Deaf and Hard-of-Hearing (DHH) people benefit from early exposure to American Sign Language (ASL). This study examines the use of ASL classifiers, early exposure, visual-spatial skills, and fluency in ASL. Using ASL handshape classifiers, which convey spatial information in ASL, this study examines the relationship between fluency and ASL acquisition age (AoA). This study used Vandenberg-Kuse Mental Rotation Tests (VKMRT), ASL Comprehension Tests (ASLCTs), and Block Design Subtests (BD). Handshape classifiers were spontaneously generated during VKMRT and BD tasks and coded. On VKMRT and BD tasks, classifiers (Bent3, BentL, Bent5, FlatO, and G) significantly improved performance. Classifiers overall significantly predicted spatial cognition tasks, depending on hearing status. Visual-spatial abilities of DHH signers were significantly enhanced by ASL classifiers, demonstrating their vital role in supporting these abilities. Deaf individuals’ spatial cognition has been demonstrated by these findings, which highlight the relationship between movement, language, and cognition.

We provide the profile of one family’s journey to bimodal bilingualism with their profoundly deaf child, ‘Holly’. Holly’s mother began to learn ASL and use it with Holly before they joined our study at 2;00. Holly developed in both ASL and English over the year of our observation, with vocabulary an area of particular strength. In ASL, her phonological accuracy increased from 69% to 86%. Analysis of English spontaneous speech put her performance in the same range as that of deaf and hearing children with deaf parents. Mother’s ASL was already strong to begin with, and improved over the year, especially in phonology and syntax. Overall, this case study supports the conclusion that early access to ASL for deaf children from hearing parents can lead to successful bimodal bilingualism.

Statistical learning has been described to be an essential part of language learning in children. Recent studies suggested that sign language exposure is an important determinant of sequence processing abilities in deaf and hard-of-hearing (DHH) children. In the present study, we investigated SL and the impact of AoA using a visual artificial grammar task (Schönberger et al., 2021), which allowed investigating transfer abilities as well. Additionally, we examined the impact of age of acquisition and the children’s ability to transfer learned rules. Our study demonstrates that DHH children show effects of statistical learning when using artificial grammar and that learning and transfer abilities are affected by AoA. Thus, early language access seems to promote efficient statistical learning.

Late exposure to first language has effects on linguistic and cognitive skills. In this study, we investigate the comprehension of zero-anaphora in native and late Deaf adult Turkish Sign Language signers to understand their reference-tracking strategies in the absence of morphological cues and see if there are differences between these groups. The results of the task conducted with 32 participants (16native-16late) emphasize the impact of acquisition age on zero-anaphora comprehension in TİD, revealing distinctive strategies employed by native and late-signers (p<0.05 (0.01310)). The position of the intransitive clause has significant effect on responses (p<0.01 (0.00807)), while thematic role did not show significance. As a follow-up, we focused on the response times of the native signers, which indicate that native signers demonstrate a strategy based on linear proximity to zero-anaphora, suggesting the influence of memory effects on retrieval. This study highlights the impact of delayed first language acquisition in syntax/semantics/pragmatics interface.

Bilingual children need regular exposure to their languages to become proficient in them. However, early exposure is highly variable and can change over time and across contexts. This study investigates transitions – changes in children’s early environments – and how they influence bilingual language exposure longitudinally. We measured environmental changes and language exposure in 53 Montreal French-English bilingual children between birth and 30 months of age, through caregiver interviews. Transitions were common, with high individual variability among children. Most transitions did not lead to changes in exposure dominance. However, with age, children experienced more changes that led to more balanced language exposure and transitions related to language strategy led to larger changes in overall exposure. These findings shed light on how environmental variability affects early bilingual exposure and what factors play a larger role, setting the path for future work looking at these changes in relation to bilingual language development.

Children must consolidate newly learned words within their lexical-semantic system, and in monolingual children, consolidation of novel words is known to require time and sleep. Yet, little is known about how bilingual children consolidate information. In line with the complementary learning systems framework, we expected children to consolidate novel words over time. In line with weaker-links account of bilingualism, we predicted that bilinguals might show weaker consolidation. Thirty-four monolingual children (Experiment 1) and 17 bilingual children (Experiment 2) learned English-like novel words. Phonological and semantic consolidation of newly-learned novel words were tested immediately after learning and after a 24-hour delay, via co-activation between newly-learned and familiar words. Monolinguals demonstrated semantic but not phonological consolidation immediately following learning. In contrast to the weaker-links hypothesis, bilinguals demonstrated phonological consolidation immediately following learning and semantic co-activation a day later. These findings suggest that bilingualism shapes within-language phonological and semantic consolidation.

