On December 8, 2025, the symposium “Infant Language Development Research in the 21st Century” was held at Fukutake Hall, the University of Tokyo. Organized by International Research Center for Neurointelligence (WPI-IRCN) in collaboration with the JEWEL Project, the event brought together leading researchers to examine infant language development from behavioral, neuroscientific, and computational perspectives.
The symposium opened with welcoming remarks by Dr. Reiko Mazuka, Principal Investigator of the JEWEL Project, IRCN Affiliated Faculty who emphasized that language development research is entering a transformative phase. Recent advances in neuroscience and artificial intelligence, she noted, now allow long-standing questions about early language acquisition to be addressed in new and integrative ways.

Janet F. Werker: "Developmental processes in infant speech perception and early language acquisition"

The keynote lecture by Professor Janet F. Werker (University of British Columbia) addressed how infants acquire native-language speech patterns during the first year of life. She reviewed evidence showing that young infants begin life with broad sensitivity to speech contrasts across languages, which gradually narrows through experience-dependent perceptual attunement. Professor Werker highlighted longitudinal findings demonstrating that early phonetic perception predicts later vocabulary growth and word learning efficiency. She further emphasized that speech perception is fundamentally multisensory and closely linked to motor and social development, even before speech production begins. Her talk framed infant language acquisition as an experience-expectant process shaped by biological constraints and environmental input.

Reiko Mazuka: "Expanding infant phonological development research to Asian languages"

Professor Reiko Mazuka (The University of Tokyo) examined how typological properties of Japanese shape early phonological development. She presented behavioral studies showing that Japanese-learning infants require more time to discriminate moraic contrasts such as vowel length and consonant gemination.
Analyses of infant-directed speech revealed substantial variability in temporal cues, suggesting that these contrasts are not easily learned from early input alone. Professor Mazuka argued that caregivers provide alternative prosodic cues that support learning, challenging strong versions of the prosodic bootstrapping hypothesis and underscoring the importance of cross-linguistic research in understanding both universal and language-specific developmental pathways.

Takao K. Hensch: "Translating critical period concepts in mouse, man and machine"

Professor Takao K. Hensch (Boston Children’s Hospital / Harvard Medical School), IRCN Director explored the neurobiological mechanisms underlying critical periods in development. Drawing on animal and human research, he explained how the maturation of inhibitory neural circuits regulates windows of heightened plasticity during early life.
He emphasized that while early experience enables efficient learning, disruptions during sensitive periods can have long-lasting effects on cognition and mental health. Translational research, he argued, offers promising avenues for identifying biomarkers of atypical development and informing early intervention strategies.

Yasuyo Minagawa: "Parent-infant social interaction and language acquisition in infants with neurodiversity: Evidence from a longitudinal cohort"

Professor Yasuyo Minagawa (Keio University) presented findings from a longitudinal cohort study examining language and social development in neurodiverse populations. Focusing on early parent–infant interaction, she showed that contingent caregiver responses and neural synchrony between parents and infants predict later language outcomes.
Using behavioral and neuroimaging measures, her work highlighted how early social experience interacts with brain development over time, illustrating the value of longitudinal approaches for understanding diverse developmental trajectories and identifying early markers of risk and resilience.

Alex Warstadt: “Baby Language Models”

Professor Alex Warstadt (The University of California, San Diego) examined how computational language models can inform theories of human language acquisition. He introduced the Baby Language Model Challenge, which aims to develop data-efficient and cognitively plausible models trained under constraints closer to those faced by infants.
His findings revealed that current models, while powerful, fail to capture key features of human learning such as age-of-acquisition effects and critical periods. These limitations, he argued, help clarify which aspects of language learning may depend on uniquely human neurobiological mechanisms.

Emmanuel Dupoux: "Can AI help understand early language acquisition?"

Professor Emmanuel Dupoux (ENS/CNRS/EHESS) explored how artificial intelligence can be used to test competing theories of early language acquisition. Focusing on statistical learning, he demonstrated that speech-based models can acquire phonetic distinctions and word-form knowledge from realistic input but do so far more slowly than human infants.
By comparing models trained on clean versus ecologically realistic input, he showed that learning outcomes depend critically on how the environment and evaluation methods are modeled. Professor Dupoux argued that these results point to the importance of inductive biases shaped by evolution, attention, and interaction, beyond statistical learning alone.

Toward an Integrated Science of Language Development

After each keynote talk, lively Q&A sessions took place, reflecting the strong interest and active engagement of the audience. Across the symposium, discussions highlighted how behavioral experiments, neuroscience, and computational modeling can jointly advance our understanding of early language acquisition. The final general discussion brought together all speakers to address shared topics that cut across the day’s presentations.
By fostering dialogue across fields, the symposium underscored both the promise and the challenges of developing a comprehensive science of language development in the 21st century.


IRCN Babylab: https://babylab.ircn.jp/en/
JEWEL Project: https://lang-dev-lab.brain.riken.jp/jewel/en/