Phonological generalization from distributional evidence


We propose a model of L2 phonological learning that proceeds not by mapping L2 inputs onto L1, but through general categorization processes where L1 knowledge serves as an inductive bias. This approach views linguistic knowledge as hierarchically organized such that the outcome of acquisition includes not only specific language knowledge, but also beliefs about how any language is likely to be structured. We test predictions regarding how two key types of information come together to drive L2 learning: distributional information and generalization bias derived from existing knowledge of language. We trained monolingual English speakers in a distributional learning paradigm on a novel contrast, segmental length, and tested them on categorization of short/long segments for trained and untrained items. Results show both learning and generalization from one class of segments to another, providing evidence for our approach to L2 learning as one of inductive inference and generalization rather than of mapping.

Back to Table of Contents