Extrinsic normalization of lexical tones and vowels : beyond a simple general contrastive perceptual mechanism

Abstract

Spoken context provides valuable information for listeners to accommodate speech variability. One example of this influence is extrinsic normalization: the finding that formant and tone ranges in preceding context constrain the interpretation of subsequent tone and formant cues. One dominant hypothesis has been that contrastive general auditory processes play an important role in normalization. A contrastive general auditory mechanism suggests that speech and non-speech contexts should have similar contrastive influences on speech perception. The present study tests this prediction across segmental (formants) and suprasegmental (tone) speech cues. Participants listened to target stimuli that were preceded by either speech or non-speech contexts. Importantly, the cues that distinguished target stimuli were contrastively related to their context. The results demonstrate that speech contexts, but not non-speech context, induced significant contrastive effects on the perception of both lexical tones and formants. In addition, we observed considerable individual difference in the size and direction of context effects. Some listeners reliably demonstrated contrastive context effects while others demonstrated assimilative effects. These results suggest that the underlying mechanism of speech normalization is more complicated than simply contrastive general auditory processes.