiments the null and overt subjects with variable expressions as antecedents, as in examples (5) to (8), and the topic/focus distinction. The tasks used involved translation and elicited production. Due to the nature of the tasks used, the accuracy scores of this study were quite variable including the results of the native speakers. In the topic/focus study, production of null subjects in topic constructions was 30.8% for the elementary, 36.5% for the intermediate, 57.1% for the advanced learners and 47.4% for the controls. In the focus stories, the percentages production of overt subjects was as follows: 60% elementary, 52.4% intermediate, 35.7% advanced and 36.8% control. Although Pérez Leroux and Glass (1999) claimed that L2 learners were sensitive to the pragmatic distribution of null/overt pronouns in topic/focus contexts, their accuracy was quite low, especially at the advanced level. Most recently, Lafond, Hayes and Bhatt (2001) tested the topic/non-topic distinction with null and overt subjects in an experiment where learners had to choose null or overt subjects in dialogue sets. Learners could make two types of errors in this study: a topic error occurred when a learner inappropriately dropped a non-topic subject (overuse of null subjects), and a non-topic error occurred when a learner inappropriately maintained a topic subject (overuse of overt pronouns.) The results showed that L2 learners do not initially drop topics (44%). Although they begin to use null subjects early, they do so indiscriminately, and there is over 40% overgeneralization of null subjects. Only advanced learners behaved like the native speakers with the use of overt subjects in non-topic contexts (.33/5 or 93.4% and .29/5 or 94.2%). In short, the results of these two studies showed that there is gradual and incremental development in the acquisition of the discourse properties of null-overt subject expression, like the topic-focus distinction.

. The present study As we have seen, investigations of the morphosyntactic aspects of null subject expression and the discourse-pragmatic distribution of null/overt subjects in Spanish have been investigated independently of each other: in different studies, with different learners, and with different tasks. In order to understand the development of these two aspects of subject expression in Spanish interlanguage grammars, our study looks at

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.7 (442-500)

Beyond the syntax of the Null Subject Parameter 

morphosyntactic and discourse-pragmatic properties of subjects together in the grammars of the same L2 learners of different proficiency levels. An oral elicitation task was used to test these two properties of subjects. Before formulating the predictions, let us reiterate that null subjects are typically not allowed in English, except for some very restricted contexts, such as coordinated structures, topic drop and progressive participle construction (Haegeman 1997). In Spanish, null subjects and postverbal subjects are allowed. However, as in Italian, overt subjects are a marked option signalling a shift of topic or focus. Thus, L2 learners of Spanish have to learn the morphosyntactic conditions that allow null subjects to be identified and recovered, and the pragmatic constraints that regulate the distribution of null and overt subjects in discourse. By contrast, subjects in English do not carry any specific pragmatic force. Focus and topicalization are achieved by word order and prosody. According to the Full Transfer/Full Access Hypothesis (White 1989; Schwartz & Sprouse 1996), transfer effects with the morphosyntax and discourse-pragmatic aspects of the parameter should be observed at earliest stages of development (beginnerintermediate) and full parameter resetting at later stages (from advanced to near-native levels). Although we know that L2 learners of Spanish, whose language is not pro-drop, learn the pro-drop characteristics of Spanish and start producing null subjects at the earliest stages of development, the percentage of null subjects should be much lower than that of more advanced learners and native speakers. Conversely, the proportion of null and overt subjects should be different from that of more advanced learners and native speakers, such that beginner or intermediate L2 learners would produce more overt subjects than null subjects. Beginner-intermediate L2 learners are also expected to produce more agreement errors than advanced learners and native speakers, and they are also expected to produce preverbal subjects exclusively. Since overt pronouns in English are not regulated by discourse principles as the Spanish subject pronouns are, at lower levels of proficiency L2 learners are expected to use predominantly overt pronouns, or null and overt pronouns indiscriminately.

. Method . Participants The participants were 48 English-speaking L2 learners of Spanish (mean age 24.3), divided into three proficiency levels on the basis of a proficiency test and an oral interview rated by two native-speaker judges (as in Montrul & Slabakova 2003). On the basis of the proficiency measures, 16 L2 speakers were classified as intermediate, 16 as advanced, and 16 as near-native speakers. In addition, 20 native speakers of Spanish who did not speak a Caribbean variety acted as the control group (mean age 32.8).1

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.8 (500-583)

 Silvina Montrul and Celeste Rodríguez Louro

. Tasks Speech was elicited with pictures narrating a particular version of the story Little Red Riding Hood. The experimenter met with the subjects individually face to face. The L2 subject was given a booklet with colored pictures of the story and was asked to narrate it. It was not always the case that the experimenter saw the pictures. In response to the concerns raised by Sorace (2004), we admit that oral production data is limited to investigate all morphosyntactic properties of the Null Subject Parameter, and to make inferences about competence from patterns of language use (see White’s comment), this task is useful and appropriate in several other respects. First, by asking subjects to retell a familiar story, it is easier to control the vocabulary in the story and it is easier for the learners to find the words that they need to express themselves, especially when different levels of proficiency are involved. Second, this task allows for an investigation of the morphosyntactic and the discourse pragmatic properties of subjects. Except for the that-t effect, which rarely occurs in production, and the structures associated with the OPC, all other grammatical properties investigated here – namely, subject-verb agreement, referential, expletive, and postverbal subjects – can be easily observed in this type of task.2 Furthermore, isolated sentences as used in GJTs (see Liceras 1988) are not always useful to evaluate the relationship between grammatical elements and their discourse function. More than one or two connected sentences are needed. Sample narratives produced by the learners and the native speakers appear in the appendix.

. Results We first present the results of the morphosyntactic aspects of the parameter: agreement errors, null and overt subjects with referential and non-referential expressions, and types of overt subjects (pronominal vs. lexical) and position (preverbal vs. postverbal). In the next section we present the results of discourse-pragmatic errors. All narratives were tape-recorded, transcribed and coded for analysis by two independent raters. Interrater reliability was 92%. In every narrative, all conjugated verbs had to be expressed with a null or overt subject in order to be considered for the analysis.3 Verbs were analyzed for finiteness and person and number agreement. Null and overt subjects were counted, as well as the types of subjects (NP vs. pronominal) and their position (preverbal vs. postverbal). The percentage production of null/overt subjects and types of overt subjects was calculated for each subject and submitted to statistical analysis.

