Syntactic Complexity from a Language Acquisition Perspective [1 ed.] 9781443893527, 9781443851770

Citation preview

Syntactic Complexity from a Language Acquisition Perspective

Syntactic Complexity from a Language Acquisition Perspective Edited by

Elisa Di Domenico

Syntactic Complexity from a Language Acquisition Perspective Edited by Elisa Di Domenico This book first published 2017 Cambridge Scholars Publishing Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2017 by Elisa Di Domenico and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-5177-9 ISBN (13): 978-1-4438-5177-0





TABLE OF CONTENTS

Acknowledgements ......................................................................................vii List of Contributors .................................................................................... viii Chapter One .................................................................................................... 1 Introduction Elisa Di Domenico Part I Chapter Two ................................................................................................. 28 On the Acquisition of Complex Derivations with Related Considerations on Poverty of the Stimulus and Frequency Adriana Belletti Chapter Three ............................................................................................... 49 Which and How Many Questions in the Acquisition of Italian Elena Pagliarini and Maria Teresa Guasti Chapter Four ................................................................................................. 63 On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition Emanuela Sanfelici, Petra Schulz and Corinna Trabandt Chapter Five ............................................................................................... 105 Auxiliaries and Verb Classes: The Complexity of Predicates in the L1 Acquisition of Italian Paolo Lorusso Part II Chapter Six ................................................................................................. 142 Syntactic Complexity and Bilingualism: How (A)typical Bilinguals Deal with Complex Structures Cornelia Hamann, Solveig Chilla, Natalia Gagarina and Lina Abed Ibrahim



vi

Table of Contents

Chapter Seven ............................................................................................. 178 Object Realization across Generations: A Closer Look on the Spontaneous Speech of Portuguese First and Second Generation Migrants Cristina Flores, Esther Rinke and Cecília Azevedo Chapter Eight .............................................................................................. 206 Cross- Linguistic Influence in the Bilingual Acquisition of Object Clitics: A Matter of Complexity? Petra Bernardini and Monica Timofte Chapter Nine ............................................................................................... 232 L2 Acquisition at Interfaces: Pronouns and Referentiality in L2 Finnish Lena Dal Pozzo Index ........................................................................................................... 255







ACKNOWLEDGEMENTS

In February 2015, two events took place at the Università per Stranieri di Perugia: the 41st Incontro di Grammatica Generativa and the workshop ‘More than one language in the brain’. Most of the contributors to this volume (at least one author for each paper) took part in these events, where the idea of this volume took shape. I am sincerely grateful to all the contributors, as well as to our anonymous reviewers. Special thanks go to Simona Matteini.



LIST OF CONTRIBUTORS

Lina Abed Ibrahim

Carl von Ossietzky Universität Oldenburg

Cecília Azevedo

Universidade do Minho

Adriana Belletti

Università di Siena/ Université de Genève

Petra Bernardini

Lunds Universitet

Solveig Chilla

Pädagogische Hochschule Heidelberg

Lena Dal Pozzo

Università di Firenze/ Pontifícia Universidade Católica do Rio de Janeiro Universidade do Minho

Cristina Flores

Maria Teresa Guasti

Zentrum für Allgemeine Sprachwissenschaft Berlin Università di Milano Bicocca

Cornelia Hamann

Carl von Ossietzky Universität Oldenburg

Paolo Lorusso

Università di Firenze

Elena Pagliarini Esther Rinke

Università di Milano Bicocca / Universitat Pompeu Fabra, Barcelona Goethe Universität Frankfurt

Emanuela Sanfelici

Goethe Universität Frankfurt

Petra Schulz

Goethe Universität Frankfurt

Monica Timofte

Universitatea Stefan cel Mare Suceava

Corinna Trabandt

Goethe Universität Frankfurt

Natalia Gagarina





CHAPTER ONE INTRODUCTION ELISA DI DOMENICO

1. Linguistic Complexity from the Acquisitional Perspective The issue of linguistic complexity is a main topic in the recent relevant literature, with researchers from different linguistic backgrounds progressing in the attempt to identify factors, measures and metrics of complexity in different linguistic sub-domains (morpho- syntax, lexicon, phonology, semantics, discourse). We may indeed look at linguistic complexity from many perspectives, ranging from historical linguistics and evolution, to formal theoretical linguistics, comparative syntax (including the comparative study of Sign Languages and Creoles), psycholinguistics, neurolinguistics and computational linguistics. 1 The peculiarity of the acquisitional perspective lies in the fact that it offers a particularly grounded and non- aprioristic starting point for any consideration on linguistic complexity, namely the simple, basic axiom in (1): (1) A more complex item is expected to be acquired later than a less complex one.2 Reading (1) in the direction suggested by its formulation, acquisitional data can be taken as evidence for formal analyses of an item as more or

 1

For a representation of this variety of approaches see, for instance, the collection of studies in Newmeyer and Preston (2014). 2 Some version of (1) is more or less explicitly assumed in all the studies here reviewed or collected, though in some cases not as an axiom but rather as a hypothesis to be demonstrated (e.g. Jakubowicz, 2005, 2011; Moscati and Rizzi 2014; but see Chapter 2, this volume). In (1) ‘item’ is to be intended as any linguistic element or process, with its related structure and the computations it involves, and is thus assumed to hold true also beyond the syntactic domain.



2

Chapter One

less complex. Reading (1) in the opposite direction, acquisitional data can be the input to formal linguistic analyses: if an item is acquired early it should not be analyzed as particularly complex, while, conversely, if it is acquired late it should be considered complex. 3 To this end, the comparative study of the comprehension and productions of children at different stages of development is a fundamental discovery tool when looking at complexity from the acquisitional point of view.4 But then the challenge will be to uncover what makes an item complex, where exactly complexity lies, what kind of complexity is involved. The basic axiom in (1) thus allows, and at the same time forces, the tight, bi- directional interplay between acquisitional findings and formal/ theoretical considerations which characterizes most studies on complexity from the acquisitional perspective, including the contributions collected in this volume, giving novel substance to the long- standing (Chomsky 1965), though not at all obvious, view of linguistic theory as a theory of language acquisition. A systematic recollection of the discoveries prompted by this approach is far beyond the scope of the present lines: in what follows I will just mention a few representative examples from the literature to illustrate this way of working. These examples, as well as the contributions collected in this volume, deal with syntactic complexity, here intended, following Hale and Keyser (1993), as s- syntax (sentence syntax) as well as l-syntax (lexical syntax). A finding that goes back to the late ‘60s, in Carol Chomsky’s (1969) seminal work, for instance, is that subject control structures with verbs of the promise type (as in (2)) are mastered surprisingly late by children, compared to object control structures (as in (3)) which are mastered earlier: (2) John promised Bill [PRO to leave] (3) John ordered Bill [PRO to leave] Carol Chomsky’s proposal to account for this data is that children strictly adhere to the Minimal Distance Principle (Rosembaum 1967), which bars subject control, and only at a later stage can master exceptions to this principle. In the same spirit, Belletti and Rizzi (2013) propose a complexity factor (intervention) at the root of subject control structures as

 3

Both directions can be pursued at the same time, as in Chapter 4, this volume . Different children, as in cross-sectional studies, or the same child or children, as in longitudinal studies. Stages of development can be identified by age or by Mean Length of Utterance (MLU).

4



Introduction

3

well as of other structures with belated acquisition (such as object relatives and passive), interpreting the acquisition of ‘exceptions’ as the capability of overcoming this complexity factor, possible once peculiar devices are available to the child. Defining intervention as the difficulty of computing a local relation across an intervener able to bear the same relation,5 object relatives are more complex than subject relatives, since in the former the subject intervenes between the head of the relative and its copy in its first merge position (4), while no intervener blocks the relation between the head of the relative and its copy in subject relatives (5):6 (4) Show me the cow that the lion is wetting (5) Show me the cow that is wetting the lion Still, object relatives are possible, though harder to process (Warren and Gibson, 2002 among many others) than subject relatives in adult language, as well as in child language after a certain age. Friedmann, Belletti and Rizzi (2009) proposed that adults and elder children are able to cope with intervention if the intervener and the target are in an inclusion relation as for their featural array,7 as shown in (6), where [+ R] is the scope discourse feature attracting the relative head, while [+NP] expresses its lexical restriction:8

 5

Following the general locality principle known as Relativized Minimality (Rizzi 1990; 2004) 6 Indeed, subject relatives are not problematic for the same population, i.e. the Hebrew speaking children aged 3.7 – 5.0 tested in Friedmann, Belletti and Rizzi (2009). The asymmetry between subject and object relatives in acquisition is a well - known observation (since Brown 1972), and equally well known is the similar acquisitional asymmetry differentiating subject and object which- questions (see e.g. Chapter 3, this volume, and the references quoted there). Besides replicating these asymmetries, Friedmann, Belletti and Rizzi (2009) showed that, in some cases, as e.g. in the object relative below, the difficulty children experience with object questions and object relative clauses disappears: (i) Show me the cow that he is wetting 7 For some differences between the featural approach to Relativized Minimality illustrated here and the one developed in Starke (2001), see Belletti and Rizzi (2013). 8 If the target and the intervener are identical as for their featural array, the resulting structure is ungrammatical in the adult grammar, as in weak wh- islands (i), while if they are in a disjunction relation the derived structure is well formed (ii): (i) * How do you wonder [who behaved ____] ?



4

Chapter One

(6) Show me the cow that the lion is wetting < the cow> [+R, +NP] [+NP] [+R, +NP] The capability of mastering the inclusion relation is thus what differentiates the adult grammar from the grammar of children below 6. A problem of intervention also arises in passive and control structures, but in these cases a different way of overcoming the problem is adopted. In passive, if, as would follow from Baker’s (1988) Uniformity of Theta Assignment Hypothesis, the external argument is syntactically projected, it intervenes in the chain created by the object in its movement to the clausal EPP position. Collins (2005) proposed that this intervention problem is overcome by the preliminary movement of a VP chunk containing the verb and the object (smuggling). The object will then move from the derived position of the VP chunk: (7) [TP

[VPV DP] by [vP DP [VP V DP]] ________ ____________

The late acquisition of passive can be thus explained assuming the complexity of the smuggling operation which needs some time to develop. Once smuggling is at disposal, however, children (as well as adults) seem to resort to it quite extensively: this is one of the main findings in Belletti and Contemori (2010) and Contemori and Belletti (2014), where elder children and adults often produced what the authors call a Passive Object Relative (POR), as (8.b), instead of the object relative (8.a), which the experimental condition was meant to elicit:9 (8) a.

b.

Vorrei essere il bambino che la mamma copre I would rather be the child that the mother covers Vorrei essere il bambino che è coperto dalla mamma I would rather be the child that is covered by the mother

The procedure that moves a verbal chunk across an intervener is not at

(ii) How do you wonder [John behaved ____] ? In the sentence in footnote 6 above, the elimination of the [+NP] feature from the intervener (he, instead of the lion as in (4) and (6)) renders the target and the intervener (the cow and he) disjunct as for their feature composition. 9 This suggests that avoiding intervention through smuggling is less complex than mastering the inclusion relation, which is required by object relatives.

Introduction

5

all confined to passive.10 As Belletti and Rizzi (2013) propose, it might be the procedure allowing subject control, which should be banned under Relativized Minimality (subsuming Rosembaum’s, 1967 Minimal Distance Principle), the object intervening between the subject and PRO:11 (9) John promised Bill [PRO to leave early] In their analysis, smuggling is a necessary ingredient of the procedure deriving the lexically decomposed (à la Hale and Keyser, 1993) structured meaning of subject control verbs of the promise type, making the subject the closest controller for PRO. They start from the assumption that promise allows the paraphrase in (10), with the light verb make + the nominal element promise, which, incorporating into the light verb would produce the verb promise itself: (10) John made Bill the promise [PRO to go] Bill is a kind of benefactive of the promise, and the benefactive relation is assumed to be mediated by a benefactive particle- like functional head (ben), yielding the structure in (11): (11) John Vmake [Bill ben [promise [PRO to go]]] The presence of ben makes the configuration not sufficiently local for the incorporation of the noun promise into the light verb to take place, and locality is restored through movement of the verbal chunk [promise [PRO to go] which smuggles the noun promise to a position suitable for incorporation: (12) John Vmake [promise [PRO to go]] [Bill ben t] The surface word order is then obtained through extraposition of the infinitive: (13) John promise + Vmake [[ tpromise tinfinitive] [Bill ben t]] [PRO to go] In (13) the object Bill does not c-command PRO, and thus the closest potential controller is the subject.

 10

See Belletti and Rizzi (2013) for a sketchy list. As the authors underline, this analysis is not committed to a movement approach to control. 11



Chapter One

6

Under this analysis a single factor (intervention) and two mechanisms to avoid it (mastering inclusion and smuggling) explain long dating and robust acquisitional findings in three distinct complex domains: object relatives, passive, and subject control.12 But the tight interplay between acquisitional findings and formal linguistic analyses, as we said, can also work in the opposite direction: the analysis of an item as complex can be confirmed by its belated acquisition. Moscati and Rizzi (2014) analyze three different types of agreement configurations in Italian (Determiner – Noun, Subject – Verb, Clitic – Past Participle) arguing that they can be ranked on a scale of complexity in terms of the movement operations that they involve and of the derived representations at interfaces. Hence they should be mastered by the child at different stages: a prediction that is confirmed both by data available from previous corpus studies and by the data experimentally collected by the authors. The three types of agreement are illustrated in (14) (from Moscati and Rizzi, 2014:68): (14) a. D-N agr:

Le Thef.plur

case housesf.plur

b. Subj- V agr:

Gianni Gianni3P.sing

parte leaves3P.sing

c. Cl- PstPart agr:

Gianni le Gianni themf.plur

ha viste has seenf.plur

The simplest agreement type is D – N (14.a) in that it involves no movement at all, and is seen by the authors as a morphological reflex of external merge putting D and N together. Subj- V agreement (14.b) involves movement of the subject from its thematic position in the vP to a functional head in the clausal architecture bearing unvalued Phi- features (labeled Subj by the authors, following Rizzi, 2006), a criterial position where the subject is frozen in place (Rizzi, 2006). The Cl- Pst Part agreement (14.c) is the most complex configuration, where agreement is checked ‘in passing’ of the clitic through the position where it triggers agreement on the past participle (an aspectual head, as in Cinque, 1999) before reaching its final destination, i.e. the clitic position in the functional structure of the clause (Kayne, 1989; Belletti, 2006). Data gathered through the experimental paradigm of the Forced Choice of Grammatical Form

 12



Plus object which- questions, see footnote 6 above.

Introduction

7

(where the child has to choose which one of the two sentences he is faced with is grammatical) confirm the predictions: while children at the age of 3 have a virtually perfect knowledge of D-N agreement, their performance is still inaccurate with Subj- Verb agreement. At the age of 4, their accuracy with Subj- Verb agreement reaches a level in which it is not significantly different from their accuracy with D-N agreement. The most problematic kind of agreement is the Cl- Pst Part, where a significant proportion of wrong answers is still present at the age of 4. Two factors of complexity are singled out by this study: the general cost associated with a movement operation, and a second factor related to the landing site of the moved constituent.13 The idea that movement (or ‘internal merge’, following Chomsky, 2001) operations are costly in acquisition is at the base of the Derivational Complexity Metric (Jakubowicz, 2005, 2011) in (15), associated to the Derivational Complexity Hypothesis, which states that less complex derivations are input convergent, i. e. emerge in acquisition, before more complex ones: (15) Derivational Complexity Metric (DCM) a. Merging Įi n times gives rise to a less complex derivation than merging Įi (n + 1) times b. Internal Merge of Į gives rise to a less complex derivation than Internal Merge of Į + ȕ. [Jakubowicz, 2011: 340] In Jakubowicz (2011), the predictivity of DCM has been tested examining the acquisition of wh- questions in French by typically developing children and by children with Specific Language Impairment (SLI henceforth, Gopnik and Crago, 1991 and much subsequent work).14 First of all, different types of root wh- questions and direct wh- questions from embedded clauses (including some types attested only in child French

 13

‘[…]Satisfaction ‘in passing’ is more complex than satisfaction at the head of a chain because checking the correctness of the configuration involves a kind of ‘reconstruction’ in the position of the trace’ [Moscati and Rizzi 2014:69]. The authors also suggest an additional potentially relevant element of complexity differentiating Cl- Past Participle agreement from Subj- Verb agreement, in that in the former case the movement chain spans over two distinct phases, in the sense of Chomsky (2001). 14 Specific Language Impairment is commonly defined as a language acquisition disorder in the absence of other mental, neurological or perceptual deficits.



Chapter One

8

and not in adult French) are analyzed and ranked according to DCM.15 According to (15.a), wh-questions involving no internal merge of the whelement, as wh- in situ questions (16), will be less complex than fronted wh-questions (17); questions that involve a smaller number of the internal merge operation (17) will be less complex than those involving a higher number of internal merge operations (19); according to (15.b), wh-fronting without V-to-C (17) will be less complex than wh-fronting with V-to-C (18): (16) Tu as vu qui? ‘You saw who(m)?’ (17) Quii tu as vu ti? Lit.’Who(m) you have seen?’

(18) Quii ask tu tk vu ti? Lit. ‘Who(m) have you seen?’ (19) Quii pense-t- elle [ti que tu as vu tì]? Lit. ‘Who thinks she that you saw?’

[Jakubowicz, 2011]

The prediction suggested by the Derivational Complexity Hypothesis is that the less complex (in terms of DCM) types of questions will emerge first. This prediction was then tested through an elicited production task (Strik, 2008 and the references quoted there), administered to three groups of typically developing children (aged 6, 4 and 3), two groups of SLI children (aged 8 and 11) and a control group of adults. Briefly summarizing the results to the point relevant for the present discussion, data revealed that in the younger typically developing groups and especially in the SLI group, the production of long distance direct questions was not consistent with adult grammar: children tried to produce these sentences but often failed to

 15

Four types of wh- question in root clauses and seven in long distance direct whquestions from embedded clauses are analyzed and ranked in this study. Of the latter group, three belong to child French: special partial movement (i), partial movement (ii) and wh- copying upstairs with wh- in situ downstairs: (i) Billy a dit [quii Grenouille a vu ti]? Lit.’ Billy said who Frog saw’ (ii) Qu’est-ce que Billy a dit quii Grenouille a vui? Lit. ‘What did Billy say who(m) Frog saw?’ (iii) Oùi Billy a dit te [que Canard a acheté le cadeau oùì] Lit. ‘Where Billy said that Duck bought the present where?’ See Jakubowicz, 2011: 341ff.



