Brian MacWhinney

Carnegie Mellon University

Title: New Technologies for Studying Language Learning

Recent advances in Machine Learning (ML), Artificial Intelligence (AI), Natural Language Processing (NLP), and web technologies provide tools that can automate and improve studies of L1 and L2 spoken language learning. We have been applying these new developments to improve the scope and power of the 14 databases in TalkBank, including databases such as CHILDES, AphasiaBank, DementiaBank, and SLABank. This presentation will explain the processes and results of six of these developments:

1. We have configured advances in automatic speech recognition (ASR) to speed and improve transcription of language samples. We use our Batchalign Python script to go quickly from raw audio to a first-draft CHAT transcript for inclusion in CHILDES, SLABank, or other TalkBank databases.
2. We have used Batchalign2 to apply Universal Dependency grammars to tag new and older transcripts in 28 languages for part-of-speech, morphological features, and grammatical dependency relations. These new resources open possibilities for deeper crosslinguistic study of language learning.
3. We have used a new Batchalign method for forced alignment of media to transcript to align hundreds of hours of transcripts in TalkBank to the related media on the utterances and word level for playback and analysis of fluency and interactional patterns.
4. The TalkBankDB system allows for systematic database queries across all materials in TalkBank
5. The Collaborative Commentary system allows for greater in-depth study and reliability checking for codes and comments on conversational patterns, error types, gestures, and learning patterns.
6. The new Phon4 program supports automatic detailed cross-tier analyses for lexical, morphological, syntactic, phonological, and fluency structures.

Next, we will consider the relation between learning methods for AI large language models (LLMs) such as ChatGPT and human language learning. Although LLMs and humans learn in markedly different ways, the resultant abilities are similar in some regards. We will consider ways in which modifications of LLMs could improve their use as models of human language learning.

Evan Kidd

Australian National University

Title: Linguistic Diversity In Language Acquisition Research

A comprehensive theory of the human capacity for language must be flexible enough to capture the fact that human children can acquire any one of the world’s 7000+ diverse languages, and more often than not acquire more than one language at a time. However, what we know about how children acquire language is heavily skewed towards English and a handful of other (mostly related) languages (Kidd & Garcia, 2022), studied primarily in monolingual contexts. In this talk I discuss the current state of language coverage in the field of child language acquisition and juxtapose it against a parallel and concerning trend: the rapid rate at which languages are becoming endangered and lost. I then argue that concentrating on studying the acquisition of threatened languages can have the dual effect of widening the number of languages for which we have acquisition data, thus leading to more robust theoretical development, and can also help stem the flow of language loss by investing in the transmission of language to future generations.

Fang Liu

School of Psychology and Clinical Language Sciences, University of Reading

Title: Using Music To Facilitate Language Acquisition In Autism

Globally, around 1% of the population is autistic, with approximately 30% of autistic individuals being nonspeaking, despite various interventions^1–4. Challenges with spoken language impact communication, social interaction, academic performance, and overall quality of life^5–8. These difficulties often lead to greater reliance on health and other services^9–12. Improving spoken language outcomes for autistic individuals can enhance their quality of life and reduce these global support costs¹³.

Research has shown that music-making and music training can improve auditory processing, sensorimotor integration, motor and imitative skills, cognitive function, social interaction, communication, and most notably, language processing¹⁴. In this talk, I will share findings from our proof-of-concept project, MAP (music-assisted programmes), funded by the European Research Council (ERC). This project explored how music can support language learning in autistic children.

We conducted a feasibility randomised controlled trial (RCT) using parent-mediated, music-based telehealth interventions to improve spoken language in autistic children aged 2-5 with minimal verbal skills^15,16. Our results showed that children in the MAP group demonstrated increased social responsiveness and better parent-child interactions compared to those in the control group, who received a social communication intervention.