We address the bilingual acquisition of optional lexical-syntactic markers with a pragmatic function, such as discourse particles, focusing on the modification of rhetorical questions (RhQs) by heritage acquirers of Italian. The topic offers a novel perspective on bilingual acquisition, as it concerns not core grammar, but the ability to appropriately exploit optional cues to achieve a desired pragmatic meaning, signalling attitude and/or discourse function. We conducted an elicitation study with 88 Italian-German bilinguals and 80 monolingual controls (age range: 6.0-10.2). Quantitatively, bilingual children used fewer markers to modify RhQs, with a strong effect of dominance. Only monolinguals, and not bilinguals showed an increased rate of modification with increasing age. Qualitatively, bilingual children exploited the same types of cues as monolinguals, with, additionally, a few cases of lexical transfer from German. Thus, bilingual children can use the optional pragmatic markers of their heritage language in RhQs, especially if receiving higher exposure.

The relationship between age of acquisition, linguistic proficiency, and brain structure in language acquisition research is significant. Previous studies remain inconclusive with variability of cortical, subcortical, or cerebellar grey matters reported, due to confounding factors brought by individual differences unrelated to bilingualism, such as, age and sex (Pliatsikas et al, 2020). Our goal is to disentangle bilingual-specific neuroanatomical variation from the common variation. To achieve this, we trained a neural network called Contrastive Variational Autoencoder (CVAE). Based on traditional Variational Autoencoders (VAEs) that learn a compressed latent representation, CVAE incorporates contrastive learning. This enables successful learning of representations by contrasting pairs from the two distinct populations, thereby isolating the variations specific to one population from the shared variations (Aglinskas et al., 2022). With a training dataset includes over 2,000 brain images, we found the common variation shared by two populations can be consistently captured by CVAE.

This study investigates the comprehension and production of subject and object relative clauses (RCs) by Mandarin-speaking children with DLD and their typically developing (TD) peers. Fifty-six children aged from 03;08 to 08;07 were tested with a comprehension task, a production task, and a forward digit span (FDS) and a backward digit span (BDS) task. The DLD children performed better on the comprehension task than on the production task, whereas the TD children did not show significant difference between two tasks. In addition, the DLD children performed much worse than their TD counterparts in RC acquisition, and these children showed different patterns in comprehension and production. The BDS correlated with DLD children’s RC acquisition, while both FDS and BDS correlated with TD children’s performance. The results are discussed within current theories of DLD and indicate structural intervention and linear intervention have a great impact on Mandarin-speaking DLD children’ RC acquisition.

This study investigates the interaction between syntactic complexity, working memory (WM) and repetition accuracy on the German LITMUS-sentence-repetition-task in 106 monolingual and bilingual children with/out SLI. Repetition accuracy decreased as a function of cumulative syntactic complexity with larger effect sizes in the SLI groups. No associations with WM emerged for the SLI groups due to the lack of inter-individual variability in the digit recall data. Performance of MoTDs on less complex structures was predicted by vocabulary, whereas BiTDs relied on both verbal-short-term-memory and L2-vocabulary, probably compensating for their weaker (L2-)language abilities by resorting to their intact WM capacities. Interestingly, performance on structurally more complex structures was only predicted by verbal-short-term-memory in both TD groups confirming that processing of computationally more complex structures necessitates more WM capacities. The findings are relevant not only for theories of SLI, but also for shortening/ameliorating the task and ensuring best diagnostic accuracy in bilingual populations.

Early (pre)verbal productions and description of language trajectories in autism are often investigated by following the siblings of autistic children who have a Higher Likelihood of having Autism Spectrum Disorder (HL-ASD). HL-ASD children who do not receive a diagnosis of ASD are also at high risk of displaying language and communicational disabilities. To compare the early (pre)linguistic development of HL-ASD children who were diagnosed with autism versus those who were not, we analyzed association between canonical babbling volubility and consonant inventory at 18 months and diagnostic outcome, expressive vocabulary, nonverbal IQ at 36 months. Overall, we found no association between diagnostic outcome groups and measures at 36 months. Siblings of autistic children who themselves receive a diagnosis and those who do not all show weaknesses in language development and seem, over this period, to follow similar language trajectories overall.