Introduction

9

do so correctly or adopted various kinds of avoidance strategies.16 This confirms the general validity of the DCM and sheds light on another aspect that characterizes the study of complexity from the acquisitional perspective: the comparison of L1 typical acquisition with other modes of acquisition. In Jakubowicz (2011), the comparison is between typically developing and SLI children acquiring their L1. Productions and avoidance strategies of SLI children prove to be in line with those of typically developing children at younger ages in this study (but see below).17 This is one of the possible outcomes of the comparison between typical and SLI acquisition, in both production and comprehension, as Belletti and Guasti (2015) note: in other cases the two modes of acquisition do not seem to proceed in a parallel way. SLI children seem to have problems in specific domains, such as verbal tense- agreement morphology (Cipriani et al. 1998, Clahsen et al. 1997, Rice and Wexler 1996) unnoticed in typical development. Specific difficulties have been interpreted as cues for the identification of different types of SLI (Friedmann and Novogrodsky, 2004, 2011). Some particularly complex domains (e. g. clitics,) are so challenging for SLI acquirers that they may be considered a clinical marker (Vender, Guasti Garraffa and Sorace, 2014 a. o.). In some complex domains, furthermore, (e.g. wh- questions, Hamann, 2006, and the above discussed Jakubowicz, 2011, or object relative clauses, Friedmann, Yachini and Szterman, 2015, Hamann and Tuller, 2015 a. o.) the difficulty may persist over the years in SLI speakers, leading to stagnation: this may reveal, as Belletti and Guasti (2015) note, that there might be some linguistic properties that need to be acquired at a critical time. Some studies (e.g. Håkanson and Nettelblatt, 1996; Paradis, 2004) point to similarities between SLI and second language acquisition, while other studies (Vender, Guasti, Garraffa and Sorace 2014) also highlight substantial differences in the two acquisition modes (see Chapter 6, this volume, for extensive discussion and new data). When we come to second language acquisition, different, interrelating conditions, such as age of onset of acquisition, cross- linguistic influence (or ‘transfer’ from the L1) and, possibly, input differences, contribute to partially differentiate this mode of acquisition from typical L1 acquisition as well.

 16 Adjunction strategies, parataxis, indirect questions and the three question types described in footnote 15 above. 17 Jakubowicz (2011:340) claims indeed that the Derivational Complexity Hypothesis (with its associated Metric): ‘[..] applies to different conditions of language acquisition: L1, L2, SLI, etc.’



Chapter One

10

A clear way in which all these conditions are connected to the issue of linguistic complexity is that their effect is assumed to emerge particularly in linguistically complex domains. A case in point is represented by object clitics. Seminal work by Hamann and Belletti (2006), for instance, brought to light a peculiar production error pattern that differentiates child and adult L2 acquirers, on one side, from monolingual typical and SLI children on the other, in the acquisition of object clitics in French. 18 Romance object clitic pronouns are delayed in all acquisition modes, and this delay is traced back by the authors to their complex derivation: they are first merged as maximal projections (XPs) and end up as heads in a position high in the clausal architecture. Hamann and Belletti (2006) note that while omissions of object clitics (in turn reflecting their complexity) are found in all the acquisitional modes examined, placement errors of the type in (20) were found only in (child and adult) L2ers: (20) a. c’est à moi, le it’s to me, him/he ‘it’s mine, that one’

Elisa, German- French child L2er

b. ça a m’etranglé that has me strangled ‘that strangled me’ It is worth underlining that in the case just described the differences between L2ers on one side, and (typical and SLI) L1 acquirers on the other, do not question the complexity of object clitics, which are and remain a complex domain in all the acquisition modes examined.19

 18

Following Hamann and Belletti (2006), we will distinguish bilingualism (or 2L1, i.e. the simultaneous acquisition of two languages from birth) from child L2 acquisition (where a second language starts to be acquired after a first one in the early ages), from adult L2 acquisition. Child L2 acquisition may be further differentiated in early and late child acquisition. Some authors prefer to distinguish simultaneous and sequential (child and adult) bilingualism. Whatever the terminology adopted, the differentiating factor in this typology is age of onset of acquisition of the second language, which interacts with cross- linguistic influence: the latter is not expected to occur in (simultaneous) bilingualism (given the hypothesis that the two linguistic systems develop separately, although simultaneously, in the bilingual mind) unless input from one of the languages can be analyzed through the grammar of the other language (Hamann and Belletti 2006:45, following Hulk and Müller, 2000). 19 As we will see in Chapter 7, this volume, object clitics are a complex domain for heritage speakers as well.A ‘heritage language’ is a first language (L1) which is not



Introduction

11

A less clear case is a domain at the syntax – pragmatics interface, where answering strategies, which seem to be in place early on in first language acquisition (Belletti 2007) are subject to protracted transfer from the L1, even at advanced stages, in L2 acquisition.20 Under the assumption that cross-linguistic influence manifests itself in complex domains, the syntax – pragmatics interface is thought of as a complex, or ‘vulnerable’ domain in L2 acquisition (Hulk and Müller, 2000, Sorace and Filiaci 2006, White 2011 a. o.; see also Chapter 8 and Chapter 9, this volume, for discussion and new data).

2. Syntactic Complexity in Different Modes of Acquisition The contributions collected in this volume all deal with syntactic complexity from an acquisitional perspective. The chapters in Part I examine peculiar syntactic aspects in typical L1 acquisition, while in Part II other modes of acquisition are also considered, whether singularly or comparatively: SLI and heritage L1 acquisition, bilingual typical and SLI acquisition, child and adult L2 acquisition. Novel factors of complexity are proposed and characterized (while the role of other factors is put into question); some complex domains are comparatively analyzed and possibly ranked in terms of complexity; new elicitation procedures concerning complex structures are proposed and tested; new data concerning the way in which different populations of acquirers deal with some complex domains are provided. The languages considered, sometimes comparatively, range from Italian, to German, French, Romanian, European Portuguese and Finnish.

 the dominant language in the speaker’s life and environment. Heritage speakers acquired their first language at home, but live in an environment that speaks another language. 1st generation heritage speakers acquired their L2 as adults (e.g. they migrated to another country). Their heritage language is typically acquired, but can be subject to attrition from the dominant language. 2nd generation heritage speakers are, strictly speaking, bilinguals or child L2 acquirers (see footnote 18 above). Their L1, the heritage language, can suffer from limited input and use, as well as from cross- linguistic influence from the dominant language. 20 Tsimpli (2014) assumes that early acquired phenomena (in turn related to macroparameters) can differentiate between simultaneous and (early) successive bilingualism with an advantage for the former group, while late and very late acquired phenomena reveal very similar (high or low) performance across bilingual groups, differentiating them from monolinguals.



12

Chapter One

2.1 Insights from Typical L1 Acquisition In Chapter 2, Adriana Belletti faces the issue of the factors determining complexity in acquisition through a peculiar, second level, kind of data stemming from the comparison of recent findings from typical L1 acquisition of Italian with adult Italian data in the same experimental settings, as well as with corpora of adult Italian (hence with what may be considered children’s input) in three different empirical domains (relative clauses, various types of passive, post-verbal subjects with unaccusatives). Adhering to the idea that what is more complex is acquired later, the author shows, through this comparison, that pre-theoretical complexity factors such as frequency in the input or number of words in a sentence do not seem to play a role in acquisition, challenging approaches to language acquisition framed in terms of analogy or imitation: children don’t do what adults do in the same experimental setting nor do they do what they hear most, but they sometimes do, instead, what might be thought of as more complex in terms of pre-theoretical complexity factors. As discussed in Section 1 above, for instance, PORs are used as a way to avoid intervention when an object relative is elicited, by adults and by elder children (Belletti and Contemori, 2010; Contemori and Belletti, 2014).21 When detailing the kind of PORs used in elicited production, sicausative passives (as (21)) are the most frequent kind of POR used by children in experimental conditions (Contemori and Belletti, 2014), though they are very rare in naturalistic corpora (Belletti and Chesi, 2014), and also very rare, in fact absent, in adults’ elicited productions: (21) Il bambino che si fa pettinare dalla mamma The kid that makes himself(cl) comb by the mum The kind of passive mostly resorted to by adults is a reduced POR (as (22)), not resorted to by children: (22) Il bambino che è/viene pettinato dalla mamma The kid that is/comes combed by the mum As for number of words, reduced PORs should be thought of as less complex than their non- reduced counterpart. Furthermore, reduced PORs are more frequent then non- reduced ones in what might be thought of as



21 PORs in general are instead quite rare in naturalistic corpora (Belletti and Chesi, 2014)



Introduction

13

the children’s input,22 so they should be preferred by children, contrary to fact. Then the task for the linguist is to uncover which formal factors make reduced PORs complex for children, and which ones make si- causative PORs simpler, and hence preferred by children: principled reasons (e.g. labeling, Belletti, forthcoming) may favor access to the latter, while the complexity of the reduction operation may disfavor access to reduced PORs. In the last part of the chapter another seemingly complex, but early acquired phenomenon is considered: young children’s sensitivity to the Definiteness Effect on post- verbal subjects of unaccusatives (Vernice and Guasti, 2015). As the author details, this constraint is quite complex to induce from the Italian input data, which are rather opaque in this respect.23 Children’s sensitivity to this constraint shows that they single out the unaccusative verb class from early on (see also Chapter 5, this volume, and the references quoted there) and from early on they master the definite/ indefinite distinction. This distinction appears to be well rooted in their internal grammar, suggesting that it does not need to be learned. This points to the conclusion that internal grammatical factors play a crucial role in language development. In Chapter 3, Elena Pagliarini and Maria Teresa Guasti discuss (and characterize as for its developmental pattern) a novel locality mechanism to account for the acquisitional asymmetry between subject and object whichquestions (see footnote 6 above), which they name ‘Agree intervention’. They start from the observation that in Italian, in object which- questions, the subject does not, in fact, intervene between the wh-object and its copy since it is typically realized post- verbally: (23) Quale fatina stanno spingendo le signore? Which fairy are pushing the ladies? Which fairy are the ladies pushing? Still, object which-questions are challenging for Italian speaking children, as they are for English (or Hebrew) speaking children (De Vincenzi at al. 1999). Following Guasti, Branchini and Arosio (2012), Pagliarini and Guasti argue that the locality violation (intervention) occurs when the Agree relation between the inflectional head and the subject in its thematic

 22 Though PORs are rare in naturalistic corpora, in child directed speech reduced PORs are overwhelmingly the most attested kind of POR used by adults. 23 While post-verbal subjects are widespread in Italian, the Definiteness Effect (i.e. the requirement that the subject be indefinite) only holds for post – verbal subjects of unaccusatives under particular discourse conditions, i.e. in all new sentences.



14

Chapter One

position is established, and the copy of the moved wh-object intervenes: 24 (24)[CP Quale fatina stanno [AgrOP spingendo] [vP le signore]]? Which fairy are pushing the ladies? Agree intervention is sensitive to different features with respect to what they call ‘Argument intervention’ (the kind of intervention discussed in Section 1 above for object relatives, passive and control, according to Belletti and Rizzi, 2013): formal features such as, e.g. case, as opposed to features involved in establishing reference, e.g. [+ NP]. To this end, which- and how many- questions should be equally challenging for children because the featural difference they entail (which- questions asking for identity of the referent (DP), how many- questions for its numerosity (QP)) is not relevant for Agree intervention. These predictions established, the paper presents and discusses two experimental studies on Italian speaking children, one comparing the comprehension of which- and how many- questions, and one comparing object questions with pre and post-verbal subjects. Results confirm the predictions: the subject/ object asymmetry holds in the same fashion for which- and how many- questions, despite the fact that they ask for different referents; the scores obtained by children in which- questions with pre- verbal subjects doubled the scores obtained with which- questions with post- verbal subjects. Two different locality mechanisms, sensitive to different features, and with different developmental patterns, are active in grammar: one concerned with the formation of chains (Argument intervention) and one valuing the inflected verb (Agree intervention). Two types of relative clauses in German (verb final (25.a) and V2 (25.b) relative clauses) and their acquisition are the topic of Chapter 4, by Emanuela Sanfelici, Petra Schulz and Corinna Trabandt: (25) a. Hier gibt es zwei Frauen, die den here there-is EXPL two women PRON:NOM the:ACC Präsidenten getroffen haben president met have ‘Here there are two women that met the President.’

 24



Agree intervention does not arise when the subject is pre-verbal: (i) Quale fatina le signore stanno spingendo? Which fairy the ladies are pushing?

Introduction

15

b. Hier gibt es zwei Frauen, die haben here there-is EXPL two women PRON:NOM have den Präsidenten getroffen the:ACC president met ‘Here there are two women that met the President.’ V2 relative clauses (iV2s) are one of the specific syntactic environments in which V2 is licensed in subordinate clauses in German.25 Previous analyses of iV2s (e.g Gärtner 2001 a/b) claimed that they are not an instance of subordination, but are rather main clauses which are paratactically coordinated with a main clause containing a presentational or existential predicate. Acquisitional evidence seemed to support this analysis, as several studies (e.g. Brandt 2004, Diessel and Tomasello 2005, Brandt et al. 2008, but not e.g. Clahsen 1990, Rothweiler 1993) have reported that iV2s are the first type of relative clauses in the spontaneous production of children up to 4. The authors investigated whether 3-year-old children prefer verb final relative clauses or iV2s in a controlled experimental setting (using a picture- supported delayed repetition task) where the syntactic and semantic conditions allowed for both structures. Results show a robust preference in children for verb final relative clauses over iV2s: while adult controls correctly repeated iV2s, children showed a strong tendency to change an iV2 into a verb final relative clause This findings, the authors argue, contradict the acquisition pattern described in Brandt (2004), Diessel and Tomasello (2005), Brandt et al. (2008), and are also a challenge for the coordination analysis of iV2 clauses. They outline an analysis of iV2s whereby the latter are considered a case of embedded root phenomena. Namely, iV2s are CPs in which the embedded verb has moved to C0 and the d- pronoun (die in 25.b) is a resumptive (topic) pronoun, an analysis which is carefully shown to capture the syntactic properties as well as the semantic restrictions of iV2s. As for the conflicting acquisitional data, the authors argue that examples interpreted by Brandt et al. (2008) as instances of iV2s, are not proper iV2s, but should rather be analyzed as instances of left dislocations (Grewendorf, 2002). So, the authors conclude, it is unclear whether iV2 structures are the first instance of relative clauses even in spontaneous production: acquisition, again, meets linguistic theory.

 25

V2 is the order in main clauses in German, while subordinate clauses are verb final. The verb placement parameter is acquired early by German speaking children, who correctly place the verb in matrix and embedded contexts at the age of 3 (Sanfelici, Schulz and Trabandt, this volume, and the references quoted there).



Chapter One

16

In Chapter 5, Paolo Lorusso considers how the complexity of verb classes (i.e. their lexical (l-) syntax, and specifically the argument structure they project) affects the production and comprehension of auxiliaries in child Italian. A solid data on the acquisition of auxiliaries in Italian is that have appears later than be. As well- known since Burzio’s (1986) seminal work, unaccusatives (26.a) select the be auxiliary in Italian, while unergatives (26.b) and transitives (26.c) select the have auxiliary:26 (26) a. Gianni è arrivato G. is arrived b. Gianni ha parlato G. has spoken c. Gianni ha comprato un libro G. has bought a book Assuming Hale and Keyser’s (1993) analysis of unaccusatives, unergatives and transitives, Lorusso proposes that verbs that project an internal argument (i.e. unaccusatives and transitives) are less complex than verbs that project only an external argument (i.e. unergatives) as for their aspectual entailment. Of the former, unaccusatives are the least complex in that they project only an internal argument in the lower VP shell. 27 In the acquisition of constructions with auxiliaries in Italian, the higher complexity of verbs that project an external argument in the vP shell is predicted to play a central role.28 This prediction is tested through a study of spontaneous production and two experiments. Analyzing the spontaneous production of four children aged 18 – 36 months, Lorusso finds that while they regularly select the right auxiliary, children use more passato prossimo forms with unaccusatives and transitives than with

 26

Auxiliaries in Italian are mainly found in compound tensed constructions, such as passato prossimo (as in (26)), which have a perfect aspectual reading. 27 The eventive relation can be thus be determined entirely within the VP shell. Wide evidence from the literature is discussed by the author, showing that children single out the unaccusative verb class from very early on (see also Chapter 2, this volume) 28 Another factor singling out unergatives from transitives and unaccusatives is dicussed, namely lexical aspect (Aktionsart). Unaccusatives and transitives are mainly telic predicates, while unergatives are mainly atelic, since no direct object is involved in the event they denote. The passato prossimo gives a perfective entailment to all verbs it applies to.



Introduction

17

unergatives. Passato prossimo forms with unergatives, appear later in children’s productions. The first experiment is designed to test the pattern of production of perfective and imperfective forms along ages and verb classes (telic transitives and atelic unergatives). Results show a systematic general tendency to attribute passato prossimo to telic transitives. Children under the age of 5, however, strongly prefer to use imperfective forms with atelic unergatives. Elder children (5 – 7) are able to use passato prossimo with both telic transitives and atelic unergatives in a proportion similar to that of adult controls. The second experiment, a comprehension task, wants to discover whether children give a complete/incomplete reading to passato prossimo with different verb classes. Results show a systematic completed reading for telic transitives in the passato prossimo in younger children, elder children and adults. As for atelic unergatives, results show that a completed reading is not available until the age of 7. Taken together, the results of the three studies confirm that unergatives are more complex than transitives. If, as early sensitivity to unaccusatives confirms, unaccusatives are the least complex class of predicates, have is more complex (and hence acquired later) than be for the predicates that it selects. Among the predicates selected by have, unergatives are more complex than transitives since their aspect cannot be retrieved directly by an overt direct object: lexical aspect is not mapped in a one- to- one fashion with aspectual morphology below the age of 7.

2.2 Insights from Other Modes of Acquisition As we have seen in Section 1, comparing different modes of acquisition sheds light on the issue of complexity in many important respects. In studies comparing different modes of acquisition, typical L1 acquisition usually constitutes, using Belletti and Guasti’s (2015) words, a sort of baseline against which other modes of acquisition are compared. This is not entirely so in Chapter 6, where Cornelia Hamann, Natalia Gagarina, Solveig Chilla and Lina Abed Ibrahim investigate, through ample discussion of the relevant literature, as well as with a vast amount of novel data, the case of bilingual SLI children dealing with complex structures, with the purpose of identifying critical factors able to disentangle the two conditions of acquisition.29 A first research question, disentangling bilingual typical development

 29

As the authors argue, besides its theoretical relevance, this issue has also a practical relevance connected to the need to establish clinical diagnosis (and therapy) of SLI in bilingual settings.



18

Chapter One

from monolingual SLI development, is investigated through two studies. The first study compares older monolingual typically developing children and older bilingual typically developing children (from roughly 8 to 10 years); the second study compares younger monolingual typically developing children, younger bilingual typically developing children and age- matched monolingual SLI children (from roughly 5; 6 to 7;6 years). A Sentence Repetition Task including four types of complex structures (passive, subject relatives, object relatives without intervention and object relatives with intervention) was administered to the children. A second research question, disentangling bilingual typical development from bilingual SLI development, is dealt with through the comparison of four groups of children: monolingual typically developing, monolingual SLI, bilingual typically developing and bilingual SLI. The Sentence Repetition Task employed included the sentence types described above, plus bare and [+NP] wh- clauses, topicalization, finite complement clauses.30 As for the first research question, results showed that bilingual and monolingual typically developing children essentially pattern alike. For older children, intervention is clearly the critical factor, since object relatives with NP intervener are significantly more difficult than the other three configurations; long passives, subject relatives and object relatives without intervener are equally mastered. In younger children a similar pattern emerges, with object relatives with intervener as the most problematic clausal type, even though passives and object relatives without intervener appear more difficult than subject relatives. A different picture characterizes SLI children, which had great difficulty with all sentence types: the kind of complexity involved in passive (movement and smuggling), in subject relatives and object relatives without intervener (embedding and movement) is already problematic for these children, and thus the effect of intervention cannot be singled out. As for the second research question, results obtained comparing correct identical repetitions for all sentence types together, show that typical monolingual and bilingual children score alike, and the same holds true for monolingual and bilingual SLI children. Significant differences were found for monolingual SLI and typically developing bilingual children, as well as for bilingual SLI and bilingual typically developing children. Furthermore, results importantly show that bilingualism and SLI do not have a cumulative effect: bilingual SLI children do not perform significantly worse than monolingual SLI children.