I will discuss the potential mechanisms behind MAP's effectiveness in enhancing language acquisition in autism. Additionally, I will outline our plans for a global project aimed at evaluating the efficacy of an AI-powered music app in supporting language learning in autistic children, focusing on both English and Mandarin speakers.

References
1 Hampton, L. H. et al. Journal of Intellectual Disability Research 60, 444–463 (2016)
2 Oono, I. P. et al. Cochrane Database Syst Rev CD009774 (2013)
3 Tager-Flusberg, H. et al. Autism Res 6, 468–478 (2013)
4 Zeidan, J. et al. Autism Research 15, 778–790 (2022)
5 Billstedt, E. et al. J Autism Dev Disord 35, 351–360 (2005)
6 Moss, P. et al. J Autism Dev Disord 47, 1830–1837 (2017)
7 Howlin, P. et al. J Child Psychol Psychiatry 45, 212–229 (2004)
8 Howlin, P. et al. J Child Psychol Psychiatry 55, 49–58 (2014)
9 Buescher, A. V. S. et al. JAMA Pediatr 168, 721–728 (2014)
10 Taneja, A. et al. Value in Health 20, A712 (2017)
11 Tinelli, M. et al. European Psychiatry 66, e76 (2023)
12 Bieleninik, Ł. et al. Brain Sciences 11, 340 (2021)
13 Gotham, K. et al. Autism 19, 794–804 (2015)
14 Kraus, N. et al. Neuroscientist (2016)
15 Williams, T. I. et al. Pilot and Feasibility Studies 7, 182 (2021)
16 Williams, T. I. et al. Autism 13623613241233804 (2024)

Lucy Zhao

University of Cambridge

Title: Anaphora Resolution In L1 And L2 Acquisition Of Mandarin Chinese

The talk explores anaphora resolution in the grammars of Mandarin monolingual adults, monolingual children and L2 speakers through two independent picture judgment tasks. Anaphora resolution has aroused much interest in L1 and L2 acquisition due to its complexity. Although there are many studies on anaphora resolution in Romance pro-drop languages, studies on Mandarin Chinese are scarce (e.g., Zhao, 2014; Su, 2020). Mandarin Chinese is of particular interest to the understanding of anaphora resolution as it differs from Romance pro-drop languages in terms of the interplay of syntax, semantics and discourse. Both the overt pronoun and the null element analysed as a syntax-discourse interface category topic has a subject preference in forward anaphora due to the discourse-prominent nature of Mandarin, while the overt pronoun is constrained by a language-specific syntactic constraint Principle C’, resulting in the obligatory disjoint interpretation in backward anaphora (Huang and Lin, 2021; Zhao 2014). Most studies on L1 and L2 acquisition of anaphora resolution in Romance pro-drop languages have found that L1 children and L2 adult learners have more difficulty in interpreting overt pronouns. which has been attributed to processing difficulty with reference to the Position of the Antecedent Strategy (Carminati, 2002). Our studies seem to show that Mandarin anaphora resolution poses different challenges to L1 and L2 learners. Highly proficient English-speaking L2 learners have acquired the interpretation of the null element but showed divergent resolution patterns for the overt pronoun. This indicates that L2 learners can acquire syntax-discourse interface categories but may have problem with language-specific syntactic constraints. By contrast, L1 children acquired the interpretation of overt pronouns at 8-10 despite the lack of convergence at 4-6. Their interpretation of the null element was still developing at 8-10. We argue that syntax-discourse interface categories may be subject to even further delays than language-specific syntactic constraints in L1 acquisition, which are acquired later than language universal constraints (Su 2020).

References:

Carminati, M.N. (2002). The processing of Italian subject pronouns. PhD Thesis, University of Massachusetts Amherst.
Huang, C.-T. & Jo-Wang, L. (2021). Quantificational binding without surface c-command in Mandarin Chinese. In Si, F. .& L. Rizzi (eds) Current Issues in Syntactic Cartography: A crosslinguistic perspective (pp.183-216), John Benjamins.
Su, Y.-C. (2020). Backward/forward anaphora in child and adult Mandarin Chinese, Language Acquisition 27, 187-216.
Zhao, L. X. (2014). Ultimate attainment in anaphora resolution, Second Language Research 30, 381–407.

