Effects of the mora and segmental features on word duration in Japanese

Abstract

Word duration in Japanese is determined by number of moras when words are pronounced at a normal speaking tempo in focus position in carrier sentences [Port et al., J. Acoust. Soc. Am. 81, 1547–1585 (1987)]. Still, variation in segmental makeup has some effect on work duration. It was expected that, when out of focus position or when spoken at faster tempo, durations for words with a given number of moras would show more variation. A set of Japanese words of two, three, and four moras having various syllabic templates (e.g., tokki, tooki, etc.) was chosen. They were produced in both focus position and nonfocus position (by changing a nontarget word from sentence to sentence) at two tempos. Words with same mora length exhibited less word‐to‐word variation at slow tempo than fast, and less variation in the focus position than when nonfocused. To factor out segmental effects, subjects were asked to produce each sentence reiterantly by replacing each mora of the target word with ta. As expected, reiterant results showed almost no variation due to original segments and much less effect of sentence position and tempo. The results suggest that competing constraints affect Japanese timing: a mora constraint (“make all moras equal”) plus intrinsic segmental constraints (different for each segment type). Mora timing dominates at slower tempos and in focus position, but can be overpowered by constraints of intrinsic segmental duration elsewhere. [Work supported by NSF, DCR‐85‐18725.]