. Morphosyntax Table 2 presents the percentage of accuracy with person and number verbal morphology.

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.9 (583-632)

Beyond the syntax of the Null Subject Parameter 

Table 2. Percentage accuracy on person and number verbal agreement (raw counts in parentheses) % agreement errors native speakers (n = 20) near-native speakers (n = 16) advanced learners (n =16) intermediate learners (n = 16)

0.49 0.13 2.28 12.55

(5/1019) (1/794) (15/659) (61/468)

As can be seen, the intermediate learners were more inaccurate than the other three groups, producing more than 12 % of errors (one-way ANOVA and post-hoc Tukey F(3,64) = 14.958, p < 0.0001). Figure 1 presents the overall production and proportion of null and overt subjects by group. Here, the intermediate learners produced significantly more overt subjects (close to 70%) than the three other groups. The advanced and control groups were not different from each other, but produced more overt subjects than the near-native group (one-way ANOVA and Tukey post hoc procedure F(3,64) = 32.020, p < 0.0001). Figure 2 shows the percentages and proportions of referential and non-referential null subjects. The control and the near-native groups produced comparable percentages of expletives, while the advanced and intermediate groups produced much less (one-way ANOVA and Tukey post hoc procedure F(3,64) = 11.274, p < 0.0001). No ungrammatical overt expletives were observed in the data. Figure 3 looks at the types of overt subjects with respect to position (preverbal vs. postverbal). The control, near-native and advanced groups produced comparable percentages of preverbal and postverbal subjects (almost 25% postverbal). These three groups were different from the intermediate group who produced no postverbal subjects at 100 90

31.7

80 70

57.2

56.3 68.4

60

null subjects

50

overt subjects

40 30 20 10 0 control

near-native

advanced

Figure 1. Percentage null and overt subjects

intermediate

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.10 (632-664)

 Silvina Montrul and Celeste Rodríguez Louro 100 90 80 70 60 50 40

90

89.7

94.2

control

near-native

advanced

98.6

intermediate

non-referential referential

30 20 10 0

Figure 2. Percentage types of null subjects 100 90 80 70 60 50 40

100 79.8

79.7

74.1

postverbal preverbal

30 20 10 0 control

near-native

advanced

intermediate

Figure 3. Percentage types of overt subjects

all (0%) (one-way ANOVA: F(3,64) = 5.870, p < 0.001). The sentences produced by the intermediate learners with preverbal subjects were not ungrammatical, but they clearly underused the stylistic possibility of expressing sentences with postverbal subjects (especially with unaccusative verbs). As a result, the narratives of the intermediate learners were less natural than those of more advanced learners. Figure 4 looks at the types of overt preverbal subjects. As Figure 4 shows, no significant differences among the four groups in the production of pronominal and lexical subjects were found (F(3,64) = .509, p < .678). Figure 5 displays the types of postverbal subjects according to the argument structure of the verb (unaccusatives, transitives, unergatives). The controls, the near-natives and the advanced produced the majority of postverbal subjects with unaccusative verbs

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.11 (664-679)

Beyond the syntax of the Null Subject Parameter 100 90 80 70 60

pronominal lexical

50 40

84.6

85.6

80.5

76.8

control

near-native

advanced

intermediate

30 20 10 0

Figure 4. Percentage types of overt preverbal subjects 100 90 80 70 60

other unaccusative

50 40

82.9

81.54

control

near-native

71.19

30 20 10 0 advanced

0 intermediate

Figure 5. Percentage types of overt postverbal subjects

(around 80%). Advanced learners produced postverbal subjects with more variety of verbs than the control and near-native groups (F(3,64) =‘54.730, p < 0.0001). To summarize, the intermediate learners show evidence of having reset some morphosyntactic properties of the null subject parameter, because they produced null subjects and they did not produce overt expletives. Nevertheless, they are still constrained by their L1 as compared with the other more advanced groups. For example, they made over 12% agreement errors, they produced more overt than null subjects, and they did not produce a single postverbal subject. In contrast, the morphosyntactic aspects of the Null Subject Parameter have successfully been acquired by the advanced and near-native groups, since the agreement error rate is quite low, and these groups did not differ quantitatively or qualitatively from the native speaker control group in



JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.12 (679-763)

 Silvina Montrul and Celeste Rodríguez Louro

all the structures examined. Let us now turn to the discourse pragmatic rules affecting subject expression.

. Discourse-pragmatics Overt and null subjects were examined to see whether they violated discourse rules. If an overt subject (pronominal or lexical) did not introduce a new referent in the narrative or was not used for emphasis, it was considered redundant. Similarly, when it was clear from the story line and the agreement morphology on the verb that there was a switch of reference, but the speaker did not use an overt subject, the null subject was considered illicit. There were also cases where the verb was in the third person but it was not clear who the subject was: the wolf, little Red Riding hood, or the grandmother. That is, the use of null subjects led to ambiguity. These results are summarized in Table 3. As can be seen in Table 3, there is incremental learning in the acquisition of discourse-pragmatic properties of overt and null subjects. With overt subjects, the intermediate group produced 22.9% of redundant subjects, a percentage that decreased to 7.6% in the advanced group and was almost negligible with the near-natives. The near-native speakers performed like the native speakers (one-way ANOVA and Tukey, F(3,64) = 354.930, p < 0.0001). With respect to null subjects, unlike native speakers, near-native and advanced learners also produced some illicit null subjects, in cases where there was a change of referent (F(3,64) = 57.890, p < 0.001). Although the intermediate group underproduced null subjects, when null subjects were produced (mostly in coordinated structures) they were pragmatically correct (this was only a 1% error rate). Since English allows null subjects in these situations (Haegeman 1997), this result is not surprising. However, errors of overuse of null subjects were most significant with the advanced group, who produced 8.4% of illicit null subjects, while the near-natives produced 5.5%. Although there were no statistical differences between the advanced and the near natives in this respect, the differences between the native and near-native speakers bordered on significant (p < 0.064).