Angel Chan

The Hong Kong Polytechnic University

Title: Supporting Bilingual Children With Developmental Language Disorder: Identification and Intervention

Speech Therapy Unit; Research Centre for Language, Cognition, and Neuroscience, Department of Chinese and Bilingual Studies, The Hong Kong Polytechnic University; The Hong Kong Polytechnic University – Peking University Research Centre in Chinese Linguistics

Developmental Language Disorder (DLD) is a specific language learning disability that affects 2 children in every class of 30, making it more prevalent than Autism. It negatively impacts an individual's academic outcomes, mental health, and career development. However, when identified and provided with appropriate language intervention, DLD has been shown to respond well to treatment.

Identifying DLD in bilingual children has been a major challenge in speech pathology and clinical linguistics. One needs to distinguish between neurotypically developing (NTD) bilingual children who are not as proficient in the target language due to reduced exposure despite full language potential, and children with genuine DLD who are not fully proficient in any language because of processing and learning limitations. Furthermore, since bilingual children with DLD have language difficulties in both languages, and children often need to learn two or even three languages when they enter school, dual language intervention supporting the development of their home language and school language would be beneficial. However, speech pathologists and educators often find it challenging to effectively support children in their development of these two languages.

In this talk, I will present my recent research and plans that emphasize the importance of setting a reference standard in identifying DLD in bilingual Chinese-speaking children, as well as the development of progress monitoring tools and intervention materials to support dual language intervention for bilingual Cantonese-English speaking children with speech, language, and communication needs. Importantly, although the studies will be presented in a Chinese and bilingual context, these instruments (assessing narrative comprehension and production, lexical comprehension and production, nonword repetition, and morphosyntactic abilities) and the dual language intervention program could be further adapted to other languages and cultures, to support the identification of and dual language intervention for bilingual children with DLD in other target languages.

Xiangjun Deng

Shenzhen University

Title: Input Frequency Effects In The L1 Acquisition Of English Quantifiers: A Corpus-based Study

This study examines the emergence and frequency of English existential quantifiers a(n), some, also, and again, and universal quantifiers all, every, only, and always in child language, based on longitudinal data from 10 English-speaking children aged between 1;6 and 5;2 and their caregivers. We found that input frequency plays a significant role in the acquisition of these quantifiers. The children resemble their caregivers in using the quantifiers in terms of frequency not only at the concrete lexical level but also at the abstract level (i.e. syntactic distribution). Input frequency parallels with the earlier emergence of existential quantification than universal quantification (as instantiated by a slightly earlier than all in most of the children), the earlier emergence of existential D(eterminer)-quantification than A(dverbial)-quantification (a slightly earlier than again in most of the children), and may explain the emergence and frequency of most quantifiers in child language (except for the underuse of a and only, and the early emergence of again). However, there are quantitative and qualitative differences between the children and adults in using quantifiers. For instance, the children show a stronger tendency to put A-quantifiers again, also and always at the periphery of a clause than adults. The most striking difference is children's non-target production, such as overpluralization (*Put some mores in) and agreement errors (*A strawberrys). Children’s non-adult-like patterns may be attributed to their developing grammatical system and cognitive factors.