 30



Examples are given in (8) to (15) of Chapter 6.

Introduction

19

In Chapter 7, Cristina Flores, Esther Rinke and Cecília Azevedo examine the heritage language production of second generation heritage speakers in the complex domain of object realization. They analyze spontaneous production data from two groups of heritage speakers (first and second generation) of European Portuguese living in Germany, and compare them with spontaneous production data from two groups of age (and education) matched monolinguals of European Portuguese living in Portugal. Object realization is a complex domain in European Portuguese, a language which has strong and clitic object pronouns. Clitic pronouns may vary in form and placement depending on the syntactic context in which they occur.31 Furthermore, European Portuguese allows specific null objects. German, the dominant language, does not have object clitics and does not allow specific null objects, though sentence initial topic drop is possible. First of all, the authors identify and classify the kinds of object realization in the four corpora, marking as ‘norm deviation’ every deviation from the expected standard norm.32 Then they examine the frequency of the different options of object realization in the four corpora. This comparison reveals that 2nd generation heritage speakers use all the possibilities of object realization, as the other three groups. The frequencies of the younger generation of heritage speakers resemble those of their age matched monolinguals: both produce less clitic pronouns and produce more object omissions than their older counterparts (though these tendencies are more marked in the 2nd generation heritage speakers). The latter differ from the other groups in that they use more demonstratives in object position: 2nd generation heritage speakers avoid clitic pronouns whenever they can use demonstratives or null objects. The dominant language, German, does not have null objects so cross-linguistic influence from the dominant language cannot be invoked as a source,33 while it could be involved in clitic omissions. Clitics, however, could also be avoided because they are complex. Norm deviations, furthermore (as extensive use of enclisis with respect to proclisis), are attested in all groups with no significant differences, and are thus considered a typical feature of the colloquial register. Finally, 1st generation heritage speakers do not show significant differences as compared to aged matched monolinguals: hence,

 31

(In)appropriateness, in the experimental corpus, was checked by two native speakers of European Portuguese. 32 Norm deviation, as the authors underline, may not necessarily reflect competence deviations, but also variation in the colloquial register. 33 Null object constructions in the corpus do not display the typical properties of topic drop in German.



20

Chapter One

the authors conclude, the input given to the second generation cannot be considered incomplete. Clitic omission thus appears to be the only clear difference in the spontaneous production of 2nd generation heritage speakers: limited use of clitics may have long term effects, leading to unstable knowledge. This in turn might explain the poor results obtained by 2nd generation heritage speakers in the same domain, but obtained through different methodologies, such as the Grammaticality Judgment Task employed by Rinke and Flores (2014). Object clitics are again the domain examined in Chapter 8, by Petra Bernardini and Monica Timofte, in bilingual children, through the comparison of two elicited oral production studies on French/Italian (Bernardini and Van de Weijer, under review) and Romanian/ Italian (Bernardini and Timofte, 2014) simultaneous and successive bilingual children. Under the assumption that cross-linguistic influence manifests itself in complex domains (see Section 1 above) the authors examine what kind of complexity is involved in object clitic constructions in French, Romanian and Italian. They compare two kinds of complexity: Interface Based complexity (Müller and Hulk, 2001) and Derivational Complexity (Jakubowicz, 2005, 2011), further distinguishing two subtypes in the latter: DC1 (number of moved constituents) and DC2 (number of movement operations). While French clitic constructions are more complex in terms of DC1, Italian and Romanian clitic constructions are more complex in terms of DC2. In Romanian, furthermore, Interface Based complexity is also involved. 34 The authors then compare the results of the French/Italian study and of the Romanian/ Italian study as for general accuracy and placement errors. Productions in general are more accurate in Italian than in French in the French/Italian study, but as accurate in Italian as in Romanian in the Italian/Romanian study. Placement errors occurred in the modal + infinitive context in the French/Italian study, interestingly revealing crosslinguistic influence both from Italian to French and from French to Italian. In the Romanian/Italian study, placement errors occurred only in Romanian, confirming the higher complexity that clitic placement entails in this language. Chapter 9, by Lena Dal Pozzo, presents three studies concerning the syntax of pronominal subjects and the resolution of ambiguous anaphoric dependencies in the adult L2 acquisition of Finnish. Finnish is a partial null subject language, allowing first and second person null subjects but not third

 34

In Romanian the position of the object clitic is conditioned by gender. Assuming gender to be a lexical feature, the syntax - lexicon interface is involved in Romanian clitic constructions.



Introduction

21

person null subjects.35 In answers requiring the identification of the subject (i.e. when the subject is a new information focus), null subject languages can adopt a VS strategy, which is not allowed in non- null subject languages.36 Different strategies to new- information focalize a subject are adopted in non- null subject languages, such as SV or clefts. The first study reported in Chapter 9 investigates answering strategies in adult L1 and L2 speakers of Finnish. SV was the prevailing strategy in both groups, (though L1 speakers of Finnish also adopted other strategies , such as e.g. XPVS and clefts, not resorted to by L2 speakers) suggesting a pattern akin to non-null subject languages. No sign of transfer from the L1 is attested in L2ers. A second study was then designed to test the hypothesis that L2ers could have assimilated Finnish to a non – null subject language. A Picture Description Task (requiring oral as well as written production) was administered to L2 Finnish adult speakers with various L1s, to elicit first and second person subjects (which, as we said, can be null in Finnish). Results show that in oral production overt forms are more extensively produced, while in written production null forms are more attested, both by the experimental subjects and by controls, i.e. L1 speakers of Finnish. This reflects, the author argues, an ongoing change in colloquial Finnish, where an extensive use of overt pronominal forms is attested. L2ers are thus successfully facing the complex task of computing two different varieties of Finnish with different morphosyntactic properties. In the third study, the interpretation of the personal pronoun hän ‘s/he’ and of the demonstrative pronoun tämä ‘this’ in L2 Finnish are tested through an off – line Picture Verification Task. Results indicate that L2ers show a strong preference for tämä to be co-referent with the object of the main clause and a fluctuation for hän to be co-referent with the subject or the object of the main clause. The same pattern is observed in the control group, contradicting the results of Kaiser and Trueswell (2008), where hän is shown to refer to the subject of the main clause. The results of the three studies suggest that in the complex domain of pronominal reference, a domain at the syntax – pragmatics interface, L2ers show patterns similar to those of L1ers, reflecting a tendency in the oral variety to treat Finnish as a non- pro drop language. L2ers differ from native speakers in that they only adopt the less complex possibility, among the different possibilities offered by the language they are acquiring.

 35 Except in some embedded clauses and with weather verbs. See examples (2.a) and (2.b) in Chapter 9. 36 Under the analysis of Belletti (2001, 2004) in a null subject language the subject can occupy a new information focus position in the vP periphery, pro occupying the high subject position in the clausal architecture.



22

Chapter One

References Baker, Mark (1988), Incorporation: A theory of grammatical function changing. Chicago: Chicago University Press. Belletti, Adriana (2001), “Inversion as focalization,” in: Aafke Hulk and Jean Yves Pollock (Eds.), Subject Inversion in Romance and the theory of Universal Grammar, 60-90. New York: Oxford University Press. —. (2004), “Aspects of the low IP area,” in: Luigi Rizzi (Ed.), The structure of CP and IP. The cartography of syntactic structures, Vol. 2, 16–51. New York: Oxford University Press. —. (2006) “(Past) participle agreement,” in: Martin Everaert and Henk van Riemsdijk (Eds.), The Blackwell Companion to Syntax, Vol. III, 493 – 521. Oxford: Blackwell. —. (2007), “Answering strategies. A view from acquisition,” in: Sergio Baaw, Frank Drjikonigen and Manuela Pinto (Eds.), Romance Languages and Linguistic Theory 2005, 19-38. Amsterdam, Benjamins. —. (forthcoming), “ Labeling (Romance) Causatives”. Belletti, Adriana, and Cristiano Chesi (2014) “A syntactic approach toward the interpretation of some distributional frequencies: Comparing relative clauses in Italian corpora and in elicited production,” Rivista di Grammatica Generativa 36,1-28. Belletti, Adriana, and Carla Contemori (2010), “Intervention and Attraction. On the Production of Subject and Object Relatives by Italian (Young) Children and Adults,” in: João Costa, Ana Castro, Maria Lobo and Fernanda Pratas (Eds.), Language Acquisition and Development: Proceedings of GALA 2009, 39-52. Newcastle: Cambridge Scholars Press. Belletti, Adriana, and Maria Teresa Guasti (2015), The Acquisition of Italian: Morphosyntax and Its Interfaces in Different Modes of Acquisition. Amsterdam: Benjamins. Belletti, Adriana, and Luigi Rizzi (2013), “ Ways of avoiding intervention: Some thoughts on the development of object relatives, passive and control,” in: Massimo Piattelli- Palmarini and Robert Berwick (Eds.), Rich Languages from Poor Inputs, 115- 126. Oxford: Oxford University Press. Bernardini, Petra, and Monica Timofte (2014), ”Bilingual acquisition of object clitics in Romanian and Italian” Structure, Use, and Meaning in Intercultural Settings, 17- 30, Brasov: Editura Universitatii Transilvania. Bernardini, Petra, and Joost Van de Weijer (under review), "On the



Introduction

23

direction of crosslinguistic influence in the acquisition of object clitics in French and Italian,", Ms. University of Lund, Språk- och litteraturcentrum. Brandt, Silke (2004), The acquisition of relative clauses in German and English. The very first steps. Leipzig: University of Leipzig unpublished Master’s thesis. Brandt, Silke, Holger Diessel and Michael Tomasello (2008), “The acquisition of German relative clauses: A case study,” Journal of Child Language 35, 325-348. Brown, H. Douglas, (1972), “Children’s comprehension of relativized English sentences,” Child Development 42, 1923–1936. Burzio, Luigi (1986), Italian syntax: a government-binding approach. Dordrecht: Reidel. Chomsky, Carol (1969), The Acquisition of Syntax in Children from 5 to 10. Cambridge, MA: The MIT Press. Chomsky, Noam (1965), Aspects of the theory of syntax. Cambridge, MA: The MIT Press. —. (2001), “Derivation by Phase,’ in: Michael Kenstowicz (Ed.), Ken Hale: A Life in Language, 1- 52. Cambridge, MA: The MIT Press. Cinque, Guglielmo (1999), Adverbs and Functional Heads: a CrossLinguistic Perspective. Oxford: Oxford University Press. Cipriani, Paola, Piero Bottari, Anna Maria Chilosi and Lucia Pfanner (1998) “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 33, 245- 280. Clahsen, Harald (1990), “Constraints on parameter setting: A grammatical analysis of some acquisition stages in German child language,” Language Acquisition 1, 361–391. Clahsen, Harald, Susanne Bartke and Sandra Göllner (1997), “Formal Features in Impaired Grammars: A Comparison of English and German SLI Children,” Journal of Neurolinguistics 10, 151-171. Collins, Chris (2005), “A Smuggling Approach to the Passive in English” Syntax, 8/2, 81 – 120. Contemori, Carla, and Adriana Belletti (2014), “Relatives and Passive Object Relatives in Italian speaking children and adults: Intervention in production and comprehension,” Applied Psycholinguistics 35(6), 1021-1053. De Vincenzi, Marica, Lisa Arduino, Laura Ciccarelli and Remo Job (1999), “Parsing strategies in children comprehension of interrogative sentences,” in: Sebastiano Bagnara (Ed.), Proceedings of the European



24

Chapter One

Conference on Cognitive Science, 301-308. Rome: Istituto di Psicologia del CNR. Diessel, Holger, and Michael Tomasello (2005), “A new look at the acquisition of relative clauses”, Language 81, 1-25. Friedmann, Naama, Adriana Belletti and Luigi Rizzi (2009), “Relativized relatives: Types of intervention in the acquisition of A-bar dependencies,” Lingua 119, 67-88. Friedmann, Naama, and Raama Novogrodzky (2004), “The acquisition of relative clause comprehension in Hebrew: A study of SLI and normal development,” Journal of Child Language 31, 661- 681. —. (2011), “Which questions are most difficult to understand? The comprehension of Wh- questions in three subtypes of SLI,” Lingua 121, 367- 382. Friedmann, Naama, Maya Yachini and Ronit Szterman (2015), “Relatively easy relatives. Children with syntactic SLI avoid intervention,” in: Elisa Di Domenico, Cornelia Hamann and Simona Matteini (Eds.), Structures, Strategies and Beyond. Studies in honour of Adriana Belletti, 303 – 320. Amsterdam: Benjamins. Gärtner, Hans Martin (2001a), “Are there V2 relative clauses in German?” Journal of Comparative German Linguistics 3, 97-141. —. (2001b), “On the force of V2 declarative,” ZAS Papers in Linguistics 23, 103-109. Gopnik, Myrna and Martha B. Crago (1991), “Familial aggregation of a developmental disorder,” Cognition 39, 1-50. Grewendorf, Günther (2002), “Left Dislocation as Movement,” Georgetown University Working Papers in Theoretical Linguistics 2, 31-81. Guasti, Maria Teresa, Chiara Branchini and Fabrizio Arosio (2012), “Interference in the production of Italian subject and object whquestions,” Applied Psycholinguistics 33, 185-223. Håkansson, Gisela, and Ulrike Nettelblatt (1996), “Similarities between SLI and L2 children. Evidence from the Acquisition of Swedish Word Order,” in: Caroline Johnson and John Gilbert (Eds.), Children’s Language, Vol 9, 135-151. Mahwah, NJ: Erlbaum. Hale, Kenneth and Samuel Jay Keyser (1993), “On argument structure and the lexical expression of syntactic relations,” in: Kenneth Hale and Samuel Jay Keyser (Eds.), The view from Building 20. A Festschrift for Sylvain Bromberger, 53–109. Cambridge, MA: The MIT Press Hamann, Cornelia (2006), “Speculations about Early Syntax: The production of Wh- questions by normally developing French children and French children with SLI,” Catalan Journal of Linguistics 5, 143 – 189.



Introduction

25

Hamann, Cornelia, and Adriana Belletti (2006), “Developmental patterns in the acquisition of complement clitic pronouns. Comparing different acquisition modes with an emphasis on French,” Rivista di Grammatica Generativa 31, 39- 78. Hamann, Cornelia and Laurie Tuller (2015), “Intervention effects in the spontaneous production of relative clauses in (a)typical language development of French children and adolescents,” in: Elisa Di Domenico, Cornelia Hamann and Simona Matteini (Eds.), Structures, Strategies and Beyond. Studies in honour of Adriana Belletti, 321 – 342. Amsterdam: Benjamins. Hulk, Aafke, and Natascha Müller (2000), “ Cross- linguistic influence at the interface between syntax and pragmatics,” Bilingualism: Language and Cognition 3, 227 – 244. Jakubowicz, Celia (2005), “The Language Faculty: (Ab)normal Development and Interface Constraints”. Presentation given at Generative Approaches to Language Acquisition (GALA) 2005, University of Siena. —. (2011), “ Measuring derivational complexity: new evidence from typically developing and SLI learners of L1 French,” Lingua 121/3, 339- 351. Kaiser, Elsi and John Trueswell (2008), “Interpreting pronouns and demonstratives in Finnish: Evidence for a form-specific approach to reference resolution,” Language and Cognitive Processes 23(5), 709748. Kayne, Richard (1989), “ Facets of Romance past participle agreement”, in: Paola Benincà (Ed.), Dialect Variation and the Theory of Grammar, 85-104. Dordrecht: Foris Publications. Moscati, Vincenzo, and Luigi Rizzi (2014) “ Agreement configurations in language development: A movement- based complexity metric,” Lingua 140, 67 – 82. Müller, Natascha and Aafke Hulk (2001), "Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages," Bilingualism: Language and Cognition 4, 1-21. Newmeyer, Frederick J., and Laurel B. Preston (Eds.) (2014), Measuring Grammatical Complexity. Oxford: Oxford University Press. Paradis, Johanna (2004), ” The relevance of specific language impairment in understanding the role of transfer in second language acquisition,” Applied Psycholinguistics 25, 27 – 82. Rice, Mabel, and Kenneth Wexler (1996), “Toward Tense as a Clinical Marker of Specific Language Impairment,” Journal of Speech, Language and Hearing Research 39, 1236-1257.



26

Chapter One

Rinke, Esther, and Cristina Flores (2014), “Heritage Portuguese bilinguals’ morphosyntactic knowledge of clitics,” Bilingualism. Language and Cognition 17 (4), 681 - 699. Rizzi, Luigi (1990), Relativized Minimality. Cambridge, MA: The MIT Press. —. (2004), “Locality and the left periphery,” in: Adriana Belletti (Ed.), Structures and Beyond: The Cartography of Syntactic Structures, Vol. 3, 223 – 251. Oxford/ New York: Oxford University Press. —. (2006), “ On the form of chains: Criterial positions and ECP effects,” in: Lisa Cheng and Norbert Corver (Eds.), Wh- Movement: Moving On, 97- 134. Cambridge, MA: The MIT Press. Rosenbaum, Peter S. (1967), The Grammar of English Predicate Complement Constructions. Cambridge, MA: The MIT Press. Rothweiler, Monika (1993), Der Erwerb von Nebensätzen im Deutschen. Eine Pilotstudie. Tübingen: Max Niemeyer. Sorace, Antonella, and Francesca Filiaci (2006), “Anaphora resolution in near native speakers of Italian,” Second Language Research 22(3), 339-368. Starke, Michal (2001), Move dissolves into Merge. PhD Dissertation, University of Geneva. Strik, Nelleke (2008), Syntaxe et acquisition des phrases interrogatives en français et néerlandais: une étude contrastive. PhD Dissertation, University of Paris 8. Tsimpli, Ianthi M. (2014), “ Early, late or very late? Timing acquisition and bilingualism,” Linguistic Approaches to Bilingualism 43, 283-313. Vender, Maria, Maria Teresa Guasti, Maria Garraffa, and Antonella Sorace (2014), “Bilingualism and Specific Language Impairment: Similarities and Differences,” in: Carla Contemori and Lena Dal Pozzo (Eds.), Inquiries into Linguistic theory and Language Acquisition. Papers offered to Adriana Belletti, 217-228. Siena, CISCL Press. Vernice, Mirta, and Maria Teresa Guasti (2015), “The acquisition of SV order in unaccusatives: manipulating the definiteness of the NP argument,” Journal of Child Language 42/1, 210-237. Warren, Tessa and Edward Gibson (2002), ” The influence of referential processing on sentence complexity,” Cognition 85, 79 – 112. White, Lydia (2011), “Second language acquisition at interfaces,” Lingua 121 (4), 577-590.



