Processing Grammatical Features by Italian Children

Fabrizio Arosio, Flavia Adani and Maria Teresa Guasti
Università degli Studi di Milano-Bicocca

In this study we investigated how Italian children process relative clauses and how developmental data shed light on processing theories. We tested 92 Italian monolingual children aged 5, 7, 9 and 11 divided into 4 age groups in a picture selection task with 18 unambiguous SUBJECT relatives (SR), 18 unambiguous NUMBER marked OBJECT relatives (OR), 18 unambiguous POSITION marked OBJECT relatives. Our results show that ORs are harder than SRs at any age. On NUMBER marked ORs there is a difference between 5-year-olds on the one hand and 9-, 11-year-olds and adults on the other. There is also a significant difference between 7-year-olds vs. 11-year-olds and adults. For POSITION marked RCs there is a difference between 5-year olds and adults. We considered the predictions of a number of processing theories and we observed that The Minimal Chain Principle (De Vincenzi, 1991) accounts for children data.

The Early Processing of Number Agreement in the DP: Evidence from the Acquisition of Brazilian Portuguese

Letícia Sicuro Corrêa, Marina Augusto and José Ferrari-Neto
Pontifícia Universidade Católica do Rio de Janeiro/LAPAL

The processing of number agreement in the acquisition of Brazilian Portuguese (BP) is investigated. BP exhibits two co-existing number systems and the crucial information concerning plurality in both systems is morphologically expressed in the determiner. It is argued that in order for children to identitify this information, it is necessary that they perceive morphophonological variation within a closed class and that they parse the DP as a phrase in which agreement relations have been established. A picture-identification task with pseudo-nouns and invented pictures was used, which aimed at verifying the extent to which two year olds identify a target multiple-token picture among pictures of singletons solely on the basis of the morphology of number. The results suggest that children are sensitive to the presence of a number morpheme, that they distinguish the morphological expression of number in BP from other possible ones and that they treat in a similar way the two co-existing BP number systems. These results are compatible with a theory of language acquisition in which an independent computational system functions as a language acquisition device as soon as its operation is bootstrapped by information in the phonetic interface.

Is Language Processing Identical in Monolinguals and Early, Balanced Bilinguals?

Cassandra D. Foursha, Jennifer B. Austin and Gretchen A. Van de Walle
Rutgers University

This study investigated grammatical competence in early, balanced bilingual adults. We examined both the accuracy of our participants' grammaticality judgments in English, and also their reaction time in judging grammaticality to determine if processing speed is affected in early bilinguals. In addition, we examined cross-linguistic interference by comparing performance on two types of sentences. In one type, Spanish and English grammatical rules converged; that is, the sentence would be judged either grammatical or ungrammatical if spoken in either language. In the other sentence type, the grammars of the two languages conflicted; that is, the sentence would be considered correct if spoken in one language, but incorrect if spoken in the other, which may create cross-linguistic interference. We found that although they were just as accurate as monolinguals, early, balanced bilinguals were slower across the board than monolinguals in making grammaticality judgments, pointing to a global language processing difference. However, this difference was not related to sentence type. The bilingual and monolingual groups were indistinguishable in their patterns of performance on conflicting and converging sentences. Thus, we found no evidence of syntactic interference in our bilingual subjects' grammaticality judgments. Possible explanations for the lack of cross-linguistic influence are discussed.

Linking Object Names and Object Categories: Words (But Not Tones) Facilitate Object Categorization in 6- and 12-Month-Olds

Anne L. Fulkerson 1, Sandra R. Waxman 2 and Jennifer M. Seymour 3
1 Department of Psychology and Office of Institutional Research, The University of Toledo, 2 Department of Psychology, Northwestern University, 3 Department of Psychology, The University of Toledo

Recent studies have revealed a precise link between object naming and categorization in infancy. In the present study, we pursued the genesis of this early link, examining the influence of either words or tones on object categorization at 6 and 12 months. We find that for infants at both ages, words -- but not tones -- facilitate object categorization. Infants hearing a novel noun for a set of distinct objects successfully formed object categories. Infants hearing a sequence of pure tones for the same set of objects did not. These results reveal that word-learning and conceptual organization are linked from the start. This link, which specifically supports word-learning and conceptual organization, is an important building block in early language acquisition. Results are interpreted within a developmental theory of lexical acquisition.

