Studies in computational Optimality Theory, with special reference to the phonological system of Malagasy

This dissertation examines Base-Reduplicant Correspondence Theory ( BRCT) and, to a lesser extent, various approaches to phonological opacity via a two-pronged approach.; The first half of this dissertation develops a computational model of Optimality Theory which is capable of encoding BRCT analyses. The model may be taken as an extension or reworking of the weighted finite state models of Ellison 1994b; Eisner 1997c; Albro 1998a, in which the candidate set of Optimality Theory is modeled as a finite state machine and the constraints as weighted finite state machines. Whereas the constraint component of Optimality Theory is still modeled by finite state methods, the candidate set induced by GEN is represented by a higher member of the Chomsky hierarchy, Multiple Context Free Grammars (Seki et al. 1991). Also presented is a simple but complete representation of candidates and constraints that allows for an order-of-magnitude increase in efficiency over earlier representations.; The second half attempts to validate the model, presenting the full-scale phonological analysis of the Merina dialect of Malagasy, the primary language of Madagascar. Malagasy has an extensive system of reduplication and a highly opaque stress system. The analysis shows that Base-Reduplicant Correspondence Theory does not work as a model of Malagasy reduplication. Instead, a combination of Kiparsky's (2000) LPM-OT framework and a morphological doubling account similar to that of Inkelas & Zoll (2000) accounts for the patterns of Malagasy reduplication and opaque stress without requiring any computational power beyond the finite state.