Schedule for BUCLD50

Previous studies report that children often misinterpret differential comparatives with numerals (e.g., “X has two more apples than Y”) as absolute quantities (“X has two apples”) or simple comparatives (“X has more apples than Y”). This study investigates whether such non-adult-like interpretations extend to comparatives with non-numeral scalar quantifiers (e.g., sukosi ‘a few’). Competing accounts yield different predictions: Arii et al.’s (2017) grammar-based account predicts various ad hoc strategies depending on the quantifier’s semantics, while Rosenstein et al.’s (2023) incremental more account predicts absolute interpretations regardless of quantifier type. Using a Truth Value Judgment Task with 20 Japanese-speaking children (4;6–6;9) and 23 adults, we found that children rarely assigned absolute interpretations to non-numeral differentials, favoring simple comparatives instead. These results support the grammar-based account and suggest that interpretive strategies depend on the scalar properties of the differential phrase—absolute interpretations require numerals, while vague quantifiers like sukosi favor simple comparative readings.

Children’s ability to learn new words depends not only on linguistic input but also on their ability to focus on relevant information while filtering out distractions. In real-world settings, word learning often takes place in the presence of auditory and visual distractions. Visual attention plays a critical role in this process, allowing children to establish links between words and their referents. Yet, little is known about how attention is deployed during word learning in environments that more closely resemble the complexity of natural learning contexts. This study investigates how young children allocate visual attention during word learning across four experimentally manipulated learning conditions: No Distractor, Auditory Distractor, Visual Distractor, and Audio-Visual Distractor. Using eye-tracking, we explore how distraction influences children’s gaze behavior and how these attentional patterns relate to their success in learning novel words.

Prior research has shown that preschool children sometimes interpret a positive disjunction (e.g., P or Q) similar to a positive conjunction (e.g., P and Q). The literature has considered three possible explanations: 1. Non-adult-like pragmatic strengthening; 2. task artefacts; and 3. non-linguistic default interpretation of unknown connectives. More specifically, Truth Value Judgment Tasks (TVJTs) with two alternatives and trivially true disjunctions have been shown to increase children’s conjunctive interpretations. We tested Bengali-speaking children on their interpretation of positive conjunction and disjunction in a novel give-item task that avoids previously discussed experimental artefacts. We find that some children still interpret positive disjunctions similar to a conjunction. Our study provides support for the hypothesis that conjunctive interpretations of disjunction are due to default biases on the interpretation of an unknown or difficult-to-comprehend coordination, potentially exacerbated by task demands.

Quantifiers such as “more” operate over various domains of magnitudes. While many studies have documented children’s comprehension of “more” in the domains of number and area, very little is known about their quantification of time. Children are slow to map words onto temporal magnitudes (e.g., “minute”), which could suggest that applying the meaning of “more” to the temporal domain may pose a challenge in acquisition. To investigate this possibility, the current study investigated children’s understanding of the comparative “more” for temporal durations, as well as numerical quantities. We tested 60 children (ages 2 ½ to 4) and found that children as young as 3 years old successfully quantified duration, despite struggling to map lexical items onto temporal magnitudes at this age. However, children quantified numerical quantities better than temporal durations, leaving open the possibility that they may not simultaneously apply “more” to all the dimensions it quantifies.

Infant’s language development is inherently tied to their linguistic input, but the majority of research investigating infant’s input has focused on adult caregiver speech (e.g. LENA measuring only adult word counts). Here, we measure the frequency and content of speech from children for infants with older siblings in two types of naturalistic recordings: daylong recordings from Homebank, and in-lab play sessions. We find that 37% of infants with older siblings’ input from high-volubility portions of daylong recordings, and in-lab play sessions, comes from children. In triadic interactions between an infant, older sibling, and parent, only ~20% of total speech is infant-directed. However, overheard speech to and by siblings shares properties with directed speech, younger siblings may be able to learn from overheard speech more readily than eldest children, who are more likely to experience overheard speech between adults.

Languages consistently distinguish between agents (doers) and patients (receivers) of actions, raising the question of whether these roles reflect innate conceptual understanding or are learned through language exposure. Prior research suggests these categories may be conceptual universals. If so, they should appear in individuals without access to linguistic input. To test this, we conducted a concept-learning task with homesigners—deaf individuals who lack access to a conventional language—and compared their performance to that of English-speaking 4–5-year-olds, who have access to a language marking these roles but little formal education. In the task, participants viewed animated events where one character acted on another and inferred a rule about which character (agent or patient) would be selected. Both groups successfully learned and generalized the rules to familiar event types, but only children showed consistent generalization to novel event types. These findings suggest that agent-patient concepts may emerge independently of language.

Recent studies suggest that a third language (L3) can regressively influence a second language (L2) and enhance L2 representations when the two languages are similar. The present study examines how L3 Spanish affects the acquisition of L2-English causative psych verbs (e.g. worry) which are present in Spanish and English but absent in Thai. In Thai, only periphrastic causatives (e.g. make someone worried) are allowed, the structure that also exists in English and Spanish. Participants included L1-English, L1-Thai L2-English, and L1-Thai L2-English L3-Spanish speakers, who completed an acceptability judgment task; the L1-Thai groups also completed a vocabulary translation task. While no statistically significant difference between the Thai groups emerged, descriptive data revealed a trend of cumulative L3 effects: L3 speakers rated lexical causatives higher than periphrastic ones aligning with data from the baseline pattern. The L3 group also outperformed the L2 group in translating causative psych verbs.

Disfluencies in natural speech often precede low-frequency words and occur in complex utterances. While previous research has shown that disfluencies influence language development and processing, the specific roles of child-produced and adult-produced disfluencies in naturalistic settings remain underexplored. This study analyzed 44 monolingual English-learning children over 1.25 years (age = 1;9–3;0) to examine disfluencies and language development in preschoolers during naturalistic adult-child play. Results indicate that both child- and parent-produced disfluencies significantly enhance children’s lexical development (e.g., verbs-per-utterance, parent disfluency: β=0.17, p<0.05; child disfluency: β=0.06, p<0.05), with their interaction amplifying verb production and overall vocabulary diversity. These effects vary developmentally, showing stronger impacts at certain age points. Moreover, the distribution of disfluency types differed between adults and children, highlighting their different roles in communicative functions. These findings provide novel evidence for the facilitative role of disfluencies in early vocabulary development, with important implications for understanding the mechanisms of language acquisition.

We asked whether goals affect how children, as well as adults, determine when an event ends. In Exp.1, 4-5-year-olds and adults read an agent’s goal (e.g., “Jessie wants to eat an orange with her breakfast to make it healthier” or “Jessie wants to peel the orange”) and then answered a question about an image that appeared (“Did she peel the orange?”). Adults, but not children, were more likely to accept a partly complete outcome as indicating a culminated event if the outcome satisfied the agent’s goal. In Exp.2, participants watched videos with two people engaged in a social interaction. A person performed an action (e.g., partly cleaning a table) that was either helpful or unhelpful with respect to the goal of the other person (setting a small book on a table). Our results show that even young children use intentionality to place boundaries on simple physical events.

We evaluated whether a computational model trained on prosodically-segmented utterances can effectively distinguish between stimuli with high and low phonotactic probability, using behavioral data from English-speaking 5-month-olds—who successfully make this distinction—as a benchmark. Using audio recordings of infant-directed speech from the Providence Corpus, we created three corpora: one with gold-standard word segmentation, one with prosodic segmentation, and one unsegmented. Prosodic boundaries were identified by automatically extracting prosodic features within utterances. We applied Bayesian logistic regression models with k-fold cross-validation to compare how accurately phonotactic probabilities from each corpus distinguished the experimental stimuli. Only phonotactic probabilities from the gold-standard word-segmented corpus outperformed the null model; those from the prosodically segmented and unsegmented corpora did not. We will compare these results based on automatic prosodic annotation with results obtained from human tagging of large prosodic boundaries to identify the sequences from which 5-month-olds extract phonotactics.

This study investigated whether newborns show different neural responses to their native versus a rhythmically distinct non-native language, and whether this differs for infant-directed (IDS) and adult-directed speech (ADS). EEG was recorded from 60 Czech-learning newborns during sleep as they listened to stories in Czech and Russian. Neural tracking was assessed via total power and inter-trial phase coherence (ITPC) in delta (word-level rhythm) and theta (syllable-level rhythm) frequency bands. Results showed greater total power for native speech, especially in the delta band, suggesting stronger word-level tracking of the native language. There was also a trend toward higher ITPC in the theta band for non-native speech. These findings suggest that language-specific neural tracking is present at birth, supporting early sensitivity to native-language prosody.

Is early age of exposure crucial for successful ultimate attainment? To answer this, we tested 30 Turkish heritage speakers (HS) in the US, as well as 24 second language (L2) learners, 34 Turkish-American returnees and 30 Turkish monolinguals residing in Turkey. All participants completed an Acceptability Judgement Task and a Sentence Repetition Task on passives, relative clauses and anaphora in Turkish. Results showed that i) returnees patterned with monolinguals outperforming HS and L2ers in both tasks, and ii) correlations between accuracy, age of return to Turkey, length of residence in Turkey, and contact with English/Turkish after return of returnees and L2ers revealed that the earlier they returned to Turkey, the better they performed in both tasks. This suggests that complex morphosyntax is still malleable post-puberty, and that early age of exposure is crucial but not sufficient for successful ultimate attainment; maximal input is also required (either before or after puberty).

Previous work has found that children interpret the words “two” and “three” with a same-kind bias. Importantly, child’s knowledge of verbal numerals predicted bias strength. This invites important follow-up questions, including whether the same-kind bias 1) is specific to small digit words, 2) is specific to production, and 3) whether learning verbal numeral syntax predicts bias strength. To answer these questions, we devised a 2AFC (two-alternative forced-choice) task using Lookit, a virtual platform for asynchronous developmental research. We recruited 69 native-English speaking children (Mage=6.45 years, range=3.3-9.9 years old) to participate in our pre-registered study (https://osf.io/sbf7g). We found that children selected same-kind sets above chance for digits (t(68)=2.32, p=.02, d=.40), multipliers (t(68)=3.95, p<.001, d=.67), and complex phrases (t(68)=3.14, p=.002, d=.53); however, we failed to detect a difference in bias strength based on age. This suggests the same-kind bias pervades all number-related linguistic categories regardless of syntactic mastery.

We examined comprehension of complex sentences in adolescent Spanish heritage speakers, investigating to what extent gender and number features of the NPs facilitate comprehension of relative clauses in Spanish. In a picture-sentence matching task with sentences presented auditorily, we compared subject relative clauses (SRCs) with object relative clauses (ORCs) and passive relative clauses (PRCs) distributed in four grammatical feature conditions. Twenty English-dominant adolescent Spanish heritage speakers were tested as well as their Spanish-speaking parents. The results indicated that comprehension of SRCs was more accurate than PRCs and ORCs. For both adolescents and adults, ORCs with number mismatch in the NP were comprehended more readily than those with matching number features, while gender features did not make a difference. Results on PRC showed effects of grammatical feature dissimilarity in the NP for both mismatch in number and gender for the adults, whereas the adolescents did not show the same effects.

An extensive literature links the development of Theory of Mind (ToM) with language acquisition in early childhood. This study examined whether these domains remain closely associated in adulthood and whether they are predictive of individual differences in online social reasoning. College students at a nonselective public college (N = 89, M age = 20.6) completed the Peabody Picture Vocabulary Test as an indicator of receptive vocabulary knowledge, a nonverbal ToM comic strip task, and the Youth Media Literacy Inventory and Parasocial Relationships Scale as indicators of online social reasoning. A path model revealed a strong association between vocabulary and ToM. Higher vocabulary and ToM scores were associated with greater online media literacy and fewer and less intense parasocial relationships. These findings underscore the continued significance of language skills in facilitating ToM in adulthood and their cascading influences on disparate facets of online social reasoning.

This study investigates whether age of acquisition (AoA) affects the use of referential loci conveyed through pointing signs in Turkish Sign Language (TİD). 866 pointing signs were annotated in ELAN across syntactic, morpho-phonological, and prosodic tiers and categorized into four grammatical categories: Locatives, Pronouns, Demonstratives, and Clitics (Weak Demonstratives). Pose Estimation was used to calculate the Euclidean distance covered by each pointing sign. While overall pointing frequency did not differ significantly between native and late signers, the categorical distributions did, with Clitics produced more frequently by late signers. Additionally, native signers used loci more often in Demonstratives and articulated them with greater spatial distances. The more restricted spatial use in Demonstratives by late signers further support a tendency towards phonologically weaker forms and thus explain the higher rates of Clitics. Overall, the results imply that late language exposure may affect how referential systems such as loci are employed in TİD.

This study examines developmental changes in fluency during early second language (L2) acquisition, focusing on the relationship between productivity and disfluency in child L2 English learners. Using longitudinal data from the CHILDES English-L2 Paradis Corpus, we tracked 17 children aged 3-7 across multiple collection rounds. Fluency was analyzed through breakdown disfluencies (e.g., filled pauses, repetitions, repairs) and productivity measures: Mean Length of Utterance in words (MLU(w)) and Type-Token Ratio (TTR). While MLU(w) generally increased with exposure, disfluency rates did not consistently decline and often peaked at intermediate stages. Children producing longer utterances were often more disfluent, highlighting planning challenges during periods of rapid growth. Clustering analyses revealed distinct developmental profiles, including children who remained highly disfluent despite limited MLU(w) growth. These findings align with dynamic systems and usage-based models, emphasizing that increased productivity does not necessarily lead to greater fluency in early stages.

Both sensorimotor exploration and caregiver-child interactions support language and learning in young children (e.g., Yuan et al., 2019; Tamis-Lemonda et al., 2014). While exploration and caregiver-child interaction are typically studied separately, they often co-occur during shared play. We hypothesized that elements of the caregiver-child interaction, such as language input and gesture, would influence exploration itself. To test this, we recorded caregiver-child interactions (N = 25) while they explored an exhibit in a children’s museum. We transcribed all caregiver utterances, identified instances of pointing, showing, attention-getting, and physical guiding, and coded the depth of exploratory behavior (on a 0-3 scale) in 30 second intervals. We observed that within an interval, showing by both caregivers and children predicted deeper exploration, while caregiver utterances predicted more shallow exploration. Showing may therefore indicate that an object is worth exploring, while language is either interfering with exploration or unnecessary for encouraging it.

We examined how non-adjacent dependencies are represented and used in early child grammar. In Experiment 1 French-learning 14- and 19-month-olds were familiarized with French sentences containing a HAVE-auxiliary followed by a novel word suffixed with the frequent past-participle -é, e.g., Elle a migé (“She has miged”). Test stimuli were the novel words in non-suffixed forms in new syntactic contexts: Grammatical trials (e.g., Tu miges “You mige”) versus Ungrammatical trials (e.g., Le mige “The mige”). Results show that both ages discriminated the test trials, indicating that they used the non-adjacent dependencies for verb categorization. When the past participle of the familiarization sentences was changed to a pseudo-suffix -ou, 14-month-olds failed to categorize the novel verbs (Experiment 2), whereas 19-month-olds succeeded in the categorization task (Experiment 3). Thus, younger infants tracked specific frequent non-adjacent dependent items; older infants went beyond specific items and represented them as dependency classes, revealing sophisticated structural knowledge.

Acquiring multiple meanings for a word is often difficult. However, that difficulty may depend on the meanings’ relationship: adults and 4- to 7-year-olds learn multiple meanings better when words’ multiple meanings are related (i.e., polysemous, like “cap”) than unrelated (e.g., “bat”). However, it remains unclear how learners infer polysemous meanings when encountering these words in referentially ambiguous contexts, as they might in everyday experience. In two studies, we examined children’s and adults’ polyseme learning under uncertainty, using artificial stimuli from prior work (Study 1) and attested non-English polysemes (Study 2). We find developmental differences in the ability to infer polysemous meanings across exposures: while adults do so successfully for both artificial and natural polysemes, children fail except in cases of clear, shape-based similarity.

This study investigates third-person pronoun use in monolingual and bilingual Russian-Hebrew-speaking children, focusing on null versus overt forms. Russian broadly allows null arguments due to discourse-drop properties, while Hebrew permits both syntactically-licensed and discourse-driven argument omission, albeit in a more restricted manner. We examined whether bilinguals show cross-linguistic influence and differ from monolinguals. Ninety-seven children and forty adults were tested via a sentence repetition task and a pronoun elicitation task. Bilinguals scored lower in morphosyntactic ability and exhibited asymmetric influence: they overused overt pronouns in Russian but patterned with Hebrew monolinguals. Age predicted increasing use of overt subjects in Hebrew and null subjects in Russian. Monolingual children in both languages overused overt subjects, supporting the Interface Hypothesis regarding syntactic-discourse integration challenges. This study highlights prolonged acquisition paths in both monolingual and bilingual children and uniquely explores bilinguals navigating two null-subject languages governed by distinct referential constraints.

During the word learning process, children map phonological forms to meanings, often relying on surface-level cues in the syntactic environment in which a word is found in order to narrow down its possible meanings—a process referred to as syntactic bootstrapping. For certain grammatical categories, like adjectives, a given frame ranges in its informativity, such that learners benefit from tracking an adjective across multiple environments. Recent research by Gotowski & Syrett (2024) with the Human Simulation Paradigm (HSP) found that adult learners successfully recruit clusters of frames to learn novel adjectives. We build on their study by adapting the HSP for children 7-9 years of age. We find that children pay attention to distributional information in order to successfully subcategorize, and routinely offer frame-compliant guesses for novel adjectives, mirroring what has been reported for the adult population.

The Theory of Lexical Refraction claims that the size of children’s overall lexicon is related to how well they draw scalar implicature interpretations. Findings from child Spanish and child English are consistent with this finding. We explore whether the same relationships will hold in a typologically distinct language, Imbabura Kichwa, a northern Quechuan language spoken in Ecuador. 48 bilingual 4- to 8-year-old Kichwa-Spanish-speaking children are given a Truth-Value Judgment Task. We also calculate a Number of Different Words (NDW) lexical measure and a Mean Length of Utterance, measured in words (MLUw), morphosyntax measure, both from a spontaneous language sample. Even the youngest child Kichwa-speakers in the sample generate scalar implicatures. Neither the morphosyntax measure nor the lexical measure significantly associated with implicature interpretations. We discuss these results in light of the distinctions among English, Spanish and Kichwa and how quantity is expressed in each language.