The Acquisition of Japanese Focus Particles: dake (only) and mo (also)

Kazumi Matsuoka 1, Nobuhiro Miyoshi 2, Koji Hoshi 1, Masanobu Ueda 3, Izumi Yabu 3 and Miki Hirata 3
1 Keio University, 2 Asahikawa Medical College, 3 Hokkaido University

Focus items mo and dake in Japanese show different syntactic behavior, which led Japanese researchers to classify them into two different categories: K-particle (mo) and F-particle (dake). Nevertheless, our research showed that young Japanese-speaking children go through a similar developmental path as they learn the focus items of different types. The path does not seem to be completely uniform, though: when children interpret sentences which include a focus item, one group of children assigned the subject-oriented interpretation, while the other exhibited the object-oriented interpretation. A total of 120 children from Sapporo and Osaka (4;7-6;10, mean: 5;10) participated in our study. The Truth-Value Judgment task (Crain and Thornton 1998) was conducted. 48% of the dake subjects (mean:5;11) gave adult-like responses. We found a larger number of children who gave the subject-oriented responses (30%, mean:5;10), as opposed to the ones who gave the object-oriented responses (2%, age 5;9). Our results, combined with the observations in Crain et al. (1992) and Huttner et al. (2004), indicate that both subject-oriented and object-oriented interpretations of focus items are universally allowed options in language development.

No Ambiguity About It: Korean Learners of Japanese Have a Clear Attachment Preference

Mari Miyao and Akira Omaki
University of Hawaii and University of Maryland

This study investigates the transfer/learning of processing strategies by L2ers, using relative clause attachment ambiguity (1):

(1) Someone shot the servant of the actress who was on the balcony.
High attachment (HA) analysis: "The servant of the actress was on the balcony"
Low attachment (LA) analysis: "The actress was on the balcony"

Based on the L1 psycholinguistic findings (e.g., Cuetos & Mitchell, 1988) that there are cross-linguistic differences in terms of which ambiguity resolution strategies are preferred (HA or LA), we conducted two experiments (off-line and on-line) to test whether Korean-Japanese L2ers and Japanese natives use similar processing strategies in parsing sentences like (1). The results indicated that Korean L2ers prefer HA in off-line processing and LA in on-line processing, whereas Japanese natives prefer HA in both off-line and on-line processing, just as Lee & Kweon's (2004) Korean natives did. Taken together, we suggest that (a) our L2ers show transfer/learning of processing strategies in off-line processing, but (b) their LA preference in on-line processing results from the limited capacity of the developing L2 parser, which prefers a strategy that imposes less processing burden.

The Genitive of Negation Construction in Russian-English Bilinguals

Nadya Modyanova
Massachusetts Institute of Technology

We examine what properties of L1 are attrited or not learned when a child primarily hears and uses L2. In the Genitive of Negation construction (GN), genitive case marks nonspecific direct objects (DO) of negated verbs, being obligatory with bleached unaccusatives (e.g. be) and optional but preferred with transitives. In contrast, Russian specific DO behave like English objects and receive nominative case for unaccusatives and accusative case for transitives. We investigated GN via elicited production in eleven Russian-English bilinguals (5;3-10;6 years). Comparing the results against monolingual GN data (Babyonyshev et al. 2001), we show that bilinguals use genitive case more than younger monolinguals (3;0-6;6 years) for bleached unaccusative objects (73% vs. 47%), but much less for transitive nonspecific objects (36% vs. 73%). Following Tsimpli and colleagues (2002), the Russian encoding of the interpretable nonspecificity feature in GN seems to become optionally unspecified as a result of near-fluency in English.