Keywords: existential quantifier, universal quantifier, D(eterminer)-quantification, A(dverbial)-quantification, input frequency, overpluralization

Virginia Yip^1,2,3

The Chinese University of Hong Kong

Title: Reflections On Language Acquisition Research: From Bilingualism To Multilingualism

¹ Childhood Bilingualism Research Centre, Chinese University of Hong Kong
² The Chinese University of Hong Kong - Peking University Joint Research Centre for Language and Human Complexity
³ The Chinese University of Hong Kong - Beijing Language and Culture University Joint Research Centre for Chinese Linguistics and Applied Linguistics

Established in 2008, the Childhood Bilingualism Research Centre at CUHK has contributed toward the study and documentation of children’s language development in diverse contexts. We have investigated how children acquire language(s) in monolingual, bilingual, multilingual and heritage contexts. In the early years, we primarily focused on bilingual children exposed to Cantonese and English in the Hong Kong community. In the last decade, our work has extended from bilingualism to trilingualism in children acquiring Cantonese, English and Mandarin which shows greater complexity in terms of input factors, language dominance and typological contrasts between the languages.

We consider this shift from bilingualism to trilingual development a step toward diversifying child language acquisition research. In terms of language combinations, we look at Cantonese as well as Mandarin paired with English as a triplet. Cantonese and Mandarin are increasingly acquired in a bilingual and multilingual context in the homeland and as a heritage language in overseas communities. Bilingual and multilingual development of Chinese languages provides a lens through which to view the interaction of languages and identify vulnerable domains that would go unnoticed if only monolingual acquisition was observed. Examples of cross-linguistic influence and vulnerable domains will be discussed.

Compared to bilingual development, multilingual development is largely uncharted territory which holds tremendous fascination for researchers. Despite the daunting challenges and formidable complexity inherent in the multiple variables, exciting discoveries will be made as we conduct systematic studies of multilingual children acquiring languages in diverse contexts.

Yao Lu

Peking University

Title: An Investigation Of Mandarin Tone Perception Among Korean-Chinese Bilinguals In Yanbian District: A Comparative Study Based On Different Age Groups

This study employs a classic categorical perception experiment paradigm to investigate the perceptual characteristics of Mandarin tones among elderly Yanbian Korean-Chinese bilinguals and compares the results with previous studies on younger participants. The experimental results indicate that the elderly participants cannot distinguish between the Mandarin rising tone (T2) and falling-rising tone (T3), but can distinguish other Mandarin tone categories. Compared to the younger group, the elderly participants exhibit significantly weaker categorization of other Mandarin tone continua, characterized by wider boundary widths (Wbc) and lower peak values (Ppk). The experimental results, combined with a language background survey, demonstrate that language experience is the primary factor leading to the weaker tone categorization degree of elderly Yanbian Korean-Chinese bilinguals compared to younger ones. Additionally, the influence of aging may also contribute to the reduced categorization of Mandarin tones among elderly participants. The low categorization level of T2-T3 perception is consistent with research findings on Mandarin learners with different native language backgrounds, indicating that the similarity in acoustic features between these two tone categories poses greater difficulties for second language learners in acquisition and perception.

Haoyan Ge

Hong Kong Metropolitan University

Title: Bilingualism Effects On Autistic Children: Evidence From The Production Of Focus

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by deficits in social interaction, along with a propensity to engage in repetitive behaviours or have restricted interests. It is commonly assumed by some professionals and parents that exposure to two languages imposes an additional burden on children with ASD. However, there is a lack of empirical evidence to support or refute this assumption. With the prevalence of autism and an increasing number of children growing up bilingual, it is urgent to understand how bilingual exposure interacts with autism. This talk will report bilingualism effects on the production of focus in 5- to 9-year-old Cantonese-English bilingual autistic children’s first language Cantonese. A production task in the form of a game was conducted to test how bilingual autistic children use different levels of linguistic knowledge to produce contrastive information in real conversations, compared to their monolingual autistic peers and typically developing children matched in language abilities, nonverbal IQ, working memory and maternal education. It was found that bilingual autistic children performed similarly to typically developing children in general, and even better than monolingual autistic children. The findings suggest a bilingual advantage in autistic children in conveying constative information. Parents are thus encouraged to engage children with ASD in rich bilingual environments. Clinicians, educators and other professionals may consider incorporating bilingualism in training programmes to support families raising bilingual autistic children.

