The mora and the syllable in KiMvita (Mombasa Swahili) and Japanese.

This thesis deals mainly with aspects of the phonology of KiMvita, the Swahili dialect spoken in Mombasa, and has special reference to moraic nasals. The KiMvita analysis is then compared to that of Standard Japanese. The framework of moraic theory that is employed is based on Hyman's (1985) "Weight Theory". The theories of Feature Geometry (FG) and Lexical Phonology (LP) are also employed in the analysis. Nasal+Consonant (N+C) sequences occur in two ways in KiMvita: (i) a sequence of a moraic nasal and a consonant; (ii) a prenasalized obstruent. The analysis of the varying expressions of nasality, either as a moraic segment or as an element of a complex segment shows considerable dependence upon the morphology concerned. In addition to N+C sequences, the analysis of Consonant+Glide (C+G) sequences turns out to be great relevance; these two different types of composite segment differ in underlying representation as well as in surface syllabification. Here too LP enables us to distinguish two distinct surface forms (light diphthongs and complex consonants) in terms of lexical vs. post-lexical levels. Syllable construction in this study crucially requires both an onset and a nucleus. Processes of syllabification will be discussed based on this theoretical requirement together with the following two assumptions: (i) strictly left-to-right syllabification; (ii) priority of the Onset Creation Rule. This study proposes that the accent bearer both in KiMvita and Japanese is not the syllable, which is generally claimed in the literature, but the mora - though this may be associated with a syllable node. Moraic nasals are generally associated with the second mora of a bimoraic syllable word-intemally in both KiMvita and Japanese. However, there is one significant difference in the status of the second mora in these two languages: it may bear accent in KiMvita, while it may not in Japanese. As far as these two languages are concerned, the phonetic evidence suggests that the actual segment duration could explain why such a difference occurs