Eye-gaze as a window into word level variability in toddlers

Haley Weaver and Jenny Saffran

Individual differences in early vocabulary have lasting impacts on later language development. We investigated variability in toddlers’ understanding of a small set of familiar words using an eye-gaze paradigm with multiple trials and exemplars for each word and compared this word recognition data to caregiver report of vocabulary. We found that on average, when collapsed across items, caregiver report of the number of target words toddlers said predicted word recognition accuracy on an eye-gaze measure. However, we found variability in the alignment between eye-gaze data and caregiver report when we examined word comprehension at the item-level for an individual child. Some toddlers recognized very few words, and their caregiver’s report overestimated their comprehension abilities. Others recognized all of the target words, and their parent’s report accurately reflected these abilities. Quantifying individual differences at the word-level provide new insights into the variability in early language.

Phonological variation in child-directed speech is modulated by lexical frequency

Eon-Suk Ko and Jongho Jun

This study investigates the hypothesis that child-directed speech (CDS) contains a higher proportion of canonical pronunciation than adult-directed speech (ADS) with focus on Korean noun stem-final obstruent variation The accommodation hypothesis of CDS might predict a higher ratio of canonical forms in CDS than their variants. Another factor influencing phonological variation, though largely overlooked in CDS research, is lexical frequency, which would predict a greater proportion of canonical forms in high frequency words in morpho-phonological phenomena. We conduct three experiments comparing the proportion of the canonical vs. variant forms to test the accommodation hypothesis of CDS. Our results indicate that CDS provides children with a higher ratio of canonical forms, which is mediated by the use of high frequency word items in CDS. Our study thus highlights the importance of considering the usage factors (and morpho-phonological phenomena in investigating the function and nature of CDS.

Resolution of backward and forward anaphora in Chinese

Lucy Zhao, Maria Teresa Guasti, Shijia Yang and Francesca Foppolo

This study investigates Mandarin-speaking adults’ and children’s resolution of the overt pronoun subjects (ta ‘he/she’) and null subjects in forward and backward anaphora configurations. While null subjects allow coreference to the subject antecedent in both anaphora configurations, overt subjects do so only in the forward configuration. Results from a picture selection task showed that unlike adults and children aged 8-10, children aged 4-6 did not differentiate between the null element and ta in either forward or backward anaphora. The younger children allowed subject antecedent at a much lower rate in both anaphora configurations. The older children did not yet show adult-like interpretations, which contrasts with the results for overt subjects. This raises a question as to whether null and overt subjects are acquired in tandem at age 6-8. We explore a possible explanation in the nature of the null subject as a null anaphor.

Adapting infant looking time paradigms for the web

Shengqi Zhong, Daoxin Li and Kathryn Schuler

The head-turn preference procedure (HPP: Nelson et al., 1995) has been an important method to the study of language acquisition in infants, but remained restricted to the lab due to the highly controlled lab settings it requires. In the present study, we provide the first demonstration that HPP can be adapted successfully for web data collection. To illustrate, we conducted an online replication of Shi, Cutler, Werker and Cruickshank’s (2006) Experiment 1, which employed the central-fixation adaptation of HPP to determine whether functors like ‘the’ facilitate the extraction of novel nouns in 11 month olds. We leveraged the Lookit platform (https://lookit.mit.edu/) to recruit and run this experiment online. Our results indicate new opportunities for infant studies during the pandemic and in the future.

Verb morphology is syntagmatically and paradigmatically simpler in contingent than in non-contingent CDS

Morgane Jourdain and Sabine Stoll

Little is known about how children acquire languages with highly complex morphology. We hypothesize that contingent child-directed speech (CDS) might support the acquisition of morphology. Here we aim to determine whether the morphology of contingent compared to non-contingent CDS is simpler. For this we analyze data from a longitudinal study of a polysynthetic Sino-Tibetan language (Nepal). The verbal morphology of Chintang is highly complex with more than 4800 forms per verb). We analyzed 44,183 verb forms produced by adults from the longitudinal Chintang corpus. Using syntagmatic, i.e., number of morphemes, and paradigmatic complexity, i.e., number of forms per verb, we find that from age 0;6 to age 4;4, contingent CDS is consistently simpler than ADS, while non-contingent CDS is not. We propose that this simplification in contingent CDS provides children with direct input that is more manageable for them, helping children acquire languages with complex morphology.