PART I

CHAPTER TWO ON THE ACQUISITION OF COMPLEX DERIVATIONS WITH RELATED CONSIDERATIONS ON POVERTY OF THE STIMULUS AND FREQUENCY∗ ADRIANA BELLETTI

1. Introductory Considerations This article is dedicated to some reflections on the complexity of syntactic computations, prompted by recent results on the acquisition of Italian in different empirical domains such as relative clauses, types of passive, types of post-verbal subjects with unaccusatives.1 To the extent that we adhere to the idea that if some syntactic construction is properly mastered by young children, this construction should be valued as not especially complex as for the computations that it involves, we will review some results from typically developing Italian speaking young children which indicate that constructions that may look pre-theoretically complex – as they involve e.g. various steps for their derivation or more words than other relatively close ones – should probably not count as such in the relevant formal sense since young children appear to master the necessary computations from relatively early on2. We will also point out that, in some of the cases that we will review, appeal cannot be made to the frequency in the input of the constructions investigated as these constructions appear not to be that frequent in naturalistic corpora and are



∗ The research presented here was funded in part by the European Research Council/ERC Advanced Grant 340297 SynCart – “Syntactic cartography and locality in adult grammars and language acquisition”. 1 See Belletti & Guasti (2015) for detailed discussion of most of the results and data presented here. 2 Constructions are in fact the result of the combination of a number of formal computations. The term is used here in its descriptive sense, as is usual practice.

Poverty of the Stimulus and Frequency

29

thus likely not to be present in the children’s primary data in a critical amount. These findings thus stimulate some poverty of the stimulus considerations in a novel fashion, based on data not previously appealed to in this context.3 The findings to be (re)discussed here are also relevant for the poverty of the stimulus considerations that they prompt in another complementary respect, as they indicate that sometimes children make use of constructions that are not only rather rare in the linguistic input, but also not resorted to by adults in the same elicitation experimental conditions; conversely, children do not make use of constructions that are widely resorted to by adults in the same conditions. This point will be especially clear in discussing use of types of passive in relative clauses. We will see that complexity is at stake here in a non- trivial way, as (passive) constructions which look pre-theoretically complex may not be actually so, as children master them quite early and relatively easily, and, conversely, (passive) constructions which may look simple as, e.g., shorter, may instead count as complex at the appropriate level of analysis as they involve some reduction operation still demanding for the (young) developing children’s grammar. These results contribute new evidence for the conclusion that children do not always do what they are likely to hear most nor do they do what looks intuitively, pre-analytically, simpler. We will attack the complexity/poverty of the stimulus issues through the (re)consideration of previous results from the literature in three domains: 1. Presence of infrequent Passive Object Relatives/PORs in children’s productions in place of the elicited active Object relatives. 2. Presence of the infrequent si-causative passive in young children’s early productions of passives. 3. Early sensitivity to the Definiteness Effect/DE in the VS order with unaccusative verbs. We will start by considering point 1 and 2 in the following section 2. Section 3 is dedicated to point 3. Section 4 summarizes the discussion and draws some general conclusion.

 3 See Berwick et al. (2011) for recent reconsideration of the poverty of the stimulus issue.

Chapter Two

30

22. Infrequen nt Constructions in C Children’s E Elicited Pro oductions 22.1. PORs in n Children’ss Elicited Prroductions Whereass Subject relattives (SRs) arre both producced and comp prehended by Italian sspeaking childdren already at around agge 3 and 4 (e.g. 79% correct SRss productionss in the age range 3 – 33;11 in the elicitation e experiment of Belletti & Contemori 2010; furthher similar results r in Contemori & Belletti 20114), Object rellatives (ORs) are hardly pro oduced in the same exxperiments. Although A we will w focus onn production here, h it is worth mentiioning that coomprehension n is also ratheer poor; in thee average ORs are noot comprehennded beyond d chance by young child dren (e.g. Contemori & Garraffa 20010 for Italian n, in the age sspan 3 to 5;5; see also Adani 2011,, and the revieew in Belletti & Guasti 201 5, chapter 5). The obviious question to ask is: Wh hat do childrenn produce instead when they are askeed questions eliciting e an OR R? Results indiicate that they y typically deliver varioous types of non-target n pro oductions, espeecially in the youngest ages4; as theey grow olderr, they tend to o resemble addults and prod duce more and more P PORs (see refferences quoteed for detailedd quantificatio on of the data). As diiscussed in deetail in particu ular in Bellettii (2014), Con ntemori & Belletti (20114) adopting thhe system dev veloped in Frieedman et al. (2 2009), the crucial factoor leading to the t production n of a POR innstead of an active a OR can be ideentified in locality, moree specificallyy featural Relativized Minimality (Rizzi 1990, 2004). In a nutshell, interrvention of a lexically restricted suubject is consiidered problem matic in the dderivation of a headed object relatiive due to thhe inclusion of o a nominal [+NP] featurre of the intervener (tthe subject of the relative cllause) within tthe feature com mposition of the target (the relative head) h as is illustrated in (1): (1)

OR R:

(…) [Il bam mbino [che la mamma m petttina +R,,+NP +NP the kidd that thee mum com mbs

]] (…) (

4 As is the case in severaal other languaages also inveestigated throug gh similar experimental designs, mosstly from Cost Action/A33; among severral others, French/Délagge (2008), EPorrtuguese/Costa et al. (2011), Hebrew/Novog grodsky & Friedmann (22006); Friedmaann, Yachini an nd Sztermann (2015) on chilldren with syntactic SLI.

Poverty of the Stimulus and Frequency

31

In contrast, in the derivation of a POR, in which passive is derived through so-called smuggling (Collins 2005), with movement of a chunk of the verb phrase containing (at least) the verb/past participle and the direct object, no intervention locality problem arises anymore. The relevant steps of the derivation of a POR are illustrated in (2): (2)

POR: Relativization step

Il bambino bi che h [(__) [( ) è [VP pettinato tti t ] bi da [vP la mamma ]

the kid

that

is

Passivisation step/smuggling / combed

Thus, passive completely eliminates intervention in the derivation of a POR; hence, no issue of inclusion of relevant features arises in this case at all, as the (lexical) relative head is raised from the smuggled position. In a corpus study of Italian undertaken by Belletti & Chesi (2014), the very clear conclusion has been reached that PORs are rather rare in the naturalistic input in Italian.5 On the other hand, a robust set of experimental results presented in the reference quoted above (and also in the further adaptation of the same elicitation experiment in Belletti & Chesi 2014) has indicated that Italian-speaking adults systematically resort to PORs when an (active) OR is elicited. The conclusion must then be reached that it is not the frequency of the construction that leads to the production of a POR instead of an active OR in the elicited production. In the spirit of the references quoted, we endorse the view that resort to POR is favored by grammatical factors, ultimately the complete elimination of the intervention configuration illustrated above. Interestingly, the same grammatical factor appears to play a role for young children. Belletti (2012) points out that initially very young children try to produce active ORs, but, as soon as they start mastering passive, they start producing PORs, and more and more so as they grow older. The following table

 5

Corpora analyzed in Belletti & Chesi (2014): files (children and adults’ child directed speech) from the CHILDES database (MacWhinney & Snow 1985); SUT corpus (Siena University Treebank, Chesi et al. 2008); CIT corpus (Corpus di Italiano Televisivo, Spina 2005). See Belletti & Chesi (2014) for all further details on the corpora utilized and the analysis performed on them.

32

Chapter Two

adapted from Contemori & Belletti (2014) illustrates the development of PORs in children of different ages (results from a Picture Description task, adapted to Italian from Novogrodsky & Friedmann 2006); in (3) PORs are divided according to the type of passive that they contain: (3) (Types of) PORs in children’s productions

Table (3) is interesting in various respects; we indicate two of them. First of all it shows the main point under discussion so far: as children grow older they tend to produce more and more PORs, thus approaching adults’ performance, which reaches up to 90% production of PORs in similar experiments. Hence, although passive may be a relatively complex computation per se, as also revealed in (3) by the fact that young children initially produce passive object relatives only to a very limited extent, yet as soon as it is properly mastered, the passive computation is adopted more and more. We may then reasonably propose that it is favored over a computation involving the problematic intervention configuration of active ORs illustrated in (1). Hence, despite their infrequency in the naturalist input mentioned above, also relatively young children tend to opt for the production of a POR, the optimal derivation as far as locality is concerned (Belletti 2012, 2014). Table (3) is also interesting in another respect concerning the type of passive resorted to by children. We now turn to a more detailed discussion of this aspect in the following section.

2.2. Si-causative Passive in Children’s Productions and the Comparison with Adults Table (3) indicates that the first type of POR produced by children already in the earliest ages contains a si-causative passive, i.e. a passive in the causative voice. An example of this type of passive is given in (4)a, and the example of a POR containing the si-causative passive is given in

Poverty of the Stimulus and Frequency

33

(4)b: (4) a

b

Il bambino si fa pettinare dalla mamma the kid makes himself(-Cl) comb by the mum Il bambino che si fa pettinare dalla mamma the kid that makes himself(-Cl) comb by the mum

Si-causative passive is not frequent in the naturalist input in Italian. This is very clear to the intuition of any native speaker6. Interestingly, a quantitative measure confirms the intuition. A first counting of the occurrences of types of passive in Italian spontaneous conversations (from the LIP Corpus)7 indicates the following distribution of occurrences in the same conversational context: -

Copular passive: Venire passive: Si-fa causative passive:

443 296 22

Passive with auxiliary be is the most commonly used type of passive; passive with auxiliary venire is also very commonly used8; in contrast, sicausative passive is extremely rare. What we want to focus the attention on here is the striking fact that (even very) young children appear to access with ease a construction that is fairly rare indeed.

 6

This is a difference with French. We do not have any quantitative data on French available, however the intuition of French native speakers is very clear in this respect (see Délage 2008): se-faire passive is a very common type of passive in French and does not require any clear causative meaning (cfr. Labelle 2002), as is instead the case in Italian. For space reasons we will not further elaborate on the French/Italian comparison here. On the early access to si-causative passive by Italian speaking children tested through a different syntactic priming technique, see Manetti & Belletti (2015). 7 LIP, Lessico di Frequenza dell’Italiano Parlato, 1990-1992 (corpus collected under the direction of Tullio De Mauro). 8The verb venire (come) can be used as a passive auxiliary in Italian; an example is given in i.: i. La porta viene chiusa dal controllore the door comes closed by the controller Venire passive is only possible with non-compound tenses; hence, this limits its use and reduces its possible occurrences. On the restrictions in the use of auxiliary venire in the Italian passive and on its development, see Belletti & Guasti (2015, chapter 4).

34

Chapter Two

The rarity of the construction is also confirmed by the fact that the types of passive in the PORs produced by adults are rather different from those resorted to by children. This is clearly visible from the figures in (5) which compare the types of passives used by children (as described in Table (3)) with those used by adults – as reported in Contemori & Belletti (2014) (from a Preference task adapted to Italian from Novogrodsky & Friedmann 2006) 9: (5)

Whereas si-causative is the most common type of passive in children’s PORs, there is no such passive at all in the adults’ productions. Hence, once again we can conclude that children do not select the type of construction they use on the basis of what is likely to be most frequently present in their input data: si-causative passive is indeed likely to be rather rare in the naturalistic input data accessed by children, and it is also very rare, in fact absent, in adults’ elicited productions (see 2.2.1 for further relevant considerations). Note that the new question opened by this comparison as to why children should select the rare si-causative passive in their PORs is an independent question, which we cannot address here in any detail. For some first speculations see Manetti & Belletti (2015) where early access to si-causative passive is confirmed also in simple declaratives, not involving any relative clause computation. In Belletti (forthcoming) the proposal is made that causativization of the Italian/Romance type involves a smuggling operation of the same kind described above for passive; it is furthermore proposed that this operation may be crucially triggered by the fundamental requirement of labeling of the syntactic structure (Chomsky 2013, 2015; Rizzi 2015a,b). The fundamental nature of this requirement,

 9 30% PORs overall in children’s production (in Picture description task), when an OR was elicited, distributed as indicated. 90% of PORs overall in adults’ productions (in Preference Task), distributed as indicated. For all detailed results the reader is referred to Contemori & Belletti (2014).

Poverty of the Stimulus and Frequency

35

which is essential in general terms for the interpretability of syntactic structures may be a crucial factor in determining the early accessibility to this type of passive in young Italian speaking children10. We leave this central issue here as it would divert too much from the main focus of the discussion. We rather underscore the relevance for the present discussion, of the comparison in (5), which illustrates early accessibility of young children to the rare si-causative passive. 2.2.1 Further Comparing Children’s and Adults’ PORs The comparison offered by (5) is interesting and revealing in other respects as well, directly relevant for the issue of complexity and the relation with frequency and poverty of the stimulus considerations. Looking at children’s and adults’ performances one next to the other a further striking observation can be made: children have a kind of mirror behavior with respect to adults, as indicated by the “u” shape of the figures in (5). Not only do they do what adults do not do, i.e. producing sicausative passives, they also do not do what adults do most, i.e. producing reduced relatives. Indeed, the majority of adults’ PORs are in the reduced form, of the type illustrated in (6): (6) Il bambino che è/viene pettinato dalla mamma the kid (who is) combed by the mum In the example in (6) the missing part of the reduced relative clause is just simply erased. The reader should be warned that this is not an analysis of what the process of reduction should amount to; it is just a way to graphically picture the missing elements of the reduced relative (all assumed to be at the present tense for concreteness). Reduction, as is implicit in the term itself, leads to a shorter string than the unreduced one; in terms of number of words, there are clearly fewer words in the reduced form than in the unreduced one. If we go beyond the string, however, and consider the structural computations leading to reduction, it is likely that

 10

Another relevant factor may be presence of reflexive si. Although the precise role played by the reflexive in this construction needs to be clarified in further detail (Belletti forthcoming for a first proposal), it is a well known fact that children have early access to the interpretation of reflexive anaphors (Hamann 2011 for an overview). This issue is left open here, as it would take the discussion too far afield in the context of the present article.

36

Chapter Two

the overall process of reduction has some cost in terms of its complexity11. The fact that the children of our reviewed experiments almost never adopt reduced relatives in their PORs suggests this conclusion even more strongly. An analysis of what reduction should amount to in the case of reduced relatives is beyond the scope of this article and will not be undertaken. The relevance however of the children-adults comparison in (5) is at its very core: Absence of the reduced relatives in children’s productions indicates that one shallow factor that one may think could be crucial in determining complexity, i.e. number of words/overall length of a string, does not in fact enter into the relevant evaluation that children “instinctively” (Pinker 1994) do while selecting the structure to produce, i.e., in this case, the type of passive to use in their POR. This is an interesting conclusion, as it rules out any simple minded, pre-theoretical approach to define syntactic complexity12. The comparison in (5) is all the more interesting and potentially revealing if the following further considerations are also made. In the counting of the various naturalistic English corpora that they report, Roland et al. (2007) have provided a typology of different types of relative clauses in different corpora. One interesting finding in the context of the present discussion is that Roland et al.’s (2007) data indicate that reduced PORs are present, especially in written corpora, and that they are more frequent than unreduced PORs; this also holds in spoken corpora: 3118 unreduced PORs vs 10730 reduced PORs in the written British National Corpus; 1729 unreduced PORs vs 2886 reduced PORs in the spoken British National Corpus. See Belletti & Chesi (2014) for recent review and critical discussion of these data. We can assume that Italian is probably rather similar to English, with reduced PORs more frequently present than unreduced ones. This is also consistent with the adults’ behavior in the elicited production task illustrated in (5) as well as with the results

 11

Be it a process or just the reflex of use of a reduced clausal structure (Siloni 1997 and, more recently, Harwood 2015, for a proposal couched within phase theory). 12 Children may disfavor use of reduced sentential structures more generally in non-root clauses (in contrast with root clauses, as in the so called “truncation stage”, Rizzi 2006). This is indicated for instance by their preference (age 3 to 5) for full sentential complements over raising-to-object/ECM type structures, as in e.g. European Portuguese (Santos et al. 2015). The virtual absence of reduced relatives in the results reviewed in the text from children’s productions may be a further manifestation of this general late access to reduced clausal structures. This is a topic currently under further investigation

Poverty of the Stimulus and Frequency

37

reported in Belletti & Chesi (2014) from an experiment adopting the same type of elicitation design, manipulating the “animacy” feature: in all conditions reduced PORs have been the most frequently produced structures by adults. The conclusion is also consistent with the counting of some naturalistic corpora that Belletti & Chesi (2014) also report from adult (spoken) Italian: although PORs are relatively rare in these corpora as mentioned in 2.1, yet when they are present they are realized in the reduced form in the vast majority of cases: e.g. 1 unreduced POR vs 78 reduced ones in the adult child directed speech from the analyzed files of the CHILDES database (see the reference quoted for detailed results).13 Overall, all these considerations lead us to conclude that reduced PORs are relatively well represented in standard Italian, as they also occur, to a limited extent, in child directed speech. Hence, these results complement the conclusions reached in the previous section: children do not necessarily access first those structures that appear to be present - though to a relatively limited extent - in their primary input data, such as e.g. reduced relatives. Furthermore, a first new counting in the same files from CHILDES (footnote 14) of si-causative passive has indicated no occurrence at all (at most one unclear case) in the adult child directed speech14. To the extent that this corpus is a representative sample of some primary data available to Italian speaking children, we can conclude that reduced PORs are more frequent than si-causative passives (78 vs 0, or 1 if the unclear case is counted). Then, once again, it is not plain frequency that children are sensitive to; grammatical factors and more generally formal factors entering the computation of the different structures must play a crucial role for developing children. The challenge then becomes that of understanding what these factors are. The considerations developed in this section represent a first attempt in this direction: principled reasons may favor access to si-causative passive (e.g., among others, labeling, Belletti forthcoming), whereas the complexity of the reduction operation may lead to later access to PORs in their reduced form.

2.3. Concluding Discussion (1) In closing this section 2, a number of conclusions can be drawn and a number of lines of research can be identified, which are listed below. It seems that they should all be taken into account in trying to make precise

 13

All available Italian files in the CHILDES database have been analyzed, corresponding to a total of 177 files. Thanks to Cristiano Chesi for providing all relevant information. 14 Not surprisingly, some active causatives are present.

