chapter  17
BIPOLAR DISTRIBUTION OF A WORD AND GRAMMATICALIZA- TION IN THAI: A DISCOURSE PERSPECTIVE
ByShoichi Iwasaki
Pages 16

The recent research on grammaticalization in Thai by Matisoff (1991), Bisang (1996), Diller (2001a), and others has greatly advanced our understanding of the diachronic linguistic process in this language. It is found repeatedly that Thai, an isolating language, shows many instances of transformation in which lexical words such as nouns and verbs have come to take on grammatical functions and have become, or are on the way to becoming, classifiers, auxiliaries, prepositions, complementizers, and conjunctions. Drawing on such previous research, this chapter attempts to provide a new perspective on grammaticalization by examining language changes that are observable in discourse and the actual language use of speakers. The discourse perspective is both possible and necessary. It is possible because language changes are always on-going and observable synchronically. It is necessary because changes in linguistic structures do not happen in a vacuum, but rather they proceed through actual deployment in discourse. This theoretical stance echoes the spirit of the ‘usage-based model of language’ (Langacker 1988, 2000; Kemmer and Balow 2000) and promotes the proposal that it is crucial to seriously take into account the language users’ cognitive involvement in discourse in the study of language change. Based on such assumptions, the current paper concerns how discourse profiles of lexical items, or, more specifically, repeated patterns of their placement in discourse, lead to a structural reanalysis and to eventual grammaticalization into function words. The study first calls for attention to a particular distributional pattern of some words, termed here as the ‘bipolar distribution of a word,’ and presents the theoretical framework in which this pattern can be analyzed. It then examines in detail actualdiscourse environments for the bipolar distribution of two words, KDA\ ‘give’ and O(#(Z ‘finish,’ using the proposed framework.1