Coping with segmental–prosodic incongruity in spoken word recognition in Japanese
Abstract
A spoken word consists of two phonological cues: segments (phonemes) and prosody (e.g., lexical pitch accent). While both segments and prosody contribute to spoken word recognition, it is unknown how listeners interpret a spoken word when these cues contradict each other. For example, a word *tansu HLL is logically interpreted as both a segmental mispronunciation of dansu HLL "dance" and a prosodic mispronunciation of tansu LHH "wardrobe". The present study investigated how Japanese listeners interpret those segmentally and prosodically incongruent words in spoken word recognition using a cross-modal repetition priming paradigm. When ISI = 0 ms, listeners' lexical decision was delayed more for prosodic mispronunciation (prime: *tansu HLL; target: tansu LHH "wardrobe") than for segmental mispronunciation (prime: *dansu LHH; target: tansu LHH). In contrast, when ISI = 750 ms, listeners’ lexical decision was delayed more for segmental mispronunciation than for prosodic mispronunciation. These results demonstrate that lexical access to segmentally and prosodically incongruent words is a dynamic process, in which listeners first try a prosody-oriented interpretation and then shift to a segment-oriented interpretation.