It is common for heritage languages in the diaspora to be lost within three generations. Yet the Old Order Amish in the U.S. have successfully maintained languages descended from varieties brought by their ancestors from German-speaking Europe 200-300 years ago. These languages–Pennsylvania Dutch and Shwitzer–are only rarely written down, not taught in school, nor subject to governmental support or protection. In addition to being fluent speakers of their heritage language(s), Amish adults are typically fully literate and fluent in English. Meanwhile, little is known about Amish children’s dual language development, nor about the language practices in their homes that support it. Based on recent fieldwork with Amish communities in Indiana, we report data from a bilingual object naming task conducted with 23 children, aged 5–17 years, and parental questionnaires/interviews with related caregivers. These observations demonstrate early and (mostly) stable bilingualism and exemplify the community practices supporting it.

This study investigates how individual differences (ID) shape online heritage language (HL) processing in adolescent Mandarin-English bilinguals, focusing on three Mandarin pronouns: PRO (ta), SE (ziji), and SELF (taziji). Using a visual world eye-tracking paradigm, 50 participants listened to sentences with these pronouns while viewing scenes with potential referents. Measures of working memory (WM), inhibition, and language exposure/use were collected. Group-level results showed LD preferences for PRO and SE, and local preference for SELF. WM positively predicted LD interpretations of PRO and SE, while inhibition also facilitated LD binding of SE. In contrast, SELF processing was only influenced by Mandarin exposure/use. These findings suggest that processing syntax-based dependencies (SELF) relies more on input, whereas dependencies requiring integration of syntax, semantics, and pragmatics (PRO, SE) benefit from stronger cognitive resources. The study highlights how child-internal and external factors differentially impact HL processing across linguistic domains in adolescence.

The current study directly compares the strength of a hierarchical predisposition within a controlled artificial language-learning task as well as a non-linguistic, puzzle-solving task. In a 2×2 design, participants completed a pattern deduction task, in which they attempted to “escape from an escape room”. Participants learned sequences of either words (language context) or shapes (shapes context). Participants were shown transformed sequences consistent with one of two constituency patterns: a linguistically attested pattern (hierarchical pattern), or a systematic but linguistically unattested one (linear pattern). We find a hierarchical bias in both language and shapes contexts, suggesting domain-general preference for such patterns. However, we found the bias is stronger in a non-linguistic context than in a linguistic context. It’s possible that a stronger hierarchical bias for language emerges when a learner is linking forms and meanings. Future work will explore whether introducing meanings influences the biases of participants in a language-learning context.

Infants make great strides in learning to map continuous speech signals to discrete phonemic categories by 12 months old. The proto-lexicon account of phonemic category acquisition posits that this is predominantly driven by word learning. We test this theory by building a computational model representative of it. The model’s objective is to keep different word exemplars apart from one another in psychoacoustic space, except when it is less costly to store any of them as homophones. Phonemic categories then emerge as boundaries between words that generalize well to new words. We tested this model on unseen words on opposite sides of phonemic contrasts and compared its performance to Whisper, a state-of-the-art transformer model trained on 680,000 hours (~78 years) of labeled speech data. When our model has acquired about 50 words, it begins to outperform Whisper on the task and phonemic categories appear to solidify.

Over the past three decades, Iceland has transformed from a homogeneous, predominantly monolingual nation into a diverse, multinational society due to rapid societal changes and increased immigration. Consequently, bilingual and multilingual language acquisition has become significantly more common, with 15.5% of students aged 6-16 in Icelandic schools being bilingual in 2023, up from 3.7% in 1998. Increased exposure to English, partly through electronic devices, has created a complex linguistic environment where bilingual children learn both Icelandic and English simultaneously. Studies indicate that bilingual children in Iceland lag behind their monolingual peers in Icelandic proficiency, particularly in vocabulary, grammar, and syntactic complexity. This project aims to investigate the Icelandic language proficiency of sequential bilingual children in primary school and assess the effectiveness of the “Talking Partners Primary” intervention method. Preliminary results show significant improvements in vocabulary, morphological inflection, and sentence structure, highlighting the benefits of targeted language support for bilingual children.

This study examines the influence of non-salient Differential Object Marking (DOM) and canonical word order on the real-time processing and comprehension of Spanish relative clauses—comparing center-embedded and right-branching structures—among heritage speakers and L2 learners. Grounded in the Good-Enough Processing framework, we test whether bilinguals rely on shallow interpretations when faced with complex or non-canonical structures. Sixty adults—heritage speakers, late L2, and L1 speakers—completed a self-paced reading task. Results showed stronger processing effects in center-embedded relatives (DOM: β = 0.19, p < 0.001; DOM × Word Order: β = -0.22, p < 0.001) compared to right-branching clauses. Heritage speakers outperformed L2 learners in comprehension (OR = 0.48, p < 0.001), though neither group matched native performance. DOM enhanced accuracy overall (OR = 2.10, p < 0.001). Despite processing difficulties, offline comprehension did not differ significantly by branching type (OR = 1.02, p = 0.769), suggesting reanalysis after initial shallow parsing.

Convenience sampling in developmental science is well documented, yet barriers to broader participation remain under-investigated. This study examined structural and individual factors influencing research participation among 310 families recruited from diverse avenues in [location] between 2022–2025. Families were invited to enroll in a multi-session study examining SES variation in language development. Fifty-nine families completed onboarding and selected from activities varying in anonymity and effort. Recruitment yields differed by avenue: the university-database and preschool were highest, and ads and Head Start centers were lowest. Higher-SES families were more likely to participate across activities, particularly those requiring more time or effort, while race/ethnicity was unrelated to task completion. Most families completed face-recorded sessions, suggesting that related privacy concerns were not a barrier to participation, while less families completed parent interviews and LENA recordings. Findings demonstrate that structural-individual interactions generate a gradient of research participation, raising questions about convenience sampling limitations and procedures.

Cross-linguistic structural priming (i.e., exposure to a syntactic structure in one language facilitates a similar structure in another language) has been proposed as a mechanism driving contact-induced language change. This study investigates such priming in Canadian French-English bilinguals, focusing on innovative French reciprocal structures missing the reflexive se (e.g., *Ils (s’)embrassent ‘They kiss’), which are grammatical in English. Seventy-three participants completed a self-paced-reading task, where French innovations were preceded by either structurally similar or structurally alternative (control) English primes. Participants also completed production tasks before and after the experiment. Results showed a significant priming effect, with faster reading times for the French innovation after similar English primes, but only in the first half of the experiment. Although more French innovations were produced after the experiment, this tendency was not significant. Overall, cross-linguistic priming might facilitate the processing of innovations (at least at initial stages), but production might require more time.

Children organize concepts into categories, which supports vocabulary growth. Previous studies yielded conflicting evidence on whether category representation is affected in children with language delay (i.e. Late Talkers, LTs). Here, we captured repeated parental reports of vocabulary in 151 toddlers every three months from 18- to 30-months of age. We assigned children to either LT group, late-blooming group (LBs), or typically developing (TDs) group. We constructed semantic networks and measured their modularity (a direct measure of how well a network can be divided into categories). While all three groups organized their vocabulary into increasingly structured categories across time, category representation was most differentiated in TDs, followed by LBs, and is least distinct in LTs. This effect is most evident during early stages (vocabulary sizes of 0-200 words). Overall, we found evidence that semantic representation in LT children is structurally different, marked by less differentiated semantic networks, less coherent categories.

Researchers have long debated whether the difficulty learners have with English third-person-singular (3SG) -s is attributable to processing factors or representational deficits. Among processing accounts, the Sentence Position Principle (SPP; Barcroft & VanPatten, 1997) argues that a morpheme’s position influences its perceptual salience. Sundara et al. (2011) reported that sentence-medial -s is harder for L1 children than sentence-final -s, but no studies have probed for similar effects among L2ers. This study examined real-time perception in an eye-tracking task and receptive knowledge in a fill-in-the-blank task. Eye-tracking results from 24 L1-English controls and 23 adult L1-Korean L2-English learners showed both groups were slower to perceive sentence-medial -s than sentence-final -s, an outcome supporting the SPP. While L2ers had slower morpheme perception and lower picture-selection accuracy than L1 speakers, they performed at ceiling in the fill-in-the-blank task, suggesting their difficulty with 3SG -s stems from processing considerations rather than a knowledge deficit.

Both definite and demonstrative descriptions mark uniqueness, but they differ in the domain of uniqueness computation: definites apply to the whole discourse, whereas demonstratives are restricted to the salient situation. Ionin et al. (2012) demonstrated that L1-Korean L2-English learners failed to distinguish these determiners in certain contexts, attributing this difficulty to L1 transfer—speakers of Korean, an article-less language, treat L2 definite and demonstrative descriptions as equivalent to their L1 demonstrative. To investigate whether this phenomenon constitutes true L1 transfer, this study replicates Ionin et al. by testing L2-English learners from two different L1 backgrounds: Japanese, an article-less language with demonstratives, and Spanish, a language with both articles and demonstratives. The results of an elicited production task suggest that Spanish speakers differentiate the definite and demonstrative descriptions in a target-like manner, whereas Japanese speakers do not. These findings align with Ionin et al., indicating L1 influence on L2 determiner choices.

Languages consistently distinguish between agents (doers) and patients (receivers) of actions, raising the question of whether these roles reflect innate conceptual understanding or are learned through language exposure. Prior research suggests these categories may be conceptual universals. If so, they should appear in individuals without access to linguistic input. To test this, we conducted a concept-learning task with homesigners—deaf individuals who lack access to a conventional language—and compared their performance to that of English-speaking 4–5-year-olds, who have access to a language marking these roles but little formal education. In the task, participants viewed animated events where one character acted on another and inferred a rule about which character (agent or patient) would be selected. Both groups successfully learned and generalized the rules to familiar event types, but only children showed consistent generalization to novel event types. These findings suggest that agent-patient concepts may emerge independently of language.

This study investigates how English native speakers and L1-Korean learners of English (L2ers) interpret empty subjects in adjunct control constructions (e.g., Susan paid Matthew [Ø to get tickets]). While native speakers consistently show a strong object-control preference across offline and online tasks, Korean L2 participants exhibit more variable patterns. In offline comprehension (Experiment 1), L2ers mirror native speakers’ object-control bias. However, in self-paced reading tasks (Experiments 2 and 3), their interpretations shift, with increased subject-control choices and greater variability, particularly when the object and the infinitival clause are separated by an adverbial phrase. These findings support the Shallow Structure Hypothesis, which suggests L2 learners underutilize syntactic information during real-time processing. Notably, even highly proficient Korean L2 participants showed reduced sensitivity to structural cues, indicating persistent challenges in L2 processing. Ongoing research will test lower-proficiency L2 learners to further explore how proficiency influences referential interpretation.

Despite increasing diversity in UK schools, black children remain underrepresented in children’s literature, with only 7% of books published in 2023 featuring ethnic minority main characters (CLPE, 2024). This study examined whether exposure to representative literature impacts self-esteem and reading enjoyment.

This study investigates how native (L1) and second language (L2) readers process garden-path sentences and whether contextual cues help resolve ambiguity in real time. In a self-paced reading task, 6 L1 English speakers and 24 L1-Chinese learners read temporarily ambiguous sentences preceded by either matching or mismatching contextual information. Reaction times at disambiguation regions and follow-up comprehension accuracy were measured. Results showed that L1 readers used contextual cues during online processing, consistent with the Constraint-Based Model. In contrast, L2 readers showed no significant context effect, aligning with accounts of cognitive limitations in L2 processing. Neither group showed improved accuracy in offline comprehension, suggesting persistent misinterpretation. These findings highlight differences in context integration between L1 and L2 readers and support the need for instructional approaches—such as Processing Instruction—that target more effective use of processing strategy during sentence comprehension.

Some research suggests that children’s early vocabulary develops gradually, while others show that words can be acquired discreetly. To explore whether this divide reflects the tools used to assess vocabulary, we examined whether a graded continuum underlies binary parent report by comparing ratings on the standard CDI and a continuous adaptation. In the continuous measure, parents rated their certainty (1–6) that their child understands and/or produces 30 CDI words. Parents of 81 monolingual and bilingual children completed both forms. Results showed that comprehension ratings were more graded than production, and parents had a high certainty threshold to endorse comprehension and production on the CDI. Findings suggest that binary measurement may obscure underlying graded knowledge, particularly for word comprehension. Moreover, variation in parents’ certainty thresholds may partially explain cross-linguistic and cross-cultural differences in CDI vocabulary sizes. Ongoing analyses are investigating how binary and continuous vocabulary ratings relate to looking-while-listening performance.

Infants’ vocalizations are influenced by social context, but less is known about how this manifests in blind infants, who lack access to visual social cues. We compared 10 blind and 10 sighted infants (ages 6–15 months) using daylong audio recordings and manually annotated randomly sampled segments. Vocalizations were analyzed across contexts with and without adult speech. Crying did not differ significantly between groups. However, blind infants vocalized less overall and showed a larger change in vocalizations when others were present, compared to sighted infants. In detailed annotations, sighted infants produced canonical babbling at consistent rates across contexts. In contrast, blind infants were less likely to babble canonically when alone, but this likelihood increased in the presence of others. These findings indicate that blind infants’ vocal behaviors are more sensitive to the presence of audible social cues, pointing to differentiated engagement with their social environment even in the absence of vision.

Previous research suggests that children do not always effectively use the lexical context to identify noun referents. We explored this issue by examining Mandarin-learning 30-month-olds’ interpretation of novel nouns co-occurring with familiar adjectives or numerals. Toddlers participated in an intermodal preferential looking experiment, where they were asked to identify the reference of novel nouns after hearing the nouns co-occurring with either familiar adjectives (‘big’/‘small’) or numerals (‘one’/‘two’). Results: Toddlers failed to identify novel noun referents cued by familiar adjectives, but succeeded with the numeral ‘two’, though not with ‘one’. Adjectives may be difficult to exploit because their semantic interpretation requires access to a comparison standard; and ‘one’ may be harder because it can be grammatically ambiguous (i.e., indefinite determiner or numeral), whereas ‘two’ functions as an unambiguous numeral. Overall, these findings suggest that children’s conceptual-semantic and grammatical knowledge modulates their ability to use familiar words in resolving novel noun reference.

Children typically prefer basic-level (e.g., dog) over subordinate (e.g., dalmatian) meanings in word learning. Learning subordinate nouns is known to be facilitated by contrast; for example, when shown different exemplars with different labels, learners infer narrower meanings for each label. We tested whether this reflects a pragmatic inference on the expectation for speakers to be more informative given relevant alternatives, versus simple lexical heuristics. In Experiment 1, 4-5-year-olds and adults learned novel words, manipulating the presence of contrasting alternatives at the subordinate level. Both groups showed more subordinate generalizations when alternatives were present and linguistically negated, even without competing labels. In Experiment 2, Participants watched scenarios which manipulated the speaker’s visual access to the contrasting alternatives before labeling the target. Both age groups made more subordinate inferences when alternatives were also visible to the speaker. Findings show children’s ability to integrate contrast inference with a epistemic reasoning during word learning

The quality and quantity of parental input are key predictors of early bilingual development. Understanding how early input shapes language outcomes is critical for supporting bilingual development, yet the effects of parental language mixing and separation strategies remain unclear—especially in culturally mixed households. This online study examines how parental language use and beliefs relate to lexical development in bilingual children aged 18–36 months in Paris, France. Monolingual and bilingual families completed the MacArthur-Bates Communicative Development Inventories (M-CDI), with bilingual families also reporting on language practices and attitudes. Preliminary data indicate that parents of bilingual children did not report significantly lower M-CDI production scores compared to monolingual peers. Families reported wide variation in parental code-switching behaviors, challenging assumptions of uniform bilingual input. Planned analyses will test whether children’s language mixing correlates with parental practices, contributing insight into the types of language environments that may best support vocabulary development in bilingual children.

Children distinguish belief verbs (know, think) from desire verbs (want, need) by age 3, but it is not yet clear which syntactic/semantic cues they use to do so. Using corpus data from European Portuguese and English, we examine two potential strategies: (a) tracking whether the attitude verb complement has features resembling declarative main clauses (a signature of belief verbs), and (b) tracking the temporal orientation of the complement relative to the attitude (as previously proposed for distinguishing epistemic/non-epistemic modals). We find that in European Portuguese, belief verb complements (e.g., for saber ‘know’) are not always differentiated from desire verbs in terms of main clause features, casting doubt on this account. In contrast, temporal orientation provides a robust cue across the two languages: Desire verbs are consistently future-oriented, while belief verbs have flexible orientation. This suggests that cross-linguistically, temporal orientation may be a more robust cue to the belief/desire split.

Visual statistical learning (VSL) correlates with literacy skills in hearing populations, and within-individual variability between linguistic and nonlinguistic SL domains is seen. However, current research in Deaf individuals focuses on nonlinguistic SL, with limited work on concurrent reading skill relations. We ask how deafness impacts linking and retention of linguistic and non-linguistic regularities in visual artificial languages. 22 Deaf and 18 hearing signing adults (CODA) were matched with hearing non-signers on literacy and nonverbal intelligence. All completed linguistic (letter) and nonlinguistic (image) VSL tasks. Deaf individuals didn’t differ from hearing-matched counterparts in either domain, and CODA only outperformed hearing-matched counterparts in non-linguistic. Only the Deaf group showed tight coupling across VSL domains, modulated by higher English literacy and earlier sign language acquisition. Thus, deafness alone does not affect VSL performance, as literacy interplay is seen. Instead, deafness uniquely connects linguistic and non-linguistic visual regularities, and visual language experience strengthens this.

This study tested whether early-acquired communicative functions of negation, like rejection and prohibition, help scaffold children’s acquisition of the concept of negation. We examined the emergence of logical denial negation in transcripts of children from CHILDES aged 16-30 months speaking Spanish, English, German, and Hebrew, which have varying mappings of negation words to early-acquired negative functions. If early negative functions support children learning the concept of negation, we predicted faster acquisition of denial in Spanish and English, in which one word expresses both those early functions and logical denial, then German, then Hebrew. If, instead, learning negation words is merely limited by their semantic transparency, we predicted faster acquisition in Spanish and Hebrew, in which the more transparent particle negator shares a lexeme with the more opaque sentential negator. We found more early denials in Spanish and Hebrew, supporting a linguistic, not conceptual, limitation on children’s acquisition of negation.

