Infants' Statistical Learning Ability is Related to Real-Time Language Processing

Lany, J.

University of Notre Dame

Infants' statistical learning (SL) ability may be a key underpinning of language development, yet little is known about sources of variation SL. We tested the hypothesis that individual differences in infants' ability to encode speech in real time are related to SL, as skilled encoding may promote learning.

In two experiments, 15-month olds were tested on well-established SL tasks using the Head-Turn Preference Procedure. Experiment 1 assessed word segmentation, and infants heard a syllable stream in which words and part-words occurred with equal frequency, but differed in their TPs (1.0 vs. .5). Experiment 2 tested word-order patterns, and infants were familiarized to phrases in which words from different categories co-occurred selectively. In both experiments, speech-processing efficiency (SPE) was assessed by testing how quickly infants found the referents of familiar English words (e.g., doggie). Vocabulary size and grammatical development were assessed with the MacArthur-Bates Communicative Development Inventory (MCDI).

Infants' SPE was strongly related to performance on both SL tasks, even when controlling for MCDI scores and age. These data suggest that SPE may support SL or vice versa. Thus, future work studying the nature of the relation between SL and SPE will contribute to a precise, mechanistic account of language development.