Diphthongization and Coindexing

Abstract

The tree model of segment structure proposed by Clements (1985) is an important innovation in phonological theory, making possible a number of interesting and arguably correct predictions about the form of assimilation rules, locality of rule application, and the organization of the distinctive feature system. Clements's proposal has given rise to an expanding literature, including Sagey (1986), Schein and Steriade (1986), Archangeli and Pulleyblank (forthcoming), and McCarthy (forthcoming). In this paper, I argue that the tree model as it stands faces a serious empirical shortcoming: it fails to provide an adequate account of diphthongization rules, here defined as rules that convert a segment (vowel or consonant) into a heterogeneous sequence. I propose a revised tree model, which for clarity and explicitness uses coindexation rather than association lines to indicate temporal association. I argue that my proposal solves the diphthongization problem, and that it also makes it possible to restrict the power of segment structure theory in the following way: the "feature- bearing units" (Clements 1980) for any feature are always elements of the prosodic tier, and not nodes in the segment tree.