Seenku displays a complex pattern of tonal alternations on syntactic heads triggered by the internal argument, dependent upon phonological, morphological, and syntactic information. This talk presents a preliminary analysis of L1 acquisition of Seenku’s sandhi system, drawing on data from four children between the ages 1;8 to 4;0. Remarkably, even the youngest child (1;8) shows evidence of sandhi alternations. The rate of errors across each obligatory construction (irrealis verbs, postpositions, inalienable possession) is largely comparable, but even older children consistently fail to apply sandhi in an exceptional alienable construction. We also note that the error rate with phonologically light arguments outpaces that of heavy arguments, which have more regular tonal alternations. We hypothesize that young children learn postpositions, inalienable nouns, and verbs as a network of surface allomorphs, as they never appear without an argument, while alienable nouns appear more commonly outside of possessive constructions, thus reinforcing their lexical tone.

Heritage speakers (HSs) grow up in dual-language environments with variable exposure to the heritage language. Building on evidence that linear distance between agreement elements and Age of Onset of Bilingualism (AOB) modulate sensitivity to gender agreement violations in HSs (Keating, 2022; 2025), we conducted an eye-tracking study. English-dominant HSs (n = 53) and monolingual Spanish speakers (n = 32) read sentences containing gender agreement violations under two conditions: No Cue (an intensifier increased distance) and Cue (a gender-marked determiner provided an early cue). Early and late measures of agreement processing were extracted. In the No Cue condition, greater sensitivity in later measures was found for early sequential HSs (AOB 4–6) but not simultaneous HSs (AOB 0–3), with proficiency modulating effects. In the Cue condition, determiners elicited earlier detection across groups, and proficiency predicted sensitivity in total reading time. These findings highlight how bilingual trajectories and reliable morphosyntactic cues influence processing.

We investigated whether 7-11 year old Spanish and English monolingual and Spanish-English bilingual children interpret pronouns (him) and reflexives (himself) in the two languages as adults, in NP and PP complements. Since pronoun interpretation problems are common in English, this study asked whether knowledge of Spanish clitics would boost comprehension of English pronouns? Results of a picture-sentence matching task that the interpretation of strong pronouns in PPs was more variable than in NP for all groups in the two languages. Age effects were found in Spanish but not in English, and heritage speakers were more accurate than L2 learners. In stimuli contrasting pronouns and reflexives there was more variability in English than in Spanish, suggestive of a pronoun interpretation problem in English. Intriguingly, the L2 children showed higher accuracy in English than other groups, suggesting that learning Spanish clitics and pronouns may boost comprehension of English pronouns in this group.

Processing negation is challenging for children because of various semantic and pragmatic factors. Here, we explore the further possibility that negation is hard to process because rejecting propositions requires more effortful control than accepting them. We presented 3- to 8-year-old children with a memory task in which they heard a novice make a statement about a novel group of animals (e.g., “Modis roar loudly”), and an expert confirm or deny these statements (“Yes, that’s right”, or “No, that’s wrong”). Later, children were asked about each of these statements (“Do modis roar loudly?”; yes/no). Children disproportionately misremembered denied statements as true both immediately after they learned these facts, and when they completed a second session two months later. Children with higher executive function capacities were better able to remember the denied, but not affirmed statements. Rejecting propositions might thus tax still-developing executive function capacities, making processing negations particularly hard for children.

Wh-ex-situ questions (e.g., English) are considered more difficult than wh-in-situ questions (e.g., Mandarin) due to different movement types. Moreover, subject wh-questions in both English and Mandarin involve short-distance dependencies and are similar in processing demand; however, their object wh-questions involve longer dependencies and may invoke stronger processing in wh-in-situ than wh-ex-situ languages. Adopting Intermodal Preferential Looking, we tested comprehension of subject and object wh-questions in 17 English and 18 Mandarin-speaking toddlers, matched on age and vocabulary size. Results showed that both language groups understood subject and object wh-questions. However, Mandarin-speaking toddlers looked significantly longer at the match than their English-speaking counterparts, hearing object but not subject wh-questions. These findings reveal early comprehension of subject and object wh-questions across languages. Moreover, the data reflect the subtle influences of movement type and the distance effect, leading to the processing advantage for wh-in-situ questions and the cross-language subject-object asymmetry of children’s wh-question processing.

We investigate the functional neural activation for Hawaiian-English bilingual adults and English monolingual adults during the resolution of two different syntactic ambiguities and compare the networks that each group engages. Participants complete two tasks, one targeting a garden-path ambiguity and the other a structural prepositional phrase ambiguity. We find that conflict resolution areas, like LIFG and pre-SMA, are more active in the ambiguous compared to the unambiguous and filler conditions in both tasks. Further, each task engages a slightly different network, suggesting that they target a different processing mechanism. We also find that the bilingual adults engage the conflict resolution network in a different way than the monolingual adults. Our findings are consistent with the body of literature that claims that bilinguals and monolinguals utilize different conflict resolution networks. It also addresses our goal of broadening the scope of Hawaiian language research to promote a bilingual Hawai’i.

This study investigates whether neural language models can simulate the patterns of developmental trajectory of human second language (L2) learning by evaluating their sensitivity to typical learner errors. Using a set of error-correction sentence pairs categorized by proficiency level, we evaluated two XLM-based language models: the JA-EN model (pretrained on Japanese followed by English) and the EN-Only model (trained only on English). For each sentence pair at each stage of English learning, we calculated perplexity and treated the sentence with lower perplexity as the model’s predicted correct form. Accuracy was defined as the rate of selecting the sentence. The ANOVA demonstrated that the JA-EN model achieved significantly higher accuracy for lower proficiency groups than for advanced groups, reflecting a stepwise improvement through learning. Conversely the EN-Only model did not exhibit this pattern. These results suggest that language models can simulate the developmental trajectory observed in human L2 acquisition.

Children learning different languages show variability in describing physical motion events in both speech and gesture. We ask whether these differences extend to metaphorical motion events and, if so, when in development the patterns become evident. We studied the speech and gestures produced by 100 children learning English or Turkish (n=50/language)—equally divided into 5 ages: 3-4, 5-6, 7-8, 9-10, 11-12 years—when describing metaphorical motion events (e.g., idea runs across mind) and physical motion events (e.g., girl runs across park). We found cross-linguistic differences in the expression of metaphorical motion events in both speech and gesture, akin to cross-linguistic differences observed in the expression of physical motion events. But language-specific patterns emerged later in metaphorical than in physical motion descriptions, both in speech and in co-speech gesture. Gesture and speech form a tightly integrated, language-specific system in the expression of metaphorical concepts, as they do in the expression of physical events.

We asked whether goals affect how children, as well as adults, determine when an event ends. In Exp.1, 4-5-year-olds and adults read an agent’s goal (e.g., “Jessie wants to eat an orange with her breakfast to make it healthier” or “Jessie wants to peel the orange”) and then answered a question about an image that appeared (“Did she peel the orange?”). Adults, but not children, were more likely to accept a partly complete outcome as indicating a culminated event if the outcome satisfied the agent’s goal. In Exp.2, participants watched videos with two people engaged in a social interaction. A person performed an action (e.g., partly cleaning a table) that was either helpful or unhelpful with respect to the goal of the other person (setting a small book on a table). Our results show that even young children use intentionality to place boundaries on simple physical events.

This study investigates the role of bareness —i.e., absence of lexical NP restriction— in the computation of featural Relativized Minimality (fRM). To this end, we compared the acquisition of Italian elliptical indirect wh-questions, i.e. sluicing, featuring an inclusion configuration (i.e., using quale+NP ‘which+NP’) with those involving a disjunction configuration. Disjunction was obtained either using syntactically bare wh-elements (i.e., chi ‘who’) or elements bare only at Phonetic Form (PF) due to nominal ellipsis (i.e., quale ‘which’). A yes/no-question task was administered to 82 Italian-speaking children (aged 3;00–5;11) and 26 adult controls. Results showed that children performed better with chi-object sluices than with quale+NP-object sluices, indicating amelioration effects in disjunction configurations. However, quale-sluices, despite being bare at PF, did not show similar amelioration, and were the most difficult overall. The findings suggest that only syntactic bareness counts for fRM, which is thus a structural principle insensitive to surface PF-properties.

Phonetic accommodation (speech modification to resemble an interlocutor) is well-documented in monolinguals but understudied in bilinguals, particularly early sequential bilinguals. This study examined whether accommodation in a speaker’s L2 can induce phonetic drift in their L1, focusing on changes in production of voice onset time (VOT) in 24 early bilingual speakers of Indian English. Participants, L1 speakers of either Hindi or Telugu, completed baseline L1 and L2 production tasks, an L2 accommodation task involving shadowing an American English speaker, and a post-accommodation L1 production task. Results of mixed-effects regressions on VOT measurements revealed significant accommodation (lengthening) in L2 VOT in both groups, but no significant changes in L1 VOT from baseline to post-accommodation. Contrasting with previous findings on late bilinguals, these findings suggest that, despite flexibility in their L2 production, early bilinguals may maintain stable L1 phonetic categories, highlighting the resilience of the L1 system in early bilinguals.

We investigate child comprehension of relative clauses (RCs) in Marathi (a split ergative language spoken in India: ergative-absolutive in the perfective; nominative-accusative in the imperfective). Moreover, stative verbs exhibit quirky case, with dative (DAT) subjects (widely considered true subjects, Wali, 2004) and nominative (NOM) objects. Marathi has two relativization strategies: a gapless correlative construction, available for any combination of aspect and verb-type, and a gapped RC which is more frequent in usage but limited to certain combinations of verb-type and aspect. This study examines child comprehension of gapped RCs, finding that rather than a subject advantage, Marathi children show success in RC comprehension when encountering nominative arguments, regardless of whether they are subjects or objects.

The current study investigates gender assignment strategies in 60 Russian-Hebrew bilingual children, an understudied population exposed to two gendered languages. Building on prior research identifying four categorical strategies (default, translation-equivalent, insertion, and shape-based) we also examine the gradient approach to gender assignment under the Gradient Symbolic Computation framework (Putnam et al., 2018; Uygun & Felser, 2023). Children completed two auditory acceptability judgment tasks with Hebrew or Russian as the matrix language and embedded nouns from the other. We manipulated gender transparency and congruency. Participants also completed vocabulary and nonce word gender production tasks. Results show stable use of the masculine default in Russian and shape-based cues in Hebrew, but variable strategies in Russian judgments, suggesting a gradient integration of gender cues. These findings underscore individual variability in bilinguals’ gender processing and support a shift from categorical to probabilistic models of gender assignment.

While past research has shown mixed results when linking executive functions directly to second language acquisition, this study introduces cognitive multitasking as a new variable capable of predicting performance in additional language learning.

Dominican Spanish is a dialect characterized by high SPE variability and variable person/number verbal affix omission (coda-final /s/ lenition). In this paper, we highlight significant past contributions from the acquisition of the Null Subject Parameter (NSP) and investigate linguistic and extralinguistic factors involved in the stages after the NSP. We analyzed spontaneous speech from children (5;1–6;9) and their caregivers in Santo Domingo. Main inferential and descriptive results show that children produced adult-like rates and patterns overall, though differences emerged: verb ambiguity significantly predicted SPE for children, while verbs of cognition predicted SPE for caregivers. Reflexivity influenced both groups, with non-reflexive verbs increasing SPE. Unlike other Spanish varieties, switch reference was not a significant predictor. These findings highlight the late acquisition of pragmatic uses of expressed pronouns with cognitive verbs, children’s reluctance to produce multiple markings, and language acquisition when input is variable.

Can young children use a situation model to guide reference resolution? In an eye-tracking comprehension task, 36-month-olds heard stories while viewing pictures of story participants. In the critical sentence, a protagonist asked one animal to play with her (“Maisy asked the frog…”). We varied whether stories introduced two potential referents for the target noun (two frogs), and whether one referent left the scene. In ambiguous trials, children looked equally at the target and distractor animals. In unambiguous trials, they preferred the target animal both in trials with only one potential referent (one frog) and trials in which one of two potential referents left the scene. Importantly, this depended on language skill: Past language learning trajectories predicted target looks across conditions, while current language ability interacted with story condition. These findings provide clear evidence that 36-month-olds leverage situation models to support reference comprehension and this ability varies with language learning trajectories.

There is debate about how learners assign grammatical gender to French nouns. This study seeks to contribute to this issue by examining the hypothesis that the final syllable of a noun can be the basis for a rule for gender assignment (i.e., assign masculine to open syllables and feminine to closed syllables). To test this, the final syllables and gender of nouns from the CDI (Wordbank) and from child-direct speech (Lyon corpus) were analyzed. The Tolerance Principle (Yang, 2016) was used to determine whether the number of candidates for the rule outweighed the number of exceptions (e.g., open syllable nouns that are indeed masculine versus those that are not) in terms of learnability effectiveness. Results suggest that it would be more effective to apply a rule by which nouns ending in an open syllable are masculine, but nouns ending in a closed syllable would be learned on a piecemeal basis.

Late talkers (LTs) are 18-35 month olds with limited expressive vocabulary of unknown cause. Research on LTs focuses on production delays with limited work directed to language comprehension. This study assessed spoken word recognition in late talkers and two control groups: one matched on age and another matched on expressive vocabulary. Participants completed a simplified visual world paradigm task where they heard a target word and saw two images (target image and competitor image). Targets were paired with either a phonological competitor, semantic competitor or an unrelated object. Eye movements were recorded and fixations to targets and competitor images were calculated. All groups demonstrated phonological and semantic competition. LTs showed greater competition compared to age-matched controls. Competition effects in LTs and younger, expressive-language controls were similar. Findings suggest that late talkers exhibit reduced efficiency in spoken word recognition and lexical processing differences may be part of a broader symptom profile.

Children with Autism Spectrum Disorder (ASD) often struggle with interpreting deictic pronouns such as I and you, in direct and indirect speech reports. This study examined whether (a) pronoun interpretation differs between direct and indirect speech in children with ASD, (b) performance relates to Theory of Mind (ToM) or Working Memory (WM), and (c) bilingualism enhances pronoun interpretation. A total of 404 monolingual and bilingual children (TD and ASD, ages 3–12) completed a sentence-picture matching task in their first language (English, French, German, Italian, or Spanish). Results showed autistic children were more accurate in direct than indirect speech. Pronoun accuracy was positively associated with age, ToM, and WM. ToM especially supported interpretation in indirect speech for autistic children. Exploratory analyses suggested that more balanced bilingual autistic children were quicker in interpreting pronouns. Findings indicate that both cognitive capacities and bilingual language exposure contribute to how autistic children interpret deictic pronouns.

We present an adaptation of the MacArthur–Bates Communicative Development Inventories (CDI) for Azerbaijani, with the aim of addressing the lack of standardized language assessment tools for this language. Using the Turkish CDI as a model, we piloted Words and Gestures and Words and Sentences forms with caregivers of 8 to 36-month-olds. Based on feedback, we revised forms to include culturally-specific vocabulary and Russian loanwords and improve instructions. Items underwent further testing and were classified by difficulty thresholds to ensure balanced coverage across item categories and age of acquisition. In parallel, we collected 20 hours of daylong home recordings for each child to further validate CDI items, using AI to generate first-pass transcriptions. The home recordings help to diversify language research by providing the first publicly available Azerbaijani-language corpus. The resulting CDI, grounded in naturalistic data and automated processing, may be used to facilitate identification of language delays in Azerbaijani children.

Infants learn words through interactions with environment and engagement with objects, places, and people. These experiences provide rich multisensory information that supports word learning. However, how such multisensory input in home settings shapes children’s productive vocabulary remains understudied. This study examined how multisensory input during object naming events contributes to infants’ productive vocabulary. We analyzed home video recordings of 18- to 23-month-old toddlers (N = 28, M = 20.5 months) and their caregivers, coding five sensory modalities—visual, manual touch, feel, proprioceptive, and mouthing. Visual input was most frequent (70%), followed by manual touch (22%), feel (11%), proprioceptive (10%), and mouthing (2%). Mixed-effects models showed that both multisensory richness and word frequency significantly predicted infants’ productive vocabulary. Additional models revealed that manual touch and proprioceptive input uniquely predicted word production, while visual, feel, and mouthing did not. These findings highlight the role of embodied multisensory experiences in shaping early vocabulary development.

Previous studies have documented perceptual challenges and production distortions related to the [l]-[n] merger in Chinese speakers’ native dialects (e.g., the Hubei dialect). However, little is known about whether this confusion affects their perceptual performance and production in nonnative languages. The present study investigated Hubei dialect speakers’ perception and production of L2-Mandarin and L3-English words contrasting [l] and [n] and examined the role of linguistic and social factors in their performance. The results showed reduced perceptual sensitivity to nonnative words contrasting [l] and [n] and non-significant acoustic differences between the two sounds, indicating evidence of the [l]-[n] merger. Participants’ performance was modulated by vowel contexts and the presence of a nasal coda in the stimuli. Moreover, less frequent dialect use and earlier acquisition of Mandarin and English significantly improved participants’ performance. Therefore, the [l]-[n] merger in L1 negatively impacted speakers’ perception and production in their L2 and L3.

Across languages, passive voice poses acquisitional challenges for children. Two competing accounts explain how children generalise grammatical knowledge: gradual abstraction, where generalisation occurs only after sufficient exposure, and early abstraction, where children rapidly extend knowledge from input irrespective of exposure. Author (xxx), the baseline study, revealed Korean monolingual children’s sensitivity but delayed mastery of passive-voice heuristics, necessitating prolonged exposure and usage-based learning given multiple competing cues, thereby supporting moderate versions of each account. The present study extends the baseline study by examining Korean monolingual children’s real-time processing of the suffixal passive using webcam eye-tracking (WebGazer.js). Results from three-to-six-year-olds’ comprehension of this construction suggest a limited yet present role of passive-voice heuristics in processing Korean suffixal passive, likely constrained by competition from more robust active-voice knowledge. This further supports moderate positions of both accounts, also elucidating children’s online processing dynamics, while mitigating sampling biases and methodological limitations in the field.

