The role of coarticulatory variation in speaker recognition

Résumé

The current study explores whether listeners encode talker-specific coarticulatory information. Listeners completed two perception tasks in which they were presented either only coarticulated vowels or non-coarticulated vowels across both tasks. Listeners are more accurate at word completion in the coarticulated vowel condition for speakers who produced greater nasal coarticulation and this led to greater voice recognition accuracy for these talkers than for speakers who produced less nasal coarticulation. These findings are consistent with theoretical claims that enhanced coarticulatory patterns are useful and informative for listeners in making decisions about the lexical content of the speech signal. Yet, we extend the usefulness of coarticulation to the tracking and learning of talker-specific coarticulatory patterns, which might be particularly useful because it allows listeners to predict how a given talker is going to produce a sequence of phonemes in a given context.