Title

Government relations in Korean phonology

Date of Completion

January 1996

Keywords

Language, Linguistics|Speech Communication

Degree

Ph.D.

Abstract

One goal of linguistic theory is to uncover universal principles which constrain the class of attainable grammars. Recent work in phonology has pursued this goal in terms of the principles and parameters framework. One line of this work is found in Government Phonology (Kaye, Lowenstamm & Vergnaud, 1985 and 1990), where it is proposed that syllables are bound together in terms of government. This thesis explores phonological phenomena in Korean such as tensing, neutralization, voicing, umlaut and palatalization within Government phonology, to test its proposals and extend its coverage to a new range of data.^ In the testing of hypotheses within this theory, the analysis of Korean exposes problems pertaining to the Projection Principle of Government Theory, which requires that government is defined in lexical representations and remains constant throughout a derivation. However, in Korean, government appears to be established during the derivation. It does not hold in lexical representation. I raise the question of whether Government Theory requires the Projection Principle and explore replacing it with a derivational account of government. Reconsidering the analysis of Tigrinya which motivates the Projection Principle, I argue that it does no work there. I then investigate how a derivational account of government handles the Korean phenomena.^ The second issue of this thesis is how we can account for phonological derivations in a principled way within Government Theory. In Harris' (1990) theory of segmental complexity, where the ultimate constituents of segments are unary elements, alternation is given by two operations: composition and decomposition. However, Harris does not explain why one applies rather than the other when either can achieve a government relation, nor can he predict how many internal elements, or which ones, are affected by any operation. To address this issue, I extend lexical representations in the Government framework along the lines of the Markedness Theory approach (Michaels 1989) and use economy (Chomsky 1991, 1993 and Chomsky & Lasnik 1991) to choose between derivations. Markedness Theory accounts for why operations affect certain elements. Economy explains why a specific government relation is established when more than one such relation is possible. ^