Table 3. Percentage of pragmatically redundant overt subjects and illicit null subjects

control (n = 20) near-native (n = 16) advanced (n = 16) intermediate (n = 16)

Overt Subjects correct redundant

Null Subjects correct illicit

100 99.7 92.4 77

99 94.5 91.6 99

0 0.3 7.6 22.9

1 5.5 8.4 1

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.13 (763-825)

Beyond the syntax of the Null Subject Parameter 

. Discussion In line with previous studies of the morphosyntactic aspects of the NSP have shown, L2 learners in this study also appear to have reset the parameter from the English to the Spanish value, as observed from patterns of language use. Errors of agreement are relatively few, and none of the L2 learners produced overt subjects exclusively (that is, 100%), or overt expletives. Still, the intermediate learners did not evidence in their production all properties of the parameter: they made significantly more agreement errors than the other groups and they did not produce any postverbal subjects. Although lack of use may not necessarily indicate lack of knowledge, the comparison with the more advanced groups is revealing in itself. The developmental trend observed suggests that some target or non-target linguistic knowledge is implicated in the production of subjects in non-native Spanish. As predicted by the Full Transfer/Full Access Hypothesis (White 1989; Schwartz & Sprouse 1996), L1 influence is evident with both the morphosyntactic and the discourse-pragmatic aspects of the parameter, which can be observed by the performance of the intermediate group. In response to White’s (1989) comment, it appears that once L2 learners handle the morphosyntactic characteristics of the parameter, they then learn to apply the discourse-pragmatic rules on subject expression. Our results suggest is the following: (1) intermediate learners were less accurate with morphosyntax and did not appear to be sensitive to the pragmatic distribution of null/overt subjects; (2) advanced learners have the morphosyntactic aspects of the parameter in place, but show residual indeterminacy with the pragmatic distribution of subjects (7.6% redundant overt subjects and 8.4% illicit null subjects); (3) near native speakers also have the morphosyntactic aspects of null subject expression in place, and the pragmatic aspect of the phenomenon is almost native-like (except for a residual 5% of illicit null subjects). Recall that the pragmatic distribution of subjects adds another layer of complexity to the acquisition task, and lower proficiency L2 learners cannot rely on their L1 grammar in this respect since in English overt subjects are the norm and they are not regulated by discourse factors. That is, parameter resetting and discourse rules eventually catch up in advanced interlanguage grammars (see also results of Pérez-Leroux & Glass 1999). It is clear from our results that although the intermediate learners have acquired the fact that Spanish has null subjects, they are still highly constrained by their L1: they made over 12% verbal agreement errors, produced more overt subjects than null subjects, used more pronominal subjects than all the other groups, and did not produce a single postverbal subject. However, when one looks at intermediate learners’ behavior in the pragmatic domain, there are fewer mistakes than with the advanced or near-native speakers in their production of null subjects. The intermediate learners’ behavior is similar to the performance of controls not because the learners are proficient pragmatically but because they are still unaware of the subtleties of Spanish discourse constraints. In fact, when they used null subjects, these were mostly in coordinated structures. This is also confirmed by the fact that the intermediate learn-

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.14 (825-865)

 Silvina Montrul and Celeste Rodríguez Louro

ers made 22.9% of errors with redundant overt subjects. When it comes to advanced and near-native speakers, it is clear that the morphosyntactic aspects of the parameter are in place. Advanced speakers made a comparable number of errors with illicit use of null subjects and redundant uses of overt subjects. By the near native level, redundant uses of overt subjects are practically non-existent, while there still remains a 5% use of illicit null subjects. We would now like to comment briefly on this last finding in light of recent work by Sorace (2004). In looking at a variety of bilingual situations (L2 acquisition, L1 attrition in near-natives, bilingual acquisition by children, etc.), Sorace (2004) has stressed the unidirectionality of pragmatic deficits with overproduction of overt subjects but crucially not with null subjects. Overuse of overt subjects result from the underspecification of the [topic shift] feature. That is, upon losing the [topic shift] feature, overt subjects freely occur in topic and non-topic contexts. Since null subjects do not have the topic shift feature, they should continue to occur in non-topic contexts. However, in this study (and see also Montrul 2004), we showed that both null subjects and overt subjects are affected in advanced developing grammars. Indeed, a 5% overproduction of null subjects in topic shift contexts persists at the near-native level. As Sorace (2004) herself pointed out, overproduction of null subjects in this study and in Montrul (2004) could be an artifact of the research methodology and story involved: even though the learners were asked to retell the story prompted with pictures, a large part of the story was known to both the experimenter and the subjects. However, the findings of a study by Lafond, Hayes and Bhatt (2001) severely mitigate this possibility. Lafond et al. found similar overproduction of null subjects in switch reference or non-topic contexts by beginner, intermediate, advanced and near-native speakers using a completely different task. The task presented learners with a series of dialogues and the learners were asked to choose which sentence (with an overt or null subject) followed the last sentence or question in the dialogue. Like the results of the study reported here, the results of Lafond et al. also showed a progressive decrease by the near-native level (beginners 2.2/5 or 44% errors, intermediate 1.89/5 or 37.8% errors, advanced 1.7/5 or 34% errors and near-natives .78/5 or 15.6% errors, native speakers 0% errors), although the error rate of the near-natives was much higher in their study (15.6%) than in our study (5%). Since both the work by Lafond et al. and the present study show that this type of error decreases by the near-native level of proficiency, it might not end up as a prototypical feature of indeterminate and incomplete grammars to the extent that the opposite error pattern – overproduction of overt subjects – appears to be. However, this clearly requires further investigation using a variety of research methodologies. Finally, the Full Transfer /Full Access Hypothesis (Schwartz & Sprouse 1996) can explain the developmental effects observed with the morphosyntactic and pragmatic aspects of the NSP in Spanish L2. The English-speaking L2 learners are unable to rely on their L1 knowledge and may, in this respect, show late acquisition of the distribution of null and overt referential pronouns in Spanish due to the added discoursepragmatic complexity (see Sorace 2004). To what extent a 5–10% inaccuracy at the

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.15 (865-954)

Beyond the syntax of the Null Subject Parameter 

near-native level is considered “experimental noise”, a performance error, or a true sign of linguistic deficit is debatable and should be addressed in further research.

Notes . Caribbean varieties are moving away from the pro-drop option and exhibit a high percentage of overt subjects in topic contexts (Toribio 2000; Otheguy 2005). . Recall that some researchers have questioned the that-t effect as related to the Null Subject Parameter (Chao 1981; Safir 1982). . For example, in a sentence like Esta es la historia de una niña que se llamaba Caperucita Roja “This is the story of a girl whose name was Little Red Riding Hood”, the finite verb se llamaba was not counted because it cannot be used with a null or an overt subject.

