Distributional structure in language: contributions to noun-verb difficulty differences in infant word recognition

Abstract

What makes some words easy for infants to recognize, and other words difficult? We addressed this issue in the context of prior results suggesting that infants have difficulty recognizing verbs relative to nouns. In this work, we highlight the role played by the distributional contexts in which nouns and verbs occur. Distributional statistics predict that English nouns should generally be easier to recognize than verbs in fluent speech. However, there are situations in which distributional statistics provide similar support for verbs. The statistics for verbs that occur with the English morpheme -ing, for example, should facilitate verb recognition. In two experiments with 7.5- and 9.5-month-old infants, we tested the importance of distributional statistics for word recognition by varying the frequency of the contextual frames in which verbs occur. The results support the conclusion that distributional statistics are utilized by infant language learners and contribute to noun-verb differences in word recognition.

Keywords: Language acquisition; Statistical learning; Verb learning; Word recognition.

Fig. 1

Infants looking time (ms) to Familiarized and Unfamiliarized Verbs in all four experiments.

Fig. 2

Average bigram probabilities for the target words and their preceding and following contexts for the nouns in Jusczyk and Aslin’s (1995) study and the verbs from our experiments.