Name:Erica Flaten

School/Affiliation:University of British Columbia

Co-Authors:Cristina Conati, Janet Werker

Virtual or In-person:In-person

Abstract:

Infants learn many of their first words from caregivers labeling objects in infant-directed (ID) speech, which is acoustically more ‘musical’ than adult-directed speech, including being more rhythmic. Infants show enhanced attention, learning, and neural tracking to ID than adult-directed speech, but whether rhythmicity, specifically, drives this remains unclear. We posit that rhythmicity engages infants’ cognitive and neural processes in real time, enhancing word learning. To test this, machine learning (ML) techniques utilize infants’ multiple signals (e.g., eye-tracking, EEG, facial expressions) together to predict learning outcomes. Nine- to eleven-month-old infants (data collection is ongoing) from English-speaking homes were familiarized with two novel objects one at a time, each paired with a pseudoword (e.g., ‘Bap’ &‘Det’), while eye-tracking, EEG, and facial expressions were recorded. Words were repeated over an intonational phrase manipulated into patterns that were either rhythmically regular (fixed inter-onset-intervals [IOIs]) or irregular (jittered IOIs). At test, infants heard the taught pseudowords one at a time while viewing both objects. Looking times to the correct vs. incorrect object indexed learning. Analyses are ongoing, but initial results suggest infants may learn word-object associations in both regular and irregular conditions. However, neural tracking appears stronger for the regular than irregular phrases. With a full sample, we hypothesize stronger neural tracking in the regular condition will predict word learning. Additionally, ML models will predict infants’ learning outcomes, starting with the eye-tracking data during familiarization. This work will inform how rhythmicity shapes early language acquisition through the combination of behavioral, neural, and computational approaches.