References Alexiadou, A. & Anagnostopoulou, E. (1998). “Parameterizing Agr: Word order, v-movement and EPP checking”. Natural Language and Linguistic Theory, 16, 491–539. Al-Kasey, T. & Pérez-Leroux, A. T. (1998). “Second language acquisition of Spanish null subjects”. In S. Flynn & G. Martohardjono (Eds.), The Generative Study of Second Language Acquisition (pp. 62–185). Mahwah, NJ: Lawrence Erlbaum. Chao, W. (1981). “Pro-drop languages and non-obligatory control.” University of Massachusetts Occasional Papers in Linguistics, 7, 46–74. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Duarte, M. E. (2000). “The loss of the “Avoid Pronoun Principle” in Brazilian Portuguese”. In M. Kato & E. Negrão (Eds.), Brazilian Portuguese and the Null Subject Parameter (pp. 17–36). Madrid/Frankfurt: Verveut Iberoamericana. Haegeman, L. (1997). “Register variation, truncation and subject omission in English and in French”. English Language and Linguistics, 1, 233–270. Hertel, J. (2003). “Lexical and discourse factors in the second language acquisition of Spanish word order”. Second Language Research, 19, 273–304. Jaeggli, O. (1982). Topics in Romance Syntax. Dordrecht: Foris. Lafond, L., Hayes, R., & Bhatt, R. (2001). “Constraint demotion and null subjects in Spanish L2 acquisition”. In J. Camps & C. Wiltshire (Eds.), Romance Syntax, Semantics and L2 Acquisition (pp. 121–135). Amsterdam: John Benjamins. Liceras, J. (1988). “Syntax and stylistics: More on the pro-drop parameter”. In J. Pankhurst, M. Sharwood Smith, & P. van Buren (Eds.), Learnability and Second Languages (pp. 71–93). Dordrecht: Foris. Liceras, J. (1989). “On some properties of the pro-drop parameter: Looking for missing subjects in non-native Spanish”. In S. Gass & J. Schachter (Eds.), Linguistic Perspectives on Second Language Acquisition (pp. 109–133). Cambridge: Cambridge University Press. Montalbetti, M. (1984). After Binding. Doctoral dissertation. Massachusetts Institute of Technology, Cambridge, MA. Montrul, S. (2004). “Subject and object expression in Spanish heritage speakers: A case of morpho-syntactic convergence”. Bilingualism: Language and Cognition, 7, 125–142.

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.16 (954-1039)

 Silvina Montrul and Celeste Rodríguez Louro

Montrul, S. & Slabakova, R. (2003). “Competence similarities between native and near-native speakers: An investigation of the Preterite/Imperfect contrast in Spanish”. Studies in Second Language Acquisition, 25, 351–398. Müller, N. (Ed.). (2003). (In)vulnerable Domains in Multilingualism. Amsterdam: John Benjamins. Otheguy, R. (2005). “Avances en el proyecto CUNY sobre el español en Nueva York: Variación, cambio e identidad en el uso variable del pronombre en seis comunidades hispanohablantes de la Gran Manzana”. Plenary talk, 20th Conference on Spanish in the US, Chicago, IL. Paradis, J. & Navarro, S. (2003). “The use of subjects by a Spanish-English bilingual child: Crosslinguistic influence or the influence of the input?”. The Journal of Child Language, 30, 371–393. Pérez Leroux, A. T. & Glass, W. (1999). “Null anaphora in Spanish second language acquisition: Probabilistic versus generative approaches”. Second Language Research, 15, 220–249. Phinney, M. (1987). “The pro-drop parameter in second language acquisition.” In T. Roeper & E. Williams (Eds.), Parameter Setting (pp. 221–238). Dordrecht: Reidel. Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris. Safir, K. (1982). Syntactic Chains and the Definite Effect. Doctoral dissertation. Massachusetts Institute of Technology, Cambridge, MA. Schwartz, B. & Sprouse, R. (1996). “L2 cognitive states and the full transfer/full access hypothesis”. Second Language Research, 12, 40–72. Serratrice, L., Sorace, A., & Paoli, S. (2004) “Transfer at the syntax-pragmatics interface: Subjects and objects in Italian-English bilingual and monolingual acquisition”. Bilingualism: Language and Cognition, 7, 183–206. Sorace, A. (1999). “Initial States, end-states and residual optionality in L2 acquisition”. In A. Greenhill, H. Littlefield, & C. Tano (Eds.), Proceedings of the 23rd Boston University Conference on Language Development (pp. 666–674). Somerville, MA: Cascadilla Press. Sorace, A. (2000). “Differential effects of attrition in the L1 syntax of near-native L2 speakers”. In S. C. Howell, S. A. Fish, & T. Keith-Lucas (Eds.), Proceedings of the 24th Boston University Conference on Language Development (pp. 719–725). Somerville, MA: Cascadilla Press. Sorace, A. (2004). “Native language attrition and developmental instability at the syntaxdiscourse interface: Data, interpretations and methods”. Bilingualism: Language and Cognition, 7, 143–145. Toribio, A. J. (2000). “Setting Parameter limits on dialectal variation in Spanish”. Lingua, 10, 315–341. Toribio, A. (2004). “Convergence as an optimization strategy in bilingual speech: Evidence from code-switching”. Bilingualism: Language and Cognition, 7, 165–174. Tsimpli, I., Sorace, A., Heycock, C., & Filiaci, F. (2004). “First language attrition and syntactic subjects: A study of Greek and Italian near-native speakers”. International Journal of Bilingualism, 8, 257–277. White, L. (1989). Universal Grammar and Second Language Acquisition. Amsterdam: John Benjamins.

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.17 (1039-1138)

Beyond the syntax of the Null Subject Parameter 

Appendix Sample narratives produced by the L2 learners and native speakers

Intermediate La Caperucita Roja vive con su abuela. La Caperucita va a el bosque. Un día un lobo va a la casa de la abuela y el lobo come la abuela. Cuando la Caperucita Roja va a la casa el lobo es en la cama de la abuela. Él hace de cuenta que él es la abuela. Y...y el lobo salta a la Caperucita Roja y (pro) come a la Caperucita y su abuela. Y un cazador va a la casa y mmm...(pro) ve el lobo muy gordo. El cazador corta el lobo y (pro) salva la abuela y ella. El cazador pone piedras en el estómago del lobo. Y...la abuela dice adiós y todas las personas son muy felices.

