R. L. Starr

Prosodic structure of Japanese: evidence from text-setting


My colleague Stephanie Shih and I have been investigating the role of the syllable and the mora in Japanese prosody, using corpus and experimental evidence from text-setting, the pairing of language and music in song. Japanese is traditionally described as a mora-based language, meaning that the mora rather than the syllable is the most salient prosodic unit. Previous studies of Japanese song have characterized Japanese text-setting style as moraic (e.g., "kaizen" would be set as ka-i-ze-n rather than kai-zen).

Our corpus studies compare text-setting in translated and native songs; we have found that syllabic text-setting is prevalent in Japanese songs, indicating that the syllable is a salient prosodic unit in Japanese. We also find that syllabic setting is more common in Foreign stratum words than Sino-Japanese words.

We are conducting experiments testing the perception of native Japanese speakers and Japanese learners. These experiments make use of the singing synthesis software Vocaloid, perhaps best known as the voice of Hatsune Miku. So far we have found that Japanese listeners are okay with both moraic and syllabic settings, but disprefer syllabic settings when moraic settings are possible. Learners of Japanese, in contrast, equally prefer moraic and syllabic settings. We are continuing to investigate the question of how learners acquire knowledge of the salience of the mora in Japanese.