Regularization and probability-matching have been regarded as complementary mechanisms accounting for child-adult differences in language learning. We propose that regularization and probability-matching share the same underlying principle: generalization. Specifically, learners regularize when they form one generalization and probability-match when they form multiple. We exposed 120 adults to artificial languages differing only in the generalizations supported by the input: one supports the generalization of one plural marker and the other for two. We test items seen in the plural (observed) and items seen only in the singular (hidden), to assess generalization to novel contexts. As in prior work, adults appear to probability-match for observed items. The hidden items reveal the underlying generalizations: participants used both markers when the input supported two generalizations, but when the input supported only one, they were significantly more likely to use that form exclusively, appearing to regularize like children.

This study investigates the production of relative clauses (RC) of Georgian-, Italian- and Yoruba-speaking children and adults. The experiment is designed to elicit 4 types of RCs: subject, object, indirect object and locative. The result reveals that across these languages, children demonstrated a tendency to transform non-subject RCs into subject RCs, further extending the well-attested subject advantage. Among the other RC types, there seems to be mixed results and no clear relation to the Accessibility Hierarchy can be spotted. Compared to adults, children produced resumptive elements more frequently, though the form varied across languages. We stress that while the Subject advantage may be universal, its linguistic effects are language-specific. The transformation to S and the use of resumption are further explained by a theory whereby speakers sometimes facilitate difficult RC derivation by generating an aboutness-topic in the RC and relativizing it instead of the original internal head in situ.

Developmental Language Disorder (DLD) is characterized by difficulties in using and learning language. Deficits in spoken word recognition have been suggested to underlie the broad challenges in DLD. Previous work has shown that children with DLD show less activation for target words and maintain activation for phonological competitors (i.e., CAT-CAB) longer. The current study examines whether these differences persist into adulthood and if early language ability can predict individual differences. Participants heard a word and clicked the corresponding image from a field of four: the target (e.g., “candy”), a phonological competitor (e.g., “candle”), and unrelated items. Adults with DLD showed similar activation of targets and competitors compared to their typical language peers. Results suggest that the between-group differences in spoken word recognition observed in childhood are resolved in adulthood. However, early language ability predicted individual differences in word recognition 30 years later. An explanation for these findings is discussed.

During the word learning process, children map phonological forms to meanings, often relying on surface-level cues in the syntactic environment in which a word is found in order to narrow down its possible meanings—a process referred to as syntactic bootstrapping. For certain grammatical categories, like adjectives, a given frame ranges in its informativity, such that learners benefit from tracking an adjective across multiple environments. Recent research by Gotowski & Syrett (2024) with the Human Simulation Paradigm (HSP) found that adult learners successfully recruit clusters of frames to learn novel adjectives. We build on their study by adapting the HSP for children 7-9 years of age. We find that children pay attention to distributional information in order to successfully subcategorize, and routinely offer frame-compliant guesses for novel adjectives, mirroring what has been reported for the adult population.

Parental calibration of speech to fit the unique linguistic needs of their child has been shown to improve vocabulary outcomes. However, the relationship between fine-tuning and language growth in late-talking children is unknown. In this study, we examine whether morphological alignment between the mother and child within conversational discourse can predict recovery or persistence of expressive language delays in identified late-talkers. We found that differences in mother-child morphemes per turn was a significant predictor of language outcomes at just 36 months, where increased alignment increases the probability of persistence for language delays. Our findings may suggest that systematic intervention for late-talking children should emphasize the role of enrichment (the volume and consistency of input) rather than linguistic fine-tuning (adjustment of input) during parent-child interactions.

Our study demonstrates a systematic effect of syntactic position on children’s interpretation of negation. Looking at German, which exhibits different surface positions for sentential negation, we investigate to what extent learners have more difficulty with ‘verb-distant’ than with ‘verb-adjacent’ negation across development, while controlling for the truth/falsity of negatives. Typically-developing German-speaking children (N=61) were tested three times (T1=3;9 years; T2=4;7 years, T3=5;8 years). Comprehension was assessed via Truth Value Judgment. The factors Neg-Position (adjacent/distant) and Truth (true/false) were crossed, resulting in four conditions. By age 5, distant negation is still harder than adjacent negation for true negatives. The difficulty of ‘distant’ negation can be explained syntactically via the need to reconstruct scrambled DPs within the domain of the negator. Alternatively, under a processing account, the ‘surface lateness’ of the negator leads to the activation of the alternate positive state of affairs that must suppressed when the negator is encountered.

Brain-based predictors of expressive language (EL) development—particularly for infants growing up in under-resourced contexts—remain understudied. We examined whether EEG in infancy can predict later EL in children from households with limited socioeconomic resources. We recruited infants from an urban pediatric primary care clinic primarily serving families with low- to mid-level income and parental education. Resting EEG was collected at 2-12 months; EL was assessed at 24 and 36m. EEG activity in infancy predicted later EL, particularly among late talkers. Steeper developmental increases and elevated power in frontocentral higher-frequency bands were associated with poorer EL outcomes, potentially reflecting delayed neural pruning or reduced specialization. These findings highlight the potential of EEG as an early, objective marker of risk for language delay and underscore the need to understand how socioeconomic influences shape early neurodevelopment and contribute to vulnerability for language delay and/or disorder.

The current study examined whether Mandarin-speaking three-to-four-year-old children are sensitive to semantic properties in interpreting double-object construction, focusing on how the presence of the Numeral-Classifier form (NumCl), an indefinite marker, may affect children’s comprehension of the construction. 145 three-to-four-year-old subjects participated in a picture selection task, assigned to one of the three conditions: No-NumCl (bare noun IO/DO), NumCl-DO (bare noun IO; NumCl DO), and NumCl-IO (NumCl IO; bare noun DO). If children were sensitive to the indefinite marker, they should display a better performance in the NumCl-DO condition than in the other two conditions. The prediction was borne out: the comprehension accuracy in the NumCl-DO condition was significantly above chance and higher than in the other two conditions. Our results show that three-to-four-year-old Mandarin-speaking children can exploit indefiniteness to correctly assign the theme role to DO, providing further evidence for the role of (universal) semantic properties in learning argument structure.

McDonough (2006) observed priming effects in L2 English learners’ production of dative constructions for prepositional datives (PD) but not for double objects (DO). We hypothesize that crosslinguistic influence (CLI) from learners’ L1 may have influenced these results, as (i) over half of the participants were L1-Mandarin speakers, and (ii) Mandarin translation equivalents of the verbs most frequently produced with DO in English allow only DOs in Mandarin. We conducted a written picture description task with L1-Mandarin learners of English and included English alternating dative verbs that have Mandarin translation equivalents allowing only DO (‘DO-only verbs’) or disallowing DO (‘non-DO verbs’). Results show that (a) learners produced more DOs with ‘DO-only’ than with ‘non-DO’ verbs across all task phases, and (b) priming occurred only for ‘non-DO’ verbs, not for ‘DO-only’ verbs. These findings support the hypothesis that CLI affects L2 production and priming and align with error-based learning accounts of priming.

This study focuses on how learners of Mandarin acquire the difference between attitude verbs by looking at know, think and wonder verbs that select different types of complement clauses. Wh-in-situ languages like Mandarin present a challenge for learning attitude verb subcategorization because there are no word order cues indicating the scope of wh-elements. Huang et al. (2022) confirmed the prevalence of this ambiguity in input through corpus, and also suggested that children may use speech acts as clues to distinguish between different types of attitude verbs. This study tests this hypothesis through experiments. Our result from a Question-Statement Task conducted in 52 Mandarin-speaking children (3;5 – 6;7, Mage= 4；11) shows that 3-yro Mandarin-speaking children indeed make use of the correlation between speech acts and corresponding syntax to determine the semantic difference between think and know verbs, which support the argument that learners use speech acts to learn verb subcategorization.

Children learn language through social interactions. Less is known about how children’s social biases may impact language learning. We investigated how children selectively learn words from adults of different gender. Study 1 tested 7- to 9-year-olds’ (N=82) endorsement of pseudoword labels associated with novel objects presented in gender-conforming and gender non-conforming voices. We found that children were more likely to endorse pseudowords produced by gender-conforming voices than gender non-conforming voices. Study 2 examined pseudoword imitation in 4-year-olds (N=72) with stimuli produced by gender-conforming and non-conforming adults. Children imitated pseudowords less accurately from gender non-conforming adults than from gender-conforming adults. Across both studies, female children strongly favored gender-conforming adults of their gender, while male children showed no preference. Together, these studies provided empirical evidence that children are actively evaluating their social partners during language learning, and that they favor gender-conforming adults when learning words.

The study investigated whether preschool-aged children with Developmental Language Disorders (DLD) differ from their typically developing (TD) peers in the timing and degree of lexical activation during the comprehension of sentences with unexpected endings.

Preschoolers struggle with deploying modal concepts in behavioral studies and often fail to distinguish between possibility and necessity modals, although younger children start producing various modal words at around age 2. To investigate whether early use of modal words indicates an early ability to think about mere possibilities, we investigated children’s comprehension of the contrast between “if” and “when” in conditionals. We hypothesized that how long children spend trying to open a puzzle box would depend on whether an adult said “if you open it…” or “when you open it…”. We find that both 3- and 4-year-old children try longer in the when condition than in the if condition. Our results suggest that children under age 4 can represent that a state of the world is certain or uncertain (merely possible), and understand which alternative a speaker believes depending on their use of “if” or “when”.

This study investigated whether children’s word meanings begin too broad or too narrow by having 208 children (ages 2;11-8;8) sort images into categories matching common nouns like “dog.” Each sorting task included target images (actual dogs), similar foils (cats, raccoons), and dissimilar foils (unrelated objects like baskets).

The shape bias, the tendency for children to generalize new words based on shape, is well-documented in English but reportedly weaker in languages like Korean. We test whether lexical statistics underlie this bias across languages. In Study 1, English-speaking adults rated 300 early-acquired nouns (from 16 languages) on shape, solidity, and countability. While most nouns were rated as solid and countable, shape-based organization was less consistent. Nonetheless, higher shape ratings predicted earlier acquisition, even after controlling for word frequency and concreteness. In Study 2, 2 independent samples of English and Korean adults completed similar rating tasks. Ratings of solidity and countability were consistent across groups, but Korean participants rated more nouns as shape-based than English participants, contrary to reported developmental biases. Results suggest that while lexical statistics predict age of acquisition, they don’t explain cultural variation in shape bias. Cross-linguistic measurement is needed to better understand how children generalize words.

An extensive body of research finds that bilingual individuals may have enhanced executive functioning (EF) due to constant training in using one and inhibiting another language on a daily basis. Recent studies showing null evidence to this phenomenon suggest that there might be more nuance to the bilingualism-EF relationship. Given the vast individual variation across bilinguals, it is important to consider more fine-grained differences in mechanisms of how language experiences might affect EF from early childhood to adolescence. The study reveals important neurocognitive mechanisms (measured with fNIRS) of EF-elicited activation in the higher-order language network and how these mechanisms vary across participants with monolingual/bilingual experiences from early childhood to adolescent years. The naturalistic setting (i.e., museum) of data collection by the means of using fNIRS in outside-of-lab environment provides a unique opportunity to obtain ecologically valid prefrontal and auditory cortex activation in a widely diverse sample of participants.

This study examines whether executive function (EF) abilities (attention, inhibition, visual/auditory working memory) or linguistic skills (morphosyntax, lexicon) influence third-person singular Subject Personal Pronoun (SPP) expression in switch-reference contexts in child Spanish. While prior research (Shin & Cairns, 2012) suggested working memory limitations may influence children’s use of overt SPPs in switch-reference contexts, this study investigates broader cognitive and linguistic influences. Eighty-two monolingual Spanish-speaking children (ages 4–8) completed sociolinguistic interviews, a narrative task, and standardized EF and language assessments. Mixed-effects regression, correlations, and multiple regression analyses revealed no significant association between SPP expression and EF abilities. Instead, morphosyntactic development strongly predicted SPP use (p < .001). These findings challenge the working memory hypothesis, suggesting that grammatical complexity, rather than pragmatic or cognitive constraints, drives overt SPP production. The results highlight morphosyntax as the key factor in children’s mastery of pronoun variation. Methodological differences from prior experimental studies are discussed

In contrast to priming studies that focus on structures with similar meanings, the current study uses a picture selection task to examine whether a particular interpretation of a globally ambiguous sentence (e.g., “The horse chooses the rabbit with the marker”; instrument interpretation: [chooses the rabbit][with the marker], NP modifier interpretation: chooses [the rabbit with the marker]) can be primed and whether that priming can overcome a verb’s bias. Despite a baseline instrument preference for all verb biases, children that produced modifier interpretation primes were significantly more likely to select modifier targets than those that produced instrument primes, but only if the verb was equi- or modifier-biased (equi:β=0.99,p<0.001; modifier: β=0.43,p<0.05; instrument: β=0.47,p<0.05). Ambiguity of the prime also had an effect, but only for equi-biased verbs. Children have a strong underlying instrument preference that is difficult to overcome when the verb is also biased in this direction.

Previous research has found that young children (under 8 years old) produce inflectional morphology more accurately than they understand it. In contrast, adult bilinguals present the opposite pattern, showing greater accuracy in comprehending than in producing inflection. We asked whether children ages 7-14 years pattern more similarly to younger child bilinguals or adults on selection versus production tasks, and how the performance of child heritage speakers (HS) and second language learners (L2L) compares. Our participants were 78 bilingual children enrolled in DLI who received 90% of their schooling in Spanish in grades K-2, and 50% in grades 3-8. We found that most of the children produced the expected gender morphology more consistently than they selected it on the forced choice task, patterning like younger bilingual children. In addition, our results did not reveal any significant differences between HS and L2 learners in their accuracy in producing or selecting clitic gender.

Function words are hypothesized to serve as anchoring cues that help learners extract the syntactic structure. The present study tested Turkish-learning infants aged 6–12 months using an artificial grammar learning task in an online format. As a consistently functor-final language with agglutinative morphology and a non-Indo-European typology, Turkish offers a valuable context for evaluating the generalizability of frequency-based bootstrapping mechanisms. Infants were familiarized with an artificial language stream composed of alternating frequent and infrequent syllables and then presented with sequences structured as either functor-initial (FI) or functor-final (IF). Looking times were analyzed offline. Preliminary data from 18 participants revealed a marginal preference for IF sequences, t(17) = 2.10, p = .051, consistent with Turkish word order and prior findings from other OV languages. These preliminary results suggest early sensitivity to distributional cues and highlight the feasibility of using online paradigms to broaden methodological and typological diversity in infant language research.

Adults and children are frequently exposed to accented speech, yet little is known about how accented speech influences word learning. In two experiments, we ask how monolingual English-speaking adults (Experiment 1) and school-aged children (Experiment 2) learn and retain novel words when exposed to accent variability across time scales. All participants learned ten English-like novel words presented in a native-English or Spanish-accent in a cross-situational word learning paradigm. Participants were randomly assigned to one of two test conditions: immediate or a short 6-minute delay. Across age groups, accents, and delay conditions, performance at test was above chance. Overall, adults and children did not demonstrate significantly better learning for native English vs Spanish-accented speech, contrary to what might be expected based on the accent processing literature.

Referring to an object using ‘one’ indicates a category construal available in the discourse/referential context, while the referring expression ‘thing’ lacks such presupposition and could mean that the category is irrelevant or unknown. Thus, despite both being underspecified, these expressions activate construals differently. In the present study, 2-year-olds saw either a familiar or unfamiliar target object and were asked to find a match among a basic-level match, a superordinate-level match, and two distractors. The target was labeled either with ‘one’, ‘thing’, or a non-referential control expression. While toddlers generally showed an above-chance basic-level preference, basic-level choices were influenced by familiarity condition and noun. With unfamiliar objects, 2-year-olds made more basic-level choices only upon hearing ‘one’ (vs. control). With familiar objects, they made more basic-level choices with ‘thing’ (vs. control). The findings suggest that 2-year-olds differentiate between underspecified nouns, but their use of pragmatic information is immature.

Caregiver talk plays a key role in supporting children’s exploration, especially in informal learning spaces like museums. This study examined how parents verbally guide children during shared exploration and whether talk types cluster into meaningful patterns. Parent-child dyads were filmed for 10 minutes while exploring a children’s museum exhibit modeled after a grocery store. All parental utterances were coded using ELAN into 11 talk types (e.g., scaffolding, labeling, commenting, directives). Scaffolding, wh-questions, labeling, and commenting were most frequent, though no single talk type dominated overall. A principal component analysis revealed that talk types loaded onto four components, explaining 67% of the variance. These results suggest that parents use a broad and varied range of talk strategies when guiding their children. This study extends previous research on parental input in homes and labs to real-world, ambulatory exploration settings. Future work will explore how these talk types relate to children’s language development.

The study investigates how Romanian-speaking five-year-olds interpret conditional disjunctions—sentences like “Go to bed early or you’ll be tired”, where ‘or’ conveys a conditional meaning ([¬p → q]). While previous research has examined children’s understanding of disjunction and conditionals separately, little attention has been paid to conditional disjunctions. 30 children and 59 adults participated in a task judging whether plausible and implausible conditional sentences using ‘or’ and ‘and’ make sense. Adults performed well on both types, while children showed high accuracy with conditional ‘and’ but struggled with conditional ‘or’. Despite incorrect responses, many children offered justifications consistent with a conditional structure, indicating partial understanding. The findings suggest that children initially interpret ‘or’ as ‘and’, aligning with the Meaning First Approach, which posits a preference for direct form-meaning mappings. Children struggle when meaning is implicit, that is, when both the conditional and the negation are covert.

Many deaf children are exposed to multiple languages yet have large swaths of time with no access to the language in their environment. We conducted structured interviews of hearing parents of deaf children (N=53, child age 4–11 years) using the Language Access Profile Tool (Hall & De Anda, 2022). Parents provided details on the amount and type (e.g., spoken English, ASL, cued speech) of language input for each year of the child’s life. We found that many life events were associated with consistent shifts in language input. For example, hearing aids or cochlear implants, COVID-19, and new siblings were all associated with increases in spoken English. Caregiver enrollment in ASL classes, deaf mentor programs, and changes to the adults in the household were all associated with increases in the amount of ASL in the child’s environment. These data highlight the shifting nature of deaf children’s early language environments.

