Browse Abstracts 2019

E. Baldimtsi, E. Peristeri, I. Tsimpli, S. Durrlemann

This study investigated whether bilingualism boosts Theory of Mind (ToM) and executive functions (EF) in ASD, and the links between ToM and EF in bilingual children. Thirty-four bilingual children with ASD (Mean age: 11; 7) were compared with age- and nonverbal intelligence-matched monolingual children with ASD, along with monolingual and bilingual TD controls. ToM was measured via an online, non-verbal false-belief task, attention-shifting via an online global-local attention task and working memory (WM) via a 2 digits-back task. Bilinguals with or without ASD were more accurate than monolinguals on WM, and ASD bilinguals were better than their monolingual peers both at shifting attention to the global level and ToM. Regression analyses revealed stronger relationships between EF and ToM in ASD bilinguals than ASD monolinguals, although these links did not emerge in TD. Bilingualism improves EF in children, including with ASD, and ASD bilinguals seem to capitalize on EF for ToM.

Are infants sensitive to informant reliability in word learning?

A. Tripp, N. Feldman, W. Idsardi

Infants are sensitive to epistemic trust, but models of infant word learning assume learners trust data from all informants equally. We model the Switch task, an audiovisual habituation procedure wherein infants are exposed to novel objects paired with labels and tested on these familiarized associations. Eighteen-month-olds outperform fourteen-month-olds on this task when the labels are minimal pairs (e.g. “”buk”” and “”puk””), but fourteen-month-olds’ performance is improved by hearing habituation stimuli produced by multiple different talkers. Here we adapt a model to show that the benefit of multiple speakers could reflect aspects of infants’ social cognition. Previous accounts attributed this improvement to the effect of acoustic variability on the infants’ speech representations, whereas our account demonstrates that it can be predicted even when assuming that infants’ phonetic perception is adult-like. Moreover, it suggests that learners may use inferences about epistemic trust during word learning earlier than previously supposed.

Tamil-speaking children do not prefer iconic adverbial sentences over non-iconic ones

L. de Ruiter, V. Priyadharshini, A. Etz, S. Kuppuraj

Five-year-old English- and German-speaking children process adverbial sentences more easily when these are in iconic order, that is, when the order of events in the sentence maps onto the order in the real world (e.g., He drives away fast before he shouts out loudly). It has been suggested that children expect language to map iconically. Using a forced-choice picture selection task, we tested if this iconicity preference is also present in Tamil-speaking five-year-olds. Tamil is typologically different from English and German and strongly prefers subordinate-main orders. For all four adverbials tested (pin ‘after’, mun ‘before’ and (n)al ‘if/because’), Tamil-speaking children’s performance was better with the highly frequent subordinate-main orders, even when the sentences were non-iconic. We conclude that the iconicity preference is not universal, but modulated by language-specific constraints. We also discuss individual differences among children.

Child cross-linguistic influence and adult L1 transfer: same or different?

S. Berends, A. Hulk, J. Schaeffer, P. Sleeman

This study reports the results of an experimental study on the acquisition of the Dutch quantitative pronoun (ER) by English-Dutch bilingual children, and compares these to a study on adult L2 Dutch speakers testing the same phenomenon (Berends, Schaeffer & Sleeman, 2017). A generalized linear mixed-effects logistic regression analysis shows that, despite similar scores on the proficiency task, the bilingual group’s responses contain significantly fewer target responses than the monolingual children’s responses. An error analysis reveals different types of non-target responses in both groups (full NP, ER-omission, doubling, irrelevant), and that ER-omission is significantly higher in the bilingual group. These results provide evidence for the Cross-linguistic Influence Hypothesis in child bilingualism and suggest a similar mechanism in both child bilingual and adult L2 acquisition (Berends et al. 2017): the lack of a quantitative pronoun in English negatively influences the acquisition of Dutch quantitative ER constructions.

The development of pragmatic reasoning from multiple information sources

M. Bohn, M. Tessler, M. Merrick, M. Frank

Language is learned in complex social settings. The messages that speakers convey go beyond the literal meanings of words, and listeners must reconstruct a speaker’s meaning using pragmatic reasoning. During language acquisition, pragmatic reasoning can help children to learn the meaning of novel words. Yet, pragmatic inferences require integrating information about the current utterance with information stored in common ground that is typically built up over time. Here, we tested information integration in a series of experiments with three to five year old children (N = 219). We formalised the integration process using the Rational Speech Act (RSA) model to generate a priori predictions about how information could be integrated. Results show that children flexibly integrate utterance and common ground information, a process which was accurately captured by the RSA model. In ongoing work, we use RSA models to explore how children integrate semantic knowledge with common ground information.