38

Chapter Two

the interplay among different factors in development, such as formal/ grammatical factors and frequency factors, and the role they play in shaping the poverty of the linguistic input and ultimately in determining the complexity of different syntactic computations. -

-

-

-

-

Intervention locality plays a crucial role in determining the complexity of the active OR structure with a lexical head and an intervening lexical subject, following Friedmann et al. (2009) and much subsequent work. These structures are not produced in elicitation experiments by both children and adults in favor of PORs, i.e. relatives in the passive. As PORs are not frequent structures in the naturalistic input, plain frequency does not seem to be a crucial factor in accounting for children’s (and adults’) linguistic behavior (Belletti & Chesi 2014 for further relevant discussion). Resort to PORs in place of active ORs suggests that the computation at play in passive, which eliminates intervention of a lexically restricted subject, is favored hence, in the relevant sense, it is less complex than the computation of active ORs involving intervention of a lexically restricted subject. Si-causative passive in the causative voice does not represent a particularly difficult computation for (even young) children as it is accessed from very early on. This suggests that Italian/Romance type causativization (itself involving an instance of smuggling of the same type as passive), and its combination with other computations involved in passive (such as e.g. raising of the internal argument into the subject position) does not constitute a complex set of operations for children in the relevant sense. Frequency factors do not appear to play a crucial role in determining children’s access to different constructions in both directions: a relatively rare construction like the si-causative passive is accessed with no special difficulty in the first PORs, whereas the reduced PORs which are present (though to a limited extent) also in naturalistic corpora of child directed speech are not resorted to by children in their PORs, in sharp contrast with adults’ (elicited) productions. Children appear not to do what they hear most nor what would seem to be simpler since e.g. shorter (as shown by the case of reduced relatives). Even in their target consistent behavior, i.e. when they produce PORs similarly to what adults do, they typically have a behavior which contrasts with that of adults (as in their use

Poverty of the Stimulus and Frequency

39

of the si-causative passive). This is a strong challenge for any approach to acquisition in terms of general notions of analogy or imitation (Tomasello 2003; Yang 2012 for critical considerations in this domain). The challenge for the formal/grammatical approach consists in trying to tease apart the crucial relevant formal/ grammatical ingredients involved each time (and in formulating hypotheses to account for development).

3. Children’s Sensitivity to the “Definiteness Effect/DE” and the Status of the VS Order In a recent article Vernice & Guasti (2015) have provided interesting evidence indicating that children around age 5 (4;2-5;11) appear to be sensitive to the so called Definitness Effect/DE (Belletti & Guasti 2015, chapter 6 for detailed presentation). It is impossible to summarize here the rich literature on DE, which dates back to the late seventies starting with Milsark’s (1974, 1977) seminal work (Belletti & Bianchi 2016, for an overview and some new thoughts on the issue). Here I only mention those aspects of the phenomenon, which will allow us to set the background for the discussion of some of Vernice & Guasti’s results; these results are relevant to the main focus of this article dealing with complexity, poverty of the stimulus and frequency considerations, since these issues manifest themselves in a rather peculiar way in this domain. Across languages the internal argument (IA) of unaccusative verbs, which characteristically ends up being the preverbal subject of the clause, is subject to an indefiniteness requirement when it remains in the IA position; for instance in English a sentence like (7)a15 is judged definitely better than (7)b; (7)c is fine with either a definite or indefinite subject: (7) a There appeared a man on the screen b *There appeared the/this man on the screen c A/the man appeared on the screen Similar data hold in French and other languages as well, including Italian. Since the post-verbal noun phrase can be a preverbal subject, the sentences

 15

Belonging to a special style in English. Interestingly, even at this peculiar stylistic level native speakers have clear judgments and make the relevant distinctions.

40

Chapter Two

in (7)a,b are often referred to as sentences containing a post-verbal subject. The possibility of sentences with a post-verbal subject is characteristically limited to sentences containing an unaccusative verb in non null-subject languages, as in the there sentence (7)a in English. As already discussed in Belletti (1988), in Italian the DE phenomenology is obscured by the fact that, it being a null subject language, Italian admits post-verbal subjects with all verb classes, yielding a widespread occurrence of the order VS. A post-verbal subject appearing with transitive and intransitive verbs is not subject to any DE constraint; typically it is interpreted as the focus of new information (Belletti 2004, and the related discussion in Belletti & Bianchi 2016; a new information post-verbal subject is also possible with unaccusatives; with a new information focus post-verbal subject no DE is at work with unaccusatives as well. When the post-verbal subject is narrow focus, there is no constraint on definiteness on it since there is no such constraint affecting the relevant new information focus position. With unaccusatives the order VS may thus either correspond to the order derived through first merge, with S in the IA position (the same position as the direct object of a transitive verb; Burzio 1986) yielding DE and the sentence is associated with an all-new interpretation, or it can correspond to a sentence with a new information post-verbal subject, with S in the relevant new information focus position cartographically yielding the relevant interpretation16 and no DE. Hence, only in the former case is DE manifested. Finally, the order VS is possible in all-new sentences with all verb classes, unaccusative, transitive and intransitive (with DE arising only with unaccusatives in the way described). In conclusion, the order VS is possible in Italian with all verb classes in different guises; S can be new information narrow focus and no DE is at work with unaccusatives, as with any other verb class; in all new sentences, DE on the post-verbal subject is manifested with unaccusatives only, as S remains in the IA argument position of first merge (Belletti 1988, Belletti & Bianchi 2016; this does not happen with the other verb classes, since the post-verbal subject never was an IA with transitives and intransitives; and DE, as mentioned, is a property of the IA position. The above essential description of a very articulated phenomenology (for more details the reader is referred to the references quoted) is sufficient to appreciate its complexity; ultimately, the status of post-verbal subjects is in principle rather opaque in a language like Italian; only some

 16

Distinct from a left peripheral contrastive/corrective focus, Rizzi 1997 and subsequent work.

Poverty of the Stimulus and Frequency

41

post-verbal subjects are constrained by DE, those of unaccusatives, but only under particular discourse conditions, i.e. in all new sentences. Intuitively, pre-theoretically this should not be an easy area to master for children. Specifically, DE with unaccusative verbs seems a rather complex constraint to induce from the Italian input data, where VS frequently appears with all verb classes, with the difference in interpretation discussed, crucially allowing for the presence of a definite new information post-verbal subject with unaccusatives as well (when it is new information narrow focus). Yet, young children do not seem to experience special difficulties in distinguishing between the appropriate use of a post-verbal definite or indefinite subject with unaccusative verbs. This is what Vernice & Guasti’s results from a repetition experiment clearly indicate. In the repetition task designed in their experiment which was set in an all new context, children (4;2-5;11) better repeated sentences with a post-verbal subject when the noun phrase was indefinite and the verb was unaccusative. Moreover, children experienced difficulties in providing identical repetitions of definite post-verbal subjects in that context, with unaccusatives. Vernice & Guasti (2015) tested all new contexts with unaccusatives and intransitive verbs. Concentrating on unaccusatives, in that context, the order VS is felicitous only if S is indefinite, whereas it is very marginal if S is definite, as indicated by the following two examples from the experiment: (8)

Context: C’è un bel sole nel bosco. Poi…. (there is a beautiful sun in the wood; then…) a

Esce un orsetto con i suoi amici comes out a little bear with his friends

b

*Esce l’orsetto con i suoi amici comes out the little bear with his friends

(Recall that a definite post-verbal subject is possible with unaccusatives only in a different type of context, when it is new information narrow focus). Children were able to perform identical repetitions of sentences like (8)a in a significantly wider number of cases than for sentences like

42

Chapter Two

(8)b (52% vs 31%).17 The ability to perform identical repetitions is considered a good test of what the speaker’s grammar can do (Friedmann 2007, Lust 2005). This result is thus revealing in various respects: first of all it shows that the unaccusative class is set apart by young children with one of its crucial defining properties clearly identified: the indefiniteness requirement on the internal argument of the unaccusative verb (see also Lorusso 2014 for converging evidence in younger children); second, it also shows that young children master the definite/indefinite distinction from very early on (see again Lorusso 2014 in this respect). Both conclusions are very relevant for the issues addressed in this article. Let us spell out the reasons why it is so in some better detail. The early identification of the unaccusative class through the indefiniteness requirement on its post-verbal subject suggests a grammar driven acquisition: since post-verbal subjects are possible with all verb classes in Italian, as reminded in the description above, the naturalistic input is rather opaque to be plausibly able to drive the selection of the correct sentences containing a post-verbal subject with unaccusative verbs. Furthermore, the fact that young children master the definite/indefinite distinction from early on also suggests that this distinction is very well rooted in their internal grammar and does not really need to be learned. Intuitively, the definiteness/indefiniteness distinction is not a simple distinction to express; it is indeed a classical topic in formal semantics (Heim 1982, references cited therein and much subsequent work; Belletti & Bianchi 2016, for relevant discussion). Yet young children master the distinction from early on, as their sensitivity to DE with unaccusatives clearly suggests.

3. 1 Concluding Discussion (2) Children’s early sensitivity to DE with unaccusatives reviewed in the previous section allows us to reach a number of conclusions relevant to the complexity/frequency/poverty of the stimulus issues central to the present work, which are summarized below:

 17

As presented in detail in Belletti & Guasti (2015), the order SV was preferred so that sentences were often produced as non-identical repetitions with both unaccusatives and intransitives, pre-posing the subject even when it was postverbal in the target. However, only when the subject was indefinite and only with unaccusatives children were able to perform identical repetitions of VS. A PP following the post-verbal subject is present in the stimuli to ensure that the post-verbal noun phrase is indeed internal to the VP, in its IA position of Merge (Belletti 1988 on the relevance of these configurations).

Poverty of the Stimulus and Frequency

-

-

43

What looks intuitively, pre-teoretically complex is not necessarily complex in development; the definiteness/indefiniteness distinction is mastered by young children and obviously, no direct teaching or explanation is provided to them on this distinction. Hence, the relevant complexity is not an intuitive notion. What may look hard, or anyway not trivial, can be relatively easy for young children, presumably because their language faculty is prepared to master the relevant distinction; children are prepared to recognize the lexical items, i.e. the class of unaccusative verbs, in which it plays a role, due to the structural position of their internal argument. Hence, the relevant lexical, semantic and syntactic properties appear to be simultaneously mastered by young children’s grammar and to interact in the target way from early on. Intuitively, the order VS is fairly opaque in the Italian naturalistic input, as it corresponds to different discourse pragmatic conditions and interacts with lexical properties of different verb classes (singling out unaccusatives in the way discussed). The word order VS in its different values is thus rather frequent in standard Italian.18 However, due to the different discourse pragmatics it may correspond to, it is associated with different structural representations and syntactic computations.19 Young children do not seem to be disturbed by these a-priori opaque input data, as witnessed, among other things, by the results discussed in this section 3.20

4. Summary and Some General Conclusion We have reviewed here a number of results from experimental work on the acquisition of some Italian (a-priori) complex constructions, which indicates that children’s early linguistic behavior is not always easily predictable on the basis of frequency considerations; all of this suggests that complexity does not apply in the vacuum and it is not a pre-theoretical notion. Furthermore, non-trivial considerations on the status of the

 18

A precise corpus study is yet to be done in this domain (but see Lorusso 2014). However, it is very clear to the intuition of any native speaker of Italian that the order VS is as “familiar” as the order SV in the language. 19 Which are well represented cartographically. Belletti & Bianchi (2016) and references cited therein. 20 And also in Lorusso (2014) corpus study quoted. See also Belletti & Contemori (2012) for relevant considerations based on the appropriate use of pronominal post-verbal new information subjects in young children’s elicited productions.

44

Chapter Two

impoverished input that is (likely to be) available to children in their primary linguistic data are also suggested by these findings. On the one hand young children make (relatively) early use (in elicited production) of constructions which are not particularly frequent in naturalistic corpora of spontaneous production, such as PORs; on the other hand, children accessed the infrequent POR structures not through the kind most commonly found in (adult) naturalistic corpora nor in the analyzed child directed speech, i.e. reduced PORs, but through a kind which makes use of a very rare type of passive, seldom found in naturalist productions in standard Italian, i.e. si-causative passive. We have proposed that grammatical considerations having to do with intervention locality combined with other factors to which we have alluded to, such as the labeling requirement, may play a crucial role in interpreting the unexpected children’s linguistic performance. Furthermore, early access to subtle properties of different verb classes that we have reviewed, such as the indefinite requirement on the IA of unaccusative verbs that Italian speaking children appear to master since their young age, constitutes a fairly inspiring and revealing indication that children are equipped from early on with the grammatical tools necessary to compute rather complex notions such as the definite-indefinite distinction on noun phrases, beside singling out the appropriate lexical class to which different verbs belong. This is so despite the rather opaque status of the input they are likely to be exposed to, in which the same word order corresponds to different discourse values and consequently different structures and related computations, as in the case of the VS order in Italian. All of this clearly points to the conclusion that internal grammatical factors play a crucial role in development. It constitutes an encouraging indication of the insights that can be gained from a detailed grammarbased study of (stages of) acquisition whereby subtle but not for that matter less important empirical distinctions manifest themselves in often unexpected and unexplored ways in children’s early productions. This often happens despite the rarity of the relevant structures investigated and their intuitive, pre-theoretical complexity.

References Adani, Flavia (2011), “Rethinking the acquisition of relative clauses in Italian: towards a grammatically based account,” Journal of Child Language 38, 141-165. Belletti, Adriana (1988), “The Case of Unaccusatives,” Linguistic Inquiry

Poverty of the Stimulus and Frequency

45

19, 1-35. —. (2004), “Aspects of the low IP area,” in: Luigi Rizzi (Ed.), The structure of CP and IP. The cartography of Syntactic Structures, Vol. 2, 16-51. New York: Oxford University Press. —. (2012), “Considering the complexity of relative clauses and passive from the Italian perspective,” in: Sandrine Ferré, Philippe Prévost, Laurice Tuller and Rasha Zebib (Eds), Selected Proceedings of the Romance Turn IV Workshop on the Acquisition of Romance Languages, 2-26. Newcastle: Cambridge Scholars Press. —. (2014), “Notes on Passive Object relatives,” in: Peter Svenonius (Ed.), Functional structure from Top to Toe. The Cartography of Syntactic Structures, Vol. 9, 97-114. New York: Oxford University Press. —. (forthcoming), “Labeling (Romance) Causatives”. Belletti, Adriana, and Carla Contemori (2010), “Intervention and Attraction. On the production of Subject and Object Relatives by Italian (young) children and adults,” in: João Costa, Ana Castro, Maria Lobo and Fernanda Pratas (Eds.), Language Acquisition and Development: Proceedings of GALA 2009, 39-52. Newcastle: Cambridge Scholars Press. —. (2012), “Subjects in children’s object relatives in Italian,” Revue Roumaine de Linguistique 57, 117-142. Belletti, Adriana, and Cristiano Chesi (2014), “A syntactic approach toward the interpretation of some distributional frequencies: Comparing relative clauses in Italian corpora and in elicited production,” Rivista di Grammatica Generativa 36, 1-28. Belletti, Adriana, and Maria Teresa Guasti (2015), The Acquisition of Italian. Morphosyntax and its interfaces in different modes of acquisition. Language Acquisition and Language Disorders series, John Benjamins Publishing Company. Belletti, Adriana, and Valentina Bianchi (2016), “Definitness Effect and Unaccusative Subjects: An Overview and some New Thoughts,” in: Susann Fischer, Tania Kupisch and Esther Rinke (Eds.), Definitness effects: Bilingual, typological and diachronic variation. Cambridge: Cambridge Scholars Press, 14-65. Berwick, Robert, Paul Pietroski, Yankaman Beracah and Noam Chomsky (2011), “Poverty of the stimulus revisited,” Cognitive Science 35, 1207-1242. Burzio, Luigi (1986), Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel. Chesi, Cristiano, Gianluca Lebani and Margherita Pallottino (2008), “A Bilingual Treebank (ITAǦLIS) suitable for Machine Translation: what

46

Chapter Two

Cartography and Minimalism teach us,” Studies in Linguistics (StiL) 2, 165Ǧ185. Chomsky, Noam (2013), “Problems of projection,” Lingua 130, 33–49. —. (2015), “Problems of Projection: Extensions,” in: Elisa Di Domenico, Cornelia Hamann and Simona Matteini (Eds.), Structures, Strategies and Beyond. Studies in honour of Adriana Belletti, 3-16, Amsterdam: Benjamins. Collins, Chris (2005), “A Smuggling approach to the passive in English,” Syntax 82, 81-120. Contemori, Carla, and Adriana Belletti (2014), “Relatives and Passive Object Relatives in Italian speaking children and adults: Intervention in production and comprehension,” Applied Psycholinguistics 35(6), 1021-1053. Contemori, Carla, and Maria Garraffa (2010), “Comparison of modalities in SLI syntax: A study on the comprehension and production of noncanonical sentences,” Lingua 120, 1940–1955. Costa, João, Maria Lobo and Carolina Silva (2011), “Subject-object asymmetries in the acquisition of Portuguese relative clause: adults vs. children,” Lingua 121/ 6, 1083-1100. Délage, Hélène (2008), Evolution de l’heterogeneité linguistique chez les enfants sourds moyens et legers: Étude de la complexité morphosyntaxique. Doctoral dissertation, Université de Tours. Friedmann, Naama (2007), “Young children and A-chains: the acquisition of Hebrew unaccusatives,” Language Acquisition 14, 377-422. Friedmann, Naama, Adriana Belletti and Luigi Rizzi (2009), “Relativized relatives: Types of intervention in the acquisition of A-bar dependencies,” Lingua 119, 67-88. Friedmann, Naama, Maya Yachini and Ronit Sztermann (2015), “Relatively easy relatives,” in: Elisa Di Domenico, Cornelia Hamann and Simona Matteini (Eds.), Structures, Strategies and Beyond. Studies in honour of Adriana Belletti, 303- 320. Amsterdam: Benjamins. Hamann, Cornelia (2011), “Binding and coreference. Views from child language,” in: Jill De Villier and Thomas Roeper (Eds.), Handobook of Generative Approaches to Language Acquisition, 247-290. Berlin: Springer. Harwood, William (2015), “Reduced Relatives and extended phases: A phase-based analysis of the inflectional restrictions on English reduced relative clauses,” Lingbuzz. Heim, Irene (1982), The semantics of definite and indefinite noun phrases. PhD dissertation, University of Massachusetts at Amherst. Labelle, Marie (2002), “The French non canonical passive in se-faire,”

Poverty of the Stimulus and Frequency

47

in: Shosuke Haraguchi, Bohumil Palek and Osamu Fujimura (Eds.), Proceedings of Linguistics and Phonetics 2002. Tokyo: Charles University Press and Meikai University. Lorusso, Paolo (2014), Verbs in Child Grammar. The Acquisition of the Primitive Elements of the VP at the Syntax-Semantics Interface. Doctoral dissertation Universitat Antònoma de Barcelona. Lust, Barbara (2005), Child Language: Acquisition and Growth. Cambridge: Cambridge University Press. MacWhinney, Brian, and Catherine E. Snow (1985), “The child language data exchange system,” Journal of child language 12/ 2, 271- 296. Milsark, Gary (1974) Existential Sentences in English. PhD Dissertation, MIT. —. (1977), “Toward an explanation of certain peculiarities of the existential construction in English,” Linguistic Analysis 3, 1-29. Manetti, Claudia, and Adriana Belletti (2015), “Causatives and the acquisition of the Italian passive,” in: Cornelia Hamann and Esther Ruigendijk (Eds.), Language Acquisition and Development. Proceedings of Gala-2013, 282-298. Cambridge: Cambridge Scholars Press. Novogrodsky, Rama, and Naama Friedmann (2006), “ The production of relative clauses in SLI: A window to the nature of the impairment,” Advances in Speech-Language pathology 8, 364- 375. Pinker, Steven (1994) The Language Instinct. New York: Harper Perennial Modern Classics Rizzi, Luigi (1990), Relativized Minimality. Cambridge, MA: The MIT Press. —. (1997), “The Fine Structure of the Left Periphery,” in: Liliane Haegeman (Ed.), Elements of Grammar, 281-337. Dordrecht: Kluwer. —. (2004), “Locality and the left periphery,” in: Adriana Belletti (Ed.), Structures and Beyond: The cartography of syntactic structures, Vol. 3, 223-251. Oxford/ New York: Oxford University Press. —. (2006), “Grammatically-Based Target-Inconsistencies in Child Language,” in: Kamil Ud Deen, Jun Nomura, Barbara Schulz and Bonnie D. Schwartz (Eds.), The Proceedings of the Inaugural Conference on Generative Approaches to Language Acquisition–North America (GALANA), Honolulu, HI, 19-49. University of Connecticut Occasional Papers in Linguistics 4. —. (2015a), “Cartography, Criteria, and Labeling,” in: Ur Shlonsky, (Ed.), Beyond Functional Sequence: The Cartography of Syntactic Structures, Vol. 10, 314- 338. New York: Oxford University Press. —. (2015b), “Notes on Labeling and Subject Positions,” in: Elisa Di

48

Chapter Two

Domenico, Cornelia Hamann and Simona Matteini (Eds.), Structures, Strategies and Beyond. Studies in honour of Adriana Belletti, 17-46. Amsterdam: Benjamins. Roland, Douglas, Frederic Dick and Jeffrey L. Elman (2007), “Frequency of basic English Grammatical structures: A corpus analysis,” Journal of Memory and Language 57/3, 348-379. Santos, Ana Lúcia, Anabela Gonçalves and Nina Hyams (2015), “Aspects of the acquisition of object control and ECM-Type verbs in European Portuguese,” Language Acquisition, Published online: DOI: 10.1080/10489223.2015.1067320. Siloni, Tali (1997), Noun Phrases and Nominalizations: The Syntax of DPs. Dordrecht: Kluwer. Spina, Stefania (2005), “ Il Corpus di Italiano Televisivo (Cit): struttura e annotazione,” in: Elizabeth Burr (Ed.), Traduzione e innovazione. Il parlato: teoria – corpora – linguistica dei corpora. Proceedings of the VI SILFI Conference, 413 – 426. Firenze: Franco Cesati. Tomasello, Michael (2003), Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Vernice, Mirta, and Maria Teresa Guasti (2015), “The acquisition of SV order in unaccusatives: manipulating the definiteness of the NP argument,” Journal of Child Language 42/1: 210-237. Yang, Charles (2012), “Ontogeny and phylogeny of language”. PNAS: www.pnas.org/cgi/doi/10.1073/pnas.1216803110.

