Perception boundary between /s/ and /ts/ in Japanese at various speaking rates

To examine effects of speaking rate on a perception boundary between voiceless fricative /s/ and affricate /ts/ in Japanese, a stimulus continuum of /s/-/ts/ at various speaking rates was constructed by changing rise and steady+decay durations of these phonemes. The stimuli in the continuum were presented to 33 Japanese native speakers, and their response ratio to /ts/ was measured. Logistic regression analysis of the response ratio revealed that the perception boundaries systematically differed as a function of speaking rate. However, when the rise and steady+decay durations were normalized by a logarithm of the averaged mora duration, the boundaries at different speaking rates nearly coincided. These results suggest that the speech perception system discriminates /s/ and /ts/ at various speaking rates with reference to a single boundary calculated with the normalized variables.