Function word acquisition is crucial for early language development, aiding in grammatical structure, word class acquisition, and word segmentation. Few studies have been done on the function word acquisition in atypically developing populations, particularly those with hearing loss. The current study analyzes the development of “a” and “the” articles in hearing loss populations utilizing the Ambrose corpus on CHILDES database. Our results show a significant delay in the production of “a” but not “the” in children with hearing loss, with the delay disappearing at 36 months. HL parents’ use of “thi” may explain the increased production of “the” in the HL population. Children with cochlear implants had a delayed production of both “a” and “the,” but by 36 months the production of “the” is no longer delayed. The study highlights the importance of input quality in hearing loss population’s function word acquisition and the need for further research.

This study investigates distributional language learning in neural networks when given child-directed speech, with a focus on semantic representations for words. Our simulations reveal a propensity in these models to form ‘entangled’ representations due to the statistics and forward-predictive nature of language, potentially impeding lexical inference. However, we also propose that the structure of child-directed speech can mitigate this entanglement by fostering more ‘atomic’ lexical representations, because speech to younger children facilitates category learning. We test this hypothesis in a set of simulations using artificial languages and the CHILDES corpus, showing models trained on language that starts out mimicking speech to younger children perform best. Our results underscore the significance of studying architecture-input interactions and the value of developmentally plausible language models. The study contributes to our understanding of how child-directed language structure may facilitate learning, mitigating risks associated with entangled representations.

One type of complex syntactic knowledge children acquire concerns constraints on allowed wh-dependencies, sometimes called syntactic island constraints (Ross, 1967). Past computational cognitive modeling work (Dickson et al., 2022) has investigated how children might acquire this knowledge, and found that knowledge of syntactic island constraints can emerge automatically from learners who try to identify an efficient representation of the wh-dependencies in their input in general. Notably, this prior work assumed an idealized representation of children’s input, with perfect memory of the wh-dependencies. Here, we investigate how this learning strategy fares with an imperfect, noisy representation of the input that may better reflect the impact of children’s cognitive constraints on their input processing. We find that even highly-forgetful modeled learners are able to succeed, generating many observed empirical patterns about syntactic island knowledge. Our findings support the plausibility of this efficient-representation learning strategy by children, who have memory limitations.

A phonotactic restriction against syllable-final voiced obstruents in Dutch manifests in voicing alternations in some noun paradigms. For instance, in [bɛt] “bed” ~ [bɛdən] “beds” the stem-final obstruent is voiced in the plural, but unvoiced in the singular, where it occurs in syllable-final position. In other paradigms, the obstruent is voiceless throughout—e.g., [pɛt] “cap” ~ [pɛtən] “caps”. Alternating obstruents are often analyzed as being underlyingly voiced. However, despite children showing productive knowledge of many morphophonological processes, Dutch-learning children show no evidence of productive knowledge of the voicing alternation at 2;6-3;6, struggling to produce or interpret the singular form of alternating nonce nouns compared to non-alternating ones. We propose a learning-based explanation for this lack of productivity by applying a cognitively-motivated computational model to child-directed speech. The results suggest Dutch-learning children may lack this productive knowledge because they are able to form effective morphological generalizations without it.

Distributional learning mechanisms are important for language acquisition. But what are the units over which distributional learning operates? Claims about distributional learning mechanisms rest on which units are being used. To address this issue, we conducted two studies: (1) predicting child vocabulary data using statistics derived from child-directed speech, and (2) employing distributional semantic models built from child-directed speech to predict semantic category development. Critically, in both studies we manipulated the corpus used to calculate distributional statistics, varying whether words were left whole, or morphologically parsed with inflectional morphemes (e.g., -s, -y, -ed, and -ing) retained as distinct units in the corpus. Our findings consistently demonstrated that models performed significantly better when the inflectionally parsed corpus was used. These studies highlight the importance of considering what units are being used in distributional learning, and demonstrate the importance of sublexical morpheme-level units in the distributional learning of word meanings.

