Second-language speech perception: The modification of automatic selective perceptual routines

It is well established that late second language (L2) learners show significant difficulties in categorizing L2 consonants and vowels that do not occur in the native language (L1), or that occur, but differ in their phonetic realization in L1 and L2. These perceptual problems underlie (some) L2 production difficulties (accented L2 speech); lead to L2 comprehension difficulties, especially in nonoptimal listening environments; and may persist even after years of immersion in the L2 environment. However, it has been shown that intensive laboratory training (listening with feedback) can lead to improved perception of difficult non-native contrasts that: (a) generalizes to novel speakers and stimulus materials, (b) persists over months, and (c) produces improvements in production of trained segments. In the model presented here, changes with experience in patterns of phonetic perception are conceived of as reflecting the modification of language-specific (L1) selective-perceptual routines used to extract phonologically relevant information efficiently from continuous speech input. Theoretical and methodological issues will be discussed regarding (a) the conceptualization of (changes in) the structure of phonetic/phonological categories, (b) empirical assessment of phonetic perceptual learning, (c) operational definitions of mastery of L2 phonological categorization, and (d) implications for improving phonetic perceptual training techniques. [Work supported by NIDCD.]