This study examines how the Incompleteness Effect (IE)―a Mandarin-specific phenomenon where some monoclausal sentences feel incomplete even when an aspect marker is present―affects subsequent referential choices in L1/L2 Mandarin. IE can be canceled through additional predication/modification of the subject (Tsai, 2008). Our experiment had Mandarin native speakers (MNSs) and L1-English L2ers of Mandarin (ELMs) complete a written sentence-continuation task, crossing IMPERFECTIVE MARKER (‘complete’ *zai* vs. ‘incomplete’ *-zhe*) with CONTINUATION CUE (null vs. *ranhou* ‘and then’). More continuations reprising the referent of the prompt’s subject (“same-subject continuations”) after *-zhe*-prompts vs. *zai*-prompts indicate grammatical-aspect information led to IE cancelations; more same-subject continuations across conditions indicate discourse-level topic maintenance. Results show MNSs had both patterns, despite an overall same-subject preference, but ELMs had only the latter. Nevertheless, L2 proficiency correlated positively with same-subject continuations in the [*-zhe*, null] condition, suggesting that―like MNSs―higher-proficiency ELMs utilized grammatical-aspect information to cancel IE and override discourse-level information.

In children’s word learning and use, the “basic” level of specificity (“dog”) is known to be privileged over the broader, superordinate level (“animal”). This basic-level bias has often been argued to emerge from the perceptual naturalness of basic-level categories and/or children’s conceptual difficulties with the superordinate level of abstraction. Here, we propose instead that this bias reflects at least in part children’s maturing competence in pragmatics – namely, the ability to identify the conversationally appropriate taxonomic level of a linguistic description in terms of its informativeness and relevance. In two experiments, we show that four-to-five-year-old children recognize superordinates to be under-informative for naming single, familiar objects but also flexibly adjust this evaluation depending on the relevant alternatives within a task. These findings offer a reinterpretation of previously observed difficulties with superordinate nouns in young learners and provide early evidence for children’s success with pragmatic informativeness and relevance computations.

How do children develop an adult-like syntactic tree? According to the prominent Growing-tree approach (Friedmann et al. 2021), it is built bottom-up by stages (Fig. 1). For example, lower structures in Stage 1 emerge earlier than those in Stage 2, which in turn precede structures in Stage 3. To test this hypothesis, we examined child utterances of Japanese sentence final particles (SFP), standardly assumed to be quite high in the CP. We first concur with previous literature (Clancy 1985, i.a.) that SFPs (1–3) emerge early. Importantly, we make the new observation that higher SFPs (above ForceP) can occur even earlier than lower SFPs (FinP), contrary to predictions of the Growing-tree approach.

This study investigated statistical learning (SL) in autistic (ASD) and neurotypical (NT) adults matched on language abilities. Participants completed an SL task measuring online (reaction time acceleration) and offline (2-AFC accuracy) learning, with stimuli varying in attentional focus. Results showed autistic individuals exhibiting greater reaction time acceleration and higher overall accuracy, particularly for unattended triplets, suggesting enhanced implicit learning. A significant correlation between accuracy and language fluency (Woodcock-Johnson scores) was found in the ASD group, indicating that SL better explains some language skills. Though the correlation was not significantly stronger in the ASD group than in the NT group, the trend suggests autistic individuals may rely more on SL to support language development. These findings suggest that implicit learning is a preserved and potentially enhanced mechanism in autism, explaining how some autistic individuals achieve language abilities comparable to those of their NT peers.

How is child perception related to production? Do children perceive like adults and simply vary in their speech motor skills? Do they learn their own speech patterns or perhaps child-typical speech patterns? Preschoolers (N=80) named pictures and then tried to choose the named picture from a set of 4 (e.g. wing, ring, sun, thumb) as they heard their own namings or another child’s. Pointing accuracy was equivalent for self than the other speaker. Pointing was more accurate for child speakers transcribed as more adultlike. Additionally, children showed ambivalence on word pairs where a child merged sounds, e.g. ring and wing both sounding like wing. Both correctly-produced (accurate wing) and erroneously-produced (ring –> wing) words reflected uncertainty, indicating awareness of child mergers. Children understand other children’s speech as a function of adultlikeness, and may have learned certain sounds that are merged (conflated) in child speech, akin to accent sensitivity.

Although word learning has been focused primarily on how children map forms to meaning, there are cases where a form—such as an expletive—is included for purely syntactic reasons. Previous research suggests children produce and comprehend expletives relatively early. Do children have early knowledge of expletives because they have expectations that such forms exist, and this is part of their innate knowledge? Or have children learned about expletives by simply paying attention to the input? Here we contribute to this discussion by evaluating language models as a proxy for what can be learned from paying attention to the input alone. We find that BabyBERTa, which was trained on data that more closely approximates the input a child receives, diverges from how children and adults behave. We argue children’s knowledge of expletive forms is not entirely input-driven.

It is well-known that children acquiring richly inflected languages acquire agreement earlier than children acquiring poorly inflected languages (e.g., English, Hyams, 1992; Spanish, Mariscal, 2009; a.o.). We investigate agreement in Marathi, a null argument language with a rich case system, but an unusual agreement system: gender agreement occurs on the verb with the highest argument that is unmarked for case. Thus, the relationship of each argument to the verb is signaled either through overt case, or if case is morphologically null, then agreement on the verb. This experiment specifically tests children’s ability to identify null arguments using only agreement. All arguments are left null, and we test whether children can reconstruct the null arguments, their case, and then use agreement on the verb to correctly interpret sentences. We find that children as young as age 4yrs show remarkable control over agreement, despite complications posed by verb semantics, alignment and case.

We introduce a brief parent-report word exposure survey as an alternative to the resource-intensive naturalistic recordings for estimating infants’ early language input. Two studies examined whether parents’ estimates of how often their infants heard specific nouns predicted their infants’ home recording noun frequency and vocabulary development. Study 1 followed 44 English-learning infants longitudinally, collecting monthly audio and video recordings (6-17mo.), alongside parent surveys on vocabulary knowledge and word exposure at 6 time points. Study 2 replicated the survey components with a cross-sectional sample (n=264) matched by age and gender to each longitudinal infant at each time point. Results showed that parent-reported word exposure reliably predicted noun frequency in home recordings (Study 1) and CDI comprehension and production (Studies 1 & 2). Both studies demonstrate that the parent surveys can effectively capture individual variation in language environments and vocabulary outcomes, offering a practical tool for researchers and practitioners.

Mazes are speech disfluencies such as filled pauses, revisions, and repetitions that do not contribute to a speaker’s intended message. They have been found to play a critical role in speech self-repair and planning. Language ability has been demonstrated to be a moderating factor in maze frequency in monolinguals, but previous analyses of mazing in bilinguals have yielded conflicting results. It is unclear how language-specific skills and the overall robustness of the language system contribute to maze usage in bilinguals. Preliminary results suggest no effects of English skills or overall language skills on mazing in Spanish speakers’ L2 narrative retells, a pattern that aligns with findings on mazing in L1 retell tasks in this population. These results indicate that bilinguals may be less sensitive to the effects of language ability on maze usage than monolinguals, and that language-specific and overall language skills may not differentially affect the use of mazes.

The viability of cross-situational learning for language has been discussed in psycholinguistic literature (Frank et al., 2009; Quine, 1960; Yu & Smith, 2007). With respect to verbs, while events often occur within half a minute of uttered verbs (Liu et al., 2019; Rodrigo et al., 2020; West et al., 2022), many other events do as well (Gleitman, 1990; Quine, 1960). We analysed uttered verbs in two mother-infant interactions in the Rollins corpus (Rollins, 2003; MacWhinney, 2000). Around half of the verbs referred to events that happened within 30 seconds of the verb utterance. However, so did large numbers of other events. Most verbs were relatively poorly associated with their corresponding event type and more strongly associated with dozens of other event types. This is consistent with proposals that infants must be able to sharply filter the hypothesis space of possible meanings for verbs.

Ad-hoc implicatures enrich an utterance’s literal meaning by negating a set of contextually salient alternatives. We investigate children’s ability to draw ad-hoc implicatures from sentences such as “Mary bought {the/a} striped sweater.” In a context in which there is a sweater with stripes and a sweater with stripes and spots, adults reject the definite description more than the indefinite, suggesting the definite triggers more ad-hoc implicatures than the indefinite. On some accounts, the definite (but not the indefinite) triggers local exhaustification within the object DP, in order to satisfy uniqueness. In a Truth Value Judgment Task experiment, we observe that adults and children display opposing behavioral patterns: adults reject (i.e. compute ad-hoc implicatures) more for “the” than “a”, while children reject “a” more than “the”. We propose that children and adults differ in how they treat the definite description, and in their preferred parses (exhaustifying locally vs. globally).

This study investigates children’s spontaneous interpretation of disjunction without proffering possible meanings. Previous studies have shown that children often interpret logical terms like “or” differently than adults, but they typically present children with multiple scenes or interpretations to assess, which may influence their responses. In naturalistic speech, listeners infer the speaker’s intended meaning without being explicitly offered alternative interpretations. To explore this, we tested 35 children (ages 4-5) using an interactive touch-screen task. After hearing a sentence like “The box lit up when Lucy put the cow or the duck on top” children were asked to make the box light up and then assess the monkey’s choice. We found that children preferred placing a single item, indicating a spontaneous preference for one-disjunct true interpretations. However, when an inclusive meaning was proffered, they accepted it and showed no evidence of a conjunctive interpretation.

How does individual variation in children’s exposure to accent variation affect their sociolinguistic competence, especially in distinguishing accents? In two experiments, Canadian 6- to 12-year-olds (Experiment 1: N=242; Experiment 2: N=699) heard two English speakers: one locally-accented (Canadian) and one with either a different regional (British/Southern US/Australian) or L2 (Brazilian-Portuguese/Mandarin/Japanese) accent. On each trial, children identified which speaker was non-local. In line with previous studies, children were overall more accurate at identifying L2 than regional accents as non-local, but contrary to our predictions, increased exposure to accent variation was associated with less accurate identification of non-local speakers. In Experiment 3, we are testing whether increased accent exposure decreases perceptual sensitivity to accent variation (as suggested by Experiments 1 and 2), or if perhaps children with exposure to more accent variation do not use accent in the same way as children experiencing less accent variation to determine whether someone is “local”.

Current word-learning accounts do not quantify how much or what kind of exposure to new words and referents children need to add them to their lexicons. We tested 40 caregiver-child dyads (18-mos;22-mos) to investigate whether toddlers could learn 3 novel words after 2 weeks of at-home book-reading, and how parent-child interactions support word learning. We annotated recordings and conducted a looking-while-listening eyetracking task.

This study investigated how gender and social context shape early language environments in Korean families using daylong LENA recordings from 141 parent–child dyads (children aged 7–30 months). Mixed-effects modeling of over one million audio segments showed that mothers spoke more than fathers, particularly in triadic (mother–father–child) settings, and children were most verbally active in triadic or peer interactions. Girls and their caregivers produced longer, more frequent speech, especially in multi-participant contexts. Three-way interactions emerged between speaker gender, interactional context, and children’s language ability. Higher language ability was linked to increased speech for girls in triadic contexts and boys in father-son dyads, while female adults spoke less with higher-ability girls in triadic settings. These findings underscore the nuanced influence of gender and social dynamics on early language exposure, highlighting the need to consider both family structure and gender when examining variability in early language development.

Our study investigated transfer of existing sign language expertise on SPACE-TIME metaphor inferencing in novel sign languages. After watching a 4-minute weather forecast in the target Swedish sign language, participants completed a meaning assignment task for 22 target signs. Crucially, we compared native signers (N=27, L1= ASL / BSL) to hearing sign novices (N=37, L1=English). We differentiated between cognate effects for linguistic signs, i.e. L1-L2 sign similarity, and overlap of target signs with gestural repertoires.

Four-year-olds consistently interpret the pronoun in sentences like (1) She loves Daisy as disjoint from Daisy, reflecting Principle C. However, in sentences containing crossover like (2) Who does she love? children’s behavior is unclear. Principle C has been argued to explain the strong crossover (SCO) constraint: the wh-phrase trace, like other R-expressions, cannot be bound by a c-commanding pronoun. Do children recognize that the same principle applies to (1) and (2)? Previous studies show children respond consistently with SCO, but we argue that this success derives from flawed experimental design. With a novel method, we show that evidence of SCO knowledge was likely a false positive: children do not readily respect the SCO constraint.

This study investigates how L1 and L2 speakers of Mandarin Chinese process temporal information during online sentence comprehension, focusing on aspectual morphemes (-le, -zhe) and temporal adverbs (yijing ‘already’, kuaiyao ‘about to’). Using a web-based visual world eye-tracking paradigm, participants listened to sentences while viewing event images. They were asked to choose the image that best matches the sentence they heard while their real-time eye movements were tracked. Preliminary results show that although L2 speakers successfully acquire temporal representation, eye-tracking data revealed group differences: L1 speakers rapidly integrated aspectual and adverbial cues with predictive eye movements, while L2 speakers showed delayed or less distinct cue use. These findings support the RAGE hypothesis. L2 performance was possibly due to limited naturalistic input, frequency effects related to individual morphemes, and competition between cues. Further analyses will examine the role of individual factors such as working memory and language proficiency.

Artificial grammar learning (AGL) and L1 acquisition studies examine how learners generalize inconsistent patterns to new examples, specifically what factors affect how strongly learners “regularize”, or overextend, the regular (majority) pattern (producing it at a higher rate than in the input). Learning theories make predictions about the timecourse of generalization, e.g. Bayesian priors predict that the effect of a bias will weaken as learners are exposed to more data. However, few AGL studies have investigated how generalization changes at different points in learning (cf. Linzen & Gallagher 2017). By manipulating both the artificial language and training duration in an AGL study on native English speakers, we found that exposure to more input data reduces the effect of the number marking bias identified in Schumacher & Pierrehumbert (2021). This result supports learning models’ bias predictions and emphasizes the importance of the timecourse of learning on generalization.

Understanding how second language (L2) learners process sentences has long been central to L2 research. Challenging the Shallow Structure Hypothesis, this study uses EEG during the processing of long-distance scrambling by L1-Korean and L1-English L2-Korean speakers. It investigates real-time syntactically constrained anaphoric relations for pronouns, and anaphors from the clause-edge and thematic gap positions. Anaphora-linked ERP effects suggest structure-dependent referential processes for pronouns in binding and coreference, and for anaphors in binding. ERP responses indicated that both groups distinguished (1) structural constrains on (coreferential and bound) pronouns vs. bound anaphors, and (2) matrix- vs. embedded-clause binding-domain computations. Crucially, earlier statistical effects were observed in the L2 group. ERP patterns across L1 and L2 groups suggest that even intermediate-to-advanced L2ers can build complex L2 syntactic representation in real time, despite not sharing typological features with Korean in their L1. These findings challenge claims of shallow processing in adult L2 acquisition.

This study examined whether working memory (WM) capacity influences sentence comprehension in Japanese-speaking children, focusing on differences between active and passive sentence processing. Fifty-three children (aged 4–6) completed a sentence comprehension task using four sentence types (intransitive active, transitive active, full passive, short passive) and a digit span task measuring forward and backward WM. A generalized linear mixed model revealed that comprehension improved with age only in the transitive active condition. In younger children, intransitive actives were best understood; older children also performed better on transitive actives than on full passives. Another model showed that forward span predicted comprehension in active conditions but not in passives. Children with higher WM capacity showed significantly better performance on transitive actives than full passives. In contrast, backward span did not predict comprehension performance. These findings suggest that WM supports syntactically simpler sentence processing, with forward span playing a stronger role than backward span.

Previous research has shown that content words are acquired earlier than function words, but the exact trajectory and order of function word acquisition has remained relatively understudied. In this study, we use the largest available child language corpora as well as Bayesian growth curve modeling to estimate the population level onset of production for more than 100 English function words. Our estimates suggest that for the large majority of function words, the earliest age of production lies between 12-24 months. A linear regression found longer function words as measured by the number of phonemes and function words with higher Mean Length of Utterance (MLU) have later estimated onset age of production. We did not find a significant effect of frequency in child directed-speech on onset age of production for function words. Overall these results point to early emergence of abstract functional morphemes with production limitations as the main bottleneck.

Caregiver talk plays a key role in supporting children’s exploration, especially in informal learning spaces like museums. This study examined how parents verbally guide children during shared exploration and whether talk types cluster into meaningful patterns. Parent-child dyads were filmed for 10 minutes while exploring a children’s museum exhibit modeled after a grocery store. All parental utterances were coded using ELAN into 11 talk types (e.g., scaffolding, labeling, commenting, directives). Scaffolding, wh-questions, labeling, and commenting were most frequent, though no single talk type dominated overall. A principal component analysis revealed that talk types loaded onto four components, explaining 67% of the variance. These results suggest that parents use a broad and varied range of talk strategies when guiding their children. This study extends previous research on parental input in homes and labs to real-world, ambulatory exploration settings. Future work will explore how these talk types relate to children’s language development.

Shape bias—the tendency to generalize object names based on shape—supports early word learning and is well documented in English-learning children. This study examined its developmental trajectory in a less-studied language, Korean. Using data from 1,597 children, we analyzed shape-based vocabulary development using the Korean version of the MacArthur-Bates Communicative Development Inventory. A total of 225 items were classified as shape-based. These nouns were acquired earlier (median age of acquisition = 25 months) than non-shape-based nouns (median = 27 months). The proportion of shape-based nouns increased until approximately 24 months, after which an inflection was observed (95% CI [22, 27]). A Tobit regression revealed a significant interaction between developmental phase and vocabulary percentile (β = -133.815, p < .001). Beta regression further identified interactions between age, sex, and vocabulary group. These findings offer insight into how perceptual features like shape relate to age and language ability in early vocabulary development.

Foreign-accented speech, shaped by the phonemes of one’s first language, often carries a distinct pronunciation that can (1) signal group membership, (2) mark someone’s linguistic, cultural and socioeconomic background, and (3) make speech more difficult to understand. These features of accented speech are the likely contributors of the early-emerging accent bias in young children too. However, how these mechanisms contribute to the bias is less well understood. Therefore, in this study, we examined both processing and social preferences for accented speech in 4–6-year-old children cross-culturally and cross-linguistically (United States vs. Hungary). Using native and non-native accents, we explored how both familiarity and nativeness influence processing and preference. Generalized linear mixed models revealed that faster reaction times and higher accuracy predicted more positive social judgments (e.g., nice, trustworthy) in both countries. These findings suggest that the cognitive load of processing accented speech may partially explain early-emerging accent-related biases.