This paper examines the comprehension of both singular and plural NPs by adult L2-English learners whose L1, Mandarin Chinese, does not have obligatory plural parking. In Study 1, participants judged sentences as TRUE/FALSE on the basis of number (mis)matches. In Study 2 (modeled after Jiang et al. 2017), participants judged sentences as TRUE/FALSE on the basis of the spatial configuration, but number (mis)matches were also manipulated. Study 1 findings were that L2ers do pay attention to number, albeit with less accuracy than NSs; when the response was based on physical configuration (Study2), L2ers largely ignored number (mis)matches, unlike NSs. While Jiang et al. argue that L1-Mandarin L2-English learners lack automaticity with English plural marking, we suggest that the issue is more about task demands: L2ers can use plural marking to guide comprehension, but limitations on working-memory capacity make it difficult to take in multiple sources of information.

English and Chinese exhibit different interpretations for the sentence ‘Every horse didn’t jump over the fence.’ This cross-linguistic variation poses challenges for L2 learners, particularly English learners of Chinese (ELC) who must acquire the absence of the inverse scope reading in Chinese, not present in their native English. Chinese learners of English (CLE), on the other hand, may find it easier to acquire scope interpretation due to positive evidence in the L2 input. Two experiments were conducted with 64 CLE and ELC. Results revealed that ELC rated the IS reading significantly higher than Chinese speakers, while CLE showed marginal differences compared to English speakers. Eye-tracking data indicated similar patterns. Bayesian analyses confirmed the marginal significances, supporting no significant differences across conditions. Overall, CLE and ELC achieved similar levels of success in native-like interpretations. The difference in interpretation performance between the TVJT and eye-tracking tasks can be attributed to task effects.

This study investigates the effects of construction frequency in the L2 and transfer from the L1 in the acquisition of the periphrastic and se-passives in L2 Spanish. The two constructions differ in that the periphrastic passive, but not the se-passive, has an analogous structure in English, and the se-passive is at least eight times more frequent than the periphrastic passive in Spanish. We ask: How accessible are periphrastic and se-passives for L2ers in production? And do L2ers’ productions reflect target-like representations of these structures? Preliminary results from our priming task show that both L2ers and native speakers produce a comparable number of passives in both immediate and delayed priming. However, results from the AJT reveal non-target-like performance by L2ers on se-passives. This is the first study to test both passive constructions, offering insight into the particular abilities of L2ers, and highlighting the importance of using multiple methodologies.

Previous studies suggest L1 and L2 speakers of English use different real-time strategies to interpret garden-path sentences. Specifically, L1 speakers show a weaker garden-path effect when the initial misleading reading is implausible, whereas L2 speakers do not. The current study extends this line of research to a new language (Mandarin) and a new population of bilingual speakers (heritage speakers). We report the results of a self-paced listening task in which plausible and implausible garden-path sentences were presented to L1, L2, and heritage speakers of Mandarin. Preliminary results from 42 participants (30 L1, 12 heritage) showed that L1 speakers responded faster after the disambiguation point in implausible vs. plausible sentences. In contrast, heritage speakers showed no differences at the disambiguation point but slower RTs at the preceding plausibility cue in implausible sentences. This suggests that heritage speakers noticed the semantic cues but did not integrate them into online syntactic prediction.

The extra-textual talk generated during shared book reading has been found to contain a high level of structurally rich linguistic construction. However, it is unclear how toddlers process the diverse linguistic input and how additional extralinguistic cues, such as gestures, work together with linguistic input in this learning process. We conducted a dual head-mounted eye-tracking study to investigate how shared book reading may support toddlers’ learning of word-referent mappings during book-reading interactions. We found that toddlers are constantly exposed to a mixture of simple and complex utterances regardless of gesture use. Although gestures do not change parents’ speech complexity, they successfully drive the child’s attention to the named referent. Our findings suggest that shared book reading provides toddlers with a linguistically diverse language training ground. Gestures provide learners an easier pathway to associate the word with its referent, which is likely to facilitate real-time language comprehension.

Children’s earliest word learning opportunities are often ambiguous, but linguistic context can reduce this ambiguity. However, it is an open question whether a noun’s linguistic context on one exposure can guide inferences about its meaning on subsequent exposures. Here, we tested whether adults and 3-year-olds used prior linguistic contexts when selecting a novel noun’s referent on later exposures. In Study 1, we found that when informative linguistic context was presented only two exposures before learners had the opportunity to use it, only adults systematically made referent selections compatible with the prior linguistic context. In Study 2, we examined an alternative role for sentence context: verifying or disconfirming a previously hypothesized meaning. Again, adults correctly used informative sentence contexts to either confirm or reject their prior meanings, but 3-year-olds did not. This suggests linguistic context’s role in children’s noun learning is primarily local, rather than constraining children’s inferences across exposures.

