出版社:Department of Linguistics, University of Toronto
摘要:The thesis explores the nature of postlexical representation, as compared to lexical representation. In particular, focusing on prosodic representations, the question is asked as to whether the representation really alters when it appears to do so. I investigate this question through a study of the postlexical process of High Vowel Devoicing/Deletion (HVD) in Japanese. In this process, apparent consonant clusters are created, with the apparent deletion of the vowel, and of the syllable and mora units projected from the vowel in the lexical domain. Two questions are raised. First, is the prosodic unit deleted in HVD? Second, does the prosodic inventory change in the postlexical domain; do these apparent consonant clusters actually form clusters linearly with the deletion of the vowel in the postlexical representation? The literature on HVD shows disagreement. Some researchers argue that the syllable unit is maintained (e.g., Beckman 1996) while others argue for desyllabification (e.g., Kondo 1997). Some researchers argue for the deletion of the vowel (e.g., Beckman & Shoji 1984) while others argue that the vowel is not deleted but rather devoiced (e.g., McCawley 1968, Tsuchida 1997). The thesis presents the following claims. First, the syllable unit and the mora unit are maintained in the postlexical domain in Japanese. By showing that the lexical accent contrasts are maintained in pitch contours in HVD, I conclude that the syllable unit remains present in HVD. With respect to the mora, by examining places where the mora is important (e.g., poetry), I show that there is no evidence for demoraification with HVD. Second, the syllable and mora inventories do not change in the postlexical domain in Japanese. I specifically test the hypothesis that HVD derives syllabic/moraic voiceless consonants. Considering the consequences of this hypothesis for segmental processes and the learning process along with the perception of Japanese speakers, I conclude that the apparent consonant clusters are best analyzed not as clusters postlexically, and that the prosodic inventory does not alter. Rather, segments and prosodic units are all kept intact as in the lexical representation, with the vowel undergoing HVD remaining present in the representation.