We examined the development of left-lateralization for sentence processing and right-lateralization for prosodic processing in a group of 5 – 13 y children. Using a functional magnetic resonance imaging paradigm, sentence processing was examined in inferior frontal and superior temporal lobe, and prosodic processing in inferior frontal lobe. Overall we found developmental changes in lateralization in both language functions. Laterality scores in all three regions were characterized by high variability before 8 y, with lateralization increasing with both age and language proficiency. Adult-like lateralization was achieved around the same time, at about age 8 y. Our findings suggest that language proficiency is accompanied by increasing segregation of these functions in the brain.

This study investigates L2 sensitivity to covert movement constraints by comparing two L1 groups: L1-Korean (L1K) and L1-English (L1E) adult L2 learners of Japanese. Recent experimental studies revealed that in-situ argument wh-expressions exhibit complex NP (CNP) island effects in wh-in-situ languages (e.g., Japanese, Korean), contra long-standing generalization, implying that in-situ argument wh-expressions involve covert movement. This contrasts with English, in which CNP island effects are observed with overt wh-movement. L1K learners are therefore predicted to exhibit CNP island effects in Japanese, as both their L1 and L2 involve covert wh-movement for argument wh-expressions. L1E learners may not exhibit such effects, since overt wh-movement is absent in their L2. Based on the results from a formal acceptability judgment experiment using the factorial definition of island effects, we show that L1K learners, but not L1E learners, exhibit CNP island effects with argument wh-in-situ in Japanese.

Children’s language and social cognition development are deeply intertwined (e.g. kinship terms, audience design, politeness strategies) but often divorced in their studies. This corpus analysis of Korean mother-child dyads (0;8, 1;1, 2;3; N = 10 per group; Ko corpus (Ko et al., (2020))) investigates honorifics in child-directed speech (CDS) as a unique window into this interface. We argue caregivers selectively use honorifics not for static indexing of social hierarchies or for behavioral socialization (e.g. compliments, gratitude, directives), but for dynamic scaffolding of linguistic socialization. Findings establish regular honorific use in CDS, significantly more common than in other-family-directed speech, challenging static social hierarchy indexing. An increasing skew towards Questions in the speech act distribution evidence against pure behavioral socialization, nor is the distribution simply a reflection of the raw frequency of speech acts. Experimental studies can further elucidate this selective socialization’s impact on children’s socio-pragmatic understanding of honorifics.

This L2 study investigates Korean case and scrambling in L1-Mandarin/L1-Russian 8-to-11-year-olds in Korea. The Case-Production Task, assessing basic knowledge of accusative and dative case, has participants (orally) correct sentences with a misidentified direct object vs. goal. The Picture-Selection Task assesses use of accusative- vs. dative-marked nominals in the comprehension of scrambled (OSV; DatSOV) vs. baseline canonical (SOV; SDatOV) sentences. We expect: (i) case and word-order flexibility in Russian will (initially) advantage L1-Russian L2ers; (ii) perceptually-salient dative *–ekey* will be more-readily acquired than accusative *–(l)ul*. In production, lower-proficiency L1-Russian children outperformed their L1-Mandarin peers; all groups’ dative suppliance outpaced accusative. In comprehension, accuracy was target-like on canonical orders but low(er) on scrambled orders, especially among L1-Russian children, with all groups’ DatSOV outpacing OSV. Despite conflicting cross-task L1-transfer results―likely reflecting the differing nature of the tasks―unambiguous dative-over-accusative findings across tasks underscore the role of perceptual salience in acquiring Korean case and scrambling.

Error-based learning proposes that language users revise syntactic representations when prediction errors (PEs) occur. L2 learners generate weaker predictions and have limited opportunities for learning. Grüter et al. (2021) showed that encouraging prediction improved L2 learners’ production of double-object (DO) datives but did not analyze PEs directly or include a delayed posttest. We extended their work by adding a delayed posttest and analyzing match/mismatch judgments—a mismatch indicating a PE. One-hundred-five L1-Korean L2-English learners completed a structural priming task. Experimental learners predicted primes and judged match/mismatch with a native model. Results showed significant long-term gains in DO production at delayed posttest, demonstrating durable learning. Trial-by-trial analysis of PE responses revealed a gradual reduction in PEs, suggesting that learners gradually adjusted their productions to reduce PEs. This study provides direct evidence that active detection and resolution of PEs promote long-term L2 learning, supporting error-based learning and addressing gaps in prior research.

Sound symbolism—the natural link between sounds and meanings—may support early word learning, but its role across languages and developmental stages remains underexplored. This study tested Korean-learning infants’ sensitivity to the buba-kiki effect and their ability to generalize sound-symbolic associations. Sixty-four monolingual infants (14 and 28 months) participated in a Looking-while-Listening task involving sound-symbolically matched and mismatched word-object pairings. Using gaze-switch frequency as an indicator of learning, we fit a binomial generalized linear mixed-effects model, revealing a significant four-way interaction among gaze start, condition, age, and phase type. Fourteen-month-olds showed sensitivity to matched but not mismatched pairings, while 28-month-olds succeeded across all conditions. Only older infants generalized associations to novel stimuli, suggesting flexible learning. These findings suggest that sound symbolism supports early word learning in younger infants, while older infants shift toward learning more arbitrary associations. This developmental change highlights the evolving role of iconicity in language acquisition.

We examined the relationship between parental (novice L2) ASL and child (bilingual L1) ASL from three hearing mother-deaf toddler dyads, focusing on vocabulary and syntax. We measured lexical development using the ASL Communicative Development Inventory (ASL-CDI; Caselli et al., 2020) and overall syntactic development by applying the ASL Index of Productive Syntax (ASL-IPSyn; Lillo-Martin et al., 2024) to language samples of parent-child interactions. ASL-CDI scores for children and their mothers steadily increase over time or are at/near ceiling. ASL-IPSyn scores reflect expected grammatical development over time, with parents consistently scoring higher than their children. Development of ASL-specific grammatical features is more varied; while children make relatively little use of these structures, some parents increase their use over time. These findings are an encouraging indication that hearing parents can support their deaf child’s early ASL development even as they themselves are still learning the language, at least during early childhood.

Where do children think words come from? Children 4-7 tend towards nominal realism, lacking the understanding that names are socially constructed. Can these behaviors be explained through an intuitive theory of language creation and change? In two studies, we asked adults and children why things are called what they are called. Adults most frequently gave explanations using some form of generics, followed by concrete stories. Children were less likely to provide explanations, consistent with prior work, but produced generics and associations nearly equally, with almost no concrete stories. While prior work found children were unaware that names are constructed, when asked who could rename objects that were human-made, children sometimes gave answers such as “people”. However, with natural kinds, they were more likely to attribute the object’s naming to spiritual/supernatural entities. Together, we present preliminary evidence which more closely examines lay-theories of language and set the groundwork for future research.

How does knowing the meaning of a handful of words influence distributional learning and the discovery of the grammatical/semantic properties of unfamiliar words? We tested two non-exclusive hypotheses about how the presence of a small number of meaningful words impacts distributional learning: (1) learners focus their attention and learning on utterances containing meaningful words (e.g., if “modi” means doctor, what might the sentence “modi dax mipen” and the words “dax” and “mipen” mean?) and (2) learners conduct distributional analysis on all utterances to discover distributional categories; meaningful words within a distributional category are used to infer the grammatical and semantic properties of words in that category (e.g., if some words from a distributional category are animate, others from that category likely are too) [‘semantic seed’ hypothesis, 1-3]. In an artificial language learning experiment, we find strong evidence in support of the second hypothesis, but not the first.

Studies on children’s implicature computation have focused on abstract terms like quantifiers, and less on common nouns. This study examined hierarchical implicatures (HI), which arise between hierarchically organized terms (‘dog’ – ‘animal’). For example, given a dog and a mouse, and a speaker who only knows the term ‘dog’, the speaker using the superordinate term ‘animal’ might be referring to the mouse. Otherwise, they would use the more informative term ‘dog’. Children might find HI inferences easier because they involve familiar and contextually accessible terms. When shown images depicting such scenarios, 4- and 5-year-olds (n=120) compute HI only when the group for which the speaker lacked a basic level term was a heterogeneous plurality with different subkinds, not with homogeneous pluralities or singletons. This suggests that 4-5-year-olds can compute implicatures based on hierarchically organized categories when given speaker knowledge of alternative labels, but only with prototypical instances of superordinate terms.

This study examines how adult Mandarin-speaking learners acquire Cantonese nominal expressions, focusing on how definiteness is encoded. While prior L2 research has explored definiteness acquisition in article-based languages, little attention has been paid to article-less target languages. Cantonese nominals, unlike Mandarin, often use the same [classifier-noun] construction for both definite and indefinite reference, relying on context. Mandarin, conversely, has a clearer division where [classifier-noun] primarily denotes indefiniteness. This typological divergence presents a significant learning challenge. An elicited production task was administered to adult Mandarin-speaking L2 learners across three proficiency levels and native Cantonese speakers. Results show natives overwhelmingly prefer [classifier-noun] for definiteness, while L2ers favor [demonstrative-classifier-noun]. Though [classifier-noun] usage increases with proficiency, it remains non-dominant. Beginners notably used [demonstrative-classifier-noun] and bare nouns (which cannot express definiteness) at nearly equal rates. The findings reveal persistent L1 transfer effects and underscore the difficulty of acquiring context-dependent form-meaning mappings.

Language comprehension in neurotypical individuals is predictive; listeners use discourse to update expectations about upcoming words. Both the Predictive Deficit Hypothesis and the Weak Central Coherence theory propose that people with autism may rely less on predictions from discourse, but prior findings are mixed.

By 12 months, naming guides infants’ flexible representations of objects as either individuals or category members. Prior work shows that distinct names support individuation, while consistent names support categorization. Here, we examine whether infants’ individual-level representations formed in a single naming episode are robust and precise enough to support reasoning in dynamic events. In the learning phase, 12-month-old infants viewed four animals labeled with either distinct or consistent pseudowords (Distinct vs. Consistent Names Condition). At test, the most recent object was occluded and then revealed as either the same or a novel animal (No Change vs. Change Condition). As predicted, on Change trials, infants in the Distinct Names condition looked significantly longer than those in the Consistent Names condition; no differences were observed on No Change trials. These findings demonstrate that even a single naming episode enables 12-month-olds to create new individual-level object representations that support reasoning about dynamic events.

This study investigates the different reading behaviors of native and non-native English speakers across sentences that include gender-matched binary pronouns and non-binary pronouns. Two competing hypotheses were tested: L2 speakers adapt to singular “they” through social exposure and demonstrate similar processing difficulty, or they experience additional processing difficulty due to integrating new pronouns into their second language knowledge. A grammaticality maze task was conducted. Linear mixed-effects regression analyses revealed that both native and non-native groups (N = 96) showed significantly longer reading times on non-binary pronouns in sentences with a gender-biased name, such as Mary decided to treat themselves to sushi. Both language groups showed a similar magnitude of the additional processing time on non-binary pronouns. Crucially, the magnitude of the slowdown was not significantly different between groups. These findings suggest that non-native speakers could acquire non-binary pronouns through later familiarization and process them as effectively as native English speakers.

This study investigates how caregiver prompt type influences pragmatic responses in children with ASD during naturalistic interaction. Using data from the ASDBank Flusberg corpus, 180 caregiver-child turns were coded across six children (ages 4–7), categorizing prompts as Directives, Wh-Questions, or Yes/No Questions, and responses as Appropriate, Echoic, Off-topic, or No Response. The results of this study revealed a significant association between prompt and response type (χ² = 30.38, p < .001), confirmed by Fisher’s test (p = 1.88e−05). Specifically, DIR prompts elicited the most total responses, though many were echoic. QYN prompts yielded the highest proportion of appropriate responses (86%) but showed limited elaboration. QWH prompts produced the most varied responses including appropriate and off-topic utterances, suggesting higher levels of pragmatic engagement. These findings support the importance of syntactic structure in shaping language use and highlight clinical implications for prompt-based intervention strategies such as ABA or AAC.

Children’s language input is dominated by short, frequent words, which in adult speech tend to share pronunciations and form homophones. Since homophones pose challenges for word learning, it is unclear whether they are “avoided” in language input—especially in homophone rich-languages like Japanese where the learning difficulty may be heightened. We examined words from maternal speech in longitudinal corpora in English, French, and Japanese to address this question, and we found no evidence of systematic homophone avoidance: homophones in language input still cluster among short and frequent words, and they are not separated in age of occurrence, frequency, or syntactic roles more than expected by chance. Cross-linguistically, homophones are more prevalent in Japanese children’s language input, but they also show more overlap. These findings suggest that children face ambiguous input early on and raise the possibility that homophones may be learned in parallel rather than sequentially.

Our investigation explored whether 15-to-18-month-olds can learn abstract relational terms, such as “same” and “different” from structured contrastive cross-situational labeling. Though infants can represent relational structures before they turn 1, they typically don’t comprehend these terms until age 2.5–3. We hypothesized that natural language environments hinder learning due to ambiguity and lack of explicit labeling. To test this hypothesis, we assigned 74 infants to one of three conditions: Control (no familiarization), Simple Familiarization (alternating identical/distinct image sets, labelled ‘same’ or ‘different’), and Complex Familiarization (similar to Simple Familiarization, but also varying set sizes). All infants completed test trials; looking time to the target pair was measured. Only infants in the Simple Familiarization condition looked significantly more to the target, suggesting comprehension of relational terms. These results suggest infants can learn relational terms with clear input. Future research will examine if findings extend to other abstract terms and enable broad generalizations.

This longitudinal study explores early sensitivity to (1) dialectal variation (Standard Czech vs. Ostrava dialect) and (2) vowel length, a phonemic feature in Czech, in monolingual and bilingual infants aged 4 and 6 months. Infants’ discrimination abilities were assessed in two experiments using a central fixation paradigm. Both groups detected vowel length contrasts, though no differences emerged in dialect discrimination. These results may reflect methodological factors. The early sensitivity to vowel length may be due to the use of a more sensitive paradigm than in a prior behavioral study on Czech-learning infants (Paillereau et al., 2021), and is expected given the perceptual salience and prenatal accessibility of the acoustic cue of duration. In contrast to findings from Catalan–Spanish bilinguals (Bosch & Sebastián-Gallés, 1997), the absence of dialect effects in our study may result from speaker variability in the stimuli, which could have increased task difficulty.

Deictic time words like ‘yesterday’ and ‘tomorrow’ are hard for children to learn because the referents of these words are highly abstract: ‘tomorrow’ does not exist, and ‘yesterday’ is a memory. Consequently, these words can only be understood in the context in which they are produced. For example, ‘yesterday’ means something different on Monday vs. Wednesday. The abstract nature of these deictic time words also poses an interesting problem for researchers who study their acquisition. We tested 3- and 4-year-old children’s comprehension of ‘yesterday’ and ‘tomorrow’ using three tasks that were similar in structure, but differed in their reliance on children’s autobiographical experience and hypothetical reasoning skills. Results indicate that, when asked about their own experiences, 3-year-olds demonstrated an understanding of these time words, but even 4-year-olds struggled to reason about hypothetical events, suggesting that tasks in which time actually passes may be more sensitive to children’s early temporal knowledge.

Mazes are speech disfluencies such as filled pauses, revisions, and repetitions that do not contribute to a speaker’s intended message. They have been found to play a critical role in speech self-repair and planning. Language ability has been demonstrated to be a moderating factor in maze frequency in monolinguals, but previous analyses of mazing in bilinguals have yielded conflicting results. It is unclear how language-specific skills and the overall robustness of the language system contribute to maze usage in bilinguals. Preliminary results suggest no effects of English skills or overall language skills on mazing in Spanish speakers’ L2 narrative retells, a pattern that aligns with findings on mazing in L1 retell tasks in this population. These results indicate that bilinguals may be less sensitive to the effects of language ability on maze usage than monolinguals, and that language-specific and overall language skills may not differentially affect the use of mazes.

While sentence imitation (SI) is widely used in research and clinical assessment, the exact contributions of language skills and working memory remain debated. These contributions may vary with the morphosyntactic complexity of sentences and between children with and without developmental language disorder (DLD). This study examines how receptive vocabulary and working memory – nonword repetition (phonological storage) and listening span (central executive) – relate to SI performance in 63 Czech children with DLD (6;5-9;6) and vocabulary-matched typically developing children (3;7-6;7). Target structures were relative clauses and simple sentences with adjectival phrases, which differ in morphosyntactic complexity but share inflectional features in Czech, a highly inflectional language. Results indicate that better language skills and central executive functioning contribute less to improved sentence imitation in children with DLD than in TD children. Working memory contributions did not differ across sentence types, probably because morphological complexity increases demands on the central executive.

Sentence repetition is the most commonly used measure of sign language proficiency but how it relates to comprehension is currently unknown. Here we compare sentence structure comprehension using a sentence-picture matching experiment, with sentence repetition using the same mono- and biclausal structures in ASL with different lexical items. We also administered digit span tasks in sign, Visual Puzzles, and a reading assessment. Thirty-two deaf/hard-of-hearing adolescents currently educated with sign language between the ages of 10 and 17 with a mean age of 14 years participated. Sentence repetition predicted sentence comprehension showing that the learning of abstract sentence structure governs both comprehension and production in sign language, as in spoken language. Sentence comprehension predicted reading comprehension, more so for biclausal than monoclausal sentence structures. Forward digit span weakly related to sentence repetition while backward digit span was weakly related to comprehension and Visual Puzzle performance was unrelated.

Adult and child listeners rapidly use morphosyntactic cues during real-time comprehension of their first language. Later learners, in contrast, may not use cues unless they are also available in their L1. Can L1 Arabic late learners of English use number-marked verbs to anticipate an upcoming noun? Arabic has subject-verb agreement, but does not have an overt present-tense copula. Using a visual-world eye-tracking task, L1 English and L1 Arabic participants heard a sentence (there are the good apples) paired with pictures that were either different or the same in number (2 apples, 1 carrot; 2 apples, 2 carrots). Advantages in different over same trials suggest that both groups use the number-marked verb in real-time comprehension, and that L1 English participants use it in both singular and plural trials, while L1 Arabic only showed evidence in plural trials. Morphosyntactic cue use during comprehension appears possible even with relatively modest overlap between languages.