CHAPTER THREE WHICH AND HOW-MANY QUESTIONS

IN THE ACQUISITION OF ITALIAN ELENA PAGLIARINI AND MARIA TERESA GUASTI

1. Introduction Subject questions are understood better than object questions by children. This asymmetry holds cross-linguistically for which-questions beyond age 5;0 and till age 10 years, with variations depending on the age of children and likely on the language (Avrutin 2000; De Vincenzi et al. 1999; Friedmann, Belletti and Rizzi 2009; Strangmann, Slop and van Hout 2014). For languages such as English and Hebrew, Friedmann, Belletti and Rizzi (2009) have attributed the object penalty to a locality violation. Consider the English examples and their representations in (1): (1) a. Which fairy is pushing the lady? SUBJECT [+Wh, +NP] b. Which fairy is the lady pushing ? OBJECT [+Wh, +NP] [+NP]………….. In (1a) nothing intervenes in the A’-dependency between the extracted subject wh-phrase and its copy. In (1b), instead, the subject intervenes in the A’-dependency between the displaced wh-phrase and its copy and this is the source of the difficulty. Typically, an A’-dependency is ruled out by relativized minimality if the intervener Z, located between the displaced



We thank two anonymous reviewers for comments. Although the paper has been written jointly, Pagliarini is responsible for sections 1,2 and Guasti for sections 3,4.

50

Chapter Three

item, X, and its copy, Y, (1) is identical to X, (2) c-commands Y (but not X) and thus is a candidate for the same relation as X (e.g., how do you wonder who behaved ?). Under a featural implementation of Relativized Minimality (RM) (Rizzi 1990), the feature composition of the elements involved in the dependency plays a crucial role. Children and adults are constrained by the same version of RM and rule out sentences in which the intervener has the same features as the target. Children and adults differ in their ability to compute the featural composition of the elements involved in the dependency. Specifically, Friedmann, Belletti and Rizzi (2009) claim that in the case of (1b) the intervening subject may have disruptive effects for children because they fail to establish that it is distinct from the moved element. This is due to the failure to compute the subset relation holding between its features and those of the target (the which-phrase). Hence, for children an object which-question instantiates a Relativized Minimality violation. The same explanation has been offered to account for the fact that object relative clauses are more challenging for children than subject relative clauses. In object relative clauses, the embedded subject, with +NP feature, intervenes between the relative head, which has [+R, +NP] feature, and its copy, as seen in (2). Children misunderstand object relative clauses, because they may fail to compute the subset relation holding between the features of the intervening subject and those of the relative head. For ease of reference, we refer to this approach as the Argument Intervention approach. (2) Show me the boy that the lady is washing . [+R, +NP] [+NP] [+R, +NP] Costa et al. (2014) have extended the Argument intervention approach to account for the observation that not only object relative clauses, but also relative clauses headed by a Prepositional Phrase (PP) are difficult to understand for children. Consider the example in (3) from European Portuguese: (3) Mostra-me o rapaz para que o pai olha. Show-me the boy at whom daddy looks. Show me the boy who daddy is looking at. In (3), the intervening subject is a DP and what has been moved in a PP. If the categorial features of the elements involved in the dependency counted for intervention, (3) should not give rise to any intervention effect and should not be problematic for children (or difficulties should be

Which and How-Many Questions in the Acquisition of Italian

51

explained without appeal to intervention). Costa et al. (2014) opt for a unified account of the challenges experienced by children with object and PP relative clauses. They claim that what matters for intervention are features involved in establishing the reference of the copy, specifically, the presence of a +NP feature both in the DP and the PP. Given this background, let us move to Italian and to a possible challenge for the Argument intervention approach. In Italian, the subject does not intervene between the wh-phrase and its copy, because it typically stays in a postverbal position, as in (4). Yet, object which questions are challenging for Italian-speaking children, as they are for English or Hebrew-speaking children (De Vincenzi et al. 1999). (4) Quale fatina stanno spingendo le signore? Which fairy are pushing the ladies? Which fairy are the ladies pushing? The Argument intervention approach cannot account for children’s challenges with (4). However, (4) features another potential locality violation. Extending Guasti and Rizzi (2002) and Franck et al. (2006), Guasti, Branchini and Arosio (2012) have argued that the difficulties stem from a locality violation occurring when the mechanism of Agree between the inflectional head and the subject in its thematic position is applied. Under the assumption that the wh-object moves to a preliminary intermediate position, say AgrOP on its way to Spec CP, the representation of (4) is as in (5): (5) [CP Quale fatina stanno [AgrOP spingendo] [vP le signore]]? Which fairy are pushing the ladies? When the inflected verb stanno ('are') looks for the goal to agree with in its c-commanding domain, it first finds the object copy quale fatina, ('which fairy') in Spec AgrOP. Since this copy is closer than the real goal (le signore, ‘the ladies’) and has the same categorial features as the real goal (+NP), it occasionally enters into the Agree relation with the verb and is mistakenly taken to be the subject rather than the object. In turn, (4/5) is understood as a subject, rather than an object question. In other words, (4) is difficult because the object copy intervenes in the Agree relation involving the inflected verbs and the postverbal subject. We call this approach the Agree intervention approach. Although this approach is

52

Chapter Three

expressed in terms of locality and appeals to features, it is distinct from the Argument intervention approach. First, it explains facts that the Argument intervention approach does not explain. Second, it appeals to different notions and has a distinct pattern of development, as we will see in this study. In which-questions as in (4), the wh-phrase asks for the identity of the patient. How many questions, as in (6), ask for the number of individuals involved in an action. (6) Quante fatine sta spingendo la signora? How many-PL fairies is pushing the lady? How many fairies is the lady pushing? A which-phrase and a how many-phrase are similar in terms of category, both include [+wh, +NP feature]. They are referentially different, as the response to the former is expressed with a DP, while the response to the latter with a quantified phrase (a QP). In discussing the Argument intervention approach, we have just seen that categorial features do not matter for intervention, while features involved in establishing reference, i.e. [+NP], do. This fact is plausible as in the Argument intervention approach, one needs to create a dependency that ensures the interpretation of the copy. Nothing of this kind is at stake in the Agree intervention approach. The agree relation expresses a formal process ensuring the evaluation of the agreement feature and what counts is likely categorial features or formal features (e.g., morphological case). In line with these assumptions, we expect that comprehension of the how-many question in (6) will be similar to comprehension of which questions in (4), although the two questions ask for the identity and thus for the referent of a DP and for numerosity, i.e., for a QP. Testing this prediction is the first goal of our study. If the prediction is fulfilled, it would provide evidence for the claim that the Argument and the Agree intervention approaches work in distinct manners. The second goal is to reinforce this claim by showing that the effects taken care of by the Argument and Agree intervention approach have distinct developmental patterns. We will do this by comparing comprehension of which-questions with postverbal subjects, as in (4) and which-questions with preverbal subjects, as in (7). (7) Mi puoi dire quale fatina le signore stanno spingendo? To-me can tell which fairy the ladies are pushing? Can you tell me which fairy are the ladies pushing?

Which and How-Many Questions in the Acquisition of Italian

53

Arosio, Adani and Guasti (2009) and Adani (2011) tested the comprehension of two types of object relative clauses in Italian, one with the embedded subject in the preverbal position and one with the embedded subject in the postverbal position. Object relative clauses with the preverbal subject were already comprehended relatively well at age 5, while object relative clauses with the postverbal subject were still difficult at that age and were better understood at age 7 and 9. On the basis of this result, we anticipate that which-questions with the preverbal subject will be comprehended better than those with the postverbal subject. To test these predictions, we conducted two experiments, one aimed at comparing comprehension of subject and object which- and how manyquestions and the second designed to test comprehension of only object which-questions with pre- and postverbal subjects.

2. Experiment 1: Comprehension of Subject and Object which- and how many-questions This experiment aims at establishing whether object which- and how many-questions elicit distinct comprehension patterns from children or not. We will do this by examining the comprehension of a group of 5 years old children and a group of 6 years old children. This will also allow us to establish whether there is an improvement across ages.

2.1 Participants Forty-nine monolingual Italian-speaking children with no history of language delay or impairment participated in the experiment. They were divided in two groups: a group of 5 years old children (N = 25; Mean age = 5;8; SD = .4; Age range = 5;3 - 6;3; 10 female); a group of 6 years old children (N = 24; Mean age = 6;6; SD = .3; Age range = 6;0 - 7;3; 12 female).

2.2 Materials We manipulated the type of wh-element (quali-NP 'which-NP' vs. quantiNP, 'how many-NP') in questions with number morphology disambiguation on the verb. The experiment included four experimental conditions. The question was either a quali ('which') question or it was a quanti ('how many') question. Furthermore, each question was either a subject question or it was an object question. The same verb was used in a subject and in an object question, but the nouns were changed. Since we used a within-subject design, this ensured a set of materials that was not

Chapter Three

54

repetitive. This yields a 2x2 factorial design. Each condition included 6 sentence-picture pairs (24 experimental stimuli in total). (8) a.

Quali leoni t rincorrono il cavallo? (S-WHICH) Which-PL lions are chasing the horse? Which lions are chasing the horse?

b.

Quali maiali rincorre la scimmia t ? (O-WHICH) Which-PL pig is chasing the monkey? Which pigs is the monkey chasing?

c.

Quanti conigli t annusano il cavallo? (S-HOW MANY) How many-PL rabbits are smelling the horse? How many rabbits are smelling the horse?

d.

Quante pecore annusa la mucca t ? (O-HOW How many-PL sheeps is smelling the cow? How many sheeps is the cow smelling?

MANY)

All sentences included transitive reversible verbs. The wh-element was always plural. In subject questions, such as (8a), the verb was inflected for 3rd plural agreement and it matched with the plural wh-phrase. In object sentences, such as (8b), the verb was inflected for 3rd singular agreement and it mismatched with the plural wh-phrase. Crucially, the number mismatch between the wh-element and its pied-piping NP (qualiPLUR maialiPLUR) and the verb (rincorreSING) was a strong syntactic cue signaling that a subject trace could not be posited, and therefore that it was not a subject question. This holds for both quali-NP and quanti-NP questions. Sentence length was similar across items (mean length quali-NP = 49 sec, SD = 51 msec; mean length quanti-NP = 53 sec, SD = 65 msec). All the sentences were spoken along with a picture displaying three sets of characters, as in Fig. 1. Moreover, all the sentences were recorded by a female native speaker of Italian.

Which and How-Many Questions in the Acquisition of Italian

55

Fig. 1: Sample of stimuli used in Experiment 1.

For instance, on each trial, a sentence such as (8a) appeared together with the event represented in Fig. 1, involving a horse in the center that is chasing a set of lions on its left and that is chased by a set of lions on its right. Therefore, for each picture, the target response was either the set of characters on the left or the set of characters on the right (never the one in the center). Left and right were counterbalanced across pictures.

2.3 Procedure Each child was tested individually in a quiet room at school. Before the experiment started, children’s comprehension of all nouns, colors (in the case of quali-NP questions) and numbers (in the case of quanti-NP questions) used in the experiment was assessed. The experiment was conducted using a Power Point presentation. While the picture was displayed on the screen, children heard the target sentence (which was a question) and were instructed to look carefully at the picture. The participant's task was to reply to the question. The elicited answer was a color in the case of quali-NP questions, and a number (ranging from 2 to 5) in the case of quanti-NP questions. We used a within-subject betweenconditions design, where the test conditions (quali-NP and quanti-NP questions) were separated and presented in two separate sessions spaced by some phonological tasks.

2.4 Results The percentages of correct responses for the two age groups, for each type of wh-element and extraction site are presented in Table 1.

Chapter Three

56

5 YO 6 YO

S-WHICH 98 96

O-WHICH 21.3 44

S-HOW MANY 94.6 93.3

O-HOW MANY 28 48

Table 1: Percentage of correct responses for each type of questions. (S-WHICH = subject which question; O-WHICH = object which question; S-HOW MANY = subject how many question; O-HOW MANY = object how many question) for the two group of age (5 years old = 5;8; 6 years old = 6;6). The descriptive data suggest that the subject/object asymmetry holds both for quali-NP and quanti-NP questions, as more accurate responses were found in subject than object questions. The comparison of the two groups of children shows that six years old children display slightly more accurate responses on object questions, but still object questions remain more difficult than subject questions. We run mixed logit models on response accuracy using R (see Jaeger 2008, for the implementation of mixed logit models for categorical data analysis). We used a forward stepwise regression to find the best model. Our analyses included Group as a between subject factor, Extraction site (subject vs. object) and Type of wh-element (quali-NP vs. quanti-NP) as within subject factors. Subject and items were modeled as random effects. First, we determined whether one predictor significantly contributed to the model fit by comparing a model including that predictor against another that did not contain it, using a Ȥ-square test (Jaeger 2008). Afterwards, we calculated the z value, based on the Wald statistic, allowing for an estimation of the statistical significance of each predictor included in the model. In our analyses, the reference categories were 5year-olds for Group, Object for Extraction site and quali-NP for Type of wh-element. Group (Ȥ(2)=10.93, p < .001) and Extraction site (Ȥ(2)=39.5, p < .001) added significant information to the model. Type of wh-element did not add significant information to the model (Ȥ(2)= .46, p = .49). Results are reported in Table 2.

Which and How-Many Questions in the Acquisition of Italian

Predictor

Estimate SE

57

Z Wald p

(Intercept) -1.73 0.38 -4.47 < .001 Group (ref. cat=5-year-olds) Group=6-year-olds 1.58 0.48 3.29 < .001 Extraction site (ref. cat.=object) Subject 6.25 0.49 12.77 < .001 Subjects and items had SD of 2 and 0.4 respectively. Summary of the fixed effects in the mixed logit model (N = 1176; log-likelihood= 357.24)

Table 2: Results of the analysis of Experiment 1 In sum, the results show that there is a subject/object asymmetry, as participants responded less accurately to object questions. In addition, the effect of Group shows that 6 years old children are more accurate than 5 years old children.

3. Experiment 2: Comprehension of wh-object verb subject vs. wh-object subject verb The goal of this experiment is to establish whether Argument and Agree intervention give rise to different disruption effects. To examine this hypothesis, only one group of children was tested. 

3.1 Participants Thirty-four Italian monolingual children with no history of language delay or impairment participated to the experiment (N = 34; Mean age = 5;3; SD = .3; Age range = 4;5 - 6;1; 20 female).

3.2 Material We manipulated the order of the subject in quali-NP object questions. Each participant was tested in 5 quali-NP object-verb-subject questions (OVS), (9a), and in 5 quali-NP object-subject-verb questions (OSV), (9b): (9) a. Mi puoi dire quali maialiOBJ insegue il cavalloSUBJ t? To-me can tell which-PL pigs is chasing the horse? Can you tell me which pigs is the horse chasing?

58

Chapter Three

b. Mi puoi dire quuali tartarrugheOBJ la gaallinaSUBJ inseg gue t? To-mee can tell which-PL w turtlees the hhen is chaasing? Can you tell me which turtles is the t hen chasinng? All senteences includeed transitive reversible r verrbs. The quali-NP was always plurral. The verb was always inflected i for 3 rd singular agreement a and it missmatched witth the pluraal wh-elemennt. Gender, which w is morphologiccally marked in Italian, was w controlledd; in each queestion the gender of thhe subject and object NP was kept connstant. Senten nce length was similar across items (mean length h OSV questioons = 10 sec.., SD = 7 msec; meann length OVS S questions = 9.4 sec., SD = 5 msec). All A items were introduuced by a short preamble su uch as In quessta storia ci sono degli animali che si rincorronoo (‘In this storry there are soome animals that t chase each other’)). All questionns were introd duced by Mi ppuoi dire (‘Can n you tell me’) and w were spoken along with a picture dissplaying threee sets of characters, aas in Fig. 2. All A the sentencces were recorrded by a femaale native speaker of Ittalian.

Fig. 2: Samp mple of stimuli used u in Experim ment 2.

Similarlyy to Experimeent 1, on each h trial, a senttence such as (9a) was heard togethher with the evvent representted in Fig. 2, in involving a ho orse in the center that is chasing a seet of pigs on itts right and thhat is chased by b a set of pigs on its left. Thereforee, for each triaal, the target rresponse was either the set of characcters on the leeft or the set of o characters on the right (never ( the one in the ceenter). Left annd right were counterbalanc c ced across picttures. Experim mental items were interspersed with 4 fillers, which were questions abbout the colorr of an objectt displayed inn the picture (In questa

Which and How-Many Questions in the Acquisition of Italian

59

storia c’è un bambino che salta. Mi puoi dire di che colore ha la maglietta? 'In this story there is a boy who is jumping. Can you tell me what color is his t-shirt?').