In the coming years, technological advances will further increase the role of digital technologies in language learning. We explored what role social robots could play in education by testing German-learning toddlers in a word-learning eye-tracking experiment with a social-humanoid robot, in which learning was only possible if the robots’ social cues were followed. The robot turned its head/gazed towards one of two novel objects and labelled it. During the familiarization, children looked faster and longer to the labelled object, suggesting that toddlers can follow the robot’s social cues. At test, infants older than the group average did not show significant differences between the two objects, whereas younger children looked longer to the correct object than to the incorrect one. Younger children, who have not fully mastered the use of social cues and tend to fail in similar tasks involving humans, may benefit from the structured social cues provided by robots.

We present the results of a large-scale effort to document the acquisition of language varieties in the Bicol region of the Philippines. Children in this region are considered emergent multilinguals, acquiring Bikol (bik; Austronesian), alongside Filipino (the national language based on Tagalog) and English (also an official language). Despite Bikol being considered vigorous by Philippine national institutions, our >100-hour corpus from 10 family contexts reveals substantial variability in intergenerational transmission. We present various analyses of the data suggesting potential obsolescence in future generations, and the need for further attention and resources on Bikol communities.

The present study focuses on parental use of parentese: the acoustically exaggerated, clear, and higher-pitched speech produced by male and female adults across cultures when they address infants. While previous research shows that parentese enhances language learning, it is still unclear what drives the variability in the amount of parental parentese use. We report on the development of a questionnaire assessing parental beliefs, attitudes, and knowledge around parentese. We also directly relate parental questionnaire scores to their own parentese use, obtained through naturalistic daylong recordings. Results suggest that parental beliefs around parentese are variable, and that maternal and paternal links between beliefs and behavior are distinct, emphasizing the importance of studying all caregivers. A better understanding of parental beliefs, and their relation to behavior, contributes to theories of language acquisition and aids in intervention design.

Blind children acquire language without visual input. We consider whether parents of blind infants tailor speech to their children as a compensatory mechanism, by looking at the type and conversational content of the utterances the children hear. Specifically, we investigate whether they hear speech that is high in informativity (information-adding utterance types and content). We classified a corpus of daylong recordings from blind and sighted infants for utterance types: imperatives, interrogatives, declaratives, directives, and utterance content: extensions, expansions, affirmations, repetitions, initiations, and reading/singing. We tested for group differences in the proportion of each utterance type and content category, and found that the input across our measures did not differ for blind vs. sighted children’s input. Input to blind and sighted children was highly similar, neither less, nor more “helpful”, suggesting input informativity is not likely to be a central compensatory mechanism.

Our aim is to identify the underlying mechanisms explaining why contingent child-directed speech (CDS) is simpler than non-contingent CDS. We hypothesize that adults adapt the complexity of their utterance to the complexity of the child’s directly preceding utterance. We call this specific type of accommodation on an utterance-by-utterance basis “responsive attunement”. We conducted a corpus study on three languages with complex verb morphology, Chintang, Qaqet and Turkish. We compared the length of children’s utterances and verb units in number of morphemes to that of adults’ contingent response. Our results show that indeed, the length of verbs and utterances in contingent CDS is correlated with the length of the utterances and verbs in the child’s preceding production. Because the Chintang corpus also contains ADS, we tested if responsive attunement was specific to CDS. Results in contingent ADS were similar, which suggests that responsive attunement is a general property of human interaction.

Language spoken to infants differs perceptually and acoustically from speech directed to adults. We know little about how speech varies depending on context within infant-directed speech (IDS). This study asks: 1) Does parents’ pitch adjustment from adult-directed speech (ADS) to IDS vary depending on context? 2) What is the relationship between IDS pitch and vocabulary development? Fundamental frequency (f0), was extracted from the speech of 11- to 20-month-old infant-parent dyads while they completed three tasks: free-play, sorting, and reading a picture book. To elicit ADS, similar tasks were completed within a parent-experimenter interaction. We found that pitch adjustment was significantly different between tasks. Parents’ f0 adjustment was also positively correlated with infant vocabulary size. These results suggest that parents use pitch differently depending on the context and that pitch is important for infant vocabulary development.

