Devoicing of Japanese vowels by Taiwanese learners of Japanese

Three Taiwanese speakers studying in Japan were tested for application of the Japanese rule of high vowel devoicing. Three native speakers of Japanese served as controls. Five repetitions each of tokens containing two possible devoicing sites were collected from each participant at speaker determined slow and fast speeds. Of the Taiwanese participants, only one showed any significant speech rate effect (t=1.88, p<.05 for the first vowel of the tokens; t=2.29, p<.025 for both vowels combined). One other showed no speech rate effects whatsoever, instead inconsistently devoicing some tokens, while the third did not devoice any vowels within the tokens, instead only devoicing a small number of sentence final vowels. It appears that each of the three learners has begun to devoice vowels according to different processes-for one speaker, a speech rate dependent rule; for another, a speech rate insensitive rule whose environment is not yet solidified; for the third, as a rule sporadically applied to the end of a sentence. Most surprisingly, native speakers also showed a lack of significant speech rate effects. This would imply that high vowel devoicing is losing its status as a fast speech rule in the Tokyo dialect. Further, larger studies will show if this indication is correct