ラワ語音韻論の研究

Abstract

After a description of phonetic details of three dialects of Lawa, i.e. Bo Luang, Umphai and Mae Sariang, in the framework of 'taxonomic phonemics, ' several problems are discussed in the framework of 'generative phonology'. In those languages, like Lawa, which have no or only a small scale of morpho-phonological alternations, it would be natural to consider that the minimization of the number of features to be specified in the lexical representations is the only or the major ground to determine the underlying phonological representations of lexical items. It can be shown, however, that, in certain cases, such representations as determined in this way do not express intuitively correct underlying forms. The author proposes a condition on the form of grammars in the following way. The 'phonological grammar' of a language consists of the phonological base component to generate the autonomous phonemic representations of words and the phonetic rule component to convert them into the phonetic representations; the former component, in its turn, consists of the phoneme inventory in its minimal schema, the canonical form, and the selectional conditions which act as a filter to sort out the set of admissible feature matrices from the set of all possible matrices obtained by replacing the segments of the canonical form by phonemes nondistinct from them. Such phonological grammar is determined for each language by minimizing, under certain conditions on the phono-logical grammar itself, the complexity of the base component. And, it is proposed that, besides Kiparsky's alternation condition, a condition is to be imposed on grammars : the word-level phonology of a grammar must contain, as a subcomponent, the phonetic rule component as defined in the phonological grammar.