An optimality theoretical account of contemporary Cantonese rhyming based on inferential statistics

This thesis is an attempt to search for a proper linguistic account of the phonological aspect of rhyming, under the Optimality-Theoretical (OT) framework, with special reference to the following questions: (a) How to distinguish rhymes and non-rhymes? (b) What are the phonological constraints governing rhymes and non-rhymes? (c) What is the phonological explanation for the variation between perfect rhymes (self-rhyming) and imperfect rhymes (cross-rhyming)? In order to answer these questions, lyrics of more than 3,000 contemporary popular songs written in Cantonese are collected for investigation. Conclusions are drawn both on the basis of statistical evidence as well as OT reasoning. In this thesis I propose that rhyming has to be re-conceptualized. The concept of rhyming should not be categorical but has to be in gradient, forming a continuum of rhymability indicated by a scale called Rhymability Coefficient (RC). The findings reveal that the treatment of rhyming as a continuum enables us to look into rhyming relationship in a more refined way, e.g. different cases of cross-rhyming may have different degrees of rhymability, providing us richer information for further analysis. Moreover, the use of stepwise multiple regression allows us to separate significant phonological factors from the others and evaluate their influence with precision. The connection of inferential statistics with Stochastic OT makes the grammatical modeling more solid-based, fulfilling both empirical and theoretical requirements. The findings of this study also shed light on Cantonese linguistics, historical linguistics and phonology in general