We characterize the everyday speech-language environments of preschoolers with and without cochlear implants (CIs) and evaluate how the children’s environments predicted their vocal maturity outcomes. Eighteen children with CIs (31-65 chronological months; 12-54 hearing age months) were matched to two groups of typical hearing controls. Children completed a single daylong audio recording (appx. 16 hrs./child; >730 hours of observation) where they wore a lightweight recorder for an entire day. Estimates of the home language environment (e.g. quantity/loudness of caregiver input, number of conversational turns between the child and an adult), as well as child vocal outcomes (e.g. quantity, duration, loudness of hourly child vocalizations; quantity and timing of vocalizations contingent with caregiver speech) were derived for each hour, of each recording. Results show differences by hearing group on vocalization maturity and contingent vocalizations made with caregivers. Additional results relating the language environment to children’s outcomes will be discussed.

When adults segment words in a new language, they continue to apply sublexical, phonotactic cues from the L1 (Finn & Hudson Kam, 2008) and benefit from in form and meaning with L1 equivalents (i.e. cognates: English: /kraʊn/; German: /kroːnə/; noncognate: English: /skɪn/; German: /haʊ̯t/; Shoemaker & Rast, 2013). In two studies, we examine the role of developing L1 lexical and sublexical knowledge as well as phonological awareness skills, which may support L2 acquisition in young children (Hu, 2014), on German 6- to 9-year-olds’ segmentation of English speech before they receive instruction in English. Our results show that children were able to segment and subsequently recognize words, which improved with increasing phonological awareness skills. Unlike adult FL learners, there was no evidence that school-aged children use L1 phonotactic cues (Study 1) nor benefit from form-based lexical overlap with their L1 (Study 2) when they begin learning a new language.

This paper examines the acquisition of vowel epenthesis (i.e., EV) in consonant + rhotic clusters (i.e., /Cɾ/) in Spanish as a heritage language (e.g., /tɾonko/ → [təɾoŋko] ‘log’) to shed light on the robustness of the encoding of representations at the phonetic-phonology level. /Cɾ/ sequences were elicited using the pictureless book Frog, where are you (Mayer, 1969) from child and adult heritage speakers (i.e., HSs), as well as from child and adult Spanish speakers raised in monolingual environments (i.e., SpanMonoSs). Our results show that coarticulatory patterns in Spanish /Cɾ/ sequences continue to mature at ~ 8 years of age. Such late development of gestural organization, however, is not constrained by a competing representation at the phonetics-phonology interface in the HSs. In addition, our grammatical predictors showing that shared features between the consonant and the tap increase probability and duration of EV provide support for a perceptual analysis of vowel insertion.

Infants in bilingual environments must acquire two languages with different phoneme inventories. While infant-directed speech in a single-language context may highlight the differences between two languages (i.e. VOT in the case of English-Spanish), studies of adult-directed speech show that the languages are often produced more similarly in a language-mixing context. The goal of this study is to evaluate whether the effect of being in a language-mixing context differs between IDS and ADS. Spanish-English bilingual women narrate stories in IDS and ADS, using a set of six wordless picture books. In each register, stories are told in English, Spanish, and while switching between languages. VOTs of voiceless stops are measured and analyzed. Preliminary work suggests an interaction between register and context in both Spanish and English, such that mothers accentuate the difference between either voiced and voiceless stops or between English and Spanish more in IDS than in ADS.

The present paper examined preschool Mandarin-speaking children’s ability to infer plausible alternatives triggered by negation. Experiment 1 tested three- to five-year-old children’s ability to make logical inferences on negation. Participants heard ‘This is not X’, while seeing four pictures that formed different relations with ‘X’, and they were asked to choose ALL the possible alternatives. Experiment 2 further examined children’s ability to make real-world inferences based on taxonomic relations, where they were asked to choose only ONE alternative. Results showed that in Experiment 1, participants’ tendencies to choose taxonomic, thematic or unrelated alternatives did not differ from each other. In Experiment 2, four- and five-year-olds exhibited a stronger tendency to choose taxonomic elements. Our findings show that preschool children are able to make logical inferences on negation. In addition, four- and five-year-olds can make further real-world inferences based on the taxonomic relations, but three-year-olds are less able to do so.