This study examines whether Japanese learners of English (JLEs) can acquire abstract syntactic constraints, focusing on the Anti-locality constraint in English tough constructions (TCs). Anti-locality, violated in structures like John is easy to please with overt objects, contrasts with degree constructions (DegC), which permit subject readings via intervening structure. Thirty-six JLEs completed an Acceptability Judgment Task (AJT) and a Self-Paced Reading Task (SPRT). Cluster analysis of AJT results identified three learner profiles: Cluster 1 (n=12) showed native-like sensitivity to Anti-locality, Cluster 2 (n=6) rejected all TC variants but accepted DegC/EagC, and Cluster 3 (n=18) accepted all forms, showing L1-based interpretations. All groups preferred subject readings, especially in EagC. SPRT results revealed no online sensitivity to Anti-locality violations, even among accurate offline responders. These findings suggest that L2 syntactic development is shaped by both knowledge source (UG vs. L1) and processing status (competence vs. performance), motivating a 2×2 developmental framework.

Negation, though abstract, is universal across languages. Homesign systems–systems of signs and gestures created by deaf individuals who lack accessible external language–offer insight into how human cognition shapes linguistic negation. In this study, we elicit negation from adult Nicaraguan homesigners. Participants described videos showing successful or failed actions (e.g., opening a door). As predicted, participants often used negators to describe failures, and rarely to describe successes, regardless of whether they viewed the success videos or the failure videos first. All participants preferred post-predicate manual negation and tended to produce headshakes that took scope over manual negators alone. However, participants varied in how frequently they produced manual and nonmanual negation simultaneously. These findings suggest that homesign negation is relatively systematic. Some features are consistent across individuals, suggesting shared biases, while others reflect individual variation. Additional analyses will explore other nonmanual markers and negation produced by the homesigners’ communication partners.

The interpretation of the conjunction “and” and disjunction “or” in negative sentences exhibit cross-linguistic variations. The negated conjunction and negated disjunction have disjunctive and conjunctive interpretations in English and Korean, while they have opposing interpretations in Japanese and Mandarin. These variations are attributed to scope interactions between the logical connectives and negation. Previous L2 studies suggests L1 transfer in interpreting negated disjunctions. Extending them, we investigated the L3 acquisition of negated logical connectives to determine the role of L1. We examined interpretations of Japanese disjunctions and conjunctions with negation by Mandarin L1, English L2 learners of L3 Japanese with comparison to Korean L1, English L2 learners of L3 Japanese. The results show that both L3 groups were accurate in interpreting negated disjunction and conjunction, but the L1 Mandarin group responded more categorically than the L1 Korean group, suggesting that the L1 plays a role in L3 acquisition.

This study examines how native English speakers (NSs) and Japanese learners of English (JLEs) process adverbial intervention in that-trace configurations, focusing on the role of syntactic constraints and cognitive load. Across three experiments—two acceptability judgment tasks (AJTs) and a self-paced reading task (SPRT)—we tested the effects of adverb type (sentential vs. manner) and position on sentence acceptability and processing difficulty. While both groups showed sensitivity to adverb-type congruency in single sentences, adverbial insertion did not improve acceptability in embedded that-trace configurations. Instead, it increased processing cost, especially when sentential adverbs intervened. These findings suggest that although syntactic knowledge is applied in online processing, it does not extend to embedded environments. Both NSs and JLEs may rely on linear order over hierarchical structure in real-time processing. Our results challenge the assumption that adverbial insertion repairs that-trace violations and highlight complexities in embedded clause processing in L1 and L2 grammars.

This study investigates L2 sensitivity to covert movement constraints by comparing two L1 groups: L1-Korean (L1K) and L1-English (L1E) adult L2 learners of Japanese. Recent experimental studies revealed that in-situ argument wh-expressions exhibit complex NP (CNP) island effects in wh-in-situ languages (e.g., Japanese, Korean), contra long-standing generalization, implying that in-situ argument wh-expressions involve covert movement. This contrasts with English, in which CNP island effects are observed with overt wh-movement. L1K learners are therefore predicted to exhibit CNP island effects in Japanese, as both their L1 and L2 involve covert wh-movement for argument wh-expressions. L1E learners may not exhibit such effects, since overt wh-movement is absent in their L2. Based on the results from a formal acceptability judgment experiment using the factorial definition of island effects, we show that L1K learners, but not L1E learners, exhibit CNP island effects with argument wh-in-situ in Japanese.

How do toddlers manage semantic interference during word recognition? Prior work shows vocabulary size contributes: Children with larger vocabularies recognize words faster, and experience less interference from unnamed semantically-related distractors. Inhibitory control (IC) may also contribute by inhibiting related word-referents when selecting correct word meanings. We directly tested this hypothesis by asking how 24-month-olds’ IC contributes to lexical recognition under semantic interference? Toddlers were eye-tracked while probing recognition of familiar words in two conditions of semantic overlap: (1) perceptually related (PR; DOG-TABLE) and (2) taxonomically and perceptually related (TPR; DOG-HORSE). We failed to observe any effects of IC on word recognition, but showed nuanced effects of vocabulary size and word recognition under semantic interference. Specifically, our results suggest that larger vocabulary size, while contributing to more conflict in word recognition overall, may also benefit from taxonomic information that helps manage competitor activation and lexical recognition as vocabularies grow.

Discriminating between languages is essential in early development and particularly relevant for bilinguals. By 4 months, infants can distinguish rhythmically similar languages and dialects. Prior research shows that monolinguals typically shift from a familiarity to a novelty preference, while bilinguals exhibit a preference for the unfamiliar variant from early on. This fNIRS study investigated neural responses to dialectal variation in 93 Czech-learning infants (59 monolingual, 34 bilingual) at 4 and 6 months. Infants heard blocks of sentences in Standard Czech and the Ostrava dialect. Behaviorally, no group differences in attention time were found. fNIRS results revealed a language × condition interaction in the right posterior temporal regions: at 4 months, monolinguals showed stronger activation to Standard Czech compared to both the Ostrava dialect and bilinguals; by 6 months, activation increased for the Ostrava dialect. These preliminary findings reveal distinct neural activation patterns as a function of language background and age.

This study examined whether working memory (WM) capacity influences sentence comprehension in Japanese-speaking children, focusing on differences between active and passive sentence processing. Fifty-three children (aged 4–6) completed a sentence comprehension task using four sentence types (intransitive active, transitive active, full passive, short passive) and a digit span task measuring forward and backward WM. A generalized linear mixed model revealed that comprehension improved with age only in the transitive active condition. In younger children, intransitive actives were best understood; older children also performed better on transitive actives than on full passives. Another model showed that forward span predicted comprehension in active conditions but not in passives. Children with higher WM capacity showed significantly better performance on transitive actives than full passives. In contrast, backward span did not predict comprehension performance. These findings suggest that WM supports syntactically simpler sentence processing, with forward span playing a stronger role than backward span.

This study examined how maternal education and psychosocial factors relate to low-income, Latinx mothers’ speech to their infants (n = 103; Mage = 13.12 months, SD = 3.95, 47 males). Mothers completed questionnaires assessing psychosocial factors, including financial strain, mental health, parenting stress, and social support. Additionally, mothers and their infants participated in a 15-minute semi-naturalistic interaction. Results suggest that maternal education is a stronger predictor of child-directed speech than all psychosocial factors. Notably, maternal education and perceived social support interacted, such that high social support buffered the effect of maternal education on child-directed speech. Neither psychosocial factors nor maternal speech variables significantly predicted child tokens, precluding mediation analyses. These results highlight maternal education as a key driver of early language input and underscore the protective role of social support. Furthermore, findings expand our understanding of how parental factors shape children’s language environments and development, particularly in lower-income Latinx families.

Theories of second language speech acquisition agree that one’s native language (L1) substantially impacts the acquisition of second languages (L2). At early stages of learning, non-native speech sounds are assimilated or compared to native speech categories (Best, 1995; Best & Tyler, 2007; Flege, 1995). We introduce a novel computational framework for predicting L1 assimilation effects using optimal transport, a method of determining optimal coupling between two probability distributions according to a cost metric. We frame the distributions here as an acoustic distribution over sounds and a category distribution over phoneme labels and the cost metric as the negative log probability (surprisal) of a particular acoustic value corresponding with a particular category label in the L1. We show that our framework captures the qualitative empirical findings in Spanish speakers’ learning of English stops.

This study examines how native speakers (NSs) of a Double Negation (DN) language acquire Negative Concord (NC) in a second language (L2). Previous work has found object Negative Concord Item (NCI) constructions with sentential negation to be problematic for L1-English L2-Spanish learners. However, little is known about L2 acquisition of subject NCI constructions in a language such as Russian, where a negative marker is required whenever subject and/or object NCIs occur. Data collection is ongoing. So far, 12 Russian NSs and 11 L1-English L2-Russian participants (L2ers) have participated in an Acceptability Judgment Task and Picture Matching Task. L2ers are target-like in accepting grammatical NC constructions, but overaccept ungrammatical NC constructions with subject NCIs. Their semantic judgments are completely native-like. The L2ers tested so far are quite advanced; only remnant transfer effects are detected in acceptability judgments. We are recruiting less proficient L2ers in order to examine L1 transfer effects.

Goro & Akiba (2004, GA) famously showed that Japanese children interpret disjunction ‘ka’ under negation (neg-OR) conjunctively – a dominant option for English-speakers, ungrammatical to Japanese adults – subsequently proposing an input-driven Semantic Subset Principle (SSP) (Goro 2007, Crain 2012). We show that children correctly interpret neg-OR disjunctively (contra GA) when the VP is of a particular verb class (Vendler 1957) – a finding inconsistent with the SSP. We show that Japanese children have full grammatical competence, challenging the SSP account.

This study examines how Italian-speaking children develop the ability to interpret and produce distributive and collective meanings in both ambiguous and unambiguous sentences. Sentences like “the children are buying a book” are ambiguous between a collective and a distributive reading. While children often accept distributive interpretations of such sentences, they struggle with using distributive quantifiers like “each” appropriately. We tested 31 preschoolers, 44 second graders, and 32 adults using visual scenes and tasks assessing both comprehension and production. Children were generally accurate in interpreting marked sentences, and all age groups preferred collective readings for unmarked sentences. However, in production, preschoolers rarely used explicit markers, unlike older groups, who favored distributive ones. This dissociation between comprehension and production suggests that although children grasp distributive meanings early, they find it harder to express them. The results support the idea that collective readings are more accessible and may serve as the default interpretation.

Previous studies with children under 5 years of age have shown that firstborns have better language skills than secondborns (e.g., in France, Japan). This has been explained by the unequal distribution between siblings in parental linguistic inputs. Previous studies with English-speaking children under 5 years of age have shown that firstborns receive more words (Gilkerson & Richards, 2009) and are read to more frequently (Raikes et al., 2006) than laterborns. To test the universality of the unequal distribution among siblings in parental linguistic inputs, we focused on picture book reading in Japan and analyzed web-based survey data (N = 1,750). Consistent with previous findings with English-speaking parents, our findings showed that Japanese-speaking parents read more picture books to their firstborns than to their laterborns. This suggests that the unequal distribution between siblings in parental linguistic inputs is universal across cultures, and therefore the birth order may affect early language development.

In Mandarin, the head noun in [modifier clause-de-N] encodes not only the arguments or adjuncts of the clausal verbs (e.g., “kids/biscuits/plate” in the event of “kids eating biscuits”), but also non-argument/adjunct participants extended from the clausal event (e.g., “crumbs” resulting from the event of “kids eating biscuits”). This article adopts a typological analytic framework of noun-modifying clause constructions (NMCCs) and proposes a developmental hierarchy of three types of NMCCs: argument > adjunct > extended. We tested the hierarchy in two studies: an analysis of 732 tokens of NMCC produced by 369 children (1;2-6;9) in naturalistic corpora, and a sentence repetition task conducted with 121 children (3;0-4;11) in China. Our findings show that NMCC emerges surprisingly early before 2;0 in Mandarin children. Argument NMCCs show developmental primacy in onset age, frequency, and accuracy. Adjunct and extended NMCCs are acquired at a later stage, consistent with the proposed hierarchy.

Previous research has shown that content words are acquired earlier than function words, but the exact trajectory and order of function word acquisition has remained relatively understudied. In this study, we use the largest available child language corpora as well as Bayesian growth curve modeling to estimate the population level onset of production for more than 100 English function words. Our estimates suggest that for the large majority of function words, the earliest age of production lies between 12-24 months. A linear regression found longer function words as measured by the number of phonemes and function words with higher Mean Length of Utterance (MLU) have later estimated onset age of production. We did not find a significant effect of frequency in child directed-speech on onset age of production for function words. Overall these results point to early emergence of abstract functional morphemes with production limitations as the main bottleneck.

Existing vocabulary strongly predicts reading success. However, static vocabulary measures, assessing current existing vocabulary (pre-existing knowledge), are influenced by varied prior experiences, potentially obscuring true learning capabilities. Dynamic assessment (DA) of novel word learning (NWL) evaluates learning potential with less experiential bias, offering crucial insights. While some research links NWL to literacy development, establishing its unique prediction of later vocabulary, comprehension, and fluency beyond static vocabulary is vital for understanding development. This longitudinal study (N=172, 2nd-4th grade) examined if NWL capacity uniquely predicts these outcomes, independent of initial vocabulary. Second graders learned novel nonwords via tasks using phonological and rich semantic information for object pairings. Analyses show 2nd-grade NWL significantly explained unique variance in 4th-grade receptive vocabulary (6%), comprehension (WRMT-3 Passage Comprehension) (3%), and reading comprehension and fluency (GORT-5 Oral Reading Index) (5%), controlling for 2nd-grade vocabulary. These findings suggest dynamic NWL assessment provides distinct information complementing static measures.

Parental gesture and speech are critical to early language development, yet little is known about how parents adapt their input when interacting with children with developmental disorders such as Down Syndrome (DS). This study examines multimodal communication of 28 mothers interacting with their children with DS (Mage = 14.38 months) during a 15-minute semi-structured play session. We asked whether parental input varies as a function of children’s language and motor abilities. Results revealed that gesture complexity varied significantly, with mothers of children with higher language and motor abilities producing more complex gestures (MLanguage= 17.21; MMotor= 15.58) than those of children with lower abilities (MLanguage= 9.71; MMotor= 9.00; p’s<0.05). Notably, these effects did not extend to speech input, which showed no significant group differences. These findings suggest that when responding to their child’s developmental profile, parents may rely more on gesture than on speech to adjust their communication.

Do infants rely on language-specific phonotactic cues or domain-general transitional probability cues when segmenting words of variable length? Using HPP (Johnson & Tyler, 2010) we tested 5- and 8-month-olds (n=95). Infants heard an artificial language with 2 disyllabic and 2 trisyllabic words. In one language, statistical words ended in tense vowels ([i], [u]), consistent with English phonotactics. In the other, statistical words ended in lax vowels ([ɪ], [ɛ]), this is illegal in English. Contra findings on words of equal length (Katsuda & Sundara 2024), when tested with variable length words, 5- and 8-month-olds showed sensitivity to phonotactics. 5-month-olds were only sensitive to phonotactics. 8-month-olds successfully segmented, for the first time, words of variable length, only when the two cues cooperated. Our findings challenge the transitional probabilities first hypothesis (Thiessen & Saffran, 2003). Rather, infants rely on phonotactics to segment words, when transitional probability cues are unreliable as in natural languages.

While children show understanding of tense/aspect markers in finite clauses from age 2–3, it has not been previously tested whether they comprehend the temporal cues available in nonfinite clauses (which lack tense marking). Our study examined 4-year-old children’s understanding of tense/aspect markers in both finite and nonfinite embedded clauses. Participants placed a character, Mickey, in a scene from an obstacle course, based on another character’s hints. Hints were future- or past-oriented; critical trials involved finite and nonfinite clauses embedded under expect (e.g., nonfinite: “I expect Mickey {to go (future); to have gone (past)} down the slide!”). Children placed Mickey after the mentioned obstacle more often for past-oriented hints than future-oriented hints (p<0.01). This suggests that by age 4, children understand tense/aspect markers robustly across both main clauses and finite and nonfinite embedded clauses.

In a forced-choice task with preschoolers and adults, we investigate how the surface-level distribution of novel adjectives signals their meaning and the properties they denote (affective emotion states or perceptible surface-level properties of color, shape, or size). We hold the animate subject constant and make manipulation of the target properties plausible by contrasting aliens. We show that certain linguistic contexts (predicative position v. prenominal position, and presence of a syntactic complement such as a PP or a ±finite clause) conspire to draw children’s attention towards a contrast in emotions and away from a contrast between other perceptually-salient properties (color, shape, size). Our findings pave the way for a broader investigation of how language supports the acquisition of a wider range of adjectives, particularly those that take syntactic arguments and those that denote internal states, allowing us to precisify the role of syntax in the acquisition of adjectival semantics.

Conversational turn-taking (CTT) is a cornerstone of language development (1). Substantial growth in CTT skills occurs during the early childhood years, perhaps in part because children go to preschool and gain experience interacting with peers (2, 3). Preschool is a particularly critical time for deaf preschoolers because 90-95% of them do not have someone to sign with at home (4). Prior research on CTT among deaf children in the preschool setting is limited, but suggests that deaf children (across educational settings) experience fewer interactions and more failed initiations with peers (5, 6, cf. 7). Deaf signing kids (N = 8, M = 3.95 years, SD = 0.41 years) needed teachers (N = 4) to scaffold most of their interactions, and peer-to-peer interactions were infrequent. Children were prolific users of various communicative gesture, especially pointing and showing. Frequency of conversational turn counts, conversations participated in, and gestures produced are reported.

Recent research highlights the positive influence of maternal responsiveness and grammatical and lexical richness of input on expressive vocabulary development in late talkers. However, although syntactic weakness is one characteristic of late talkers, the specific contribution of maternal input quality to their syntactic development remains under-explored. This study analyzed 76 naturalistic mother–child conversational samples from late talkers (ages 2;6 to 3;6) in the CHILDES database to investigate this issue. Findings revealed that maternal syntactic complexity and lexical diversity significantly predicted children’s syntactic development. In terms of vocabulary, a marginally significant effect of maternal lexical diversity on children’s lexical gains was observed. A trend-level interaction suggested that the effectiveness of lexical input may depend on the level of syntactic complexity in the caregiver’s speech. These findings underscore the importance of targeting both syntactic and lexical features in caregiver-focused interventions aimed at supporting late talkers’ language development.

