Perception of intervocalic consonant clusters by Japanese listeners

Abstract

This study reexamines perceptual epenthesis by Japanese listeners. It has been argued that since Japanese phonology prohibits consonant clusters Japanese listeners tend to perceive an illusionary vowel /u/ between consonants in a cluster so that their percepts conform to their native phonology. As a consequence, they do not perceive difference between strings like /VCCV/ and /VCuCV/. However, consonant clusters actually occur in natural speech in Japanese as a result of vowel deletion. Additionally, it has been shown that Japanese listeners are sensitive to acoustic difference between consonant clusters with a deleted vowel and their counterparts with an undeleted vowel. These observations raise a question about the above generalization. This study hypothesizes that the perceptibility of consonant clusters for Japanese listeners is determined not only by their phonological knowledge, but also by their phonetic knowledge about how likely certain types of consonant clusters occur in natural speech as a result of vowel deletion. In order to test the hypothesis, 19 Japanese listeners were tested on discrimination between consonant clusters and their counterparts with an intervening vowel /u/. Consonant clusters were classified into 4 ranked (lowest to highest) types according to the probability of their occurrence as a result of vowel deletion: voiced stop - voiced stop (DD), voiceless stop - voiced stop (TD), voiceless stop - voiceless stop (TT), voiceless fricative - voiceless stop (ST). The result showed that performance varied depending on cluster types. However, the variation did not follow the order of probability; performance was slightly better with TD and TT types than DD and ST types. The result also showed an effect of stimuli presentation order with TT type (Figure1). These findings are discussed in the light of phonological bias and general auditory speech processing.

Figur 1