Advanced Esto es el cuento de Caperucita Roja. Un día (pro) quería ir a visitar su abuela, porque (pro) tenía que visitarla, y su madre le preparó a...un cesto lleno de agua, y dulces y fruta. Y (pro) la despidió. Y *(pro) fue caminando por un bosque donde (pro) encontró a un lobo. Un lobo hambriento. Y el lobo, mientras ella estaba caminando por el bosque, el lobo corrió antes que ella y (pro) fue a la casa de la abuelita, mientras que la Caperucita Roja estaba recogiendo flores. Y el lobo entró en la casa y (pro) comió a la abuelita. Pero cuando llegó la Caperucita Roja con sus flores, el lobo ya la estaba esperando en la cama. Preparando para comerla. La Caperucita Roja se sorprendió, y (pro) preguntó a su abuela que estaba en la cama por qué (pro) tenía sus orejas grandes, sus ojos grandes....Y el lobo la sorprendió, (pro) saltó de la cama y la Caperucita Roja se asustó y (pro) comenzó a correr. Pero el lobo la comió, y entró un cazador para salvar a los abuelos, porque él ya lo vio, y (pro) sintió mucho ruido de personas gritando, y (pro) mató al lobo, (pro) mató, no (pro) sé. Pero (pro) hizo algo al lobo, y el hombre cortó el estómago del lobo para que la abuelita y la Caperucita Roja pudieran salir del estómago del lobo. Ah...mientras que estaba durmiendo el lobo, el cazador, la abuela, la chica, la Caperucita Roja y también el perro pusieron piedras en el estómago del lobo, así que cuando se despertó el lobo (pro) tenía mucha sed, así que (pro) fue a beber en la...uy, uy cómo (pro) se llama, beber agua en el.....fuente, No. Y mientras (pro) estaba bebiendo (pro) se cayó porque las piedras pesaban mucho, y mientras (pro) estaba cayendo, mientras que (pro) estaba en el agua, el cazador, la abuelita y Caperucita Roja estaban viendo todo, y contando y riendo de lo que (pro) pasó al lobo. Y después de eso la abuelita saludó a su nieta y al cazador y....y se regresaron a casa, contentos todos. Fin.

Near-native Caperucita Roja estaba bailando en la sala. Y (pro) estaba con su abuelita, (pro) estaban muy felices. Y luego, Caperucita se salió de la casa, de una casa. Y.... (pro) se fue por el bosque. (pro) Estaba caminando por el bosque y.... (pro) había un lobo, un lobo muy cómico. Y...un hombre...enojado...También (pro) había unos animalitos. Y Caperucita no vio al lobo, (pro) siguió caminando como si (pro) no fuera nada, no (pro) había nada. Y luego el lobo llegó a una casa, una casita en el bosque. Y el lobo trataba de comer a una mujer, creo que la abuela. (pro) Estaba tratando de comerla, a la abuela. Y después Caperucita llegó a la casa, y (pro) estaba arreglando las flores en un vaso, y (pro) no vio al lobo, y el lobo estaba en la cama como si (pro)

JB[v.20020404] Prn:5/06/2006; 14:41

F: LALD4118.tex / p.18 (1138-1193)

 Silvina Montrul and Celeste Rodríguez Louro

fuera la abuela. Y entonces Caperucita Roja trataba, estaba tratando de hablar con el lobo. (pro) Estaba muy preocupada por algo. Entonces el lobo se saltó de la cama y (pro) trataba de atacar a Caperucita Roja. Y (pro) estaba muy asustada, (pro) tenía mucho miedo. Y entonces llegó el hombre, casero, y (pro) vio que el lobo estaba dormido en la cama, y su estómago estaba muy grande, como si (pro) hubiera comido a la abuela y a Caperucita Roja. Y entonces el casero le cortó el estómago del lobo y (pro) rescató a la abuela y a Caperucita Roja, y ellas estaban muy felices, hasta libres. Y luego, Caperucita Roja ayudó al casero a llenar el estómago del lobo con rocas y la abuela iba a coser el estómago para que el lobo no supiera nada. Y entonces el lobo salió de la casa con el estómago llena de rocas y el casero y Caperucita Roja y la abuela estaba riendo al lobo, *(pro) tenió muchos problemas en caminar. (pro) Estaba, era muy difícil para él caminar con el estómago lleno de rocas. Y después Caperucita Roja y el casero se salieron, (pro) se despidieron a la abuela y...a los animalitos. Al fin todo está bien.

Native speaker control (pro) Había una vez en un pueblecito, en una casita, una nena que se llamaba CR. (pro) Vivía con su madre y (pro) quería mucho a su abuelita. Y entonces un día (pro) pensó, (pro) quiero ir a ver a mi abuelita, y (pro) le quiero llevar una torta, y unos panecitos y unas uvas. Y entonces su mamá le puso su caperucita roja, porque (pro) hacía un poco de viento, y (pro) le dijo, bueno ahora (pro) vete por el camino correcto, (pro) no te pierdas, (pro) no te distraigas y (pro) vete pronto donde la casa de tu abuelita y (pro) llévale esta cesta, con toda la comida. Pero mientras tanto, cuando Caperucita iba caminando, el lobo la vio, y cuando (pro) la vio, (pro) pensó, ummmm qué rico bocado lo que (pro) me voy a comer. Pero entonces Caperucita no sabía que por allí andaba el lobo, y entonces el lobo muy vivaz se fue por un camino más corto para llegar antes a la casa de la abuelita de Caperucita. Y mientras tanto Caperucita lo que hacía (pro) era pararse a recoger todas las flores lindas que encontraba en el camino. Y entonces cuando llegó el lobo a la casa de la abuelita, uauhhhh. (pro) Te voy a comer, y (pro) se la comió. Y ahora (pro) dijo, ahora sí. Entonces, (pro) se puso toda la ropa de la abuelita y (pro) se acostó en la cama. Cuando llegó Caperucita y (pro) la vio, (pro) dijo, (pro) mira abuelita (pro) te traje todas estas cosas. Pero Caperucita no sabía que (pro) era el lobo el que estaba en la cama. Y entonces Caperucita la vio medio extraña, y (pro) decía, abuelita, qué (pro) son esos ojos tan grandes que (pro) tienes. (pro) Es para verte mejor. Y esas orejas tan grandes. (pro) Es para oirte mejor. Y esos dientes tan grandes. (pro) Es para comerte mejor. Ahhh, grita Caperucita. Y entonces por ahí iba pasando un labrador, y el labrador escuchó los gritos de Caperucita, y (pro) corrió y (pro) corrió. Pero cuando (pro) llegó (pro) era muy tarde: el lobo se había comido a Caperucita. Y entonces como el lobo estaba durmiendo, el cazador abrió a el...abrió el estómago y salieron Caperucita y la abuelita bien. Y entonces Caperucita fue a recoger piedras, y *(pro) pusieron piedras en el estómago del lobo, y cuando se despertó el lobo pues (pro) se sentía tan lleno que (pro) fue a tomar agua. Pero cuando (pro) tomó agua con las piedras tan pesado ahhhh, (pro) se fue en el pozo y (pro) se ahogó. Y entonces ahora todos felices y contentos. La abuelita se comió su pan y Caperucita regresó a su casa.