Human language learning is marked by the remarkable ability to generalize linguistic rules from minimal exposure, in contrast to the memorization-heavy statistical learning seen in Large Language Models (LLMs). This study compared humans and ChatGPT-4o using a structured artificial naming task, where two-syllable words encoded both plant type and plant color. Participants were tested on their ability to recall labels from training and generalize them to novel stimuli. ChatGPT-4o significantly outperformed humans in recalling trained items, reflecting its strong memorization capabilities. However, humans showed a clear advantage in generalization, particularly in accurately applying second-syllable rules when labeling new items, demonstrating a flexible understanding of underlying structure. These results highlight core cognitive differences between human learners and LLMs, suggesting that LLMs may struggle with deeper abstraction and linguistic flexibility. Future research should improve methodological diversity and continue benchmarking new LLMs to better understand the boundary between statistical learning and human-like generalization.

Bilingual semantic priming research typically relies on target language norms from native speakers, potentially misrepresenting L2 learners’ conceptual networks. This study examines within-language semantic priming in L1 and L2 using stimuli derived from Chinese-English bilinguals’ L1 associative networks. Following three-stage validation from the Small World of Words Chinese database and two norming studies, 92 critical word pairs were selected with associative strengths of >0.3 and high translation consistency. One hundred highly proficient Chinese-English bilinguals will complete lexical decision tasks with a 200-ms SOA—fifty in Chinese, fifty in English. Preliminary data from eleven participants reveal 9-ms priming in L1 versus 51-ms in L2. This pattern aligns with Smith et al. (2019) but diverges from earlier studies showing stronger L1 priming. The larger L2 effect may reflect lower resting-level activation of L2 targets benefiting maximally from L1-grounded semantic relationships, suggesting robust L2 form-meaning connections when conceptual knowledge transfers from native networks.

This study examines whether children acquiring a wh-scope marking (WSM) grammar display consistent behavior across comprehension and production. Building on Lutken et al. (2020), who found no such correlation in English-learning children, we replicated their study with 45 Hindi-learning children aged 3;01–6;11. Participants completed both comprehension and production tasks involving long-distance questions. Additionally, 15 children were tested with ungrammatical lead-ins to assess their tolerance for deviant structures. The results revealed a clear asymmetry: children who misinterpreted WSM constructions in comprehension frequently produced target-like forms in production. Notably, Hindi-learning children demonstrated strong grammatical conservatism—13 of 15 spontaneously corrected incorrect scope markers, and 43 of 45 produced either fully adult-like or partially target-like questions. These findings suggest that children’s grammars may differ across modalities and that non-target-like comprehension may reflect exploratory strategies rather than immature grammar. The study challenges the assumption that consistent cross-modal performance is necessary.

Research on the relative clauses (RCs) of L1-English adults indicates that island effects are stronger in subject RCs (SRCs) than in object RCs (ORCs). This study investigates whether the asymmetry holds for L2-English adults. Participants―English native-speaker controls, L1-Korean L2ers, L1-Mandarin L2ers―were distributed across two sub-studies: one on ORCs, one on SRCs, both including an elicited production task (EPT) and an acceptability judgment task (AJT). EPTs showed (a) gapped RCs were rarer in SRC islands than in ORC islands and (b) resumptive RCs significantly outnumbered gapped RCs only in SRC islands. AJTs showed it was only in SRC islands that (a) gap-trial ratings fell consistently in the lower half of the acceptability scale and (b) resumption-trial ratings were significantly higher than gap-trial ratings. These results indicate that WH-island SRCs are stronger islands than WH-island ORCs, not only for L1-English adults but also for L2ers whose L1 lacks overt WH-movement in WH-questions.

Q&A Time: Friday, November 7, 3:30 – 3:41

Bilingual children’s L2 skills are crucial for academic success and psychosocial well-being. Unlike their hearing peers, bilingual hard-of-hearing (BiHoH) refugee children face additional challenges and often have limited language exposure before migration. This study examines morphosyntactic development in German spoken language (DLS) and sign language (DGS) in five BiHoH refugee children, compared to five age-matched bilingual hearing children and children with DLD. Using German LITMUS sentence repetition tasks for spoken (DLS-SRT) and sign language (DGS-SRT), we assessed complex morphosyntax. BiHoH children faced significant difficulties in sentence repetition compared to hearing peers; by T2, they showed growth but some performed similarly to children with DLD, corroborating findings on HoH children’s marked difficulties in complex morphosyntax. Concerning DGS, BiHoH children performed within the range of late monolingual signers, with some performing better on DGS-SRT than DLS-SRT. Results further revealed heterogeneous profiles among the BiHoH children attributable to differences in cumulative language experiences.

Q&A Time: Friday, November 7, 3:41 – 3:52

Within the study of morphosyntactic acquisition, grammatical gender is often a point of interest. In a language like Irish, many factors lead to difficulties in acquisition even in L1 speakers. Alongside its status as a minoritized language in heavy contact with English, Irish grammatical gender is opaque and most, if not all, of its speakers are bilingual. Studies have shown that opaque gender systems have protracted acquisition periods, though ones on Irish have focused on children older than three and used monolingual/bilingual homes as a stand-in for input frequency. This study uses a corpus to directly compare a child’s gender agreement patterns to their parent’s with the aim to analyze both how Irish children under three utilize the grammatical gender system and how that relates to their parents’ usage of this system.

Q&A Time: Friday, November 7, 3:52 – 4:03

Everyday caregiver-infant interactions are dynamic and multimodal. The activity contexts in which they occur both vary from day to day and provide consistency that supports infants’ learning. To extend beyond existing research on how activity context influences caregiver-infant communication – which primarily focuses on the extent to which context shapes caregivers’ use of speech – we examined how caregivers’ use of multimodal infant-directed communication (IDC; including speech, action, gesture, emotion, and touch) varies from day to day and across contexts. Caregivers’ use of IDC was most similar when it involved the same caregiver engaged in the same activity on two different days, with the next strongest similarity for different caregivers participating in the same activity. However, IDC use varied markedly across activity contexts. These results suggest that, while caregivers do provide a source of day-to-day stability in everyday communication, specific activity contexts also strongly shape the caregivers’ multimodal behavior.

Q&A Time: Friday, November 7, 4:03 – 4:14

Tone has long been argued to possess greater phonological flexibility than segmental features, with properties that surpass those of segmental and metrical systems. This asymmetry is evident in non-local processes such as tone plateauing, first-last harmony, and phonological counting. This study provides the first experimental evidence for the learnability-based explanation of the tone-segment asymmetry in morphophonology. Specifically, it tests whether the typological absence of segmental counting rules arises from learning biases. We find that an unattested segmental counting rule is significantly harder to learn than its tonal counterpart in both tonal and non-tonal speaking populations, supporting the hypothesis that typological asymmetries may reflect synchronic constraints on acquisition.

Q&A Time: Friday, November 7, 4:14 – 4:25

Some studies suggest that adults parse linguistic information differently depending on how it is conveyed (e.g., asserted, conventionally implicated, presupposed), but that the way the information is packaged does not affect memory recall. Other studies report that recall of presupposed information differs across triggers. Through two experiments, we investigate whether children treat information differently depending on how it is packaged. Participants saw cartoon aliens and were presented with facts about them (which either asserted, presupposed, conventionally implicated, or conversationally implicated a piece of information). Participants were subsequently asked to recall whether the description had contained the target information. The findings suggest that children treat information differently depending on how it is packaged. They successfully add asserted information to the Common Ground and recall it accurately. Presupposed and implicated content, however, vary by trigger/expression: only certain presuppositions, and conventional implicatures, but not conversational implicatures, are treated as robustly as asserted content.

Q&A Time: Friday, November 7, 4:25 – 4:36

Stem-final Turkish stops alternate between voiced before vowel-initial suffixes and voiceless elsewhere for some but not all nouns. Elicitations of the accusative of known nouns showed that children’s alternation errors were exclusively underapplying the alternation and that those errors decrease with age. When deciding whether nouns alternate, Turkish adults show sensitivity to some but not all lexical statistical regularities: they appear sensitive to the identity of the stop and the prosodic shape of the noun, but not the quality of stop-preceding vowels, all of which are statistically predictive of (non-)alternation in a corpus. A learning-based view is that learners construct a grammar to sufficiently account for where alternation occurs and thus only track dependencies when generalization is untenable without them. This proposal—but not one where learners match the frequency of alternation for categories of words shaped by UG constraints—is consistent with Turkish child language acquisition results.

Q&A Time: Friday, November 7, 4:36 – 4:47

Using a Truth-Value Judgment Task, we examined Turkish-speaking children’s interpretations of three disjunction markers (veya, ya da, and ya…ya…). Results from 58 children and 59 adults revealed that children’s interpretations were conjunctive only with simplex forms, supporting the ambiguity account. Moreover, even when the stronger alternative (i.e., and) was salient in one of the disjunction marker’s morphological makeup (i.e., veya), no children computed exclusivity implicatures. These findings suggest that children’s non-adult-like interpretations might stem from the ambiguity of disjunction and call for a more nuanced alternatives-based account that explicitly integrates not just the salience but also the relevance of alternatives in implicature computation.

Q&A Time: Friday, November 7, 4:47 – 4:58

Some children with autism acquire foreign languages non-interactively through exposure to digital media, also termed non-interactive language acquisition. This study examined narrative abilities in English and Hebrew among 54 multilingual children aged 4;5–12, in three groups: children with autism who acquired English non-interactively (ASD-NI), children with autism who acquired English naturalistically (ASD-Nat), and non-autistic peers (TLD-Nat). The narratives were analyzed for verbosity, lexical richness, syntactic complexity, and the use of referential expressions. All groups produced comparable narratives across languages. Children in the ASD-Nat group used more ambiguous referential expressions. English narratives were more ambiguous than Hebrew. Some children exhibited cross-linguistic influence, such as pronoun omission in English. Findings suggest that non-interactive language acquisition can yield functional language skills in some individuals with autism, while subtle qualitative differences persist. These findings have both clinical and theoretical implications, which are discussed.

Q&A Time: Friday, November 7, 4:58 – 5:09

This study investigated the role of working memory (WM) in 4- and 5-year-old children’s production of relative clauses (RCs). The production of complex structures like RCs is well suited for investigating the role of WM, because it requires the integration of various linguistic components across potentially long distances within a sentence, thereby directly engaging WM capabilities. Children’s production of RCs was examined using an elicited production task and their WM capacity was assessed using a 0-back task and a nonword repetition task. We found that the 4-year-olds exhibited significant effects of the nonword repetition and the 0-back tasks, suggesting their reliance on phonological memory and central executive. In contrast, the 0-back task had no significant effect on the 5-year-olds’ production, indicating a reduced reliance on central executive when producing complex structures. We discuss the findings in relation to the dynamic interaction between different WM components and children’s grammatical knowledge.

Q&A Time: Saturday, November 8, 3:30 – 3:43

Previous studies report that while adults tend to interpret disjunction exclusively (‘A or B, not both’), English- and French-speaking children access inclusive (‘A or B, or both’) or conjunctive (‘A and B’) interpretations. This corpus study aims to better understand the evidence that children receive for disjunction in the parental input. We analyzed occurrences of the French disjunction “ou” in the Paris corpus (CHILDES) and compared the results to English data reported in Jasbi et al. (2008). The input for both languages provides ample evidence for exclusive interpretations, which is striking in light of the near absence of exclusive interpretations by children in experimental studies. Francophone parents also produce a larger proportion of exclusive disjunctions than anglophone parents, despite children displaying the same experimental profiles in the two languages. The findings further highlight the asymmetry between what children hear in the input and the interpretations they display in experimental settings.

Q&A Time: Saturday, November 8, 3:43 – 3:55

This study investigates the acceptability of bare nominals (BNs) in preverbal position in the interlanguages of Brazilian Portuguese (BrP) speakers acquiring English, French, and Spanish. Grounded in the Minimalist Program and the DP Hypothesis, it explores how the formal features of BNs vary across languages. Drawing on Roberts’ (2019) parameter hierarchy and Longobardi’s (2008) Person feature framework, we classify BrP–English differences as microparametric and BrP–French/Spanish differences as mesoparametric. A mixed-methods design using an Acceptability Judgment Task (AJT) was employed with 120 BrP-speaking learners and native controls. Results confirmed strong L1 transfer, with learners rating bare singulars as more acceptable than native speakers. No significant effects of proficiency level were found. The learners showed greater convergence with English than with French or Spanish, supporting the hypothesis that parametric proximity facilitates acquisition. Findings underscore the influence of L1 and the role of parametric distance in the L2 acquisition of articles.

Q&A Time: Saturday, November 8, 3:55 – 4:08

Multilingualism affects children’s cognitive and communicative development. Verbal Fluency (VF) tasks reflect both domains. We present results of a study measuring phonemic and semantic VF with 320 children (ages 2.8–11 years, 214 typically developing (TD) and 105 autistic (ASD) children), assessed for multilingual profiles with the Q-BEX questionnaire. Our findings reveal that ASD and TD children perform similarly in VF tasks, and increase word production with age, especially in semantic VF. For TD-children, a beneficial effect of multilingualism (operationalized as a continuous variable via cumulative exposure entropy) emerges with age, replicating earlier reports of multilingualism benefits on VF and cognitive benefits associated with high exposure entropy. For ASD-children, multilingualism has neither positive nor negative effects on VF. Our results support recent findings suggesting that multilingualism offers a host of potential benefits and to date no discernible disadvantages for children’s development, including for children experiencing communicative and cognitive challenges.

Q&A Time: Saturday, November 8, 4:08 – 4:20

Heritage language (HL) children develop their languages at different rates due to child-internal and -external factors. Cognitive skills like verbal short-term memory (VSTM) and non-verbal analytic reasoning (NVAR) influence abilities in HL and the majority language (ML). Input richness especially affects HL, which is more vulnerable due to limited resources. This study investigated 31 children’s vocabulary and morphosyntax in Farsi (HL) and German (ML). Cognitive abilities (VSTM, verbal working memory (VWM), NVAR) and input richness (via parental questionnaires) were analyzed with age and age of acquisition as covariates. Results showed that stronger VSTM predicted better HL vocabulary and morphosyntax in both languages, while NVAR predicted only ML abilities. VWM had no significant effect. Input richness positively affected HL outcomes but not ML. These findings suggest that children with weaker NVAR could benefit from diverse L2 input, and increasing HL input richness may support HL maintenance and reduce attrition risk.

Q&A Time: Saturday, November 8, 4:20 – 4:31

This study investigates whether Japanese learners of English (JLEs) construct interlanguage grammar consistent with the Uniformity of Theta Assignment Hypothesis (UTAH), a principle of Universal Grammar, focusing on unergative and unaccusative intransitives. Previous research, such as Oshita (2000), reported that JLEs produced it-V-NP structures with unaccusatives. While Oshita viewed it as an expletive, learners might interpret it as a referential pronoun. According to UTAH, unaccusatives reanalyzed as transitives by JLEs could accommodate an additional argument in the vacant VP-specifier, unlike unergatives where the Agent fills that position. To test this, we conducted an acceptability judgement task with 62 intermediate JLEs and 51 native speakers. Linear mixed-effects analysis revealed that JLEs rated it-V-NP and NP-V-NP structures similarly, with higher acceptability for unaccusatives than unergatives. These findings suggest that JLEs construe it as referential in it-V-NP structures and are more prone to reanalyzing unaccusatives as transitives, reflecting UTAH in their interlanguage.

Q&A Time: Saturday, November 8, 4:31 – 4:44

Focus particles like “only” establish a contrast between the focused element and a set of alternatives. Previous research suggests that young children often misinterpret “only”-sentences, possibly ignoring “only” altogether. This study investigated whether children truly ignore “only”, using a selection task designed to disentangle interpretations of sentences with and without “only”. Thirty-three English-speaking children (ages 4;03–6;02) and 36 adults completed a covered picture task in which they selected images based on clues containing sentences with or without “only”. On critical trials, the correct interpretation required choosing the covered picture, consistent with the exclusivity meaning of “only”. The results show that both children and adults selected the covered picture significantly more in the “only”-condition compared to the “only”-less condition. These findings indicate that children do not ignore “only”, but instead use visual alternatives to interpret it in an adult-like way, challenging previous claims of developmental delay in focus particle comprehension.

Q&A Time: Saturday, November 8, 4:44 – 4:57

Cross-linguistic research shows that subject relative clauses (SRCs) are acquired earlier than object relative clauses (ORCs) due to structural differences: SRCs follow canonical SVO order, while ORCs involve non-canonical object fronting. ORCs tax working memory (WM), which is often limited in autistic children. Bilingual autistic children may show WM advantages, raising the question of whether bilingualism supports syntactic processing. We tested 71 autistic children (38 bilinguals, 33 monolinguals), aged 3–12, speaking English, French, or German, using a character-selection task targeting SRC/ORC comprehension and a nonword repetition task indexing WM. GLMMs revealed SRC–ORC asymmetries in French and German (RQ1) and a significant WM × Clause Type × Group interaction (RQ2). Group-level models showed that WM affected comprehension differently across groups. Only in French bilinguals, higher WM reduced the SRC–ORC gap (RQ3).(RQ4) Greater cumulative L2 exposure enhanced WM-related support for ORC comprehension, underscoring the importance of treating bilingualism as a continuous variable.

Q&A Time: Saturday, November 8, 4:57 – 5:10

The study investigated whether Mandarin-speaking preschoolers could draw contrastive inferences from two connectives, danshi ‘but’ and suiran ‘though’. Thirty-two 4-year-olds, 32 5-year-olds and 32 6-year-olds participated in a modified picture-selection task. The within-participant independent variable was Connective Type (no connective, suoyi ‘so’, danshi ‘but’, and suiran ‘though’). We used the similar materials as prior studies, but modified the test sentences so as to make inferences more explicit, thus presumably reducing children’s difficulties in comprehending contrastive connectives. The results demonstrated a developmental trajectory in children’s abilities to understand contrastive connectives: 5- and 6-year-olds could draw contrastive inferences from both danshi ‘but’ and suiran ‘though’, whereas 4-year-olds could not. In addition, older groups could better contrastively infer using danshi ‘but’ than suiran ‘though’, presumably reflecting children’s earlier mastery of danshi ‘but’. We discuss these findings in relation to cross-linguistic differences, as well as the potential effect of input on different contrastive connectives.