Both children learning word meanings and adults labelling objects prefer “basic”-level terms (“dog”), over subordinate-level (“dalmatian”) or superordinate-level (“animal”) terms. We propose that the basic-level bias partly has linguistic-pragmatic underpinnings, and that the use of basic and non-basic terms crucially involves identifying the relevant level of pragmatic informativeness. We test two predictions. First, a superordinate-level term should be judged as infelicitious (but not entirely incorrect) where the basic-level one is known and relevant. Second, the basic level should be preferred in production, unless the context introduces more specific informativeness expectations. Results from a ternary rating task show that adults but not children are sensitive to the pragmatic violations (of calling a familiar dog an “animal”, for example). However, children still preferred the basic-level in a subsequent production task. We conclude that the basic-level bias can be linked to expectations about the pragmatic levels of informativeness encoded in category labels.

Belief verbs describe abstract mental states, making their semantics potentially difficult to learn. Dudley et al. (2017), analyzing know and think, hypothesize that learners might use syntax and pragmatics to learn whether a belief verb is factive or non-factive. Our study of seven prominent English belief verbs in child-directed speech (factive: know, remember, see; non-factive: think, guess, say, tell) suggests that these verbs can indeed be differentiated syntactically and pragmatically in the input. Regarding syntax, factive verbs occurred often with interrogative clauses, while non-factive verbs think, guess, say occurred more often with declarative clauses. Tell’s distribution was ambiguous, matching suggestions that tell has both factive and non-factive uses. Regarding pragmatics, factive verbs showed higher proportions of indirect questions than non-factive verbs. The exception to this pattern was see, which will be discussed. We plan to expand the analysis to Mandarin corpora, to investigate the cross-linguistic viability of these bootstrapping hypotheses.

Much existing research suggests that the acquisition of Information Structure (IS) in general, and focus in particular, is a prolonged process, pointing to non-adult-like performance at preschool ages in comprehension. Our experiment investigated preschoolers’ focus interpretation, relying on the effect that focus on negation can have on logical scope interpretation. Potentially logically ambiguous negated disjunctive sentences were placed in two SCOPE scenarios (disjunction narrow vs. wide scope) in a focused-negation and in a neutral IS condition. Both adult controls’ and children’s results showed highly significant interaction between SCOPE and IS. Focus on negation shifted acceptance rates in the same direction from the baseline by almost identical degrees in the two age groups in both SCOPE conditions. These results strongly suggest that preschool children already have the competence to access prosodically marked information structure in comprehension, and they can exploit it in an adult-like manner to guide logical scope interpretation.

A common finding in metaphor comprehension research is that children tend to favor literal interpretations. Previously, children’s literalist tendencies have been treated as indicative of delayed metaphor comprehension development. We investigate whether this observed literalism genuinely reflects difficulties with metaphor comprehension or whether children may privilege literal interpretations when available. We assessed 3-to-7-year-olds’ metaphor comprehension abilities using different novel functional, attributional, and psychological metaphors. Preliminary findings suggest that when children were not provided with literal options, they could derive metaphorical interpretations successfully (Study1). However, when literal response options were available, children predominantly chose them over metaphorical interpretations (Study2). These initial findings challenge the previous assumption that young children’s literalism is solely attributable to a deficit in metaphor comprehension. Final results will be discussed, including response time and adult sample comparisons.

Children are known to derive more implicatures when the stronger alternative is made available through contrast or contextually relevant through a story/QUD. We extend this investigation to the exclusivity implicature of disjunction in child Romanian, an understudied language, targeting the prosodically marked sau ‘or’, sau…sau and fie…fie ‘either or’. Based on results from three experiments, we show that simply hearing unrelated conjunctive statements such as The deer chose a cake and a salad is not enough to boost implicatures. Instead, implicatures increase as an effect of both access to alternatives and explicit contextual relevance (expressed through the explicit conjunctive question Did the hen push the train and the boat?). This supports a Relevance-based Account, ascribing relevance an important role in implicature-derivation. Interestingly, the boost in implicatures can be seen only for sau and sau…sau, not for fie…fie, which is ambiguous between an inclusive disjunction and conjunction.

