Normalizing talker variation in the perception of Cantonese level tones: Impact of speech and nonspeech contexts

Cantonese level tones (high level, mid level, and low level tone) are perceptually confusing if produced by talkers with different pitch ranges. External context with cues of a talker’s
pitch range is essential for resolving the ambiguity of level tones. This study investigates the impact of two contexts – speech and nonspeech contexts on the perceptual normalization of talker variation in Cantonese level tones. Four talkers with different pitch ranges are asked to produce the stimuli so that perceptual ambiguity is expected without normalization. We find unequal effects of speech and nonspeech contexts. F0 cues in speech context are efficiently used as reference for normalizing talker difference in tone perception, whereas nonspeech context carrying identical F0 cues shows no obvious effect. It suggests that the carrier of F0 cues (speech or nonspeech) affects whether the context is engaged or not in tone normalization. This finding implies that tone normalization is likely to be a speech specific process, congruent with the phonetic module of speech perception.