Phonetic experience with specific words affects categorical perception of those words.

Abstract

How does experience with specific words influence linguistic‐phonetic categories? Listeners were trained over a five‐session, listen‐and‐repeat task on a set of target words embedded in continuous speech and altered so that the initial stop consonant voice‐onset time (VOT) was 80% longer than natural. Voicing boundaries were estimated before and after training using a two‐alternative, forced‐choice perceptual task on an eight‐step VOT continuum. Stimuli were highly natural tokens by two stimulus talkers. Part 1 of the experiment asked whether exposure to lengthened forms would influence location of the voicing boundary, and, if so, whether that effect would generalize to similar forms. Results showed longer boundaries after exposure to lengthened VOTs for the trained forms, but lengthening did not generalize to new forms. Part 2 investigated voicing boundary locations as a function of lexical status (word, nonword) and usage frequency (high, low). Boundary locations indicated expanded VOT regions both for nonwords over words (opposite to the Ganong effect) and for high‐frequency words over low‐frequency words; neither lexical status nor usage frequency interacted with training. Results suggest a lexical sensitivity to low‐level speech cues, thus offering support for a rich memory language model.