Redundant morphological marking benefits child learners

S. Tal, I. Arnon

Redundancy in morphological marking is found across languages (e.g., agreement, multiple exponence). Its presence is somewhat puzzling given that it adds complexity to the language, and speakers seem to disprefer production of redundant linguistic elements (for example, optional case marking is likely to be omitted when the meaning it encodes is more predictable from context).What could be the functionality of redundant morphological marking? We propose learning as a possible explanation: redundant cues may facilitate learning, making them advantageous in the system as a whole. We test this hypothesis in an artificial language learning study with children, where either word order alone or both word order and case marking serve as cues for thematic assignment. Results show that children learned the redundant language better despite having to learn and produce an additional morpheme. This supports the idea that learning constraints help maintain redundancy in language.

Parental Language, Functional Utterance Type, and Play Context Impact Children’s Usage of an Endangered Ancestral Language

C. Lowry, P. Yuksel, P. Brooks

This study investigates the factors that supported children’s (age 15-48 month; n = 59) use of an ancestral language, Lazuri, which is being supplanted by a societal dominant language, Turkish. Results indicate effects of parental language, functional utterance type, and play context. One main finding is that caregivers’ use of labels and questions spoken in Lazuri and mixed (Lazuri+Turkish) utterances was more effective in promoting children’s use of Lazuri, whereas caregivers’ comments, demands, and invitations promoted Turkish responses. Children’s usage of Lazuri was largely restricted to one-word utterances that were labels and deictic expressions. Hence, when conversational contexts demanded more complex responses, children resorted to the dominant language. Our results illustrate the challenges of ancestral language preservation in contexts where children have greater functional competence in the dominant language, but also suggest that maintenance of the ancestral language can be supported by meaningful interactions that scaffold the child’s functional abilities.

“It’s in your box!” –– Personal pronoun comprehension in children with ASD

H. Clancy, A. He, R. Luyster, S. Arunachalam

Personal pronouns have been argued to be challenging for children with autism spectrum disorder (ASD). Pronoun comprehension includes (a) establishing the right mappings: first-person pronouns refer to the speaker, and second-person pronouns refer to the addressee; and (b) tracking referent shifts––identifying the speaker and addressee in the current round of conversation. The latter may require social-communicative sensitivity, which children ASD may have difficulty with. Here, we used a multi-converser setting to test pronoun understanding given referent shifts. Results show that both children with ASD and typically developing children know the semantic differences between first- and second-person pronouns, and they know that pronouns’ referents shift by discourse role. However, both groups are biased to interpret second-person pronouns as referring to themselves. TD children show similar patterns to the ASD group, despite better performance overall.

Multiword units predict children’s non-inversion errors in wh-question formation

S. McCauley, C. Bannard, A. Theakston, M. Davis, T. Cameron-Faulkner, B. Ambridge

Subject-auxiliary inversion in interrogatives has been a topic of great interest in language acquisition research, and has often been held up as evidence for the structure-dependence of grammar. Usage-based and nativist approaches posit different representations and processes underlying children’s question formation and therefore predict different causes for these errors. Here, we explore the question of whether input statistics predict children’s spontaneous non-inversion errors with wh- questions. In contrast to previous studies, we look at properties of the non-inverted, errorful forms of questions. Through a series of corpus analyses, we show that the frequency of uninverted subsequences (e.g., “she is going” in “what she is going to do?*”) is a good predictor of children’s errors, consistent with recent evidence for multiword units in children’s comprehension and production. This finding has implications for the types of mental representations and cognitive processes researchers ascribe to children acquiring a first language.

Joan Maling (National Science Foundation), Brett Miller (National Institutes of Health)

N. Vasilyeva, M. Ellwood-Lowe, M. Srinivasan

Children readily learn from language. We explore whether children can learn about the social world in indirect ways, from what a speaker could have said—but chose not to say—about a social group. For example, if a teacher has just observed a classroom of boys and girls and says “The boys are good at math,” they may express that the girls are worse than the boys at math, because they could have instead said “The boys and the girls are good at math”. Such inferences critically depend on the situational context: no inference about the girls would arise if the teacher had only observed a classroom of boys before making their statement. We demonstrate that both preschoolers (N=46) and adults (N=102) are capable of learning about the social world by context-sensitive “reading between the lines”, with preschoolers showing a tendency to draw stronger inferences than adults.