Shanshan Yan

Peking University

Title: Sentence-final Particles In Bilingual Adult And Child Mandarin Chinese

Mandarin Chinese sentence-final particles bear semantic, syntactic, and discourse features (Paul and Yan, 2022), which are realized by different linguistic elements in English. In this talk I will present our recent work examining cross-linguistic influence in the acquisition of four sentence-final particles (le, ma, ne, and ba) by adult L2 and heritage learners of Mandarin Chinese. Findings will demonstrate that while the complex semantics-syntax and discourse-syntax interactions are not always problematic, pure syntactic features may pose great difficulties in the course of L2 acquisition. I will further show that early and richer exposure to Chinese usually, but not always, facilitates the acquisition of the particles. Finally, I will share some preliminary results from an ongoing PKU-CUHK collaboration investigating sentence-final particles with modal meanings in Mandarin-English bilingual children.

Wenchun Yang

Xi'an Jiaotong University

Title: Expressive Narrative Competence In Left-behind Ethnic Minority Children In China: Evidence From Yi And Kam Children

Wenchun Yang¹, Angel Chan², Natalia Gagarina<sup>3

1</sup>School of Foreign Studies, Xi'an Jiaotong University, Xi'an
²Department of Chinese and Bilingual Studies, The Hong Kong Polytechnic University, HK
³Leibniz-Centre General Linguistics (ZAS), Berlin

This talk features children raised in a distinctly unique linguistic and sociocultural environment. They typically live in remote rural areas and are 'left behind' as their parent(s) go to cities for employment (SCPRC 2016). Narrative production data from Yi (a Tibeto-Burman language) and Kam (a Kam-Tai language) children were collected using LITMUS MAIN (Gagarina et al., 2019). Thirty-nine (Mage=6;4, 19 left behind vs 20 non-left behind) Yi children from Yunnan and 55 Kam (Mage=7;0, 23 two parents-left vs 32 one parent-left) children from Guangxi were recruited. Each child (re)told four picture-based stories. Narrative competence was measured by two macrostructure measures (story structure (SS), story complexity (SC)), and the use of internal state terms (IS). The first round of analysis focused on Yi children’s L2 Mandarin and Kam children’s L1 Kam. Preliminary results showed that left-behind experience as a whole negatively predicted expressive narrative competence. Left behind children scored lower in all three measures, particularly IS, than non-left behind children, and two-parents left children scored lowest. These findings offer some important implications for policies and practices pertinent to this group of underprivileged population.

Yuqing Liang

Title: Input-outcome Relations In Cantonese-English Bilingual Toddlers In Hong Kong

Bilingual toddlers have been found to have lower grammatical complexity than monolinguals due to reduced amounts of input in each language. Nonetheless, relatively few studies have examined whether and how individual grammatical structures are affected differentially by input reduction. While the proportion of input in a language was found to be a strong predictor of performance in the same language in bilingual children, how qualitatively aspects of input account for individual differences is less clear. This study filled these gaps using questionnaire data and speech samples of 34 Cantonese-English 3-year-olds and their age-matched Cantonese monolingual peers (n = 25) drawn from the Early Additive Child Multilingual Corpus (Mai et al., in prep). Results showed that the bilinguals performed lower than the monolinguals in structures that are realized through a larger number of lexical items and lack close structural equivalents in English. Grammatical complexity in the input accounted for an additional 17% variance in bilingual children’s Cantonese vocabulary size, after input proportion, SES and cognition.

