Acquisition of syntactic structures.
Relationship between production and comprehension in child language.
The effects of processing and working-memory limitations in language development.
Word segmentation by monolingual and bilingual infants.
Cross-linguistic differences in language acquisition.
Romance languages (particularly Catalan and Spanish).
Current and Past Projects
Sluicing in Early Grammars: Evidence of Syntactic Intervention (with Nina Hyams)
Part I. Sluicing: Identity Condition and Intervention
This study investigates the acquisition of sluicing, i.e. sentence ellipsis, in English to assess if three- to six-year-old children are able to comprehend sluiced sentences in an adult-like manner. We used a wh-question task in which children were asked to answer questions about an image, e.g. ‘I can see the girl is chasing someone, can you see who <the girl is chasing _>?’. We find three- to six-year olds are fully capable of reconstructing the antecedent of ellipsis and respect the identity condition. Interestingly, children performed significantly worse in the object-extracted sluices (e.g. 'I can see Ben is spraying someone, can you see who <Ben is spraying __>') than the subject-extracted sluices (e.g. 'I can see someone is spraying Ben, can you see who <__ is spraying Ben>'). We hypothesize this is due to intervention effects. This study thus contributes to theories on the acquisition of ellipsis and also to the theoretical debate about the syntactic status of the elided material. Parts of this study will be presented at GALA 13, in Palma de Mallorca, Spain, in September 7-9, 2017, and BUCLD 42, in Boston, MA, in November 3-5, 2017.
Part II. Sluicing and Relative Clauses: Intervention, Frequency, and Animacy effects
This study tests children’s comprehension of relative clauses (e.g., 'Point to the boy that the girl is pushing _') and sluiced wh-questions (e.g., 'The girl is pushing someone, can you see who <she is pushing_>?'). We find that children do better with subject than with object RCs, arguably due to intervention, and also with subject- as opposed to object-extracted sluices, consistent with analyses that posit structure at the ellipsis site. Moreover, children do better when subject and object mismatch in animacy (e.g., the boy / the train), but only with object-extracted sentences. We conclude frequency plays a role, but we cannot do without structure-based approaches to intervention and submit that [animacy] should be included in the computation of intervention. Parts of this study will be presented at NELS 49 at Cornell University, NY, October 5-7, 2018, BUCLD 43 in Boston, November 2-4, 2018, and LSA 2019 in New York, January 3-6, 2019.
Dissertation: Intervention effects in the acquisition of Raising and Control
In my dissertation I investigate two constructions that are not fully mastered until school age: Subject-to Subject Raising verbs such as ‘seem’ (e.g. John seems (to Mary) to be nice) and Subject Control verbs such as ‘promise’ (e.g. John promised (Mary) to be nice). Despite the extensive literature on the acquisition of these constructions in English, there is no consensus as to why they are late to develop. One prominent explanation holds that the delay is due to the presence of an intervening argument (the experiencer/benefactive Mary). Crucially, in Spanish an intervener is possible with ‘prometer’ (as with ‘promise’) but not with the functional verb ‘parecer’ (in contrast to ‘seem’). The results from this study show: (i) English-speaking children have difficulties with raising 'seem' until the age of 6. Contrastively, Spanish-speaking children perform well with raising 'parecer' by the age of 4. This behavioral difference is accounted for if we assume a more complex syntactic derivation for English, namely, the presence of a structurally intervening experiencer argument that forces a circumventing operation; (ii) Both English- and Spanish-speaking children have difficulties with 'promise'/'prometer' until at least 6-years-old, consistent with the idea that intervention is at play. However, (iii) we do not find a correlation in performance between raising 'seem' and control 'promise', which suggests these are not derived similarly. Finally, (iv) we find evidence for both grammatical and processing factors affecting children's performance on intervening constructions. This study has been supported by the National Science Foundation Doctoral Dissertation Research Grant (BCS-1451589). Different parts of this study have been or will be presented at LSA 2016 in Washington D.C. on January 2016; GALANA 7, in Urbana-Champaign, IL, on September 8, 2016, BUCLD 41, in Boston University, MA, on November 6, 2016, and LSA 2017 in Austin, TX, on January 6, 2017.
A cross-linguistic investigation of word segmentation (with Megha Sundara)
Part I. Monolingual English infants
Infants’ ability to segment words from fluent speech is affected by their language experience. We test four hypotheses: 1. Infants can segment words in a rhythmically-similar language (e.g. stress-timed/stress-timed) but not in a rhythmically-different one (e.g. stress-timed/syllable-timed), 2. Infants can segment words that share the predominant stress pattern of their native language (trochee/trochee), but not those that do not (e.g. trochee/iamb), 3. Infants can segment in any language by using statistical cues, 4. Infant can use stress-based cues to segment words in any language as long as it has lexical stress. In order to disentangle these hypotheses we investigated the segmentation abilities of 8-month-old monolingual English infants with Spanish trochees and iambs (Spanish has lexical stress), and French iambs (French has no lexical stress) using the classic HPP procedure at the UCLA Language Lab. We found that that monolingual English-learning infants can segment Spanish trochees and iambs given an extended familiarization time (60s but not 45s). However, they fail to segment French iambs given the same familiarization time. This shows that infants can segment (i) in a rhythmically different language (Spanish) and (ii) words that have an unfamiliar stress pattern (iambs). However, they fail to segment words in a language with no lexical stress, e.g., French. This study was presented at WILD 2013, at the Basque Center on Cognition, Brain and Language (BCBL), in San Sebastian, Spain, and has been recently published in Cognition.