Preverbal infants’ sensitivity to grammatical dependencies

M. Babineau, A. Christophe

We investigate if preverbal infants have encoded the specific co-occurrences between common content words and function words. Experiment 1 used both common nouns and verbs, while Experiment 2 used only common verbs. Using the Visual Fixation procedure, French-learning 11-month-olds were presented with two types of lists containing either grammatical or ungrammatical phrases. The grammatical lists contained the correct type of function word for each content word (e.g. tu manges ‘you eat’, des biberons ‘some bottles’), whereas the ungrammatical lists contained phrases with incorrect function words (e.g. des manges ‘some eat’, tu biberons ‘you bottle’). In both experiments, we obtained significantly longer looking times for the trials presenting grammatical phrases over those presenting ungrammatical phrases. These results demonstrate that infants as young as 11 months track the co-occurrence patterns between function words and content words (i.e. determiners before nouns, and pronouns before verbs) in their daily language input.

Hierarchical structure dependence in infants at the early stage of syntactic acquisition

R. Shi, E. Emond, S. Badri

The most fundamental principle in the universal grammar is structure dependence. This principle enables children to impose hierarchical phrase structures onto linear linguistic expressions that they hear. In an experiment with 17-month-olds, an age when syntactic acquisition is beginning, we tested this principle (i.e., hierarchical hypothesis) versus statistics based inductive learning (i.e., linear hypothesis). Infants were presented with French sentences in two structures, and the grammaticality of the subject feature agreement was manipulated. Crucially, the contrasting structures in our sentences contained the same linearly ordered feature-bearing words. Thus, the hierarchical hypothesis predicted different listening responses to the sentences of the two structures, whereas the linear hypothesis predicted comparable responses. Results confirmed the hierarchical hypothesis: infants discriminated the structures and processed the feature agreement accordingly. Our finding provides evidence for the structure dependence principle in infants, demonstrating the influence of infants’ internal system, suggesting that UG knowledge guides early language development.

Preschool children generate pragmatic inferences from both words and pictures

C. Richards, A. Kampa, A. Papafragou

We investigated whether 3- to 5-year-old children apply pragmatic principles, such as expectations of informativeness, to communicative acts beyond language. In a task adopted from Stiller et al. (2015), children were presented with 3 copies of the same object: a no-feature distractor (e.g. smiley face), a one-feature object (smiley face with glasses), and a two-feature object (smiley face with glasses and a hat). Children were either told “My friend has glasses” (Linguistic) or were told “My friend has this” and presented with a picture of glasses (Non-linguistic), and were asked which one was the experimenter’s friend. Children in the Linguistic and Non-linguistic conditions chose the one-feature object at a significantly different rate from controls, starting around age 3.5, with no significant differences between Linguistic and Non-linguistic. Our results offer the first direct evidence that children as young as 3.5 use the principle of informativeness to interpret both words and pictures.

Specificity of infant statistical learning

S. Parvanezadeh Esfahani, J. Hay

Infants’ sensitivity to co-occurrence patterns in continuous speech has been extensively documented. However, we still know very little about how infants represent the sequences that are the output of these statistical learning processes. Here we test 8-month-old infants’ (N=74) indexical (i.e., talker) and suprasegmental (i.e., stress pattern) representations of newly encountered statistically-defined words. In both experiments, infants were familiarized with a naturally-produced Italian corpus that contained two trochaic (strong-weak) high transitional probability (HTP) target words produced by a female speaker. Following familiarization, infants were tested on their ability to discriminate modified target words, either produced by a male speaker (Experiment 1) or with an iambic (weak-strong) stress pattern (Experiment 2), from foils. In both experiments infants demonstrated a significant familiarity preference, listening longer to the modified HTP words than to foils. Findings demonstrate that infants are able to generalize representations of statistically-defined words across a range of acoustic forms.

3 year-old children respect the complex-NP constraint

M. Hirzel, J. Lidz

By age 2, children understand that in wh-argument questions, a displaced wh-phrase fills an argument role, but it is unknown whether their grammars respect island constraints. Whereas 4 year-olds have been shown to respect relative clause islands and temporal adjunct islands, no research has effectively probed knowledge of islands between age 2 and 4. Here, we demonstrate that 3 year-olds respect the Complex-NP Constraint (CNPC). In a question-after-story task, we compare child responses to (1) and (2).

1. She hugged the bear with the blanket. Which blanket was it?
2. Which blanket did she hug the bear with?

22 children (range:2;8;6-4;2;25) participated in this study. Children in the wh-condition gave adverbial responses 66% of the time, compared to 45% in the control condition (t = -2.0603, df = 20, p-value = 0.05). We argue that children’s resistance to adjectival responses to (2) is best explained by knowledge of the CNPC.