Lany, J.
University of Notre Dame
Experience with statistical regularities plays a key role in word learning: Tracking sequential statistical structure (e.g., determiners typically precede nouns) promotes learning grammatical categories (Gerken et al., 2005), as well as learning word-referent mappings (Lany & Saffran, 2010). The current experiment tested the extent to which statistical learning supports vocabulary development by increasing the predictability of speech sequences and/or by promoting lexical organization (i.e., forming connections among similar words). Infants (22-month-olds, N=84) listened to an artificial language in which X- and Y-word categories were reliably marked by statistical cues: disyllabic X-words followed a-words, while monosyllabic Y-words followed b-words. Infants were then trained on pairings between words and pictures. In the Correlated-Semantics condition, X-words referred to pictures from one semantic category (e.g., animals) and Y-words referred to pictures from another category (e.g., vehicles). Thus, infants could benefit from predictability (experience with the aX bY pattern) and lexical organization (experience with statistical-semantic correlations). In the Uncorrelated-Semantics condition, X- and Y-words referred to pictures from both semantic categories, and thus infants could benefit from predictability alone. If infants are affected by lexical organization, then performance should differ across conditions. Moreover, tracking statistical-semantic correlations may be challenging, and infants' ability to do so may depend on their native-language proficiency. Performance across conditions did not differ overall. However, within the Correlated-Semantics condition more linguistically advanced infants shifted their gaze to the Target more rapidly and were more likely to maintain their gaze to the labeled picture than less-advanced infants. In the Uncorrelated-Semantics condition, infants' performance was unrelated to language proficiency. These findings suggest that statistical regularities support word learning by making speech predictable. As infants develop, they also benefit from statistical-semantic correlations, suggesting that statistical learning plays an increasingly important role in word learning across development.