Jingyao Liu

Title: Naturalistic Production Of Exclusive Focus In Mandarin-English Bilingual Children

Encoding only-type exclusive focus in discourse involves complex computation and integration of knowledge from multiple linguistic domains. We present a comprehensive analysis of syntactic, semantic, prosodic and discourse contextual features of 864 utterances with only and its Mandarin equivalents zhi(you) produced by Mandarin-English bilingual preschoolers and matched monolinguals (age 2-6), all sampled from naturalistic play interactions. The results revealed largely target-like syntactic positioning and semantic association of only and zhi(you) in both languages in the bilinguals, with cross-linguistic influence between only and zhi. Interestingly, the bilingual children, like their Mandarin monolingual peers, employed longer duration but not raised mean pitch to shift the prosodic stress to the intended focus, although both acoustic features, in addition to positional and contextual cues, were instantiated in the focus utterances in the Mandarin parental input, suggesting prolonged development in focus-prosody mapping in children independent of bilingualism.

Melody Mengyao Shang

Title: Elicited Imitation Of Noun-modifying Clause Construction In Mandarin-speaking Children

We report an experimental study on 121 monolingual children (3;0-4;11) investigating the acquisition of Mandarin noun-modifying clause construction (NMCC, [[modifier clause] de head]). We study Mandarin NMCC from a typological perspective, focusing on three NMCC subtypes (argument, adjunct, extended). We designed a sentence repetition task to test a NMCC learning hierarchy previously observed in a corpus-based study: argument > adjunct > extended (easier > more difficult). This study extends beyond the traditionally defined relative clauses to cover all NMCCs in a language with a general noun-modifying clause construction. The results show a consistent pattern with our corpus-based analysis that a syntactic dependency is less difficult than a loose “aboutness” relation in the acquisition of Mandarin NMCCs.

Xuening Zhang

Title: Oral Narrative Development In School-age Learners Of Mandarin In Hong Kong

Previous studies found the development of narrative skills in both monolinguals and bilinguals, yet the findings have mostly focused on Indo-European languages and children aged 3 to 7. This gap motivates our study exploring narrative development in older Mandarin-learning children. In a Putonghua proficiency test, timed picture-based storytelling was elicited from 158 primary school students in Hong Kong, divided into two age groups (76 younger primary students, Mage = 9.26; 82 older primary students, Mage =10.68). We found that the older group outperformed their younger peers in both macrostructure (i.e., story grammar) and general microstructure measures (i.e., total number of utterances, total number of words, number of different words, mean length of utterance in words). Fine-grained microstructure measures behaved differently in Mandarin learners’ storytelling. Two semantically complex structures-adverbial clauses and evaluative adverbs-showed the most rapid development and the strongest correlations with macrostructure. These structures fulfill discourse functions in narrative by directly encoding the logical relations between story events, which may constitute a distinctive “meso-level” linking macro- and microstructure.

Hecheng Zhang

Title: How Does Typological Distance Shape Bilingual Cognitive Advantages In Mandarin-Speaking Children

In exploring the variability of bilingual cognitive benefits, this study addresses the underexamined influence of typological distance (TD) between language pairs on cognitive outcomes in bilingual children. Focusing on Mandarin-dominant bilingual children acquiring a typologically similar language (Cantonese) or a different language (English), we aimed to assess whether and how TD interacts with bilingualism to shape executive function (EF).

17 Mandarin-Cantonese and 19 Mandarin-English bilingual children aged 3;4-6;10, with 48 Mandarin-speaking monolingual controls were recruited to participate in a series of cognitive tasks including Flanker and DCCS-B. Results show that both bilingual groups excelled in inhibition and shifting than their monolingual peers regardless of TD, with only Mandarin-English speakers outperforming Mandarin monolinguals in visual-spatial working memory, highlighting benefits from engaging with typologically distant languages. Mandarin-Cantonese bilinguals showed superior phonological working memory to Mandarin-English bilinguals at higher L2 proficiency levels, suggesting a positive cognitive effect of typologically similar languages. We also found that age has a significant positive influence on cognitive performance across all groups, with older children demonstrating higher scores in all tasks. 

Our findings shed light on the nuanced impact of TD on cognitive performance, underscoring its significance in shaping bilingual cognitive advantages across tasks and age groups.