Young Children's Use of Unaccusative Intransitives in Novel Verb Experiments

Kaya Ono and Nancy Budwig
Clark University

This study investigates whether preschool children produced unaccusative intransitives using experimental procedures that simulated everyday verb learning. Thirteen 30- to 42-month-olds and their caretakers participated in two 45-minute sessions within a one-week span. Four novel verbs were modeled in transitive constructions while participants engaged in the novel verb actions. Most of the children were able to produce novel verbs, and almost half of them were able to produce an unaccusative intransitive construction though it was not modeled. On average, children needed 27 models by adults before their first use of a novel verb in any construction. However, children needed an average of 13 models before their first use when they produced an unaccusative intransitive. The findings indicate that children learn a new verb quicker when they could go beyond input and apply unfamiliar constructions to a novel verb. The results confirm the entrenchment hypothesis that children are more inclined to overgeneralize unfamiliar verbs compared to familiar ones. Discussion focuses on implications for Tomasello's verb island hypothesis.

Learning Allophones from the Input

Christine Shea 1 and Suzanne Curtin 2
1 University of Calgary, Department of Linguistics, 2 University of Calgary, Department of Linguistics and Psychology

Learning allophonic distributions requires familiarization with cues that identify these alternations. We exposed Spanish and English Native Speakers (ENS, SNS) to Arabic allophonic variations resulting from the spread of secondary pharyngealized articulation of an emphatic consonant giving rise to conditioned alternations between [a]/[Ê]. These vowels correspond to different distributions in the native languages of the participants: ENS - two L1 phonemic categories; SNS-one phonemic L1 category. CV sequences of emphatic/nonemphatic coronal consonants with three naturally occurring Arabic vowels were presented in an ABX task. Subsequently, participants listened to one of two continua with eight CV tokens each: one varied in emphatic [d(] and the second varied the F2 of the low front vowel. Results suggest: 1) sensitization to contextual cues signaling allophonic variation in a foreign language; 2) sensitization can affect L1 phonemic boundary perception; 3) differential NS group results indicate L1 phonemic status of target allophone affects training sensitivity.

Word-Final Consonant and Cluster Acquisition in Indian English(es)

Caroline R. Wiltshire
University of Florida

Learners of English in India speak first languages (L1s) with varied phonotactics. Variations in Indian English reveal the importance of three factors in L2 acquisition: transfer, markedness, and input frequency. Data was recorded from speakers of 5 L1s, which allow no word-final Cs (Angami), a single sonorant or voiceless obstruent (Mizo, Ao), or voiced or voiceless obstruents or clusters (Gujarati, Hindi). Transcription of word-final consonants and clusters shows more frequent: 1) devoicing for speakers of L1s lacking voiced codas, attributable to transfer and markedness; 2) deletion for speakers of L1s lacking clusters, particularly in clusters violating sonority sequencing markedness. However, sonority sequencing cannot explain why deletions in /s/-stop clusters outnumber deletions in stop-/s/, if stop-/s/ clusters are more marked. I use the Graduate Learning Algorithm to show that starting from different L1s results in different grammatical stages in L2 acquisition and to illustrate that input frequency cannot override markedness, suggesting that stop-/s/ clusters are not more marked than /s/-stop in word-final position.

Syntactic Complexity and Productivity: A Study of Early Verbs in L1 Acquisition of Mandarin Chinese

Xiaolu Yang
Tsinghua University

The present study addresses the issue of whether early verbs develop along different paths, thus forming the so-called "verb islands" (Tomasello 1992). We looked at complexity and productivity of early syntax of Mandarin Chinese by drawing longitudinal data from two Mandarin-speaking children. The subjects' multi-word combinations containing verbs in the transcripts before 2 were extracted and analyzed, with a special focus on the occurrence of functional categories such as negators, modals, and focus adverbs, serial-verb constructions and the internal structure of VP. It is found that quite a number of different verbs appeared as arguments of higher predicates: negators and modals consistently appeared pre-verbally. Many different verbs appeared in serial verb constructions and in V-V compounds. It was also common for the two kids to use verbs in two or more patterns during the observed period. Our data suggest that Mandarin-speaking children's syntax before 2 is quite productive and complex, not easily susceptible to an item-based account, as proposed in Tomasello (1992, 2000, 2003).