Acoustic and articulatory correlates of Japanese devoiced vowels

In Japanese, high vowels may devoice between two unvoiced obstruents (k[u̥]kaku 'division', k[i̥]kaku 'plan'). Recent studies have shown that the devoicing of non-high vowels (k[ḁ]karu 'take', k[e̥]ta 'digit, k[o̥]tae 'answer') may also occur. This paper evaluates the role of vowel duration and vocal folds gestures involved in vowel devoicing in Japanese. An experiment using an electroglottograph (EGG) was conducted, and the results showed different duration values and different vocal folds gestural patterns, depending on vowel quality. However, different duration values were not observed for partially devoiced vowels. Therefore, it is argued that Japanese vowel devoicing is a gradual phenomenon involving the reduction of time and glottal gesture magnitude. Additionally, it is suggested that Japanese devoiced vowels emerge as a targetless vowel that cannot be characterized in terms of duration and vocal folds gestures.