Locality effects in the acquisition of nominal ellipsis in L1 Italian

Caterina Tasinato and Emanuela Sanfelici

This study investigates the comprehension of nominal (NP) ellipsis in 3-to 5-year-old Italian-speaking children and adults, asking what the role of locality is in the acquisition of NP ellipsis. We tested two structures differing in the position of the antecedent: discourse ellipsis vs. coordination ellipsis. Results from a picture supported TVJT show that, in interpreting NP ellipsis, Italian preschool children are sensitive to locality. Children overall recovered the meaning of the elided NP in coordination but not in discourse ellipsis, thereby diverging from adults. Moreover, in coordination 3-years-old children differed from the target-like interpretation exhibited by older children. This finding suggests that NP ellipsis in coordination is acquired earlier than in discourse. Children’s difficulty to comprehend discourse NP ellipsis may be accounted for in terms of intervention (Rizzi 2004) or in terms of discourse integration abilities (Berman 2009).

The role of facial cues on infant word learning

Monica Barbir, Mei Kono, Yuka Tatsumi, Samuel Recht, Yu Fujimura, Reiko Mazuka and Sho Tsuji

Faces communicate a rich source of perceptual and social cues to language. Yet, the importance of facial cues on word learning has been hard to investigate, because these cues are so ubiquitous that removing them also removes social felicity. Here, we use common clothing accessories to overcome this problem. We presented infants and toddlers with videos in which a woman introduces two novel nouns and corresponding objects while wearing a mask (mouth cues absent), sunglasses (eye cues absent), or hat (facial cues present). Our results reveal that, even in rich socially felicitous environments, facial cues, both eyes and mouth, are important for word learning. They also suggest that there are developmental periods during which facial cues play a larger role in word learning, notably in the second and third years of life. Our study provides practical insights into the potential impact of pandemic mask wearing on children’s language acquisition.

Verb dominance in Korean revisited: Verbal and nonverbal characteristics of maternal input to Korean toddlers

Anna Shusterman, Abigail Seevak, Petra Mišmaš, Franc Lanko Marušič, Rok Žaucer and Madeline Pelz

Do verbal and nonverbal characteristics of naturalistic Korean caregiver input support the verb bias in Korean? From 30-minute home video recordings of 44 mother-child dyads (14- to 20-month-olds), we obtained the token frequency of nouns and verbs by their subcategories. With short muted vignettes with a beep sound replacing the target word, 200 adults guessed what the word might be using only the visual/social context. Korean mothers produced verbs twice as frequently as object nouns. Adults correctly inferred the meaning of the target word 42.6% of the time, with higher accuracy for verbs than for nouns (p = .001). Individual differences in the total number of nouns and verbs in input positively predicted the total inferential accuracy (p= .04), Korean caregiver input may provide both linguistic and extralinguistic support for verb learning early on, and we aim to further investigate this with our full dataset (n=72) with M-B CDI-K data.

Use of null-arguments in wh-questions: The view from Malayalam and Telugu

Kazuko Yatsushiro, Aurore Gonzalez, Varsha Eturi, Chiara Dal Farra, Johannes Hein, Gayathri G. Krishnan, Vaijayanthi Sarma, Silvia Silleresi, Maria Teresa Guasti and Uli Sauerland

This study investigates the production of wh-questions by children speaking two Dravidian languages, Malayalam and Telugu. Previous studies on the acquisition of wh-questions have detected some asymmetries, both in production and in comprehension: (i) subject wh-questions are acquired earlier than object wh-questions, (ii) who-questions are easier than which N-questions ( Friedmann et al. 2009; Guasti et al. 2012; Sauerland et al. 2016). Most of those studies have focused on the comprehension and production of wh-questions in well-studied languages like English, Italian, and German, however. We observed a novel subject-object asymmetry in two Dravidian languages. Null arguments are used more frequently in subject position of the object wh-questions than in the object position of the subject wh-questions. This surprising finding from two under-studied languages sheds new light on the cross-linguistic status of the subject/object asymmetry.

