Abstract
In the vowel devoicing distribution map in NHK Japanese Pronunciation and Accent Dictionary, dialects are classified according to their prominence in vowel devoicing. It is not clear, however, what the criterion for such classification was, and also whether the classification accounts for both, word-mid and word-final vowel devoicing. Thus an acoustic study was conducted to investigate the occurrence of high vowel devoicing in the above two positions and to define factors affecting it. Material was taken from the speech database collected throughout Japan between 1986 and 1989, and involves 608 respondents (304 high school students and 304 their grandparents) from 41 regions (prefectures). The data were acoustically analyzed and devoicing rates were discussed through the surrounding sounds, regional variation, and generational variation. Results revealed that the devoicing rate in dialects belonging to the so-called prominent regions of vowel devoicing rises above 60%. By observing the consonants and vowels of the following syllable, it was found out that devoicing rates differ statistically significantly in almost all regions, following the order affricate < fricative < plosive, and high vowel < non-high vowel. Regional differences obtained confirm results from the previous studies. In addition, generational difference were also found to contribute to devoicing, namely, more devoicing is used by younger generations, the tendency being especially notable in the Tohoku area. All these findings concern high vowels that are located in word-mid position and surrounded by voiceless consonants. In the word-final position, however, where high vowels are located between a voiceless consonant and a pause, devoicing rate is shown to be considerably lower irrespective of region, which brings us to the conclusion that word-final position does not condition devoicing.
