Two types of vowel devoicing in Japanese: Evidence from articulatory data

Abstract

Japanese high vowels [i, u] become devoiced when they occur between voiceless segments: e.g., [kita] ‘‘north.’’ Vowel devoicing (VD) occurs systematically, except when a high vowel appears between two voiceless fricatives, where VD is less consistent and nonobligatory. The present study examined the laryngeal gestures during the production of Japanese vowels in various phonological contexts using fiberscopic and electromyographic techniques. Stimuli consisted of [CiCe] words, where the consonants were stops (S) or fricatives (F). The results revealed that the glottal gestures for typical devoiced vowels (those in [SiSe], [FiSe], or [SiFe]) and devoiced vowels in [FiFe] were distinct, although the two types were indistinguishable in the spectrograms. Consistent with observations in a previous study [H. Hirose, Phonetica 23, 156–170 (1971)], [CiC] sequences with typical devoiced vowels were produced with a wide open glottis accompanied with high activities of the abductor muscle. In contrast, in [FiF] sequences, the glottis was wide open in each of the fricatives, but was almost closed during the vowel, resulting in a bimodal pattern. When the glottis reached complete closure during the vowel, a voiced vowel was observed. The present study shows two distinct mechanisms for VD, suggesting that they are two independent processes.