Less-is-more: A developmental fMRI study on auditory statistical learning

Anqi Hu, Katherine Trice, Yi-Lun Weng and Zhenghan Qi

In this fMRI study, we compare the engagement of linguistic-specific (LN) and domain-general attention networks in statistical learning (SL) between children and adults to examine two competing accounts. The less-is-more account predicts greater learning-induced changes in the LN in children. The age-invariant account predicts similar learning-induced changes between children and adults. Participants completed two auditory SL tasks in the scanner – a linguistic syllable SL task intermixed with random tone sequences and a non-linguistic tone SL task intermixed with random syllable sequences. Adults showed greater overall engagements in both networks across both tasks. Children’s LN showed a greater degree of learning-induced plasticity only for linguistic SL. Greater attention network engagement during the task was associated with greater learning-induced malleability in LN for both, supporting an interplay between domain-general and domain-specific learning mechanisms underlying SL. Together, these results provide the initial neural evidence supporting the less-is-more account of language acquisition.

What predicts comprehension of temporal conjunctions in bilingual children? Age of onset, length of exposure, or something else?

Christos Makrodimitris, Petra Schulz and Ianthi-Maria Tsimpli

Our study investigates the role of LoE and AoO in bilingual children’s comprehension of sentences containing the conjunctions “after” and “before”, a phenomenon mastered very late (after age 7) cross-linguistically. Following the Timing Hypothesis, we expected LoE to be a better predictor of performance than AoO. Fifty-five typically developing children with L1-Greek and L2-German (Age: 6;0–12;9 yrs.) performed a sentence-picture matching task in German and in Greek, with order counterbalanced.
The results were first analysed with two separate GLLMs for AoO and LoE. As expected, LoE, but not AoO, showed a main effect. An additional model including LoE, Age, short-term memory, German grammar skills, and accuracy in the Greek after/before-task, revealed main effects of accuracy in the Greek after/before-task and of German grammar skills, but not of LoE. Our findings support the Timing Hypothesis, but suggest that ‘LoE’ may not adequately account for the performance of bilingual children.

Infants bootstrap grammatical categories from morphological suffixes
Megha Sundara

It has been proposed that infants use information about grammatical categories to acquire morphology. We provide experimental evidence against this claim. In two experiments with over 100 infants we show that English-learning infants distinguish homophonous 3rd singular -s and plural -s by 8-months; by 14-months, they group nonce words suffixed with 3rd singular -s, but not plural -s, with nonce verbs suffixed with -ing. Thus, not only do English-learning infants discover suffixes early (Kim & Sundara, 2021), they use morphological suffixes themselves to discover English noun and verb categories. We will discuss the implications of our findings for phonological, semantic and distributional proposals for establishing grammatical categories in infancy.

Conversation disruptions in early childhood longitudinally predict receptive language development
Amy Carolus, Margaret A. Sheridan, Liliana J. Lengua, Kate A. McLaughlin and Rachel Romeo

Qualitative measures of language input are potent predictors of a child’s language development. In particular, more frequent experience with adult-child conversational turn-taking is associated with stronger language development. However, little is understood about how disruptions to temporal contingency in turn-taking (i.e., interruptions and overlapping speech) that may impede its benefits relate to language development. In this analysis of 275 mother-child dyads, contrary to hypotheses, more conversational disruptions as a proportion of total utterances during a 7-minute free play at age 3 significantly predicted higher language ability 9 months later, as measured by the by the NEPSY-II, controlling for sex, age, and socioeconomic status. This relationship remained significant even after additionally controlling for T1 language ability, indicating that disruptions predicted language increases. Ongoing analyses seek to further classify types of temporal conversation disruptions, such as cooperative interruptions or back-channeling, to investigate how they may facilitate language development in certain circumstances.

