Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Moraic Segmentation in Japanese Revisited

P. A. Halle, Juan Segui

Laboratoire de Psychologie Experimental, CNRS-Paris V, Paris, France

Japanese naturally segment speech into moras. The possibility that the syllable also plays a role was investigated here with a syllable-monitoring task. The material was designed so as to discourage Japanese listeners from relying on orthographic representations. In experiment 1, Japanese subjects had to detect targets such as /ta/ vs. /tas/ in e.g. /ta-sa-ku/ vs. /tas-su-ru/. They could detect targets in all types of items with reasonably low error rates. CV targets were always detected faster than CVs targets such as /tas/. But CVs targets were detected faster in test items whose first syllable exactly matched the target. These results were replicated in experiment 2, where the targets used were CV vs. CVm. In experiment 3, French subjects were run with the Japanese material. They probably processed the material at an acoustic or phonetic level. Like Japanese, they also were slower to detect CVC than CV syllables but showed no 'syllabic effect1. We propose that, in addition to the mora, some closed CVC syllables, which do occur in Japanese, also have a psychological reality for Japanese listeners.

Full Paper

Bibliographic reference.  Halle, P. A. / Segui, Juan (1994): "Moraic segmentation in Japanese revisited", In ICSLP-1994, 1143-1146.