JB[v.20020404] Prn:20/06/2006; 16:33

F: LALD41IN.tex / p.1 (47-146)

Index

A A-chain , , , , , , , , , , , ,  adjectives , , , –, ,  attributive adjectives , , , , , ,  colour adjectives  agent-implying mood  AGR , , , , , , , , , , , ,  Agreement , –, –, , , , –, –, –, , –, , , , –, – pronominal agreement  Arity reduction , ,  auxiliary , , –, , –, , , , –, , , , , ,  auxiliary selection , ,  B base-generated zero topic  Basque –, , , , , , , –, , ,  bilingualism , , ,  bilingual acquisition , , , , , , , , , , , ,  Binding Theory ,  bound variable , ,  C Catalan , , –, –, , , , , –, , ,  Chinese , , , , , , , 

clitics , , , , , , , , –, –, , , , , , , , , , , , , , , , ,  clitic-left dislocation  non-reflexive clitics , ,  object clitics , , , , , , , , –, ,  reflexive clitics , , ,  subject clitics , ,  complementizer , , , , , , , –, , , , ,  agreeing complementizer , –, ,  copula , –, , , –, ,  cross-linguistics cross-linguistic interaction , , , , , ,  cross-linguistic interference , , , , , – D D-marking –, , –,  deficiency  deictic strategy ,  determiner , , , , , , , –, –, , , ,  protodeterminers  discourse reference ,  discourse-pragmatic felicity ,  discourse-pragmatic properties –, , ,  DP , , , , , , –, , , –

Dutch –, –, , , , , , , , , –, , , , , , –, –, –, –, , , , –, –

E Economy Principle –, , , , , , , ,  elsewhere elements , , , , , ,  empty topic  English , –, , , , , , , –, , , , , , , , , , , , , , , , –, , , , , , , , –, , , , , –, –, –, –, –, , , , , , , , , , , , , , , , , , –, , ,  epithets  EPP feature , , , , , , – eventivity constraint ,  evidence frames , , ,  experiencers –, –, –,  expletive , , , , , , , 

F feature checking , , , , , –, , ,  features opposition , , ,  fixed expression 

JB[v.20020404] Prn:20/06/2006; 16:33

F: LALD41IN.tex / p.2 (146-246)

 Index focus , , –, , , –, –, , –, , ,  focus movement  French , , , , –, , , , –, –, –, –, –, –, –, – G German , , , –, , , , , –, , , , , , , , –, , ,  H Hebrew , , ,  hierarchy , , , –, ,  Hungarian , , , , , , , , , ,  Hypotheses Causativity Hypothesis , ,  Configurational Hypothesis  External Argument Requirement Hypothesis , ,  Features Opposition Hypothesis ,  Full Competence Hypothesis , , ,  Full Transfer/Full Access Hypothesis , , ,  Imperative Analogue Hypothesis , – Intermediate State Default Grammars Hypothesis , ,  Maturation Hypothesis , –, , , , ,  Semantic Bootstrapping Hypothesis , – Syntactic Bootstrapping Hypothesis , ,  Truncation Hypothesis , , –, ,  Unique Checking Constraint ,  I I-marking , , , , , , 

Icelandic , , , , ,  identification , , , , , , , , , , , ,  imperative , , –, –, –, , , , ,  indicative , –, , , , , , , – infinitives , , , , , , , , , , –, –, –,  optional infinitives –, –, – root infinitives , , , , , , ,  information structure , ,  insertion , , –, , ,  intensional context  interface , –, , , , , , , , –, –, , , , , , – conceptual-intentional interface  lexicon-syntax interface –, , , – syntax interface  syntax-pragmatics interface , , ,  interference , , , , , –,  Inuktitut , , , , – irrealis , –, ,  Italian –, –, –, , , , , –, , , , , , –, , , –, –, , , , , , , , –, –, , – J Japanese , , –, , , , , , , , , , ,  K Kiche , , , , , ,  L LD-questions 

M markedness , , ,  maturation , , –, , , , , , , , , , , – middle passive  MLU –, , , –, , , , ,  modal reference effect ,  mood , , –, , , , , , – agent-implying mood  movement , , , , , , , , , –, –, –, ,  focus movement  partial movement , –, –, –, –, –, , , ,  wh-movement –, , –, , , , –, –, , –, ,  N narrow syntax , , –, , , , , , , , ,  nativeness  noun –, , , , ,  bare noun , –, , , –, , , , ,  definite noun , , ,  NP-pro , , ,  null arguments , , ,  Null Subject Parameter , –, , , ,  O objects , , , , , , –, , , –, , , , , , ,  lexical objects  null objects , ,  object clitics , , , , , , , , –, ,  object omissions , , , , , , , , – omissions copula omissions , –, , , , , –, 

JB[v.20020404] Prn:20/06/2006; 16:33

F: LALD41IN.tex / p.3 (246-336)

