thesis

Text

Ph.D.

360

http://hdl.rutgers.edu/1782.1/rucore00000002165.ETD.000064928

doi:10.7282/T3TQ60BJ

In this dissertation I argue for a general model of assimilation within Optimality Theory, with vowel harmony processes serving as a specific empirical testing ground. The model centers around agreement constraints. I demonstrate that the interaction among agreement constraints and other well-established constraints (particular members of the general markedness and faithfulness constraint families) is sufficient to account for the various vagaries of vowel harmony, and that representational devices such as crucial underspecification and autosegmental feature-sharing are consequently unnecessary. Vowel harmony processes can be either stem-controlled or dominant-recessive. Stem-controlled vowel harmony processes are the more familiar kind, where the harmonic feature value of vowels in the stem determines the harmonic feature value of vowels in subsequent affixes. In a language with a dominant-recessive vowel harmony process, on the other hand, one harmonic feature value is 'dominant' and the other is 'recessive' such that any dominant-valued morpheme vowel, stem or affix, determines the harmonic feature value of all other (otherwise recessive-valued) morpheme vowels. I offer a novel analysis of the distinction between stem-controlled and dominant-recessive vowel harmony processes. I argue that the phenomenon of stem control is due to the relatively higher rank of faithfulness constraints on the correspondence relation between stems and their affixed forms; i.e., that stem-controlled vowel harmony is a cyclic process. Dominance, on the other hand, is due to the relatively higher rank of a constraint that specifically prevents dominant-valued (here understood as 'unmarked') segments from becoming recessive-valued ('marked'). I argue that this type of constraint, a local conjunction of markedness and faithfulness, is independently required in order to explain the fact that a larger number of segments with a particular value of the harmonic feature do not ever "gang up" on a smaller number of segments with the opposite value, a universal fact that has otherwise proven to be difficult to guarantee as a general result. The factorial typology of the constraints relevant to the proposed model is thoroughly investigated, and several challenging examples (Yoruba, Maasai, Turkana and Nez Perce) are given detailed attention and analysis. Taken together, these components of the dissertation confirm the descriptive and explanatory adequacy of the proposed model.

ETD doctoral

The author owns the copyright to this work

ETD

application/pdf

application/x-tar

1423360

bae59d3e4bb48efae4fb17ce0e299ca0272d0b3a