UBC Theses and Dissertations
The syllable structure of Japanese Paradis, Johanne Catherine
The question of how to represent prosodic structure is of current theoretical interest in three dimensional phonology. Two current theories/models of representation are the onset/rime model (Kaye and Lowenstamm 1982, Kaye Lowenstamm and Vergnaud 1987 and Levin 1985) and the mora model (Hyman 1985, McCarthy and Prince 1986 and Hayes 1988). This thesis consists of a detailed investigation of the descriptive adequacy of these two theories for the Japanese language. Japanese can be considered an archetypal mora language since in the indigenous linguistic tradition it is analysed into moras. The version of each model which I am adopting is explicitly stated in a set of universal syllabification rules. This syllabification algorithm is compatible with the following assumptions: (1) a. No predictable prosodic structure is present in the underlying representation. The distribution of glides in most cases is predictable. b. Prosodic structure is built by rule and is erected around a syllabic peak which is determined by the relative sonority of segments and not by a feature [syllabic]. Furthermore, the version of the onset/rime model I propose is a paramaterized model where the unmarked setting does not include a nucleus constituent. This onset/rime model is designed to account for weight distinctions as well as the mora theory. Sample structures from both theories are given below. (2) [Diagram Omitted] A syllabification algorithm for Japanese is adapted from the general algorithm and fitted into a model of the lexical phonology of Japanese. It is shown that Japanese prosodic structure can be generated by rule, in either model, with no underlying distinctions between glides and high vowels, and with no feature [syllabic]. Therefore, it is concluded that both the onset/rime model and the mora model are adequate for describing the Japanese language. This conclusion crucially depends on the parameterization within the onset/rime model. Because Japanese is not the only language which employs the weight distinctions a Type I model represents, the parameterization is necessary for the onset/rime model to remain equal in descriptive power with the mora model.
Item Citations and Data