Tonal alignment in Tokyo Japanese
A large amount of evidence for regularities of tonal alignment in various languages has been accumulated recently. However, there is still much disagreement on the characterisation and modelling of these alignment regularities. This thesis investigates tonal alignment in Tokyo Japanese with two objectives. One is to provide a thorough description of tonal alignment in Tokyo Japanese, including a well-known phenomenon, ososagari ('peak delay'); the other is to contribute to the current understanding of tonal alignment, based on empirical data of tonal alignment in Tokyo Japanese. Three speech production experiments were performed. The first experiment examined the alignment of the F0 targets at the beginning of initial-accented words, varying the syllable/mora structures of the accented syllable. The results showed that both the F0 valley and peak were consistently aligned with specific segmental landmarks, and that the alignment of the F0 peak depended on the syllable/mora structure of the accented syllable. The second experiment explored how the alignment patterns found in the first experiment were influenced in different speaking modes; the speaking modes of interest were fast speech rate, raised voice, and local emphasis. The results showed that the orderly alignment behaviour found in the first experiment remained intact irrespective of different speaking modes, although different kinds of small effects were found. The third experiment compared the F0 peak alignment of unaccented and non-initial-accented words to those of initial-accented words. The results of unaccented words demonstrated consistent alignment of the F0 peak with a specific landmark, which is comparable to those of initial-accented words. On the other hand, the results of non-initial-accented words showed earlier alignment of the F0 peak for the pitch accent than those of initial-accented words. The results of the current study as a whole demonstrate consistent alignment of the F0 targets with specific places in the prosodic structure in a language-specific way, which are rather resistant to changes caused by differences of speaking mode. Further durational analyses, together with the alignment data, also suggest that segments and tones are mutually synchronised with each other. These findings provide further evidence that segmental anchoring is a necessary concept in accounting for alignment regularities.