Part II. Monolingual Spanish and English-Spanish Bilingual infants
The focus of this study is whether monolingual Spanish and bilingual English-Spanish infants can segment disyllabic words in Spanish and/or English. Both Spanish and English have lexical stress. In contrast to English, where 90% of disyllabic words start with a stressed syllable, stress in Spanish is much more variable. Using the Head-turn Preference Procedure, we found that while monolingual Spanish 8-month-olds fail to segment Spanish CVC.CV trochees and CV.CVC iambs, bilingual 8-mo-olds exposed to both English and Spanish successfully segment Spanish CVC.CV trochees and CV.CVC iambs, as well as English iambs, even though monolingual English 8-mo-olds fail at that same task. Thus, while word-segmentation strategies are language-specific, they may be transferred between languages, even across rhythmically-different ones. We are currently investigating whether they are segmenting the full word or only the stressed syllable (or both). Part of the results from the bilingual study were presented at WILD 2015, at Stockholm University, Sweden, Bilingualism Matters 2017 at UCR, and at the International Symposium on Bilingualism 12 (ISB12), at the University of Alberta, Edmonton, Canada.
Object clitic omission in child Spanish
This study explores the widely documented difficulty children have with object clitics in the acquisition of Romance languages. It reports on two experiments: a production task and a comprehension task. Results from the elicitation task confirm that object omission occurs at nonnegligible rates in 2- and 3-year-olds. Findings from the sentence-picture matching task show that children do not sanction a grammar with referential null objects, as suggested by previous research, and that children do not always assign a transitive interpretation to clitic constructions. Further analysis reveals that both the frequency of object omissions in production as well as the results in the clitic conditions of the receptive task are strongly negatively correlated with an independent measure of verbal working memory (nonword repetition task), consistent with the hypothesis that object clitic omission is affected by linguistic processing and short-term memory limitations
Part of this study was presented at BUCLD 38 and has been published in its entirety in Language Acquisition: A Journal of Developmental Linguistics, you can download a copy here.
The acquisition of Spanish indefinite and numeral 'un' (with Nina Hyams)
Previous studies have argued that children draw on the meanings of known scale-mates when interpreting unknown numerals. Under these accounts, the child first assumes a weak ‘at least’ meaning of one (one = a). Only once they acquire the weak meaning of two (‘at least two’), they are able to strengthen the meaning of one to ‘exactly one’ by computing a contrast-based scalar implicature. In the Romance languages, however, the phonological forms for the numeral one and the singular indefinite article a are the same, i.e. un.
We tested Spanish-speaking children aged 3-5 on their interpretation of un (a/one) and other numerals and quantifiers in a truth-value judgment task and a sentence-picture matching task. Children performed significantly better with dos 'two' and tres 'three' than with un. Unlike English-speaking children, Spanish-speaking children assign ‘un’ an ‘at-least’ reading, as they do for the quantifier ‘algunos’ (some), even after learning the ‘exact’ meaning of higher numerals, consistent with claims that ‘un’ is only an indefinite and not a numeral (a.o. Kayne, 2009). Additionally, we found that Spanish-speaking children are relatively delayed in the acquisition of the exact meaning of two and higher numerals as compared to English-speaking children. These results provide further evidence for the hypothesis that the acquisition of numerical concepts can be accelerated or delayed by the morphosyntactic use of number and number words, and that the (lack of) exact meaning for numerical concepts is at least partially bootstrapped from the language system. The results of this study were presented in GALANA 2015, at the University of Maryland, MD. You can download the pre-printed version of the Proceedings paper here.
Antilogophoricity in clitic clusters (with Isabelle Charnavel)
In some Romance languages, including Spanish and French, there is an interesting asymmetry concerning the behavior of isolated clitics and clitic clusters with respect to coreference. In the Spanish example Anai cree que se lai recomanderán para el ascenso ‘Annai thinks that they will recommend heri to him for the promotion’, the accusative clitic la ‘her’ in the embedded clause cannot corefer with Anna when a dative clitic se ‘to him’, co-occurs in the cluster. The only previous account of this constraint (Bhatt and Šimík, 2009) attributes this to a binding restriction. Based on an online grammaticality judgment task disentangling binding and logophoricity, I found that the generalization capturing the distribution of clitics clusters in Spanish is the following: an accusative clitic cannot be clustered with a dative clitic if the accusative clitic refers to a logophoric center and is read de se. Subsequent collaborative work with Isabelle Charnavel, at Harvard University, has been able to replicate the findings in French. We derive this antilogophoricity effect from perspective conflicts, which we represent as intervention effects in the presence of a single logophoric operator in the relevant domain. This analysis furthermore provides a semantic motivation for intervention effects that have been postulated for the Person-Case Constraint (PCC), which we hypothesize also derives from perspective conflicts. Parts of this study were presented at WCCFL 32, GLOW 37, and LSRL 44. You may download the Linguistic Review paper here.