[PS-2.13] Indexical and Suprasegmental Specificity of Infant Statistical Learning

Parvanezadeh Esfahani, S. & Hay, J.

University of Tennessee, Knoxville

Infants have broad sensitivities to environmental regularities. In the domain of speech perception, infants? sensitivity to co-occurrence patterns in continuous speech has been extensively documented. However, we still know very little about how infants represent the sequences that are the output of these statistical learning processes. Here we test 8-month-old infants? (N=50) indexical (i.e., talker) and suprasegmental (i.e., stress pattern) representations of newly encountered statistically-defined words. In both experiments, infants were familiarized with a naturally-produced Italian corpus that contained two trochaic (strong-weak) high transitional probability (HTP) target words produced by a female speaker. Following familiarization, infants were tested on their ability to discriminate modified target words, either produced by a male speaker (Experiment 1) or with an iambic (weak-strong) stress pattern (Experiment 2), from foils (matched in speaker sex and stress pattern). In both experiments infants demonstrated a significant familiarity preference, listening longer to the modified HTP words than to foils. Findings demonstrate that infants are able to generalize representations of statistically-defined words across a range of acoustic forms and suggest that segmental information may override less lexically relevant information at this age, which may be weak or underspecified early in development. Implications and future directions will be discussed.