Surface opacity and phonological issues in Klamath and Lushootseed

Miae Park

Ph.D. Dissertation, 2000

Abstract

In several recent phonological theories and in particular in Optimality Theory, serial derivations are eliminated entirely from underlying-surface mappings in phonological analyses. The overall goals of this dissertation are to provide arguments that serial derivations are required in underlying-surface mappings by examining cases of surface opacity found in two American Indian languages, Klamath and Lushootseed, and to provide an alternative constraint-based Lexical Phonology account in which only a restricted form of phonological derivation is allowed. I show that various leading proposals made to handle surface opacity within the non-serial versions of OT cannot characterize the opacity cases in Klamath that are created by the interaction between phonology and reduplicative morphology and by the interaction of purely phonological rules. I argue that those proposals are inadequate on both theoretical and empirical grounds and that the opacity cases found in this language provide strong evidence for the necessity of serial derivation between underlying and surface representations. I show that in Lushootseed, the diminutive allomorphy, a case of the emergence of the unmarked, provides support for a constraint-based account, whereas the opaque interaction of the diminutive allomorphy with vowel reduction and syncope, which cannot be properly handled within the non-derivational OT framework, provides evidence for a derivational account. I argue that the diminutive allomorphy and its opaque interaction with vowel reduction and syncope in Lushootseed pave the way for a constraint-based derivational account. I reanalyze the opacity cases in Klamath and Lushootseed within the constraint-based Lexical Phonology framework and show that a restricted form of serial derivation that follows from the organization of the grammar provides a sufficient and principled account of the surface opacity found in the two languages, eliminating the problematic B-R correspondence relationship, constraint parameterization and sympathy. The innovations needed in the constraint-based LP framework are (1) the base-derivative faithfulness relation, which holds between the optimal output of a cycle or level and its derivative on a subsequent derivation, and (2) constraint rerankings on a level-by-level basis.