Phonetic And Phonological Mismatch Of Tonal Mapping: Revision Of Cantonese Register Model

The study aspires to provide possible solutions to the phonological-phoneticmismatches between models of Register Theory and fieldwork linguistic data intonal research. Register Theory is one of the central theories in the realm ofAuto-segmental Phonology in the investigation of tonal languages. Via specificationof two hierarchically-related structural parameters in a binary approach, the theorydefines tonal inventories and alternations of Chinese dialects with four levels.However, mismatches between phonological feature values and phonetic tonalvalues abound in the mapping between the newly proposed four-scale registermodels and fieldwork data transcribed in traditional methods headed by Chao’sfive-scale notation system. Moreover, linguistic evidence for present Registermodels is mainly drawn from different dialectal fieldwork data that cannot becompared, which further threatens the theory itself and its explanatory power. Thepaper discovers that there exist two major conflicts underlying the intersectionbetween the phonetic tier and the phonological one. First, there is ambivalence in thespecification of contour feature under the same phonetic scale for different tonetypes. Second, register spanning is noticed within the same tone type, the contourfeatures of which in turn are endowed with ambivalent register values.Taking Guangzhou Cantonese as an example, with the help of acousticsoftware Praat and fundamental frequency normalization, the author aims toreconstruct phonetic maps of Cantonese that are more unified with lesser researcherbias than those spawned from purely intuitive scaling based on individual perception.Then, against six phonetic maps generated from the five-level and four-level scalingstrategies with equal divisions, the representations and performance of Yip’s (1980)and Bao’s (1990) models are evaluated. Analyzing the limitations of former models and meshing with a suitable phonetic background, the author proposes a newfour-level register model with embedded configuration of contour features of tworegisters so that the first conflict can be alleviated. Meanwhile, the concept ofpredominant linguistic target is proposed to define the ultimate register attribute ofcontour tone types and, hopefully, to account for the second problem. In this manner,phonological features of Guangzhou Cantonese are better captured while linguisticphenomena of the following research of Cantonese changed tones are given a moreelegant explanation.