Word-level vs. sentence-level prosody in Koshikijima Japanese

Abstract

This paper discusses the phonology of tone in Koshikijima Japanese, an endangered dialect spoken on the Koshikijima Islands in the south of Japan. Particularly interesting is the relationship between word-level and sentence-level phonology, as well as the interaction between the mora and the syllable.

Separated by the sea from its neighboring dialects, Koshikijima Japanese has developed a unique prosodic system. Like its neighbors, it has a two-pattern system, where every word takes either of two tonal patterns, Type A and Type B. Unlike its neighbors, however, both tonal types can have two pitch peaks, or two high tones (Hs): Type A has a H₁L₁H₂L₂ pattern, whereas Type B has a H₁L₁H₂ pattern. Interestingly, the first two tones of each pattern (H₁L₁) are associated with words on the basis of the syllable, while the remaining tones (H₂L₂ or H₂) are linked with particular moras. At the word level, the domain of the first high tone (H₁) can be computed only after the position of the second high tone (H₂) is determined, indicating that H₂ phonologically dominates H₁. However, this dominance relationship is reversed at the sentence level, where H₂ undergoes deletion everywhere except in the sentence-final position. This paper discusses this paradoxical relationship between word-level and sentence-level tonal patterns and other related issues.