Dissimilation, consonant harmony, and surface correspondence

In this dissertation, I argue for a theory of long-distance consonant dissimilation based on Surface Correspondence, correspondence that holds over the different consonants contained in the same output form. Surface Correspondence is posited in previous work on Agreement By Correspondence, which explains long-distance consonant assimilation as agreement driven by similarity (Rose & Walker 2004, Hansson 2001/2010). I demonstrate that dissimilation is a natural outcome of this theory of correspondence, and develop a novel and more formally explicit characterization of the Surface Correspondence relation and the constraints sensitive to it. The consequences of this theory are explored in analyses of dissimilation and agreement patterns in Kinyarwanda, Sundanese, Cuzco Quechua, Obolo, Chol, Ponapean, Zulu, Yidiny, Latin, and Georgian. The Surface Correspondence Theory of Dissimilation (SCTD) posits only constraints that demand surface correspondence, and constraints that limit it. Dissimilation falls out from the interaction of these constraints. Correspondence is only required between consonants that are similar in a specified respect; if they are not similar in the output, they need not correspond. Constraints that disfavor Surface Correspondence therefore favor dissimilation, because dissimilating is a way to avoid penalized surface correspondence structures. This interaction derives long-distance consonant dissimilation without any special mechanism like the OCP or anti-similarity constraints; it also explains certain dissimilation patterns that aren’t accounted for by previous OCP-based theories. The SCTD unites long-distance consonant dissimilation and consonant harmony under the same theory, but does not predict that they are formally identical. Agreement is based on correspondence; dissimilation, on the other hand, is based on non-correspondence – consonants dissimilate instead of corresponding. Surface Correspondence constraints therefore affect dissimilation in different ways than harmony: limiting correspondence limits agreement, but favors dissimilation. The resulting prediction is that harmony and dissimilation are related in a consistently mismatched way, and not in the matching way predicted by previous theories that link them together (MacEachern 1999, Nevins 2004, Mackenzie 2009, Gallagher 2010, a.o.). This outcome of the SCTD is empirically supported: a survey of over 130 languages shows that the typology of long-distance consonant dissimilation indeed does not match the typology of consonant harmony.Ph. D.Includes bibliographical referencesby William G. Bennet