Perceptual learning of non-native speech sounds: The importance of features from an attention to dimensions perspective.

Abstract

Research on the acquisition of non-native speech sounds is frequently conducted within the framework of theories that were originally developed to account for aspects of cross-language speech perception (e.g., PAM, NLM, and the SLM). Such theories typically focus on the influence of native phonetic categories on the perception of non-native speech, and second language learning researchers may tend to adopt this category-focused view although these theories make few strong predictions about learning. Interestingly, despite their focus on the role of categories, all of these theories incorporate the same assumption: That learning (of both first- and second-language contrasts) progresses by reweighting of phonetic features. This talk presents an overview of results from a series of studies that support this core assumption, suggesting that features, not categories, should play the central role in structuring our understanding of phonetic learning, especially with respect to such short-term adult learning as may be obtained in the laboratory or classroom. Taking such a perspective facilitates connections between theories of linguistic and non-linguistic perceptual learning, reduces dependence on category-based models of speech perception and production, and may provide a more principled basis for making predictions about what non-native contrasts should be easy or difficult to learn.