Index  object omissions , , , , , , , , – subject omissions , , ,  operator predicates , ,  P passives , –, , , , –, , , , –,  adjectival passives , , , , – eventive passives  resultative passives , , ,  past participle agreement , , ,  peripheral specifier , , – Portuguese –, , , , , , , , , , , , –,  pragmatic strategy , , , ,  principle B , , , , –, , –,  pro-drop , , –, , , –, , , , , , , , –,  pronouns A-bound pronouns –, –, ,  anaphoric pronouns , ,  clitic pronouns ,  object pronouns , , , , ,  resumptive pronouns ,  strong pronouns , , , , , ,  subject pronouns , , , –, –, , , , ,  psych verbs –, , , ,  Q quantifiers , , , , , , , ,  R raising , , , , , , , , , , 

realisation –, , , , , , ,  reciprocal , , , , –,  referential dependencies , , , , ,  reflexives , , –, , , , , , , ,  reflexive-marked , , ,  reflexivity , , –, ,  relative clauses , , –, , , , , , , –, ,  Russian , , , , 

null subjects , , , , –, , , , , , , , , –, – strong subjects  subject clitics , ,  subject omissions , , ,  subjunctive , , , , , , , – polarity subjunctive , ,  Swedish , , , , , , , 

S se , –, –, , , – anticausative SE , , – inherent SE , – SE-anaphors –, , , ,  SE-constructions , , –, –, ,  self-anaphors –, , –, ,  Sesotho , ,  sí mismo , –, – situation-bound , , , , , , ,  Slovenian , , , , , ,  Spanish –, –, –, , –, , , , , , –, , , , , , , –, –, , –, , –, , , , , , , , –, –, , , –, –, ,  specific referents  specificity , , , , –, – stage , , , –, , , , , , –, , –, , , –, –, , , , , , , , , , , , , , , , , , , , –, ,  stress shift , , , ,  subject relative clause  subject/object asimmetry , , , , , ,  subjects implied subjects 

T topic-drop , , , –, , –, , , ,  trigger , , , ,  ΦP-pro , , , – truncation , , , , –, , , –, , , ,  U unaccusatives , , , , , , , , , , , , , –, ,  underspecified form , ,  unergatives , , , , , , –,  V vulnerability , , , ,  W wh-copying , , , , –, –, – wh-in-situ , , , –, –, –, , , – wh-questions –, –, –, , , –, –, , –, , , – wh-scope marking , , , , , , –, ,  Z zich , , –, –, , ,  zichzelf , –, –

In the series Language Acquisition and Language Disorders the following titles have been published thus far or are scheduled for publication: 41 TORRENS, Vincent and Linda ESCOBAR (eds.): The Acquisition of Syntax in Romance Languages. 2006. xii, 421 pp. 40 DEEN, Kamil Ud: The Acquisition of Swahili. 2005. xiv, 241 pp. 39 UNSWORTH, Sharon, Teresa PARODI, Antonella SORACE and Martha YOUNG-SCHOLTEN (eds.): Paths of Development in L1 and L2 acquisition. In honor of Bonnie D. Schwartz. 2006. viii, 222 pp. 38 FRANCESCHINA, Florencia: Fossilized Second Language Grammars. The acquisition of grammatical gender. 2005. xxiv, 288 pp. 37 MONTRUL, Silvina A.: The Acquisition of Spanish. Morphosyntactic development in monolingual and bilingual L1 acquisition and adult L2 acquisition. 2004. xvi, 413 pp. 36 BARTKE, Susanne and Julia SIEGMÜLLER (eds.): Williams Syndrome across Languages. 2004. xvi, 385 pp. 35 SÁNCHEZ, Liliana: Quechua-Spanish Bilingualism. Interference and convergence in functional categories. 2003. x, 189 pp. 34 OTA, Mitsuhiko: The Development of Prosodic Structure in Early Words. Continuity, divergence and change. 2003. xii, 224 pp. 33 JOSEFSSON, Gunlög, Christer PLATZACK and Gisela HÅKANSSON (eds.): The Acquisition of Swedish Grammar. 2004. vi, 315 pp. 32 PRÉVOST, Philippe and Johanne PARADIS (eds.): The Acquisition of French in Different Contexts. Focus on functional categories. 2004. viii, 384 pp. 31 MARINIS, Theodoros: The Acquisition of the DP in Modern Greek. 2003. xiv, 261 pp. 30 HOUT, Roeland van, Aafke HULK, Folkert KUIKEN and Richard J. TOWELL (eds.): The Lexicon–Syntax Interface in Second Language Acquisition. 2003. viii, 234 pp. 29 FERNÁNDEZ, Eva M.: Bilingual Sentence Processing. Relative clause attachment in English and Spanish. 2003. xx, 294 pp. 28 SHIMRON, Joseph (ed.): Language Processing and Acquisition in Languages of Semitic, Root-Based, Morphology. 2003. vi, 394 pp. 27 SALABERRY, M. Rafael and Yasuhiro SHIRAI (eds.): The L2 Acquisition of Tense–Aspect Morphology. 2002. x, 489 pp. 26 SLABAKOVA, Roumyana: Telicity in the Second Language. 2001. xii, 236 pp. 25 CARROLL, Susanne E.: Input and Evidence. The raw material of second language acquisition. 2001. xviii, 461 pp. 24 WEISSENBORN, Jürgen and Barbara HÖHLE (eds.): Approaches to Bootstrapping. Phonological, lexical, syntactic and neurophysiological aspects of early language acquisition. Volume 2. 2001. viii, 337 pp. 23 WEISSENBORN, Jürgen and Barbara HÖHLE (eds.): Approaches to Bootstrapping. Phonological, lexical, syntactic and neurophysiological aspects of early language acquisition. Volume 1. 2001. xviii, 299 pp. 22 SCHAEFFER, Jeanette C.: The Acquisition of Direct Object Scrambling and Clitic Placement. Syntax and pragmatics. 2000. xii, 187 pp. 21 HERSCHENSOHN, Julia: The Second Time Around – Minimalism and L2 Acquisition. 2000. xiv, 287 pp. 20 KANNO, Kazue (ed.): The Acquisition of Japanese as a Second Language. 1999. xii, 180 pp. 19 BECK, Maria-Luise (ed.): Morphology and its Interfaces in Second Language Knowledge. 1998. x, 387 pp. 18 KLEIN, Elaine C. and Gita MARTOHARDJONO (eds.): The Development of Second Language Grammars. A generative approach. 1999. vi, 412 pp. 17 ARCHIBALD, John: Second Language Phonology. 1998. xii, 313 pp. 16 HANNAHS, S.J. and Martha YOUNG-SCHOLTEN (eds.): Focus on Phonological Acquisition. 1997. v, 289 pp. 15 BRINKMANN, Ursula: The Locative Alternation in German. Its structure and acquisition. 1997. x, 289 pp. 14 CLAHSEN, Harald (ed.): Generative Perspectives on Language Acquisition. Empirical findings, theoretical considerations and crosslinguistic comparisons. 1996. xxviii, 499 pp. 13 ALLEN, Shanley E.M.: Aspects of Argument Structure Acquisition in Inuktitut. 1996. xvi, 244 pp. 12 JUFFS, Alan: Learnability and the Lexicon. Theories and second language acquisition research. 1996. xi, 277 pp. 11 YIP, Virginia: Interlanguage and Learnability. From Chinese to English. 1995. xvi, 247 pp. 10 LAKSHMANAN, Usha: Universal Grammar in Child Second Language Acquisition. Null subjects and morphological uniformity. 1994. x, 162 pp.

