Perceptual learning of talker-idiosyncratic phonetic cues

A number of recent studies have explored “perceptual learning,” in which listeners use lexical knowledge to learn about a talker’s idiosyncratic phoneme pronunciations and adjust their perception of other tokens from that talker accordingly. In a typical perceptual learning study, listeners might hear an item that is ambiguous between “crocotile” and “crocodile” during exposure. Since only crocodile is a word, listeners would learn (following several examples) that this talker has long VOTs, and subsequently at test show a shift in their categorization of a ∕d∕‐∕t∕ VOT continuum by the same talker. The present study explored perceptual learning through cues rather than through lexical knowledge. We used a phonetic contrast (s‐th) in which there are both primary (spectral) and secondary (amplitude∕duration) cues to phonetic identity. Listeners heard tokens of minimal s‐th word pairs in which either the primary or secondary cue was ambiguous, but the alternative cue was unambiguous and thus disambiguated the phonetic identity of the word. We tested whether listeners use the unambiguous cue to learn about the speaker’s production of the ambiguous cue (even though doing so was unnecessary for lexical identification) which would then influence later perceptual identification of a series based only on that cue. 10 192