3.3 Procedure Each participant was tested individually in a quiet room at school. Before the testing session, children’s comprehension of all the nouns and colors used in the experiment was assessed. The experiment was conducted using a Power Point presentation. A puppet watched the story along with the children. Children were told that they had to teach the color names to the puppet. While watching the picture, participants heard the brief preamble and were asked to answer the question, whose answer was always a color name.

3.4 Results Percentages of correct responses on filler totalled to 100%, so all children were included in the data analysis. However, we excluded from the analysis an OVS item, since a high percentage of participants gave an incorrect response (88.24%). This bias towards an incorrect answer was probably due to a not appropriate picture. Thus, the analysis was conducted on 5 items for OSV order and 4 items for OVS order. The percentages of correct responses for OVS questions was 48.5% and for OSV questions was 76.5%. Therefore, by looking at the descriptive data, OSV order elicits more accurate responses than OVS order. We run mixed logit models on response accuracy using R. We used a forward stepwise regression to find the best model. Our analyses included Order (OVS vs. OSV) as within subject factors. Subject and items were modeled as random effects. We determined the z value, based on the Wald statistic, which allows for an estimation of the statistical significance of our independent variable. In the analyses conducted, the reference category was OSV for Order. Results are reported in Table 3. The statistical analysis confirmed that participants’ responses were more accurate in OSV order questions.

60

Chapter Three

Predictor Estimate SE Z Wald p (Intercept) 2.04 0.49 4.17 < .001 Order (ref.cat=OSV) Order=OVS -2.11 0.37 -5.71 < .001 Subjects and items had SD of 2.2 and 9.6 respectively. Summary of the fixed effects in the mixed logit model (N = 306; log-likelihood= 148.74)

Table 3: Results of the analysis of Experiment 2

4. Discussion In Italian object which-questions, the subject is placed in a postverbal position and does not intervene between the which-phrase and the object copy. Therefore, it cannot be the source of a potential featural Relativized Minimality violation and Italian which-questions should not be problematic for children, as far as the Argument intervention approach is concerned. However, Italian object which-questions are challenging for children. To account for this fact, we appealed to the Agree intervention approach, whereby the object copy on its way to Comp lands in an intermediate position. In that position, it acts as a closer goal for the Agree relation and occasionally it values the agreement features of the inflectional node, turning an object into a subject question. The Argument and Agree intervention approaches both appeal to the notion of locality, i.e., to the notion of a closer element for some given relation; yet they are two distinct constructs. The first study was set out to provide new evidence for this claim by looking at what features count for intervention in the two mechanisms. The rationale behind it is as follows. In the Argument intervention approach, the features that count are not categorial features, but features involved in establishing reference (the [+NP] feature). To recall, in a study on the comprehension of PP and direct object relative clauses, Costa et al. (2014) showed that children found PP relatives as difficult as object relatives to comprehend: in both cases the intervening subject shared features used in establishing reference with the moved element, being this a PP or a DP. By contrast, the features that disrupt an agree relation are categorial or formal features, as agree is a formal relation. In line with this observation, the first experiment established that the subject/object asymmetry holds for how many- and which-questions. This similarity stands in spite of the fact that these questions ask for different referents, a quantified Phrase (QP) and a

Which and How-Many Questions in the Acquisition of Italian

61

referential DP, respectively. In the same experiment, we showed that 6year-olds perform better than 5-year-olds. The second goal of our study was to provide evidence from language development to distinguish the two approaches. This goal was achieved by comparing the comprehension of which-object questions with a preverbal and postverbal subject. We showed that children understand object whichquestions with a preverbal subject better than object which-questions with a postverbal subject. This result is in line with the findings reported in Arosio et al. (2009) on children’s comprehension of Italian relative clauses. The scores obtained by children in the comprehension of object which-questions with a preverbal subject doubled those obtained in the comprehension of object which-questions with a postverbal subject. This result provides evidence that the disruptive effects of the Argument intervention are fading away at age 5 in Italian, while the disruptive effects of intervention on the Agree relation are still very robust. In other words, children are sensitive for a longer time to a strict version of locality when it comes to the Agree relation. In sum, we have provided empirical evidence from child language for the presence of two locality mechanisms in the grammar, one concerned with the formation of chains and one of valuing the inflected verb. These two mechanisms account for distinct difficulties in the formation of object questions and object relative clauses. They work on distinct features and they display a different pattern of development.

References Arosio, Fabrizio, Flavia Adani and Maria Teresa Guasti (2009), “Grammatical features in the comprehension of Italian relative clauses by children,” in: José M. Brucart, Anna Gavarró, and Jaume Solà (Eds.), Merging Features: Computation, Interpretation, and Acquisition, 138-155. Oxford: Oxford University Press. Adani, Flavia (2011), “Rethinking the acquisition of relative clauses in Italian: Towards a grammatically based account,” Journal of Child Language 38, 141–165. Avrutin, Sergey (2000), “Comprehension of discourse-linked and nondiscourse-linked questions by children and Broca’s aphasics,” in: Yosef Grodzinsky, Lewis P. Shapiro and David Swinney (Eds.), Language and the brain: Representation and processing, 295-313, San Diego, California: Academic Press.

62

Chapter Three

Costa, João, Naama Friedmann, Carolina Silva and Maya Yachini (2014), “The boy that the chef cooked: Acquisition of PP relatives in European Portuguese and Hebrew,” Lingua 150, 386-409. De Vincenzi, Marica, Lisa Arduino, Laura Ciccarelli and Remo Job (1999), “Parsing strategies in children comprehension of interrogative sentences,” in: Sebastiano Bagnara (Ed.), Proceedings of the European Conference on Cognitive Science, 301-308, Rome: Istituto di Psicologia del CNR. Franck, Julie, Glenda Lassi, Ulrich H. Frauenfelder, and Luigi Rizzi (2006), “Agreement and movement: A syntactic analysis of attraction,” Cognition 101, 173–216. Friedmann, Naama, Adriana Belletti and Luigi Rizzi (2009), “Relativized relatives: Types of intervention in the acquisition of A-bar dependencies,” Lingua 119, 67-88. Guasti, Maria Teresa and Luigi Rizzi (2002), “Agreement and tense as distinctive syntactic positions. Evidence from acquisition,” in: Guglielmo Cinque (Ed.), Functional structure in DP and IP: The Cartography of Syntactic Structures, Vol. 1, 167-194. Oxford: Oxford University Press. Guasti, Maria Teresa, Chiara Branchini and Fabrizio Arosio (2012), “Interference in the production of Italian subject and object whquestions,” Applied Psycholinguistics 33, 185-223. Guasti, Maria Teresa, Stavroula Stavrakaki and Fabrizio Arosio (2012), “Cross-linguistic differences and similarities in the acquisition of relative clauses: Evidence from Greek and Italian,” Lingua 122, 700713. Jaeger, T. Florian (2008), “Categorical data analysis: Away from ANOVAs (transformation or not) and towards logit mixed models,” Journal of memory and language, 59(4), 434-446. Rizzi, Luigi (1990), Relativized Minimality. Cambridge, MA: The MIT Press. Strangmann, Iris, Anneke Slomp and Angeliek van Hout (2014), “Development of Dutch children's comprehension of subject and object wh-questions: The role of topicality,” Linguistics in the Netherlands 31, 129-144.

CHAPTER FOUR ON GERMAN “V2 RELATIVE CLAUSES”: LINGUISTIC THEORY MEETS ACQUISITION EMANUELA SANFELICI, PETRA SCHULZ AND CORINNA TRABANDT

1. Introduction A vast amount of literature has dealt with the role of finite verb placement and its acquisition path in German, both in monolingual German-speaking children (e.g., Clahsen 1982, Clahsen and Smolka 1985, Weissenborn 1990, Tracy 1991, Clahsen, Penke and Parodi 1992, Rothweiler 1993) and in L2-learners (Hamann 2000, Prévost 2003, Rothweiler 2006, Tracy and Thoma 2009, Meisel 2009, among others). Much attention has been paid to detecting when and how children acquire the verb placement parameter and the ability to correctly fill the verb position, namely V2 in finite main clauses and verb-final in subordinate clauses. The data show that German children acquire the verb placement parameter very early and place the verb correctly in both matrix and embedded contexts around age 3 (see Clahsen and Smolka 1985, Wexler 1998). This conclusion is generally shared in the German acquisition literature and has been corroborated by many studies (see Rothweiler 1993, Tracy 1995).



We thank the two anonymous reviewers for their helpful comments, Adriana Belletti, Hans Martin Gärtner, Günther Grewendorf, Klaus von Heusinger, Joachim Jacobs, Horst Lohnstein, Alan Munn, Cecilia Poletto, Claudia Poschmann, Florian Schwarz, Alex Thiel, and Helmut Weiß, for their feedback on the data and/or on the ideas developed in this work. Although this article has been conceived and worked on jointly and all the proposals made are shared by the three authors, Emanuela Sanfelici takes direct responsibility of Sections 2, 3, 5, 6, Petra Schulz of Sections 4, 4.1, 4.2, 4.3, and Corinna Trabandt of Sections 4.4 and 4.5. The introductory and discussion Sections, 1 and 4.6, are in common.

Chapter Four

64

Although verb-final placement is usually taken as one of the main characteristics of subordinate clauses in German, in specific syntactic environments subordinate clauses may license V2. This option is available e.g., in some complement clauses, in weil-clauses, and in relative clauses (for an overview, see Reis 1997, Wurmbrandt 2005, among many others). Up to now, it is an open question in acquisition research how children deal with these contexts, in which V2 and verb-final are both grammatical options. The present paper addresses this question, investigating whether children prefer V2 or verb-final in a controlled experimental setting where the syntactic and semantic conditions allow for both structures. The syntactic structures we used are relative clauses (RCs). In German RCs the verb usually occupies the final position (1a), but under specific conditions, so-called integrated V2 (henceforth iV2)1 structures as in (1b) are licensed (Brandt 1990, Gärtner 2001a/b, Zwart 2005):2 (1) a.

Hier gibt es zwei Frauen, die den here there-is EXPL two women PRON:NOM the:ACC Präsidenten getroffen haben. president met have ‘Here there are two women that met the President.’ RC

b.

Hier gibt es zwei Frauen, die haben here there-is EXPL two women PRON:NOM have den Präsidenten getroffen. the:ACC president met ‘Here there are two women that met the President.’ iV2

Note that in (1), at the surface iV2 clauses and verb-final RCs are minimal pairs differing only in the position of the finite predicate. Whereas verb-final RCs are syntactically subordinate clauses, iV2 structures have been treated as paratactic coordination, i.e. as main clauses that are paratactically coordinated with the clause containing a presentational or existential predicate (Gärtner 2001a/b, 2002, Endriss and Gärtner 2005, de Vries 2006, a.o.). Previous acquisition research seems to support this syntactic analysis, as several studies have reported V2

 1 Note that previous acquisition studies have referred to iV2 structures as “V2relatives”. 2 Examples will be provided with glosses when the morpho-syntactic information is relevant for the explanation. Otherwise, we will give the direct English translation and highlight in bold the relevant item.

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

65

structures to be the first relative clause-like structures produced in spontaneous speech by children up to age 4 (Brandt 2004, Diessel and Tomasello 2005, Brandt, Diessel and Tomasello 2008). Unlike this line of acquisition studies, we investigated how children deal with the alternation in (1) in an experimentally controlled context. A picture-supported delayed-repetition task was developed to test monolingual German-speaking children between age 3 and 5. In the following we report our results comprising data from 3-year-old children (n = 23) and from the control group of adults (n = 21). Our experimental results show a robust preference in children for verb-final RCs over iV2 structures. Whereas adults correctly repeated iV2 structures with V2, children showed a clear tendency to change iV2 stimuli into verb-final RCs. These findings contradict the acquisition pattern proposed in Brandt et al. (2008) and, furthermore, challenge the syntactic coordination analysis for iV2 clauses as proposed in Gärtner (2001a/b). The paper is structured as follows. In Section 2 we outline some basic properties of German relative clauses. We then focus on iV2 clauses and their licensing conditions and briefly summarize the analysis proposed by Gärtner (2001a/b). Section 3 provides an overview of the previous acquisition studies. In Section 4, our experimental design is sketched, and the results are discussed. In Section 5, we delineate our syntactic proposal for iV2s and provide evidence in support of our derivation. Section 6 presents concluding remarks.

2. Relative Clauses in German: V2 and V-final Structures Finite RCs in German appear in post-nominal position. The verb usually occupies the final position, but as stated in the Introduction, under specific conditions iV2 structures are licensed (see example (1)) (Brandt 1990, Gärtner 2001a/b, Zwart 2005). In the following we first characterize the main properties of verb-final RCs (Section 2.1) and then focus on the properties exhibited by iV2 structures and outline the analysis proposed in Gärtner (2001a/b) (Section 2.2).

2.1. Verb-final Relative Clauses Verb-final RCs are introduced by a relative pronoun in Standard German and by an uninflected complementizer in some German varieties. In Standard German, the relative pronoun der/die/das ‘which’ must be marked for gender, number and case; this type of RC is labeled d-RC. Due to syncretism in the inflectional paradigm of German, only the masculine

Chapter Four

66

singular form of the d-pronoun differs in its overt morphological realization in nominative vs. accusative case (der vs. den), as illustrated in (2a) compared to (2b). The feminine and neuter forms of the relative pronoun are identical for nominative and accusative case as are the plural forms of all d-relative pronouns. (2) a.

b.

Der Mann, der rote Haare hat, ist mein the:NOM man PRON:NOM red hair has is my Bruder. brother ‘The man that has red hair is my brother.’ Subject RC Der Mann, den du getroffen hast, ist the:NOM man PRON:ACC you:NOM met have is mein Bruder. my brother ‘The man that you met is my brother.’ Object RC

D-RCs appear both center-embedded as in (2) and sentence final as in (3). (3) Du hast gestern einen Mann getroffen, der you:NOM have yesterday a:ACC man met PRON:NOM rote Haare hat. red hair has ‘Yesterday, you met a man that has red hair.’ Apart from d-relative pronouns, Standard German RCs can be introduced by other pronouns including welcher ‘which’ and wer ‘who’; these are reported to be rare (Fleischer 2004: 218). In some German varieties, such as Alemannic and Bavarian dialects, RCs may be introduced by complementizers such as was ‘what’ or wo ‘where’, as illustrated in (4): (4) Der Mann, wo einen Hut auf hat, ist mein Bruder. the:NOM man where a:ACC hat on has is my brother ‘The man that wears a hat is my brother.’ Concerning the area of Hesse, where our testing took place, marking of RCs via d-pronoun is reported to be the most widespread strategy of relativization (Fleischer 2004). Hence, in the following we focus on dRCs.

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

67

2.2. iV2 Structures: A Paratactic Analysis iV2 structures as in (1b), repeated below as (5), are reported to be frequent in spontaneous speech, although their distribution is restricted to specific contexts (Weinert 2012).3 (5) Hier gibt es zwei Frauen, die haben den here there-is EXPL two women PRON:NOM have the:ACC Präsidenten getroffen. president met ‘Here there are two women that met the President.’ Both theoretical (Brandt 1990, Gärtner 2001a/b, 2002) and empirical studies (Weinert 2012) agree that iV2 clauses are licensed under specific syntactic, semantic, and prosodic conditions, spelled out below: (a) Type of predicate The predicate in the main clause must be presentational or existential (Gärtner 2001a/b). Moreover, according to Weinert (2012) iV2 clauses are mainly licensed by a small set of matrix predicates, among them es gibt ‘there is’, da ist/sind ‘there be’, possessive existential haben ‘have’, and evidential existentials as sehen ‘see’, kennen ‘know’ and hören ‘hear’. (b) Type of antecedent The antecedent must be indefinite. More precisely, it must be a weak noun phrase in the sense of Milsark (1974), which is introduced by an intersective determiner, e.g., by some, by numerals, or by the indefinite article. Note that these are low risk quantifiers in Gärtner’s (2007) terms. This restriction regarding the antecedent is confirmed by the corpus data: Weinert (2012) reports that she did not find strongly quantified NPs as antecedents of iV2 structures.4 Furthermore, the indefinite must be specific

 3

Weinert (2012) performed a corpus study of spontaneous speech of both male and female speakers from many geographical areas of Germany. The corpus includes around 280.000 words from a range of informal, formal and public data. 4 There is one exception to this robust pattern, namely (i) with a definite DP. However, as Weinert herself suggests, it involves an ‘of a kind’ reading (Weinert 2012: 260). (i) es gibt die leute die na die die springen ins wasser die schwimmen immer weißte ‘there are some people who who who jump right in at the deep end and they always stay afloat you know (= things always turn out all right for

68

Chapter Four

(Gärtner 2001a/b). (c) Type of introducing pronoun The iV2 structure can be introduced by d-pronouns only. If introduced by a w-pronoun, the verb has to appear in final position (Gärtner 1998). Furthermore, Weinert’s study shows that subject pronouns dominate. More than 70% of iV2 clauses contain a subject gap. (d) Position of the iV2 clause Unlike verb-final RCs, iV2 clauses cannot be center-embedded. They only appear in sentence-final position (Gärtner 2001a/b). (e) Semantic role of iV2 structures iV2 structures can only have a restrictive interpretation (Gärtner 1998, and subsequent works). (f) Prosodic contour of iV2 clauses iV2 clauses must be prosodically integrated in the main clause (Gärtner 2001a/b, Endriss and Gärtner 2005).5 Given the properties identified for iV2 structures, both Gärtner and Weinert conclude that iV2 clauses exhibit a hybrid behavior that is intermediate between real restrictive verb-final RCs and main clauses. Like restrictive RCs, iV2 clauses restrict the reference of the NP and behave like predicates. In addition, they are prosodically integrated into the matrix clause. Like main clauses, iV2 structures show the verb in second position and always appear in sentence-final position. In order to capture these properties, Gärtner (2001a/b) suggests that iV2 clauses

 certain people)’ The only counterproposal for prosodic integration of iV2 structures we are aware of comes from Birkner (2006, 2008). The author analyzed corpus examples of what she calls the Mensch-Konstruktion ‘person-construction’ in (i) (from Birkner 2006: 222). She observes that the level of syntactic integration is overall reflected prosodically, with verb-final relative clauses tending to be integrated (53% of 53 strongly integrated) and verb-second clauses tending not to be (69% of 16 unintegrated). (i) das Is einfach n = mensch äh der gehört auch zu mir that is simply a person eh pron belongs also to me ‘he’s simply someone who is also part of my life’ However, we disregard her results due to the amount of variation observed in the data and the small number of verb-second clauses in Birkner’s study. We leave a deeper prosodic analysis for future research.

