Vowel devoicing rates in Japanese were examined at three different speaking rates. The results showed that effect of speaking rate was generally minimal, and that high vowel devoicing occurred at all tempi, but the effect of tempo varied depending on the devoicing environment: (1) type of preceding consonants and (2) devoicing condition, single or consecutive devoicing environments. There was no tempo effect in single devoicing condition, but more devoicing occurred at faster tempi in the consecutive condition. Unlike similar phenomena in other languages, Japanese vowel devoicing does not appear to be an optional process in fast speech, but more positive process partly controlled by its syllable structure.