Categorization of phonemic length contrasts in Japanese by native and non‐native listeners

Abstract

Previous studies reported that exposure to Japanese and identification training allowed English listeners to improve their identification accuracy of Japanese words with phonemic length contrasts. However, this does not necessarily mean their perceptual properties approach those of native listeners in all respects. Past perceptual studies have shown that native listeners show, for example, an extremely sharp short‐to‐long boundary and shift of the boundary to adapt to changes in the temporal context, i.e., typically speaking rate variations. To carefully investigate the effects of training on such properties, the present study analyzed the learners’ identification of stimulus continua between word pairs that minimally differed in the length of a phoneme. Overall results showed that English listeners’ boundaries tended to be sharpened by the identification training, but only to a limited extent. On the other hand, the results also showed that English listeners’ boundaries shifted across different speaking rates in ways that resemble native listeners’ adaptation tendencies. The latter result suggests that even non‐native listeners can adjust their identification boundaries according to differences in temporal context. Discussion will include both prospects and limitations of listeners’ improvement by the perceptual training in this study. [Work supported by NICT and JSPS.]