The effect of language switching on bilingual children’s novel word learning
Rachel Ka-Ying Tsui, Jessica Kosie, Laia Fibla Reixachs, Casey Lew-Williams and Krista Byers-Heinlein

Children growing up in bilingual environments are often exposed to mixed-language utterances in their language input. Patterns of language switching also vary where caregivers switch between their languages in different ways. Some bilingual children hear their languages in long blocks of a single language (a ‘one-language-at-a-time’ strategy), while others hear more frequent switching (an ‘immediate-translation’ strategy). Our study explored whether the two different switching patterns imposed different demands on bilingual children’s learning of novel nouns in their two languages. Using an online tablet task, data were collected asynchronously from 36-to-60-month-old bilinguals from French-English bilingual families in Montreal, Canada (N = 31) and Spanish-English bilingual families in New Jersey, USA (N = 20). Mixed-effects logistic regression analyses on proportion of accuracy revealed strong evidence of learning, with no differences across communities or conditions. Our findings highlight that different patterns of bilingual interaction provide equal learning opportunities for bilingual toddlers’ vocabulary development.

Intervention effects in Mandarin-speaking children’s comprehension of passives
Minqi Liu, Victoria Mateu and Nina Hyams

In Mandarin, an external argument is structurally absent in Short Passives but it intervenes in the dependency between the surface subject and its gap in Long Passives. Therefore, the Intervention Hypothesis (Friedmann et al. 2009) predicts that Mandarin LongPass should be harder for children to comprehend and this Intervention Effect should decrease when the grammatical features of the two arguments in LongPass are distinct.

These predictions are borne out in our experiments. Exp1 tested a grammatical feature (Number) by manipulating the plural vs. general classifiers in Mandarin DPs. Data from 80 monolingual Mandarin-speaking 3- to 6-year-olds showed that LongPass are harder than ShortPass and a mismatch in Number improves children’s comprehension of LongPass.

The same children did not show this match/mismatch difference with LongPass in Exp2 where the arguments/classifiers mismatched only in inherent lexical properties but not Number, supporting grammatical accounts to Intervention such as featural Relativized Minimality (Rizzi 1990).

Contrast, sufficiency, and the acquisition of morphological marking
Sarah Payne

Much work on morphological acquisition has examined how children learn to map morphological features to phonological form. But how do children learn which features are marked in their language to begin with? We propose that the child’s early segmentation of inflectional morphology allows them to make use of collisions: a single stem appearing in multiple inflected forms (e.g. walk-walked). Once the child learns which features distinguish individual collisions (e.g. ±PAST distinguishes walk-walked), we propose that they employ the Tolerance/Sufficiency Principle to determine whether there is sufficient evidence for marking of these features across the paradigm. We demonstrate that a computational implementation of such a learner matches well with developmental findings regarding vocabulary size and order of acquisition for the well-studied problem of English verbs, as well as agglutantive Spanish verbs and templatic Hebrew verbs.

Translation equivalents in bimodal-bilingual deaf children’s ASL and spoken English vocabularies
Elana Pontecorvo, Amy Lieberman, Jennie Pyers and Naomi Caselli

Children learning two or more languages often have to map multiple phonological forms onto the same concept, learning translationally equivalent words (e.g., “dog” and “perro”). Contrary to early predictions that children’s early vocabularies in each language may tend to be non-overlapping, translation equivalents have been found to be present and sometimes overrepresented. In this study, we found that translation equivalents are overrepresented in the early vocabularies of bimodal-bilingual children, deaf children who are learning both ASL and spoken English. This indicates that the preference for translation equivalents is not necessarily motivated by phonological similarity across languages.