9 8 7 6 5 4 3 2 1

ADONE, Dany: The Acquisition of Mauritian Creole. 1994. xii, 167 pp. HOEKSTRA, Teun and Bonnie D. SCHWARTZ (eds.): Language Acquisition Studies in Generative Grammar. 1994. xii, 401 pp. MEISEL, Jürgen M. (ed.): Bilingual First Language Acquisition. French and German grammatical development. 1994. vi, 282 pp. THOMAS, Margaret: Knowledge of Reflexives in a Second Language. 1993. x, 234 pp. GASS, Susan M. and Larry SELINKER (eds.): Language Transfer in Language Learning. Revised edition. 1992. x, 236 pp. ECKMAN, Fred R. (ed.): Confluence. Linguistics, L2 acquisition and speech pathology. 1993. xvi, 260 pp. EUBANK, Lynn (ed.): Point Counterpoint. Universal Grammar in the second language. 1991. x, 439 pp. HUEBNER, Thom and Charles A. FERGUSON (eds.): Cross Currents in Second Language Acquisition and Linguistic Theory. 1991. viii, 435 pp. WHITE, Lydia: Universal Grammar and Second Language Acquisition. 1989. xii, 198 pp.

The Acquisition of Syntax in Romance Languages (Language Acquisition and Language Disorders)

Language Processing and Acquisition in Languages of Semitic, Root-based, Morphology (Language Acquisition & Language Disorders)

The Acquisition of Swedish Grammar (Language Acquisition & Language Disorders)

First Language Acquisition of Morphology and Syntax: Perspectives across languages and learners (Language Acquisition and Language Disorders)

Telicity in the Second Language (Language Acquisition and Language Disorders)

Lexicon-Syntax Interface in Second Language Acquisition

Second Language Acquisition of Articles: Empirical findings and theoretical implications (Language Acquisition and Language Disorders)

Input and Evidence: The Raw Material of Second Language Acquisition (Language Acquisition & Language Disorders)

The Acquisition of French: The development of inflectional morphology and syntax in L1 acquisition, bilingualism, and L2 acquisition (Language Acquisition and Language Disorders)

Language acquisition and learnability

Language Acquisition and Learnability

Language Acquisition and Learnability

The Acquisition of French in Different Contexts: Focus on Functional Categories (Language Acquisition and Language Disorders)

Language Acquisition across Linguistic and Cognitive Systems (Language Acquisition and Language Disorders)

Developmental Theory And Language Disorders (Trends in Language Acquisition Research)

Investigations In Instructed Second Language Acquisition (Studies on Language Acquisition)

Third Language Acquisition and Universal Grammar (Second Language Acquisition)

Speaking and Instructed Foreign Language Acquisition (Second Language Acquisition)

The L2 Acquisition of Tense-Aspect Morphology (Language Acquisition and Language Disorders)

The Acquisition of Swahili (Language Acquisition and Language Disorders, Volume 40)

The L2 Acquisition of Tense-Aspect Morphology (Language Acquisition and Language Disorders)

Language Acquisition: The Age Factor (Second Language Acquisition)

Vocabulary Learning Strategies and Foreign Language Acquisition (Second Language Acquisition)

Language acquisition: core readings

Language Acquisition: Core Readings

Language Acquisition By Eye

First Language Acquisition

The Psychology of the Language Learner: Individual Differences in Second Language Acquisition (Second Language Acquisition Research)

Language Acquisition: Studies in First Language Development

Language acquisition after puberty

Language Acquisition After Puberty

The Acquisition of Syntax in Romance Languages (Language Acquisition and Language Disorders)

Language Processing and Acquisition in Languages of Semitic, Root-based, Morphology (Language Acquisition & Language Disorders)

The Acquisition of Swedish Grammar (Language Acquisition & Language Disorders)

First Language Acquisition of Morphology and Syntax: Perspectives across languages and learners (Language Acquisition and Language Disorders)

Telicity in the Second Language (Language Acquisition and Language Disorders)

Lexicon-Syntax Interface in Second Language Acquisition

Second Language Acquisition of Articles: Empirical findings and theoretical implications (Language Acquisition and Language Disorders)

Input and Evidence: The Raw Material of Second Language Acquisition (Language Acquisition & Language Disorders)

The Acquisition of French: The development of inflectional morphology and syntax in L1 acquisition, bilingualism, and L2 acquisition (Language Acquisition and Language Disorders)

Language acquisition and learnability

Language Acquisition and Learnability

Language Acquisition and Learnability

The Acquisition of French in Different Contexts: Focus on Functional Categories (Language Acquisition and Language Disorders)

Language Acquisition across Linguistic and Cognitive Systems (Language Acquisition and Language Disorders)

Developmental Theory And Language Disorders (Trends in Language Acquisition Research)

Investigations In Instructed Second Language Acquisition (Studies on Language Acquisition)

Third Language Acquisition and Universal Grammar (Second Language Acquisition)

Speaking and Instructed Foreign Language Acquisition (Second Language Acquisition)

The L2 Acquisition of Tense-Aspect Morphology (Language Acquisition and Language Disorders)

The Acquisition of Swahili (Language Acquisition and Language Disorders, Volume 40)

The L2 Acquisition of Tense-Aspect Morphology (Language Acquisition and Language Disorders)

Language Acquisition: The Age Factor (Second Language Acquisition)

Vocabulary Learning Strategies and Foreign Language Acquisition (Second Language Acquisition)

Language acquisition: core readings

Language Acquisition: Core Readings

Language Acquisition By Eye

First Language Acquisition

The Psychology of the Language Learner: Individual Differences in Second Language Acquisition (Second Language Acquisition Research)

Language Acquisition: Studies in First Language Development

Language acquisition after puberty

Language Acquisition After Puberty

Recommend Documents