Acquiring multiple meanings for a word is often proposed to be difficult for word learners. However, the degree of difficulty may depend on the meanings: Floyd and Goldberg (2020) [4] demonstrated that word-learning is easier for both adults and 4- to 7-year-olds when words’ multiple meanings are related (i.e., polysemous, like “cap”) than unrelated (i.e., homophonous, like “bat”). However, all labeling events were referentially unambiguous, and word-learning was only tested after all labeling events completed. Thus, it is unclear if polysemy 1) benefits learning in more common, ambiguous contexts, and 2) helps learners infer new polysemous meanings across exposures. Here, we address those questions by using the same sets of related (i.e., polysemous) referents from Floyd & Goldberg (2020) in a cross-situational word-learning paradigm, and found that children showed limited ability to use prior meanings to infer polysemous mappings under ambiguity.

Conceptual combination is a hallmark of human cognition. Yet even the simplest of combinations produces a structure that has an ambiguous meaning. Consider a sentence like “The restaurant is efficient.” Humans represent numerous properties as true of restaurants – properties like serving food, having electricity, plumbing, foundations, wifi, overhead lighting, chairs and tables, menus, and restrooms. But to which of these properties does “efficient” apply? Considering each interpretation would be computationally costly. Prior research demonstrates that adults possess a strong bias to interpret sentences in terms of form properties—e.g., serving food for a restaurant —which are understood with (i) stability (e.g., restaurants will always serve food) and (ii) certainty (e.g., I’m certain a restaurant will serve food). Across two studies, with both familiar and novel kinds, we provide evidence this bias is present by age four. Given its early emergence, we highlight its significance for language acquisition and complex thought.

Even before infants speak, language exerts a powerful influence on their cognition. Listening to language supports 4-month-old infants’ ability to form object categories, while non-linguistic signals (backwards speech) do not. This language-cognition link is initially abstract enough to include language in the visual modality: American Sign Language (ASL). However, by 6 months, ASL no longer boosts cognition in hearing infants. Here we test whether brief exposure to ASL permits infants to maintain the link between Sign and cognition. 5- & 6-month-old infants viewed a 2-minute video of a woman signing a story in ASL, followed by a standard categorization task with ASL. Results show that brief exposure is insufficient for ASL to boost cognition in 5- & 6-month-old infants, indicating that longer and richer exposure is needed. Follow-ups will investigate the effect of prolonged and earlier ASL exposure, and will illuminate infants’ remarkable plasticity in early language and cognitive development.

This paper investigates the syntactic properties of adverbs in Italian children’s early productions. Theoretical studies have argued that adverbs are hierarchically merged along the clausal spine in a fixed order constrained by UG. This observation finds large support cross-linguistically, in adult languages, but has never been verified in child language. We filled this gap analyzing the spontaneous speech of 17 Italian monolingual children ranging from 1;04-3;04. We investigated whether children’s production of adverbs respect the adverbial hierarchy proposed in literature, and whether the adverbial hierarchy “grows” with the syntactic tree. Our findings suggest that adverbs placement is constrained by UG from the beginning of production, since adverbs are hierarchically merged along the clausal spine in a fixed order also in Italian children’s early production. In addition, our data suggest that the hierarchy is unraveled from the bottom to the top of the tree, in line with the Growing trees approach.

In the first year of an infant’s life, caregivers use labeling, pointing, gaze and other cues to communicate about referents in their environment. While there is some evidence from lab-based studies that young infants use referent-oriented cues during communication, naturalistic studies have found that young infants do not robustly attend to or leverage these cues during dyadic interactions. Using a uniquely dense longitudinal dataset of headcam recordings, the current study examined parent and infant referent-oriented behaviors around word utterances in one infants’ life between 6-12 months of age. We found that parent-infant-referent interactions change with development. Notably, the average proportion of referent-oriented cues during word utterances was lower than in previous studies, and there was substantial variability across individual words for all cues. Referent-oriented communication differs across contexts, and there is likely no universal trajectory of parent and infant behaviors that can be applied to all words.

This study presents an initial effort to integrate real language input, automatic parsing, and learning algorithms into a computational model, aiming to achieve a comprehensive understanding of the learning mechanisms involved in first language acquisition. The computational model provides a new tool to assess the development of syntactic knowledge by parsing and learning from real linguistic data. The learning trajectory of each syntactic parameter can be evaluated based on the linguistic input, allowing for an assessment of the correlation between input and learning. This also enables the comparison of the effects of different types of input on child language acquisition and reveals the learning order and trajectories of syntactic knowledge being investigated.