5

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

69

involve paratactic coordination. More precisely, iV2 structures are claimed to be main clauses paratactically coordinated with the clause containing the presentational or existential predicate. According to this analysis, (6) has the representation in (7) (Gärtner 2002, Endriss and Gärtner 2005, de Vries 2006, a.o): (6) Maria ist ein Menschi, deri liebt seine Kinder. Maria is a person PRON:NOM loves his children ‘Maria is a person that loves her children.’ (7)

The structure in (7) accounts for the core properties of iV2 structures: The verb is in second position as in main clauses; iV2 clauses cannot be center-embedded since they are main clauses; only d-pronouns are licensed since the pronoun is an anaphor. In order to also account for the semantic and prosodic similarity of iV2 structures with verb-final RCs, Gärtner further refines the structure in (7). His starting point is the observation that iV2 clauses are syntactically main clauses, but do not have the same interpretation as simple main clauses. This is illustrated in (8). Unlike the main clause (8b), the iV2 structure in (8a) acts as a restrictor on the nominal expression in the matrix clause (Brandt 1990: 40): (8) a.

b.

Maria ist ein Menschi, deri liebt seine Kinder. ‘Maria is a person that loves her children.’ #Maria ist ein Mensch. Der liebt seine Kinder. ‘Maria is a person. That one loves his children.’

The intended interpretation in (8b) is something like ‘Maria is a person,

70

Chapter Four

a human being. As a human being she loves her children’. This interpretation fails, however, because the indefinite NP in predicate position does not provide a discourse marker and hence, the pronoun der does not have an antecedent to which it can refer (we come back to this point in Section 5.2). The different reading of (8a) compared to (8b) shows that iV2 clauses are not computed in isolation. Instead, the matrix clause and the iV2 structure form a discourse unit. Gärtner suggests that the mediating head coordinating the matrix clause with the iV2 clause hosts the feature [REL]. The head ʌREL is responsible for switching the feature specification on the weak demonstrative from [+demonstrative] to [+relative] when iV2 is interpreted.6 As a result, although iV2 clauses are main clauses, they are interpreted as restricting the indefinite pronoun. Hence, the pragmatic oddness in (8b) is not present in (8a). Other proposals minimally differ from the one in (7) with respect to the nature of the mediating head. Den Dikken (2005) for instance proposes a topic head instead of ʌREL. Common to all these accounts is that iV2 structures belong to the Discourse Grammar in the sense of Williams (1977), whereas verb-final RCs are treated as real instances of Sentence Grammar subordination. In Section 5 we return to this topic.

3. iV2 Structures in Acquisition As mentioned in the Introduction a vast body of research has systematically investigated the development of verb placement and more generally of word order in German subordinate clauses. Two main findings have emerged from these studies. First, it has been shown that main clauses are acquired before subordinate clauses, a finding that is uncontroversial also cross-linguistically (Clahsen 1982, 1990, Rothweiler 1993, among many others). Second, children learn verb placement rules quite early (Clahsen 1990, Weissenborn 1990, Clahsen et al. 1992, Rothweiler 1993, Müller and Penner 1996). According to Clahsen (1990), children correctly place the finite verb in second position in main clauses around age 3 (see also Weissenborn 1990). Rothweiler (1993) observed that German-speaking children systematically place the finite verb in clause final position as soon as they produce subordinate clauses. Based on an analysis of about 800 embedded clauses produced by seven monolingual German children, Rothweiler (1993) found only twelve

 6

As will be discussed in detail in Section 5, a classical E-type analysis of the pronoun (Kamp 1981, Heim 1982) cannot capture the observed facts.

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

71

clauses in which the verb does not occupy the verb-final position, nine of them with V2 after weil ‘because’ as in (9), and one with V2 in a RC as in (10). Note that both structures are grammatical in German and hence do not constitute a deviant pattern. (9) Weil da is kein gesich. because there is no face ‘Because there is no face’ (XI) (10) Es

(Age 5;06)

gibt

Menschen die werfen einfach dreck people:ACC PRON:NOM throw simply dirt ausm me aufm Fenste. out-of-the on-the window ‘There are people who simply throw garbage out of the window’ (XI) (Age 5;06) EXPL there-is

The example in (10) can be taken as an instance of iV2 structure since all restrictions (see Section 2) for its licensing are met (e.g., presentative/ existential predicate, indefinite antecedent, d-pronoun, sentence-final position). On the basis of the results in Rothweiler (1993), we can conclude that iV2 structures as in (10) are produced less frequently than verb-final RCs in child language. Interestingly, the findings reported by Brandt et al. (2008) suggest a different acquisition pattern. Based on a much larger database than the one in Rothweiler (1993), Brandt et al. (2008) report that young Germanspeaking children regularly produce verb-second RCs in their spontaneous speech. In fact, the majority of the first relative clauses observed are argued to exhibit V2, i.e. structures with the verb in second position. Analyzing data from one German-speaking child called Leo (available on CHILDES), Brandt (2004) and Diessel and Tomasello (2005) report that iV2 structures are especially frequent in the early speech samples. Up to the age of 2;5, 70% of Leo’s RCs exhibit the finite verb in second position, 22% show an ambiguous word order, and only 8% occur with the finite verb in final position. In the subsequent development, the proportions change radically. At the age of 5;0, 68% of Leo’s RCs show verb-final, 27% show verb-second, and 5% show an ambiguous verb order. Brandt (2004), looking at data from another German-speaking child called Simone (also available on CHILDES), observes that iV2 clauses are predominant over verb-final RCs until 4;0, the age at which the recordings of Simone ended. This suggests that verb-final relatives emerge after age 4;0. Taking these findings together, Brandt et al. (2008) conclude that the

72

Chapter Four

first type of RCs produced by German-speaking children shows the verb in second position and that children only later produce relative clauses with verb-final word order. The authors propose that two factors make the V2 structures available to the child early in development and prior to verbfinal relatives: (i) the frequency of V2 in the input, and (ii) their similarity to simple main clauses. As for (i), it is claimed that verb-second clauses are much more frequent in German child directed speech than verb-final subordinate clauses (see Stoll, Abbot-Smith and Lieven, 2009). As for (ii), iV2 structures are similar to main clauses in terms of word order. According to Diessel and Tomasello (2005), children acquire structures in a piecemeal bottom-up fashion, starting with constructions that minimally differ from simple main clauses. In their view iV2 structures play a key role in the development of German RCs. They exhibit properties of both main and subordinate clauses, which may help the child to bridge the gap between simple sentences and complex relative constructions. At first glance, this view provides further support for the syntactic analysis of iV2 clauses proposed in Gärtner (2001a/b). However, a closer look at the structures produced by the children reported in Brandt et al. (2008) calls the analysis as verb second RCs stricto sensu – see Section 2.2 – into question. The utterances produced by the children mainly consist of an isolated DP followed by a V2 clause, as illustrated in (11) (from Brandt et al. 2008: 340). In our view, these structures cannot be analyzed as proper instances of iV2. (11) Ne Scheibe, die kann man auch darunter rollen lassen. a disk PRON:ACC can one:NOM also under-it roll let ‘A disc that you can roll under there’ (Leo 4;6) Of the conditions for licensing iV2 structures (see Section 2.2), the example in (11) does not meet condition (a): Since there is no matrix clause, there is no presentative/existential predicate in the main clause. Furthermore, the majority of the structures considered by the authors as instances of iV2 clauses involve a definite description as in (12) instead of an indefinite NP (from Brandt et al. 2008: 334). Hence, besides condition (a), also condition (b) is not met in examples like (12): (12) Die Biene, die holt ein Mittagessen. the bee PRON:NOM gets a lunch ‘The bee that/she is getting lunch’

(Leo 2;4)

On the basis of this evidence we suggest that structures like (11) and

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

73

(12) should be interpreted as examples of left-dislocation along the lines proposed by Grewendorf (2002) rather than as instances of iV2 clauses. Consequently, it remains unclear whether iV2 structures are indeed the ‘first’ relative clause structures to appear in children’s production.

4. The Experiment To investigate how children treat verb-final RCs and iV2 structures, we developed a picture-supported delayed-repetition task. We aimed at testing typically developing monolingual German-speaking children between the ages of 3 and 5. The results reported in the present paper are based on the data of 23 3-year-olds and 21 adults. Our research questions were: (Q1) Do 3-year-old children prefer V-final RC or iV2 structures? (Q2) Does children’s preference differ from that of adults? Addressing (Q1), children are expected to prefer iV2 structures over Vfinal RCs based on the theoretical account proposed in Gärtner (2001a/b) and the previous acquisition studies (Brandt 2004, Brandt et al. 2008). If iV2 structures are main clauses coordinated via a discourse procedure, and V-final RCs are instances of subordination, a preference for main clause structures in young children is expected. As for research question (Q2), Brandt et al. (2008) state that children’s preference for iV2 structures holds until age 4. Given this, we expect our 3-year-olds to exhibit a clear preference for iV2, whereas adults should not show a preference for iV2 structures anymore.

4.1. Participants In the following, we report results on 23 typically developing monolingual German-speaking children at age 3. All children were recruited in kindergartens in Frankfurt am Main. In addition, 21 adults were tested as control group. The participant details are summarized in Table 1. Participants

Age range

Mean age

SD

Children

23

3;1-3;9 years

43;0 months

3 months

Adults

21

19;2-30;9 years

25;4 years

74 months

Table 1: Description of participants

Chapter Four

74

All parents gave written consent for their children’s participation in the study. A parental questionnaire ensured that none of the participants had signs of language impairment, language delay, or hearing problems. In addition, all children were administered a standardized language test (SETK 3-5, Grimm 2001), on which they performed within age-appropriate norms.

4.2. Design and Material We developed a picture-supported delayed-repetition task (see Lust, Flynn and Foley 1996), which consists of three parts: listening to the prerecorded stimulus, pointing to a matching picture, and repeating the heard sentence. The pointing task was implemented to test children’s comprehension of the pre-recorded stimulus. In addition, it served to reach a more than 3s delay between the stimulus presentation and its repetition. According to McDade, Simpson and Lamb (1982), this delay ensures that participants repeat only those sentences they comprehend. Furthermore, pictures made the task suitable for young children. The experiment consists of 24 test items and 24 fillers as well as 6 warm-up items to familiarize the participants with the experimental procedure. Each item was presented with a picture, as exemplified in (13). The main factor varied in the test items was verb placement in the RC. For each of the 24 test items, a verb-final RC as in (13a) and an iV2 version as in (13b) was constructed. Each version was assigned to one of two experimental lists. Every participant was tested on 12 test items inviting the repetition of a verb-final RC, and 12 test items inviting the repetition of an iV2 structure. (13) Example test item Hier gibt es einen Mann, here there-is EXPL a:ACC man a. der

ein gefährliches Krokodil a dangerous crocodile eingefangen hat caught has

V-final

PRON:NOM

b. der

hat ein gefährliches Krokodil iV2 has a dangerous crocodile eingefangen caught ‘Look, there is a man who caught a dangerous crocodile.’ PRON:NOM

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

75

 Figure 1: Picture paired to test item in (13)

Since many studies found children to have difficulty with object relatives (for German e.g., Diessel and Tomasello 2005, Brandt et al. 2008, Adani et al. 2012), subject RCs were used, i.e., the pronoun in both the RC and the iV2 structure was the syntactic subject of the clause. In order to rule out the possibility of interpreting the test items as pseudo-relative clauses (Grillo and Costa 2014), we used different tenses for the predicate in the main clause (present tense) and in the RC and in the iV2 clause (either perfect tense, as in (13) or future modal tense). The presence of a finite auxiliary form (e.g., hat ‘has’ or will ‘wants to’) helped to determine the position of the verb in the participants’ responses. In 6 of the RC and 6 of the iV2 test items the predicate in the clause was in the perfect tense, e.g., hat eingefangen ‘has caught’ in (13) with the picture showing the result of the performed action (Figure 1). In the other 6 RCs and 6 iV2 items, the predicate exhibited the analytic future form (e.g., fotografieren will ‘wants to photograph’). All test items were construed in accordance with the specific conditions for licensing iV2-structures (see Gärtner 2001a/b; Section 2.2). Hence, we tested sentences with presentative or existential predicates in the main clause. In line with the results of Weinert (2012), we included the

76

Chapter Four

following predicates: gibt es ‘there is’; hier ist, da ist ‘here is’; kennen ‘to know’; sehen ‘to see’, which are reported to be the most frequent ones in adults’ speech. All head nouns in the main clause were indefinite nominals introduced by the indefinite article ein/eine ‘a’ or by the quantifier zwei ‘two’ (see Sanfelici, Schulz and Trabandt, in prep. for more details on the experimental set up). In order to meet the prosodic constraints exhibited by iV2 structures (Gärtner 2001a/b), prerecorded stimuli with an integrated prosodic contour were used. The picture setting ensured that both RCs and iV2 structures had a restrictive reading only. In all pictures, two candidates were depicted as possible referents of the head noun. In (13) for instance, two men are present that differ with respect to their paired objects: a crocodile or a kangaroo. In addition, two possible candidates for the object DP are depicted, i.e. two crocodiles in (13), which differ in their paired agents, a man and a woman. An additional set of 24 filler sentences was created that was balanced between V2 and V-final to test whether children exhibited general difficulty with either verb second or verb final structures (see Sanfelici, Schulz and Trabandt, in prep. for more details).

4.3. Procedure Participants were tested individually by an experimenter in a quiet room in their kindergarten. The test session started with a familiarization phase. During this familiarization phase, the child was introduced to the “Findebuch” ‘finding book’, a book with pictures of animals and people. The child was asked to list the names of the depicted animals and people to introduce the lexical items used in the task. The experimenter and the child sat on one side of a table, and the picture book and a computer were placed in front of the child. The experiment was video-taped and audiorecorded for further analysis. After this introductory session, the experimenter outlined the instructions as follows: (14) Hör gut zu, ich sag dir was, du suchst das richtige Bild und zeigst es mir. Und dann sagst DU den Satz genauso noch mal. ‘Listen, I will tell you something, and then you look for the matching picture and show it to me. Then YOU will repeat the sentence once more exactly as it was.’ In the picture book, pictures and empty pages were alternated. After the presentation of the prerecorded stimulus, the experimenter turned the empty page and showed the corresponding picture to the child. Then, the

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

77

child pointed to the matching scene and then she repeated the sentence. Three warm-up items were presented first. If the child did not understand the instructions or the order of the tasks, the experimenter repeated the three warm-up items. After this, the testing session started. If the child was hesitant or did not follow the requested order, the experimenter repeated the respective item once. No response-contingent feedback was given to the children. Children were tested in two sessions, comprising the experiment and the standardized language test, which lasted between 25 and 40 minutes each.

4.4. Coding Scheme for RC and iV2 Test Items Repetitions of the test items were analyzed according to a two-step coding scheme: the first step focused on the type of structure produced by the participant, and the second step involved coding the verb position in the repeated utterance. The first step was included since children produced also structures other than the intended RCs and iV2 clauses respectively, as expected in this type of task with 3-year-olds. Adults showed high task conformity in their repetitions. The first coding contained several categories. Here, we report on the two categories that are relevant for our discussion: RC-STRUCTURE and MAIN CLAUSES (see Sanfelici, Schulz and Trabandt, in prep., for further details). (a) RC-STRUCTURE We coded both verb-final RCs and iV2 structures as RC-STRUCTURE when the repetition contained the main clause, the head noun, the dpronoun, and the following sentence. This code was assigned independently of the verb position. Hence, we coded a repetition as RCSTRUCTURE both if the child correctly repeated the stimulus, and if the verb placement was changed. Examples for a verb-final and an iV2 repetition are given in (13a-b), here repeated as (15a) and (15b). (15) Hier gibt es einen Mann, here there-is EXPL a:ACC man a. der ein gefährliches Krokodil PRON:NOM a dangerous crocodile eingefangen hat caught has b. der PRON:NOM

hat ein gefährliches Krokodil has a dangerous crocodile

V-final

iV2

78

Chapter Four

eingefangen caught ‘Look, there is a man who caught a dangerous crocodile.’ .

(b) MAIN CLAUSE We coded as MAIN CLAUSE syntactically complete main clauses with the finite verb in second position. (16) Ein Mann hat ein gefährliches Krokodil eingefangen ‘A man caught a dangerous crocodile.’ The second coding step focused on the verb placement. It was applied to all repetitions coded as RC-STRUCTURE. If the participant repeated the verb in the same position as in the test item, we coded the repetition as ‘correct’. If the repeated utterance showed the finite verb in a position different from that in the test item, the repetition was coded as ‘V-change’. Thus, the code ‘V-change’ was used if the child repeated a verb-final RC test item as an iV2 clause and vice versa if an iV2 test item was repeated as a verb-final RC. In addition, the label ‘unanalyzable’ was used for cases where the finite verb was missing as in (17a), or when it was doubled as in (17b). (17) Hier gibt es einen Mann, here there-is EXPL a:ACC man a. der ein gefährliches Krokodil eingefangen PRON:NOM a dangerous crocodile caught b. der hat ein gefährliches Krokodil eingefangen hat PRON:NOM has a dangerous crocodile caught has ‘Look, there is a man who caught a dangerous crocodile.’

4.5. Results In the following section we first give an overview of children’s types of repetitions regardless of the actual verb placement. Then, we present our results on the verb placement in the repetitions of type RCSTRUCTURE. Performance on the filler items reached ceiling (see Sanfelici, Schulz and Trabandt, in prep. for details).

On German “V2 Relative Clauses”: Linguistic Theory Meets Acquisition

79

4.5.1. Children’s Repetitions Table 2 characterizes children’s repetitions in the two test conditions, verb-final RCs and iV2 clauses.7 Statistical analyses revealed that the two conditions significantly differed in the rates of RC-STRUCTURES (Wilcoxon related samples, Z=2.48, p=.013), but not in the rates of MAIN CLAUSE (Wilcoxon related samples, Z=-.925, p=.355).

RC-STRUCTURE

MAIN CLAUSE TOTAL OF ANALYZED REPETITIONS

V-final 115 56.4% (35.7) 89 43.6% (27.8)

iV2 97 50.3% (31.0) 96 49.7% (25.3)

204 100%

193 100%

Table 2: Children’s repetitions of the test items as RC or main clauses in the two conditions: Raw numbers, percentages, and (SD of percentages) 4.5.2. Results in the RC-STRUCTURE Responses In order to address the two research questions participants’ RCSTRUCTURE responses were analyzed according to how often verb placement of the test item was repeated correctly and how often it was changed (V2 > V-final and V-final > V2). Seven ‘unanalyzable’ responses had to be excluded from the analysis. Among these were repetitions with structures lacking a finite predicate, as shown in (17a) and two instances of the finite predicate as depicted in (17b). Results are reported only on RC-STRUCTURE repetitions containing the finite predicate (n = 111 in V-final condition, n = 94 in iV2 condition). First, we present results on the V-final condition and then on the iV2 condition. Figure 2 summarizes the results in the V-final condition for children and adults.

 7

As stated in Section 4.4, we report on only two types of responses here, RCstructures and Main Clauses, which add up to 100% in Table 2.

Chapter Four

80

100

15

80 60 85

40

100

V-Ch hange Correect

20 0 Age 3

Adults

Figure 2: V-fi final condition: Percentage of correct c verb plaacement and verrb change

Both chiildren at age 3 and adults were w accurate iin their repetittion of Vfinal relativve clauses, altthough the nu umber of corr rrect repetition ns differs significantlyy between chhildren and adults a (U=48 3.0, Z=6.03, p