日本語における無声化音節とピッチアクセントの実現について

Abstract

In Tokyo Japanese, short high vowels devoice between voiceless consonants. There are two major views to this phenomenon : an active and an inactive hypothesis. (1) Hirose (1971) Electromyographic analysis shows that there is a glottal opening gesture particular to devoiced vowels, which suggests a voluntary movement control of the glottis. (2) Jun and Beckman (1993) Vowel devoicing occurs because of low-level overlap of consonant and vowel laryngeal gestures. The present paper shows a complex interaction between pitch accent realization and devoicing and casts doubt on the inactive view. Devoicing conflicts with the realization of pitch accent. Sugito (1971) shows a compesatory pitch elevation and a resulting abrupt pitch fall after the devoiced region when an accent falls on a devoiced syllable. Matsui (1993) confirms that the pitch elevation and the abrupt fall are the cue for the perception of the accentedness for a single devoiced syllable. Then an interesting question arises : how do speakers differentiate consecutively devoiced words with different accent position such as penultimate, final and no accent? Maekawa (1990) reports that the pitch elevation is the major cue for the perception of accent for pseudowords with devoiced accented syllables and further that the amount of the pitch elevation decreases proportional to the number of devoiced syllables between the devoiced accented syllable and the pitch-elevated syllable. His analysis predicts that the penultimate accent word will be realized with the relatively smaller pitch elevation on the following syllable than the final accent word does. However, he used a carrier sentence with no lexical accent before the target word and did not pay special attention to the pitch pattern before the devoiced region. In the present paper, an experiment using real test words and an accented carrier sentence was conducted. F0 values after the devoiced region were about the same for the penultimate accent words and unaccented words while words with the final accent had a high pitch. The distinction between the penultimate accent and no accent was realized as the difference of the slope of pitch falling. More interestingly, one of the speakers showed that there was a noticeable change in the amount of pitch fall before the devoiced region when the carrier sentence had an accented word immediately before the target word. This pre-devoicing compensation for devoiced pitch accent suggests that speakers anticipate devoicing of a following accent in executng the whole pitch contour, which suggests that the devoicing itself is not simply due to overlap of gestures but should be incorporated into the high-level planning of the utterance.