Does grammatical number scaffold the acquisition of number words?
Anna Shusterman, Abigail Seevak, Petra Mišmaš, Franc Lanko Marušič, Rok Žaucer and Madeline Pelz

Can children use syntactic information, like grammatical number, to facilitate the meanings of number words? While English only differentiates between singular and plural, other languages, like Central Slovenian, have distinct grammatical markers for referring to quantities of one, two, and more than two. The current study tested the claim that morphology scaffolds number-word learning by assessing whether young children acquire dual marking before the number word “two” and reanalyzing data from Give-a-Number and Give-a-Set tasks. We find that most children did not learn number morphology before the corresponding number word, complicating claims that number morphology could scaffold number word learning. Furthermore, we did not replicate the finding that children acquiring a dual-marked language learn “two” relatively early. Finally, we demonstrate heterogeneous clusters in the acquisition of number morphology, suggesting mechanisms inconsistent with the scaffolding hypothesis.

Person reference via pronouns or verbs has different effects on false belief performance: Longitudinal study from Czech toddlers
Anna Chromá and Filip Smolík

Personal pronouns are probably related to grammatical as well as social development, but the relation to false-belief understanding – a milestone in social-cognitive development – has not been examined. We tested whether the mastery of personal pronouns and similarly functioning inflected verb forms predicted the success in false belief tasks. Sixty-one Czech children participated in a two-wave (29 and 43 months) study investigating social-cognitive and linguistic development, including false-belief understanding, and elicited and spontaneous production of pronouns and verb inflections. Regression analyses revealed a number of significant unique effects of person reference above and beyond the effects of general language skills. Both the elicited and spontaneous pronoun data show effects of second-person mastery, while the spontaneous verb-person data show effects of first-person forms. This confirms that pronouns and person-inflected verbs in 2-year-olds are related to subsequent development of social cognition.

Children’s code-switching changes from 2;6 to 3;6 years to maximize their power of expression
Michelle K. Tulloch and Erika Hoff

Spanish-English bilingual children’s use of code-switching was investigated longitudinally from 2;6 years to 3;6 years from recorded sessions of Spanish-designated and English-designated interactions with a bilingual caregiver. Between 2;6 and 3;6 years, children became more English dominant, and their rate of code-switching to English increased while their rate of code-switching to Spanish decreased. The strongest influence on bilingual children’s frequency of code-switching was their relative proficiency in their two languages over and above proficiency in the matrix, or instructed, language. These results suggests that lexical gaps are defined by the speaker with reference to the language in which their skills maximize their power of expression.

Belief reports facilitate long-distance binding in child Japanese
Akari Ohba and Kamil Deen

Children are known to have difficulty with long-distance (LD) interpretations for reflexives such as Japanese zibun ‘self’ (Orita et.al., 2021). Children are also known to interpret ‘think’ differently from adults (DeVilliers & Pyers, 2002, a.o.): they report “John thinks dogs quack” as false because dogs don’t actually quack. Lewis et.al. (2017) show children can correctly interpret ‘think’ when the belief-report meaning of ‘think’ is salient.

We investigate whether the difficulty with ‘zibun’ is an artefact of using omou ‘think’ as the matrix verb. ‘Zibun’ is often logophoric and takes a LD antecedent whose speech/thought is reported. If children struggle with the belief-report function of ‘omou,’ they may not consider the matrix subject a licit antecedent. We highlight the belief-report function of ‘omou’ and show that Japanese children can indeed access LD interpretations for ‘zibun,’ concluding that children’s difficulty with ‘zibun’ may derive from the well-documented difficulty with the verb ‘think.’

Letitia Naigles

Children on the autism spectrum engage in social interactions with lower frequency and in different patterns compared with children who are developing typically. This variation in social engagement has afforded many explorations of the significance of social interaction as a mechanism in language acquisition, development, and use. The Longitudinal Study of Early Language is a 20+-years-long multi-domain exploration of the language of children on the autism spectrum, including their comprehension via Intermodal Preferential Looking, their production via naturalistic interactions, and their performance in numerous psycholinguistic and clinical assessments. In this talk I will highlight how comprehension findings demonstrating early acquisition of grammatical constructions, combined with wide heterogeneity in language production by these same children, indicate how social engagement plays a targeted rather than universal role in early language. Moreover, investigations of lexical semantics reveal challenges but also promote reconsideration of the definitions of social strengths and weaknesses in both autistic and non-autistic individuals.